Patent Application
20220237996
This disclosure generally relates to video-based data collection systems, and more specifically to an image, video, and sensor data capture, storage, transmission, and analysis for detection and reporting of high-priority events.
With the wide adoption of smartphones and our ubiquitous connectivity to the Internet and social networks, software apps and cameras have become common place in our daily lives for personal applications. We take pictures and videos with our smartphones of all sorts of events, items, and situations, and easily upload to cloud services and share them with friends, family, and other people who subscribe or follow our shared content.
Many products and services also exist in the smart home or automated home market segment. Security cameras around the home or business place are widely used that record either constantly or with event-based triggers, like motion sensors, and store the recorded video locally on video servers or upload the video to cloud services, either via wired connections through a home router or using Wi-Fi to connect to a home network. The recorded video is typically available for the user for a period of time and accessible in real time from software apps in smartphones or via websites. Multi-camera systems store video feeds from various cameras around the home and make the various feeds available to the user through a common user interface.
Some services provide the ability to share these videos with other users, not only via social networks, but also based on other factors. For example, Bot Home Automation, Inc. of Santa Monica, Calif., provides camera-equipped doorbell systems called Ring. Customers get access to the video from the Ring cameras via a website, ring.com. One feature of the Ring system is called “Ring Neighborhoods” (described at https://ring.com/neighborhoods). A user can set a radius around the user's home equipped with Ring cameras and automatically get notified when other users within that radius share videos in the Ring platform. Users can share any video they find may be interesting for other users in the neighborhood. However, this system requires the users to review all their video to find potentially interesting video and then upload it to share it with other Ring users within a predefined distance.
Another area where cameras are being used is in vehicles. Safety cameras for backing up or side view cameras are becoming common-place. For commercial vehicles, like taxis or other vehicle fleets, security camera systems record video from both inside and outside the vehicle for safety and management purposes. For example, Safety Track of Belleville, Mich., provides a 2-channel dash camera system equipped with a 3G/4G cellular dongle that connects to the camera system via USB for streaming video from the vehicle in real time (described at http://www.safetytrack.net/duel-lens-in-vehicle-fleet-camera-system/). However, these in-vehicle systems are not simple to install for an average consumer and lack automatic detection and reporting of high-priority events, such as for example, burglaries, vehicle break-ins, and the like. For example, if a burglary or vehicular break-in takes place and the perpetrator removes or disables the camera upon breaking in, any footage of the event may be lost and may be difficult to determine if a break in is taking place or if connectivity to the dash camera was lost for some other reason. Thus, what is needed is a video collection and sharing platform that addresses the deficiencies of the prior art.
According to various embodiments of the present invention, a video data collection and sharing platform is provided capable of detecting high-priority events and communicating alerts to users. In one embodiment, a method and system is provided for notifying a user of a high-priority event captured by a video capturing device with one or more sensor modules, including one or more video cameras for recording video footage, and a communications module for establishing a network connection. According to this embodiment, a network connection is established and input from one or more sensor modules is received. The input includes image data from the one or more video cameras. An input from the one or more sensor modules is analyzed to determine a potential high-priority event. Upon determining a potential high-priority event, conditions of the network connection are measured and an amount of data to be transmitted to a remote server through the network connection is determined based, at least in part, on the measured conditions of the network connection. The determined amount of data is then transmitted to the remote server. The transmitted data includes at least part of the image data from the one or more video cameras. Inputs from the one or more sensors is analyzed. Based at least in part on the first input and the second inputs, a positive determination regarding the occurrence of the high-priority event is made. The video data input from the one or more video cameras is used to generate a video clip and the user is notified of the positive determination regarding the occurrence of the high-priority event.
In one embodiment, the high-priority event relates to criminal activity, including for example, a burglary, a break-in, or a theft. The criminal activity may, for example, involve a vehicle monitored by the video capturing device.
According to another embodiment, analyzing the first input comprises comparing a digital sound signal from the one or more sensors to a pre-stored digital glass-breaking sound signature. In one embodiment, analyzing the first input includes analyzing a power spectral density of an accelerometer sensor signal to determine whether energy in a low frequency range exceeds a threshold. For example, the low frequency range may include frequencies between 1 and 10 Hertz.
According to another embodiment, a sensitivity level corresponding to a time limit for notifying the user is also determined. In this embodiment, analyzing second inputs, making a positive determination, generating the video clip, and notifying the user is done substantially within the time limit. For example, the second inputs analyzed may depend on the sensitivity level determined. These inputs may include, for example, a car start signal, a Bluetooth signal, a face recognition signal, or a GPS signal.
According to another embodiment, additional inputs from remote sources can also be analyzed. For example, inputs from determination of events by other video capturing devices located in the vicinity of the video capturing device may be analyzed. In this embodiment, the positive determination regarding the occurrence of the high-priority event can also be based, at least in part, on these additional inputs.
The figures depict various example embodiments of the present disclosure for purposes of illustration only. One of ordinary skill in the art will readily recognize form the following discussion that other example embodiments based on alternative structures and methods may be implemented without departing from the principles of this disclosure and which are encompassed within the scope of this disclosure.
The Figures and the following description describe certain embodiments by way of illustration only. One of ordinary skill in the art will readily recognize from the following description that alternative embodiments of the structures and methods illustrated herein may be employed without departing from the principles described herein. Reference will now be made in detail to several embodiments, examples of which are illustrated in the accompanying figures.
The above and other needs are met by the disclosed methods, a non-transitory computer-readable storage medium storing executable code, and systems for collecting and processing video data and for notifying a user of a high-priority events captured by a video capturing device with one or more sensor modules.
Referring now to
Client device 101 is connected to cloud-based system 103 via connection 107. In one embodiment, connection 107 is a cellular-based wireless packet data connection, such as a 3G, 4G, LTE, 5G, or similar connection. Connections 108a-108c between other system components and cloud-based system 103 are Internet-based connections, either wired or wireless. For example, in one embodiment, mobile device 104 may at different times connect to cloud-based system 103 via Wi-Fi (i.e., any IEEE 802.11-based connection or similar technology) and cellular data (e.g., using 4G, LTE, or the like). In one embodiment, Web-based system 105 is connected to cloud-based system 103 over the World Wide Web using a wired Internet connection, such as DSL, cable modem, or the like. Similarly, in one embodiment, auxiliary camera module 106 is connected to cloud-based system 103 via a Wi-Fi connection to a home router connected to the Internet via cable modem, DSL, or the like. Any combination of available connections can be used to connect any of the system components to cloud-based system 103 via the Internet or similar networks.
Referring now to
The client device 101 in this exemplary embodiment includes a location module 204, a wireless transceiver module 205, an audio I/O module 206, a video module 207, a touchscreen module 208, a sensor module 209, and an I/O module 216. In this embodiment, the different modules are implemented in hardware and software modules. In alternative embodiments, these modules can be hardware, software, or a combination of both. For example, alternative embodiments may be provided with one or more central processor (“CPU”) cores on an SoC also including a wireless modem, multimedia processor, security and optionally other signal co-processors, such as for example, one or more graphics processor unit (“GPU”) cores, one or more holographic processing unit (“HPU”) cores, and/or one or more vision processing units (“VPU”). In one embodiment, one or more SoC processors used to embody the invention may encompass CPUs, GPUs, VPUs, HPUs, and other co-processors, motherboard buses, memory controllers, screen controllers, sound chipsets, camera modules, on-board memory, and several peripheral devices, including for example cellular, Wi-Fi, and Bluetooth transceivers, as further described below. Alternative embodiments include modules as discrete components on a circuit board interconnected by bus 202 or a combination of discrete components and one or more SoC modules with at least some of the functional modules built-in.
In one embodiment, location module 204 may include one or more satellite receivers to receive and decode signals from location satellite systems, such as Global Positioning System (“GPS”), Global Navigation Satellite System (“GLONASS”), and/or BeiDou satellite systems. In one embodiment, location module 204 is a Qualcomm QTR2965 or Qualcomm QGR7640 receiver that connects to a GPS antenna for receiving GPS satellite signals and providing geographical coordinates (latitude and longitude) of the location of the client device 101. The wireless transceiver module 205 includes a cellular modem, e.g., compliant with 3G/UMTS, 4G/LTE, 5G or similar wireless cellular standards, a Wi-Fi transceiver, e.g., compliant with IEEE 802.11 standards or similar wireless local area networking standards, and a Bluetooth transceiver, e.g., compliant with the IEEE 802.15 standards or similar short-range wireless communication standards. In one embodiment, the wireless transceiver module 205 is a Sierra Wireless HL-7588.
The audio I/O module 206 includes an audio codec chipset with one or more analog and/or digital audio input and output ports and one or more digital-to-analog converters and analog-to-digital converters and may include one or more filters, sample rate converters, mixers, multiplexers, and the like. For example, in one embodiment, a Qualcomm WCD9326 chipset is used, but alternative audio codecs may be used. In one embodiment, video module 207 includes a DSP core for video image processing with video accelerator hardware for processing various video compression formats and standards, including for example, MPEG-2, MPEG-4, H.264, H.265, and the like. In one embodiment, video module 207 is integrated into an SoC “multimedia processor” along with processor 201. For example, in one embodiment, client device 101 includes an integrated GPU inside the Qualcomm MSM8953 but alternative embodiments may include different implementations of video module 207.
In one embodiment, the touchscreen module 208, is a low-power touchscreen sensor integrated circuit with a capacitive touchscreen controller as is known in the art. Other embodiments may implement touchscreen module 208 with different components, such single touch sensors, multi-touch sensors, capacitive sensors, resistive sensors, and the like. In one embodiment, the touchscreen module 208 includes an LCD controller for controlling video output to the client device's LCD screen. LCD controller may be integrated into a touchscreen module 208 or, in alternative embodiments, be provided as part of video module 207, as a separate module on its own, or distributed among various other modules.
In one embodiment, sensor module 209 includes controllers for multiple hardware and/or software-based sensors, including, accelerometers, gyroscopes, magnetometers, light sensors, gravity sensors, geomagnetic field sensors, linear acceleration sensors, rotation vector sensors, significant motion sensors, step counter sensors, step detector sensors, and the like. For example, in one embodiment, sensor module 209 is and Invensense ICM-20608. Alternative implementations of sensor module 209 may be provided in different embodiments. For example, in one embodiment, sensor module 209 is an integrated motion sensor MEMS device that includes one or more multi-axis accelerometers and one or more multi-axis gyroscopes. Client device 101 may also include one or more I/O modules 210. In one embodiment, I/O module 210 includes a Universal Serial Bus (USB) controller, a Controller Area Network (CAN bus) and/or a LIN (Local Interconnect Network) controller.
In one embodiment, client device 101 also includes a touchscreen 211. In alternative embodiments, other user input devices (not shown) may be used, such a keyboard, mouse, stylus, or the like. Touchscreen 211 may be a capacitive touch array controlled by touchscreen module 208 to receive touch input from a user. Other touchscreen technology may be used in alternative embodiments of touchscreen 211, such as for example, force sensing touch screens, resistive touchscreens, electric-field tomography touch sensors, radio-frequency (RF) touch sensors, or the like. In addition, user input may be received through one or more microphones 212. In one embodiment, microphone 212 is a digital microphone connected to audio module 206 to receive user spoken input, such as user instructions or commands. Microphone 212 may also be used for other functions, such as user communications, audio component of video recordings, or the like. Client device may also include one or more audio output devices 213, such as speakers or speaker arrays. In alternative embodiments, audio output devices 213 may include other components, such as an automotive speaker system, headphones, stand-alone “smart” speakers, or the like.
Client device 101 can also include one or more cameras 214, one or more sensors 215, and a screen 216. In one embodiment, client device 101 includes two cameras 214a and 214b. Each camera 214 is a high definition CMOS-based imaging sensor camera capable of recording video one or more video modes, including for example high-definition formats, such as 1440p, 1080p, 720p, and/or ultra-high-definition formats, such as 2K (e.g., 2048×1080 or similar), 4K or 2160p, 2540p, 4000p, 8K or 4320p, or similar video modes. Cameras 214 record video using variable frame rates, such for example, frame rates between 1 and 300 frames per second. For example, in one embodiment cameras 214a and 214b are Omnivision OV-4688 cameras.
Alternative cameras 214 may be provided in different embodiments capable of recording video in any combinations of these and other video modes. For example, other CMOS sensors or CCD image sensors may be used. Cameras 214 are controlled by video module 207 to record video input as further described below. A single client device 101 may include multiple cameras to cover different views and angles. For example, in a vehicle-based system, client device 101 may include a front camera, side cameras, back cameras, inside cameras, etc.
Client device 101 can include one or more sensors 215. For example, sensors 215 may include one or more hardware and/or software-based sensors, including, accelerometers, gyroscopes, magnetometers, light sensors, gravity sensors, geomagnetic field sensors, linear acceleration sensors, rotation vector sensors, significant motion sensors, step counter sensors, step detector sensors, and the like. In one embodiment, client device 101 includes an accelerometer 215a, gyroscope 215b, and light sensor 215c.
Referring back to
A mobile app on mobile device 101 provides a user interface to a user account on cloud system 103 and to client device 101. In one embodiment, mobile app includes functionality similar to auxiliary camera 106. For example, mobile app uses one or more cameras on mobile device 104 to record video events in accordance to one embodiment of the disclosure. The video recording, buffer management, and other methods and techniques described herein may be also incorporated into mobile app in one or more embodiments of the invention.
Now referring to
In one embodiment, real time camera feeds are provided to the mobile app with the same approach used for providing video clips based on a playlist file or manifest file as further described below. For real-time feeds, the playlist files are dynamically updated to include each newly generated video data object or file captured by the relevant camera. For each new video file, the file location is provided in the updated playlist and the playlist file is updated via the cloud system 103 or directly from the source of the video feed. For example, in one embodiment, playlist files for streaming video are dynamically updated as described in the HTTP Live Streaming specification (as for example described in Internet Draft draft-pantos-http-live-streaming-23 submitted by Apple, Inc. to IETF on May 22, 2017) incorporated herein by reference in its entirety. Alternative streaming techniques may be used in other embodiments, including, for example, MPEG-DASH (ISO/IEC 23009-1), Adobe's HTTP Dynamic Streaming, Microsoft's Smooth Streaming, or the like.
In one embodiment, camera pane 401b includes camera control elements 411. For example, a recording or manual tagging control element 411a is provided for the user to instruct the currently selected camera to generate a clip for the currently displayed video (as further described below). For example, if a user is involved in a video-clip-generating event, e.g., car accident, police stop, break-in, or the like, in addition to the any video clips generated through client device 101, either manually or automatically, mobile device 104 can also be used to generate additional video clips for the given event from a different angle or perspective. Further, in one embodiment, any time the mobile app is running on the mobile device 104, one or more cameras on the mobile device 104 are recording video data and manual tagging control element 411a is used to generate a manually-tagged video clip as further described below. Thus, mobile device 104 can be used as client device 101 or auxiliary camera device 106 according to this embodiment.
In one embodiment, camera pane 401b may also include additional control elements 411, such as, buttons, icons, or other selection elements or menus, to access non-live video stored in the buffer of the currently selected camera. For example, a user may remotely access an entire set of video data objects or files stored in the buffer of the user's client device 101 (e.g., video files for the preceding 24 hours) through user control elements 411. In one embodiment, based on the user input selecting a point in time from which to begin streaming buffered video, the source camera device (e.g., client 101, auxiliary camera 106, or other camera device) generates a dynamic playlist or manifest file including the video files for the next preset time period, for example, one minute, and it is progressively and dynamically updated in increments of same amount of time (e.g., every minute) with the next set of video files. The playlist or manifest files are generated as further described below with reference to video clip generation methods.
Now referring to
As noted above, the features described above with respect to the mobile app may also be provided via Web-based system 105 using conventional website programming techniques to implement the functionality described for the mobile app.
Referring back to
Preferably, client device 101 includes two or more cameras 214. For example, in one embodiment, a first “IN” camera 214a is directed at the inside of a vehicle, i.e., the cabin, driver, and passengers, and a second “OUT” camera 214b is directed at the road in front of the vehicle. In alternative embodiments, additional cameras 214 may be used, for example facing the back and/or sides of the vehicle, multiple interior areas of the vehicle, one or more top camera with a wide-angle lens providing a 360° view around the vehicle, or the like.
According to one embodiment, client device 101 is always turned on as long as it has sufficient power to operate. Cameras 214a and 214b are always turned on and recording video. The video recorded by the cameras 214 is buffered in the memory device 203. In one embodiment, memory device 203 is configured as a circular buffer. For example, in one embodiment, memory device 203 may be a 32 Gb FLASH memory device. Client device 101 manages the buffer in memory device 203 to store video data for a predetermined and programmable set amount of time. For example, in one embodiment, memory device 203 buffers video data from two cameras 214a and 214b for the preceding 24 hours.
In one embodiment, client device 101 includes software to manage the cameras 214 to control the amount of data, e.g., bytes, generated by the cameras 214 and buffered in memory 203. In one embodiment, cameras 214 record data at various selectable video modes and rates. For example, cameras 214a and 214b can be set by client device 101 to capture video at various resolutions, including for example 1440p, 1080p, 720p, 360p, 240p, and the like. In addition, the frame rate for the video collected by each camera 214 can be set by client device 201. For example, in one embodiment, each camera 214 can independently change its video capture rate from 0 to 30 frames per second.
Now referring to
Based on the inputs received, an operational mode is determined 503. For example, the possible operational modes of a vehicle incorporating client device 101 according to one embodiment may include: default, driving, recently parked, parked, armed, low battery, and very low battery. Different embodiments can provide a subset or additional modes of operation, which may also vary depending on the vehicle or other location where the client device 101 (or auxiliary camera) may be located. A status change is determined at step 504. For example, after powering up, input data is received and the operational mode is no longer in “Default” mode. Based on the determined operational mode, the camera settings (e.g., resolution and frame rate) are changed 505 to produce more or less data for the video being recorded. Once the camera settings have been changed, recording of the video is done 506 using the camera settings. This results in video data objects, records, or files of varying size to manage the buffer, storing higher quality data with more bits during operational modes with higher likelihood of capturing video for events of interest while using lower quality data with less bits during operational modes with lower likelihood of capturing video of interest.
In an alternative embodiment, as illustrated in
According to another aspect of one embodiment, the buffer management methodology used in client device 101 will optimize the memory available for buffering to ensure that video data is not stored on the memory device for longer than a preset, programmable amount time. For example, if the buffering time is set to 24 hours, client device 101 may change the camera settings to change the size of the video data objects or files to ensure that “stale” video data is written over by new video data as the 24-hour limit approaches. For example, in one embodiment, even if the vehicle operational mode is determined to be “Parked,” processor 201 may over-write the mode to the camera settings associated with the “Driving” mode to ensure that older video data is written over in the circular buffer. In the case where even when using the highest quality video and maximum frame rate available some of the video data in the buffer remains after 24 hours, the system deletes the video data.
According to another aspect of the invention, in one embodiment, the buffer management methodology further includes a learning function to further optimize the storage of video data in the device's buffer memory. According to one embodiment, the camera device monitors the use of the camera device and creates history of use data for further application to buffer management algorithms. For example, in one embodiment, the times when each mode is activated and for how long each mode is activated is recorded. The buffer management methodology then uses the mode history information to optimize use of the buffer and/or to avoid buffer overrun. For example, the percentage of the buffer used within the current 24-hour timeframe and the expected use for remaining time based on history information is considered at the camera settings changing step 505 to reduce or increase camera quality settings for a given mode. For example, after determining that the Driving mode should be set at the 20th hour of a 24-hour period, the method further checks the percent usage of the buffer and determines to have excess capacity, e.g., at 50% when historically it would be at 80%, and determines based on historical use data that for the next 4 hours of the 24-hour period it is expected to use 20% of the buffer. Since the buffer is being underutilized, the method increases the quality of video data for the Driving mode, for example, to 1440p/30 fps for both cameras.
In another embodiment, a vehicle-mounted client device 101 includes a learning algorithm for buffer management that learns the user's typical schedule for driving and corresponding modes (morning commute, parked until noon, lunch “drive”, after noon “parked”, etc.) and considers the expected use of the buffer at each given time. In this embodiment, if one day there are some unusual events causing modes that require higher quality camera settings earlier in the 24-hour period, later in the day the camera settings of lower quality settings modes, e.g., Parked mode, can be further reduced to lower resolution and frame rate than the normal settings for that mode. Alternatively, direct user input may also be provided to indicate a change in the typical operation schedule. For example, the user may indicate the use of the system for an extended road trip and the user input is used to override the expected schedule for that time frame.
According to another aspect of the invention, in one embodiment, the buffer usage history data learned by the system is further input to the operational mode determination step 503. In this embodiment, a weighting function is used to determine a probability for each operating mode based on the strength of the combination of inputs. For example, if the GPS input indicates no motion but the CAN bus input indicates some RPM, the confidence of the motion component for the mode determination is lower than if both the GPS and the CAN bus inputs both indicate no motion. Similarly, a face recognition positive input would increase the probability of the mode being “Driving Mode.” Optionally, the confidence level of any image recognition input is also use as a weighting factor for the mode determination. For example, the confidence or likelihood of a positive image recognition match (e.g., the likelihood of a positive recognition of a face, the user's face, a body, a uniform, flashing lights, etc.) is used as a multiplier to the contribution of the match to the mode determination. The determination of the operating mode is set if the various probabilities from the multiple inputs exceed a threshold. In one embodiment, the mode probability thresholds are changed based on historical buffer usage data. For example, if the buffer storage is above the expected usage level for a given time within the buffer storage period (e.g., 24 hours), a higher threshold is used to determine a mode that uses higher definition/frame rate. Conversely, if the buffer storage is underutilized based on the historical use data, the mode threshold for the same modes can be reduced.
Now referring to
In one embodiment, the video data is encrypted 605. Any encryption algorithm may be used, such as, for example encryption algorithms compliant with the Advanced Encryption Standard (AES), Blowfish, Twofish, Data Encryption Standard (DES) (e.g., Triple-DES), RSA, or the like. Preferably, the video data is encrypted 605 based on a user-specific encryption key. In a vehicle-based embodiment, an encryption key is provided for each driver authorized to drive the vehicle. For example, in this embodiment, a biometric input from the driver is required to operate the system, such as, a fingerprint recognition, a voice recognition, or a face recognition based identification is used to identify the authorized driver. For each authorized driver, a corresponding randomly generated encryption key is maintained in a data table. Any video generated while the authorized driver is determined to be driving the vehicle is encrypted 605 using the driver-specific key. Subsequently, in order to provide privacy, only the authorized driver can provide access the encrypted video using biometric identification.
In another embodiment, video encryption 605 is based on other forms of user identification. For example, in one embodiment, a mobile device unique ID (GUID), such as for example the Bluetooth ID, MAC address, system-assigned ID, or the like, associated with the mobile device 104 of an authorized user is used for identification. In one embodiment, for example, a client device 101 may display the picture or pictures of the users for which the client device 101 has recognized the presence of their associated GUIDs. For example, according to one embodiment, the client device 101/106 implements a Bluetooth beacon protocol (e.g., iBeacon, AltBeacon, URIBeacon, Eddystone, or the like) using Bluetooth Low Energy (“BLE”) to determine proximity of mobile devices 104. Any mobile device 104 with a cloud system enabled app, as for example described above with reference to
The recognized user who is driving can select his or her picture on the screen on the client device 101 and the corresponding encryption key is used for encrypting video. Alternative approaches for selecting the encryption key may be used in other embodiments. For example, a hierarchical level of authorized users may be provided, such as, an owner level versus a guest level or a parent level versus a child level, such that the encryption key for the highest level of authorized user recognized is used to encrypt the video in situations where multiple authorized users are detected. Alternatively, in some embodiments, the encryption 605 may not be user-based. For example, the encryption key may be a random key that is unique to each device. Moreover, is some embodiments the system may record video in un-encrypted form omitting step 605.
According to another aspect of the invention, several other privacy measures may optionally be provided for passengers of a vehicle with a camera device in one embodiment. For example, for ride-sharing applications, customer/passengers may want to protect their privacy from information capture by client device 101. In this embodiment, a ride-sharing mobile device app provides privacy features customizable by the user. Upon detection of the user/customer in the vehicle, client device 101 retrieves privacy settings for the detected passenger and applies them accordingly. For example, using face recognition, Bluetooth ID, or other means of recognizing the passenger, ride-sharing passengers' preferences may be applied on client device 101, such as turning certain cameras on or off, blurring video or parts of the video (e.g., faces), storing more or less of the sensor data collected, and/or enabling or disabling other features of the client device 101. In one embodiment, customers' qualifications may be required to provide access to customizable preferences, which may be accessible in different tiers, for example based on continued usage of the ride-sharing service (e.g., loyalty points/levels), payment levels, or the like.
Referring back to the method of
In one embodiment, a message is generated 608 including the metadata for each time period and the corresponding video data hash. Preferably, the message is then cryptographically signed 609 to guarantee the message payload originates from an authorized device. For example, a private key associated with a system-authorized device may be used to generate a one-way hash of the message payload. In an alternative embodiment, the private key is used to encrypt the payload of the message. In one embodiment, each client device 101, auxiliary camera 106, and mobile device 104, is associated with a unique cryptographic key-pair. The device securely stores the private key. The cloud system 103 retains access to the public keys for each device so it can verify that messages it receives come from authorized devices. For example, cloud system 103 maintains a set of records uniquely associating a device ID for each authorized device in the system with a corresponding public key that is applied to messages received from the device. For example, private-public-key cryptographic signature methodologies may be used to verify that each received message includes a signature or encrypted payload encrypted with a private key from an authorized device.
In yet another embodiment, at step 607, optionally, instead of hashing the video data object, the client device uses its private cryptographic key to cryptographically sign or otherwise encrypt the video data object itself, for example, if the actual video data object is to be sent or otherwise uploaded to another device, such as cloud system 103. This could optionally be done in conjunction with step 609 as described above.
Finally, the message is sent 610 to the cloud system. Preferably, the message is sent using a secured connection, such as for example, an SSL/HTTPS connection over TCP/IP or the like. The process then repeats for the video data and metadata captured in the subsequent time period. Preferably, the time required to perform the process of
In an alternative embodiment, the message signing step 609 is omitted. Instead, a device establishes a secured connection with the cloud system 103, such as an SSL/HTTPS connection, and authenticates itself to the server 102. For example, a device provides its device ID and a cryptographically signed version of its device ID, signed with the device's private key. The server 102 retrieves the public key corresponding to the device ID provided and verifies the signed device ID for a match. Upon authorization, the server provides the device with a session token that uniquely identifies communications from that device for a given session. Thereafter messages are sent 610 over the secured connection with the metadata and video hash and also including the server-provided token.
Now referring to
Now referring to
Now referring to
The selected video data is marked for buffering 704 for a longer period of time. For example, the video files for the selected time period are copied over to a second system buffer with a different buffering policy that retains the video for a longer period of time. In one embodiment, the selected video data being in a buffer storing video for 24 hours is moved over to a second buffer storing video for 72 hours.
Referring back to
In one embodiment, video data objects are stored on the network-accessible buffer of the camera device and the playlist or manifest files for the generated event-based video clips identify the network addresses for the memory buffer memory locations storing the video data objects or files. Alternatively, upon identifying and selecting 703 the relevant video data objects, in addition to or as an alternative to moving the video data to the longer buffer 704, the video data may be uploaded to the cloud system 103. The clip generation 705 then identifies in the playlist or manifest file the network addresses for the video data stored in the cloud system 103. A combination of these approaches may be used depending on storage capacity and network capabilities for the camera devices used in the system or according to other design choices of the various possible implementations.
In one embodiment, other system components, such as the cloud system 103 or mobile device 104, are notified 706 of the event or event-based video clip. For example, in one embodiment a message including the GUID for the generated video clip is sent to the cloud system in a cryptographically signed message (as discussed above). Optionally, the playlist or manifest file may also be sent in the message. In one embodiment, the playlist or manifest files are maintained in the local memory of the camera device until requested. For example, upon notification 706 of the clip generation, the cloud system may request the clip playlist or manifest file. Optionally, the cloud system may notify 706 other system components and/or other users of the clip and other system components or users may request the clip either from the cloud system 103 or directly from the camera device. For example, the clips pane 401a in the user's mobile app may display the clip information upon receiving the notification 706. Given that the clip metadata is not a large amount of data, e.g., a few kilobytes, the user app can be notified almost instantaneously after the tag event is generated. The larger amount of data associated with the video data for the clip can be transferred later, for example, via the cloud system or directly to the mobile device. For example, upon detection of a “Baby/Animal in Parked Car” event or a “Location Discontinuity” event, the user's mobile device 104 may be immediately notified of the tag event using only tag metadata. Subsequently, the user can use the video clip playlist to access the video data stored remotely, for example, for verification purposes.
Once a video clip is generated 705, it may be shared with other devices owned by the same user or, if authorized, the video clip may be shared with other users of the system. For example, the GUIDs for every video clip generated by a camera device of a given user may be stored in a user clip table in the cloud system 103. For example, GUIDs for the clips from all the cameras on a multi-camera client device 101, for the clips from any auxiliary camera device 106, and for the clips generated by the mobile app on the user's mobile device 104, may all be stored in the user clip table. The user may access the user clip table via mobile device 104. For example, mobile app may maintain a user clip table that is synchronized with the user clip table in the cloud system. Every time a new clip notification is received, the mobile app and cloud-based user clip tables are updated and or synchronized. Alternative synchronization approaches may be used, such as for example a periodic synchronization approach.
In addition to the GUID, in one embodiment, the user clip tables may also include other information or metadata for each clip of the user, such as for example, a name or descriptor, device ID where the video was captured, time and date information, tag or event information, thumbprint images, or the like. Further, the playlist or manifest file may also be stored or identified in the user clip table. In one embodiment, a user may access video clips through the mobile app on the mobile device 104 through the clip pane 401a. Upon selection of a clip through the clip pane 401a, the mobile app uses the clip GUID to request the corresponding playlist or manifest file from the cloud system 103, directly from a camera device (e.g., client device 101 or auxiliary camera 106). Using the playlist or manifest file, the mobile app can playback the video clip by requesting the relevant video objects using their network address identifiers. In one embodiment, if the video data objects are encrypted, the user may provide an identification (e.g., biometric ID, face recognition, user ID and password, or the like) to access the decryption key as further discussed above.
In one embodiment, for example, HTTP Live Stream (“HLS”) playlists are used. Each playlist is a combination of the 2-6 second clips, back to back. To be able to include multiple different files back-to-back in the same playlist, in one embodiment, the #tag, typically used to insert advertisements in a playlist, may be placed in front of the name/location of each video object file in the buffer, essentially designating every video object to be an advertisement. While HLS playlists typically provide different streams from a single file, using the #tag approach allows the streaming of the same video feed from separate but continuous files (of a few seconds each). This approach allows for continues smooth playback of high-quality video and provides high performance seeking through the entire buffer of video objects stored in the client device 101/106. Further, since the video objects are already stored in few-second increments (e.g., 2, 4, 6, etc.) the video objects in the client device buffer require no additional processing, avoiding any additional delay. In addition, in one embodiment, the HLS playlists are nested to allow for access to longer time periods worth of video. In this embodiment, the client device 101/106 generates playlists of playlists, to avoid overly long playlists. When scrubbing from the mobile app through time, the mobile app switches which HLS playlist is in use at any given time. Each playlist may include, for example, a few minutes to a few hours of video from the video object files in the buffer. In one embodiment, each playlist includes files for 10-20 minutes of video. This enables the user to see long periods of time worth of video, possibly multiple days, and scrub instantly across the entire time. The same approach may be used for access event-generated clips or other video accessed using playlists described herein.
According to another aspect of one embodiment, video clips generated by devices registered under the same user account are automatically shared with the user. According to another aspect of the disclosure, a user may also share video clips with other users through the system or using other Internet messaging or connectivity. For example, in one embodiment, mobile app on mobile device 104 (or website on Web-based system 105) includes functionality to link or publish video clips to social media sites or networks, such as Facebook, Twitter, Google Plus, or other sites from social networking platforms. Video clips can also be shared directly via email, text messaging, or similar electronic communication approaches, or otherwise exported to other computer systems. In one embodiment, cloud system 103 stores video data for a given video clip in a publicly accessible network storage location. Cloud system 103 may be accessed via the Internet by anyone with an event-based video clip playlist or manifest file as is known in the art. A user may share the playlist or manifest file, either directly or via a network link, such as a URL, to the playlist or manifest file stored on an Internet-accessible storage location, for example, on cloud system 103 or any other similar location.
According to another aspect of the disclosure, video clips may also be shared automatically with other users of the system. For example, upon joining the system, the user may be presented with a number of options to pre-authorize the sharing of the user's video clips with other users of the system. In one embodiment, users have the option to pre-authorize access to video clips generated by certain camera devices. For example, the user may authorize the sharing of video clips generated with video data captured by an “OUT” camera on a vehicle-based system. Optionally, the user may impose restrictions on the video clips that are shared with other users. For example, a user may only allow sharing of video clips of a certain video quality, with or without sound, or the like. For example, a user may authorize the sharing of video clips from an “IN” camera in a vehicle-based system but without any audio. Optionally, another option for pre-authorization of access to a user's video clips may be based on location. For example, the user may pre-authorize access to video clips generated by a “home” auxiliary camera 106 to other users registered in locations within a pre-defined radius, e.g., neighbors.
The location of camera devices that are part of the system can be identified by IP address lookup, GPS location (e.g., from a smartphone device, client device, or the like) or my manual entry of location. Any time a new user joins, the location of the user (e.g., a home address, preferred location, or the like) is used to determine nearby existing users. For example, in one embodiment, the distance between every pair of users is calculated and maintained in a database and a pre-defined radius or distance limit is applied to designate which users are “nearby” with respect to other users, for example by adding a flag to user's whose pairwise distances are below the pre-defined radius. In one embodiment, during the sign-in process, the system sends a consent request to existing users to share with the new users. Alternatively, in another embodiment, upon signing on to the system, every user pre-authorizes the sharing of at least some camera-specific video with “neighbors” or “nearby” users. Additionally, the user may be allowed to provide additional restrictions with respect to the video clips that may be shared with neighbors.
According to another aspect of the video clip sharing functionality, users may mark individual video clip with a sharing designation. In one embodiment, this sharing designation would overwrite any other pre-authorization, such that a user would have control of which video clips may be shared and which ones may not. Additional techniques for sharing video clips are further discussed below, such as for example, accessing of shared neighbors' video via Web-based system 105 or mobile device 104.
According to another aspect of the disclosure, detection of tagging events 702 may be done automatically by the system. For example, based on the monitored inputs, in different embodiments events such as a vehicle crash, a police stop, or a break in, may be automatically determined. The monitored inputs 701 may include, for example, image processing signals, sound processing signals, sensor processing signals, speech processing signals, in any combination. In one embodiment, image processing signals includes face recognition algorithms, body recognition algorithms, and/or object/pattern detection algorithms applied to the video data from one or more cameras. For example, the face of the user may be recognized being inside a vehicle. As another example, flashing lights from police, fire, or other emergency vehicles may be detected in the video data. Another image processing algorithm detects the presence of human faces (but not of a recognized user), human bodies, or uniformed personnel in the video data.
Similarly, sound processing signals may be based on audio recorded by one or more microphones 212 in a camera device, (e.g., client device 101, auxiliary camera 106, or mobile device 104). In one embodiment sound processing may be based on analysis of sound patterns or signatures of audio clips transformed to the frequency domain. For example, upon detection of a sound above a minimum threshold level (e.g., a preset number of decibels), the relevant sound signal is recorded and a Fast Fourier Transform (FFT) is performed on the recorded time-domain audio signal as is known in the art. The frequency-domain signature of the recorded audio signal is then compared to known frequency domain signatures for recognized events, such as, glass breaking, police sirens, etc. to determine if there is a match. For example, in one embodiment, pairs of points in the frequency domain signature of the recorded audio input are determined and the ratio between the selected points are compared to the ratios between similar points in the audio signatures of recognized audio events.
Sound processing may also include speech recognition and natural language processing to recognize human speech, words, and/or commands. For example, certain “trigger” words may be associated with particular events. When the “trigger” word is found present in the audio data, the corresponding event may be determined. Similarly, the outputs of the available sensors may be received and processed to determine presence of patterns associated with events. For example, GPS signals, accelerator signals, gyroscope signals, magnetometer signals, and the like may be received and analyzed to detect the presence of events. In one embodiment, additional data received via wireless module 205, such as traffic information, weather information, police reports, or the like, is also used in the detection process. The detection process 702 applies algorithms and heuristics that associate combinations of all these potential inputs with possible events.
The following are exemplary inputs, rules, and heuristics for corresponding events according to one embodiment of the invention in a vehicle implementation. While a set of specific examples is provided, it is understood that the present invention can be applied to a wide variety of inputs, rules, and heuristics that can identify other possible events, depending on the application. For example, a break-in or burglary event may be based on first and second sensor inputs, including Sound “IN” camera above threshold and close match to glass breaking sound signature; GPS—no motion; accelerometer—small vibrations; CAN bus—engine off; IN camera—object motion detected or unrecognized face detected; No authorized user Bluetooth ID. For a “Police Stop” event, sensor input may include any of the following: GPS—location in a freeway or highway—stop after speed above posted limit; traffic data—no slowdown/heavy traffic reported; accelerometer—small vibrations; CAN bus—low RPM (idle) or off; IN camera—flashing lights detected/police vehicle detected/uniformed personnel detected; OUT camera—road shoulder detected/police vehicle detected/uniformed personnel detected; Sound—sirens detected. An “Accident” or car crash event may be detected based on the following sensor inputs: GPS—no current motion; accelerometer—threshold deceleration exceeded; gyroscope—threshold angular acceleration exceeded; sound—specific “distressed words” identified/loud crashing sounds detected. An animal or baby left in the car event may be detected based on the following sensor-based inputs: GPS—no current motion;
accelerometer—minor motion or no motion (and for example in Recently Parked Mode); sound—possible animal sounds or baby crying; image recognition IN camera—possible optical flow indication of movement inside vehicle or human/baby face recognition. A “location discontinuity” event (vehicle transported/stolen) may be detected based on the following sensor-based inputs: prior GPS reading indicating location exceeding a maximus distance from current location upon power up; time gap from last operation exceeding maximum time limit. These combinations of events and inputs are illustrative only. Some embodiments may provide a subset of these inputs and/or events. Other embodiments may provide different combinations of inputs and/or different events. The event detection algorithms may be implemented locally on the camera device (e.g., client device 101) or may be performed in cloud servers 102, with the input signals and event detection outputs transmitted over the wireless communication connection 107/108 from and to the camera device. Alternatively, in some embodiments a subset of the detection algorithms may be performed locally on the camera device while other detection algorithms are performed on cloud servers 102, depending for example, on the processing capabilities available on the client device. Further, in one embodiment, artificial intelligence (“AI”) algorithms are applied to the multiple inputs to identify the most likely matching event for the given combination of inputs. For example, a neural network may be trained with the set of inputs used by the system to recognize the set of possible tagging events. Further, a feedback mechanism may be provided to the user via the mobile app to accept or reject proposed tagging results to further refine the neural network as the system is used. This provides a refinement process that improves the performance of the system over time. At the same time, the system is capable of learning to detect false positives provided by the algorithms and heuristics and may refine them to avoid incorrectly tagging events.
For example, according to one embodiment, detection of a high-priority event 702, such as for example a “break-in” tagging event, is done automatically by the system in two stages.
Referring now to
Based on the available network conditions, image data is processed and sent 712 to cloud system 103 with a notification of the potential high-priority event determination. For example, in one embodiment, the video data captured just prior and during the potential high-priority event is immediately transferred via network 107 when full bandwidth is available. If, for example, the LTE network signal is not strong and the available bandwidth and transmission rate is limited, only a subset of frames of the recorded video is sent. The quality and amount of video/image data transferred is proportional to the available network bandwidth and speed. In addition, depending on the network conditions, additional sensor data may be set to the cloud, such as accelerometer, sound, GPS, or other sensor parameters that may be included with the image/video data as metadata as for example described with reference to
Once the initial video/image data is sent 712 to the cloud system, a further analysis 714 of the available inputs is performed. In one embodiment, the further analysis 714 includes processing at the client device 101 and server 102. The client device 101, to the extent it is operational, continues to capture video and sensor data and uses the additional data in its input analysis 714 as further described below. Upon receiving the initial data and possible high-priority event notification at step 712, the server monitors the connection to the client device 101 for a period of time, and if disconnected, a high-priority event is determined 702B, presuming that client device 101 has been purposely disabled. For example, the server may attempt to contact the client device 101 after a few seconds or minutes, and if the client device 101 does not respond, the server determines a high-priority event 702B. During this time, in one embodiment, the cloud system 103 also collects other data to conduct its own analysis 714. For example, notifications from other client devices in the vicinity may be analyzed as further described below to determine what type of tag event caused the possible high-priority event trigger. In addition, if additional input is received indicating that there is no high-priority event, the processed may be stopped. For example, a user may cancel the process directly via user input, the client device 101 may detect the user, for example via Bluetooth ID, face recognition, or the like, and may terminate the process.
In one embodiment, a user-programmable sensitivity level is optionally available allowing the user to set how fast the system should respond to a potential high-priority event. For example, depending on the user sensitivity level setting, e.g., low, medium, or high, a different pre-set time frame and/or set of inputs is used for the second phase of the event analysis 714 and tag event determination 702B. For example, in one embodiment for a high sensitivity setting, the system takes a short amount of time, for example 10 seconds, to analyze 714 the additional inputs and provide a determination of the tag event 702B, while with a low sensitivity setting the system takes a longer amount of time, for example, 50 seconds, to conduct the input analysis 714 and make the tag event determination 702B. These time frames are illustrative but one of ordinary skill would understand that the amount of time allowed may be varied or even be programmable and variable, for example depending on event type. Accordingly, additional sensitivity levels may be provided between high and low in different embodiments, such as for example a medium level allowing for 30 seconds to complete the tag event determination 702B. Similarly, and relatedly, the input analysis 714 may vary depending on sensitivity level. For example, for higher sensitivity levels, the number of inputs considered and the rules and heuristics applied may be lower while for a lower sensitivity levels more inputs can be analyzed and considered with more or more complex rules an heuristics, taking a longer amount of time to process but producing a higher confidence in the tag event determination 702B. For example, in one embodiment, high sensitivity level processing uses the initial sound sensor input (e.g., glass break signature match), failure to detect an occupant's face in the IN camera, along with absence of registered user's Bluetooth ID to provide a determination of a high-priority event, such as a break-in, and generate a video clip 705 of the event and provide a user notification 706 within 10 seconds.
As the sensitivity level is decreased, additional inputs may be considered, such as for example, an engine start signal from the CAN bus, vibrations from accelerometer, GPS data, and the like. The different inputs may be processed and combined in different ways according to rules and heuristics to provide an ultimate tag event determination 702B. For example, in one embodiment, accelerometer signals are analyzed to determine the likelihood of being the result of damage to a vehicle. In this embodiment, the power spectral density of the accelerometer signal generated by an event is analyzed to determine whether the low frequency energy, e.g., 1-10 Hz., is above a threshold indicating likely damage to the vehicle. This “damage indication” along with a match on the glass-break sound signature and an irregular vehicle start power signature due to, for example “jumping” the ignition wires, would indicate a “stolen vehicle” high-priority event.
As the sensitivity level decreases, indicating a desire for higher confidence, even more inputs may be considered. In one embodiment, the input analysis 714 may include processing beyond the client device 101. For example, additional inputs from the user, such as text message responses or app notification confirmations from the user may be used. For example, client device 101 may send a text message or app notification requesting the user to inform the client device about the user's presence in the monitored location or vehicle. In addition, events detected in other client devices in the surrounding area may also be used in lower sensitivity settings. For example, similar “break-in” events detected by other client devices in the network in the same street may be an indication that the event under analysis is a “break-in” while if a crash is detected near-by, the current event may be related to a car crash, for example a parked car being hit as a byproduct of a collision between other cars. Common events in time in the vicinity of the current client device may be used by the system to make a tag event determination 702B. In some embodiments, at least part of the processing associated with the input analysis 714 may be performed by servers 102 in cloud system 103.
Ultimately, the number of sensitivity levels and corresponding time delays and/or inputs used for further analysis 714 may vary in different embodiments and maybe programmable and/or user selectable. Further, for events deemed to be higher priority, e.g., a break-in, the same sensitivity level may be configured to allow for a different amount of time for processing the additional inputs. The time to sensitivity level allocation may be programmable based on user settings or other inputs and may also vary based on the severity of the event detected. For example, if the event detected is based solely on motion sensor information, e.g., someone bumped on the car, even if the sensitivity level is set to high, the allowed time for providing a response may vary as compared to a detection of a glass-break sound signature. In one embodiment, the higher priority event type, i.e., glass break, will provide a user response faster than a lower priority event type, e.g., car bumped.
In one embodiment, when available, the system checks the sensitivity level 713 to determine how fast or what set of inputs to consider in the input analysis 714. After completing the input analysis 714, optionally within the time limit allowed by the pre-set sensitivity level, the system determines 702B whether a tag event was detected, including determining whether the potential high-priority event detected in the first stage 702A did occur. If no high-priority event is determined, the image/video and other data sent to the cloud at step 712 can be discarded.
Otherwise, to the extent the client device 101 is still operational, the additional data captured by the client device is preserved 704 in the buffer and a video clip is generated 705. The user is then notified 706 of the tag event or high-priority tag event. If the client device is no longer operational, after the allowed time delay optionally imposed by the sensitivity settings, the cloud system uses the image/video and possibly other sensor data to generate a high-priority event notification to notify 706 the user.
Referring back to
According to another aspect of the disclosure, in one embodiment, the detection process 702 is configured to detect a user-determined manual tagging of an event. The user may provide an indication to the system of the occurrence of an event of interest to the user. For example, in one embodiment, a user may touch the touchscreen of a client device 101 to indicate the occurrence of an event. Upon detecting 702 the user “manual tag” input, the system creates an event-based clip as described above with reference to
In another embodiment, the additional user speech may include additional commands. For example, the user may indicate the length of the event for which the manual tag was indicated, e.g., “short” for a 30-second recording, “long” for a two-minute recording, or the like.
Optionally, the length of any video clip can be extended based on user input. For example, after an initial event-based video clip is generated, the user may review the video clip and request additional time before or after and the associated video data is added to the playlist or manifest file as described with reference to
In one embodiment, the tagging process may optionally be programmable. For example, camera device may be programmed to recognize traffic signs using image recognition and a classifier and to capture and store metadata associated with the recognized sign. For example, stop signs may be detected and the speed or other sensor data may be recorded as metadata associated with the stop sign. This feature may be used by third-parties for monitoring driving behavior. For example, parents can monitor children, insurance companies can monitor insureds, employers can monitor employees, etc. Optionally, in one embodiment the camera device may provide driver feedback based on the detected signs and sensor data. For example, in one embodiment, the camera device may recognize street parking signs and notify the user regarding parking limits. For example, the device may alert the user regarding a “No Parking” zone, a limited time parking zone, and/or remind the user prior to the expiration of a parking time limit with sufficient time for the user to return to the vehicle (e.g., based on the sign image recognition, time, and location information). One of ordinary skill in the art would recognize the additional applications of driver feedback are possible within the scope of the invention, such as for example, feedback regarding speeding, traffic light/sign compliance, safety, or the like.
In another embodiment, the programmable tagging may be accessed remotely, e.g., via cellular communications module 205, to provide image queries remotely. For example, in one embodiment, license plate and/or car image queries associated with an “Amber Alert” may be provided by authorities via cloud system 103 to all camera devices in the system. According to one embodiment, standard “definitions” of image queries can be shared amongst cameras ahead of time so that all cameras can be looking for a specific object or item. Optionally, the image queries may include a timing component to specified an amount of time during which camera devices may periodically run the image query. For example, an Amber Alert may provide one or more image queries (e.g., a license plate and/or a specific vehicle brand and/or color) to be searched for some amount of time, for example during 24 hours. Optionally, in one embodiment, the user may also provide programmable tagging instructions, for example via mobile device app or Web-based interface. For example, in one embodiment, the user may schedule a tag generation event for capturing video data at a particular time, or may remotely instruct the camera device to start recording on command.
Now referring to
In yet another embodiment, a user may select an object (e.g., a car, person, item, or the like) being displayed on the screen of a camera device. For example, via a tap on a touchscreen of a client device 101 while video is being played, using voice commands, or other user input device capable of identifying objects being displayed on a video. Optionally, an object of interest can also be identified on a video automatically. For example, as part of the auto-tagging feature described above with reference to
Image processing algorithms and/or computer vision techniques are applied to identify the selected object from the video and formulate an object descriptor query. For example, the user input is applied to detect the region of interest in the image, e.g., the zoomed-in region. The data for the relevant region is processed into a vector representation for image data around detected relevant points in the mage region. From the vector or descriptor of the relevant region, feature descriptors are then extracted based on, for example, second-order statistics, parametric models, coefficients obtained from an image transform, or a combination of these approaches. The feature-based representation of the object in the image is then used as a query for matching in other video data. In one embodiment, a request for sharing video clips includes an image query for an object and metadata from the video data in which the object was detected.
Referring back to
Using the obtained metadata values, a set of relevant camera devices with video data responsive to the request, that for example may include the same object of interest or match the desired location, date, time, and/or orientation, is identified 802. In one embodiment, to respond to an image-query-based request, camera devices located within a given geographical radius at a given time frame and with cameras pointing in a desired orientation may be first identified 802. For example, if the object of interest is an unrecognized face detected inside a vehicle parked in a parking lot, camera devices from other vehicles in the same parking lot at the same time and directed at the vehicle that was broken into at the right time may be identified 802. Optionally, once the relevant camera devices are identified 802, a request for an image search query with the query for the object of interest is sent 803. The camera devices receiving this request can search their buffered video data with the image search query provided to determine if there is a match.
In one embodiment, the feature vectors for the object of interest and compared with feature vectors for potentially relevant objects identified in the video data to determine if there is a match. For example, if the object of interest is a human face, a face feature vector is provided with the query and camera devices can use image processing to identify faces in the video data, extract feature vectors for the identified faces, and compare to the face feature vector of the desired face. Optionally, the search request may provide a time frame of interest to further reduce the buffered video objects that need to be analyzed to respond to the request.
In one embodiment, the cloud system 103 monitors the user object selection process to identify selection of the same object by multiple users. Upon determining that multiple users have selected the same object, generating the same or a substantially similar image query, the system may, for example, notify the users via news pane 401c of other users with similar interests. The object query may be additionally matched based on location (e.g., same object identified by users within a maximum distance), time, and/or event type.
Responses to the search request are received 804. If no matches are found 805, the sharing request process ends 806. For example, if the search request was initiated by a user, the user may be notified that no matching video clips were found. If matching video clips are found 805, an authorization request is sent 807 to the user of the camera device responding with a match.
As discussed above with reference to
In one embodiment, the authorization request may include a dynamically generated video clip for the user to review in determining whether to authorize the sharing of the video clip with other users. In one embodiment, the authorization request may be fulfilled automatically based on pre-authorization recorded during sign-on, e.g., for neighbors, for specific cameras, or the like. Alternatively, the authorization request is fulfilled by other users. For example, a playlist or manifest file may be included with the request allowing the authorizing user to playback the relevant video objects with the matching object. As noted above, the video objects can be accessed directly from the camera device buffer, for example via the Internet or a direct network connection (e.g., Wi-Fi) between a mobile device and the camera device. In addition, if the video objects are encrypted, the authorization request may include a user identification request to obtain the required encryption key, such as for example, a biometric identification (e.g., face recognition, fingerprint, or the like). With the appropriate encryption key, the video objects are decrypted and playback to the user to obtain authorization for sharing. In addition, in one embodiment, the user may optionally request the system to obfuscate identified objects in the shared video clip. For example, any human faces, license plates, address numbers, and/or any other identifiable objects selected by the user may be automatically blurred in the video data to protect privacy upon user request. Alternatively, the system may by default obfuscate identifiable objects unless otherwise requested and/or authorized by system users.
If sharing authorization 808 cannot be obtained, the sharing request terminates 806, by for example notifying a user requesting the sharing that no clips are available. If authorization is obtained 808, for every matching video clip for which authorization is obtained is shared 809 with other users. For example, in one embodiment, if the sharing request was initiated by a user, the requesting user is notified of the availability of matching video clips. For example, the mobile app of the requesting user's mobile device 104 receives a notification from cloud system 103 and provide the notification to the user via the mobile app user interface. If the sharing request was automatically generated by a camera device of a user, for example from an auto-tagging event, the mobile app in the mobile device 104 of the user receives a notification of the availability of other video clips relevant to the user. The mobile app may then display information regarding the available video clips on the news pane 401c. Optionally, the mobile app may directly link the available video clips to the event-generated clips on the clips pane 401a. Any video clips for encrypted video data would have been decrypted through the authorization process and thus become shared video clips in unencrypted form.
In one embodiment, the video sharing request process is used to generate a virtual network of distributed cameras recording video for an event of interest. For example, the video clip generation process may include a live-stream playlist or manifest file dynamically generated and updated with additional clips being recorded for the given event. Using this approach, the system may generate a set of associated video clips for a given event, such as for example, a break-in, car accident, or the like, captured from cameras in the dynamically generated virtual network to provide views from different angles, vantage points, and/or wider or narrower views. For example, in one embodiment, interspersed still images from video captured by multiple camera devices may be used for license plate recognition purposes where video from a single camera is insufficient. In one embodiment, in addition to the license plate or if unable to recognize the license plate the color and make and model of the vehicle may be determined based on classifier-based image recognition techniques. The video sharing process of
According to another aspect of the disclosure, the video data generated by the camera devices in the system may be uniquely identifiable and verifiable to be authentic and unmodified. Now referring to
In one embodiment, if the request is determined 901 to be for a video clip, the playlist or manifest file for the video clip is accessed to retrieve 902 the list of video data objects or files in the video clip. The first video data object is selected 903. In one embodiment, if the request is determined 901 to be for a video data object, or if it is for a video clip and the first video data object has been selected 903, the metadata record associated with the video clip is retrieved 904. For example, in one embodiment, the GUID for the video data object is used to access a repository of metadata records associated with video data objects captured by camera devices associated with the cloud-based system 103. As described above, every camera device sends signed messages to the system including the metadata and a hash of the video data object for every data object recorded. In one embodiment, a metadata record includes the metadata and the hash of the video data and may be indexed by the associated GUID.
The stored hash of the video data object corresponding to the GUID is then compared 905 to a one-way hash of the video data object for which authentication is requested. In one embodiment, the authentication request includes the video data object. In that embodiment, the video data object is hashed using the same one-way hashing function used by the camera devices of the system. In an alternative embodiment, a network address for the video data object is provided in video clip file. In such an embodiment, the video data object is retrieved, for example at step 903 (or step 909 for subsequent video data objects), and it is hashed as described above. If the system is implemented based on hashing of the video data along with the metadata, the metadata retrieved 904 (if not part of the request) is used in the hashing function for the video data object being verified. The hashing function may be applied on a server, such as server 102, or may be performed on a client, such as Web-based client 105, and provided to the authentication system, for example along with the request.
In one embodiment, the result of the hash comparison 905 is used to output 906 a verification for the video data object. The verification output may, for example, provide a positive or negative result, indicating whether the video data is authentic or whether it has been modified. In one embodiment, the verification output may also include relevant metadata associated with the video data object, such as time, location, orientation, and the like. In one embodiment, if the video data object verified is not part of a video clip 907, the verification process concludes 908. However, if the video data object is part of a video clip 907, the process continues to step 909. At step 909, if the video data object that was verified was the first video data object in a video clip 909, the next video data object is selected 910 and the process repeats from step 904 for verification of the second video data object in the video clip. If the video data object is not the first in a video clip, a time analysis 911 is performed next. In one embodiment, as described above, the metadata for a video data object includes time information to identify when the video data was captured. For example, in one embodiment, atomic time from a 4G/LTE service provider is used to create a time stamp of the beginning of the video data object and either a duration or end stamp to indicate its end. In one embodiment, this time information is provided with the video object verification output 906, and used for time analysis 911. For example, the ending time of the first video data object in a clip is compared to the beginning time for the second video data object of the clip to determine if there is a gap. A gap in the time sequence between consecutive video data objects of a given video clip may for example indicate some editing to the video clip.
In one embodiment, if there are additional video data objects to be verified in a video clip 912, the process moves to step 910 and repeats through the time analysis step 911 for every video data object. Once all the video data objects in a video clip are verified 912, a video clip verification output is provided 913. For example, if all the video data objects in the clip were positively verified and the time analysis did not identify any gaps, a positive authentication for the video clip may be output 913. Optionally, the output may for example, include additional information regarding the video clip, such as, for example, time, duration, location, camera device used, user, or the like. Conversely, if any of the video clips cannot be authenticated, e.g., the hashes do not match, or a gap in the video clip timeline is found at step 911, a negative result is output 913. The output may for example, include reasons for the negative result in addition to or in place of any of the information provided for a positive result. For example, in one embodiment, a video clip consisting of 15 two-second video files generated upon detection of a car crash by a client device 101 could be uniquely verified as authentic by cloud system 103 using the approach described above.
According to another aspect of the disclosure, a process for setting up a camera device, such as a client device 101, is provided. Referring to
The user's biometric signature or signatures are stored in the camera device. In one embodiment, a cryptographic key is also generated based on a random input and stored in association with the biometric identification of the user. Optionally, if more than one user is required, for example for a vehicle with multiple possible drivers, the user set up process 1001 is repeated for each user.
Referring back to
In one embodiment, a user may prompt the camera device to pair with the user's mobile device 104. For example, in one embodiment, the user may utter a voice pairing command, provide a pairing command through a touchscreen, or through any other user input device available in the camera device. In one embodiment, the pairing process involves a Bluetooth paring process. In another embodiment, the camera device displays a unique pattern that is captured by the mobile device and sent back to the camera device via the connection to the could system 103. For example, camera device may display a randomly generated alphanumeric code, a QR code, a series of black and white screens in a random order, or some other random output. The random output is captured or entered into the mobile device by the mobile app and transmitted via a secured Internet connection to cloud system 103 along with a unique identifier of the mobile device, such as, for example a Bluetooth address, a MAC address, or the like. The random output and the mobile device input are compared. If they match, the camera device authenticates the mobile device unique identifier (e.g., Bluetooth address or MAC address) and from that point on is associated with the mobile device. In an alternative embodiment, instead of comparing the output of the client device with the input captured by the mobile device, both devices generate an output that is compared at the server. For example, each device uses a camera to perform face recognition of the user during the set-up process and their face recognition results are sent to the server for comparison to match the same user.
In one embodiment, a QR code is displayed on the display of the client device 101. The QR code encodes a device ID for the client device 101 and an encryption key (or seed for generation of an encryption key) for communicating with the client device 101. The mobile app on the mobile device 104 captures and interprets the QR code to obtain the device ID and encryption key. The device ID may for example include a telephone number, email address, or other means for electronic messaging with the client device 101. Using the encryption key, the mobile device 104 can send encrypted communications to the client device 101 as further described below to associate the mobile device with the client device, including for example, sending to the client device 101 a unique identifier for the mobile device 104, for example, telephone number, email address, Bluetooth address, MAC address, or the like. While described with the client device 101 being the device that displays the QR code, the same approach may be used with the mobile device 104 displaying the QR code and the client device 101 initiating the encrypted messaging using the encryption key provided by the mobile device 104.
Other “shared secret” approaches may be used for mobile device association 1002, include for example, a series of instructions to cause the user to move the mobile device while the mobile app records the outputs of one or more mobile device sensors to be matched with the provided instructions. For example, the user may raise or lower the device, shake the device, etc. in a random series causing accelerometer and/or gyroscope changes that match the requested motions. The series of sensor-detected motions can be provided via Internet connection for matching with the camera device instructions for association. Alternatively, in one embodiment, a user may provide a telephone number for the mobile device during a registration process, for example through the mobile device app. For the mobile device association step 1002, camera device may display a device ID on its screen. The user inputs the device ID on the mobile app and it is transmitted to the cloud system 103. The cloud system identifies the device ID and sends a message to the camera device 101/106 via Internet connection 107/108 including the telephone number for mobile device 104. The camera device sends a text message to mobile device 104 with a random code. The user inputs the random code via the mobile app for verification by cloud system 103 or camera device 101/106. If the random code matches the texted code, the mobile device is authenticated. Once the camera device and the mobile device are associated 1002, the camera device can trust the mobile device for subsequent interactions, based on a unique ID for the mobile device (e.g., Bluetooth address, MAC address, or the like). According to another aspect of disclosure, in one embodiment, the set-up process optionally includes the step of provisioning the mobile device 104 with a mobile app.
Upon interpreting the QR code, the mobile device uses its existing software (e.g., a web browser) to send 1103 an HTTP request to the web server identified through the link or URL and including the parameters encoded into the link. The cloud system 103 receives the request and creates 1104 a record for the request, including the link-encoded parameters and additional metadata and network information derived from the HTTP requesting process, including information for uniquely identifying the mobile device 104 (e.g., combination of HTTP heather metadata, TCP/IP header information, or the like). In addition, cloud system 103 redirects 1105 the mobile device to a location from where the appropriate mobile app may be obtained. For example, cloud system 103, using, for example, the “User-Agent” data from the HTTP request and/or the unique device ID for the camera device 101/106, redirects the mobile device 104 to either the Apple App Store when the User-Agent indicates the mobile device to be an iOS device or to the Google Play Store if the mobile device is determined to be an Android-based device or alternatively, to other servers capable of providing the mobile app to the mobile device over a network. Similarly, the cloud system 103 may include parameters in the redirection link to the appropriate version of the mobile app determined using the device ID of the camera device 101/106.
Once redirected, the mobile device 104 obtains 1106 the proper mobile app, e.g., the app for interaction with camera device 101/106 and cloud system 103. After the downloading and installation of the mobile app on mobile device, when executed, the mobile app contacts the cloud system 103 to access 1107 the record previously generated at step 1104. For example, the mobile app may derive a unique ID for the mobile device 104 using the same parameters, metadata, or other information available from the mobile device 104 when making an HTTP request like the one made at step 1103. In one embodiment, a time limit (e.g., 2-15 minutes) may be used between the HTTP request step 1103 and the record access step 1107 to facilitate the mobile device 104 identification. Cloud system 103 determines that the same mobile device 104 is accessing the system based on that information and provides 1108 access to the previously generated record and any other additional set up parameters that may be necessary to complete the set-up process. For example, if provided, the randomly generated number may be provided as a “shared secret” for the device association process described above. Alternatively, encryption information and/or messaging information for the camera device may be provided. Referring back to
Similar to step 1101 in
Upon approval by the primary mobile device 1209, the paring request is conveyed 1210 to the client device 101/106. The client device receiving the pairing request verifies 1211 the pairing request. For example, in one embodiment, the client device 101/106 checks that it is in “pairing mode” and/or that it can otherwise pair with a mobile device 104. In addition, the client device 101/106 checks the pairing information, including whether the Bluetooth ID for the mobile device 104 is found in the list of Bluetooth IDs scanned at step 1201 indicating that the mobile device 104 is located near the client device, within range of the short-range wireless communication radio. The client device 101/106 stores the pairing information and once the pairing request is verified 1211, it notifies the cloud system 103 that pairing has succeed 1212. According to one embodiment, the cloud system 103 updates its pairing record for the client device 101/106 to include the newly paired mobile device 104 and notifies the mobile device 104 of the pairing success.
At this point, communication between the client device 101/106 and the mobile device 104 are enabled 1213 and the paring process ends. The mobile device 104 and client device 101/106 can now, for example, post messages to each other on their respective “topics” via cloud system 103 using a messaging protocol, such as MQTT, for example. All messaging between the two devices enabling the functionality described above can be exchanged via the cloud-based messaging approach establish according to this embodiment. For example, the mobile device 104 can request a secured real-time feed of the video in the buffer memory of client device 101/106 using this approach and send and receive commands, for example, using RTC technologies as is known in the art. While all communications with the cloud system 103 may be encrypted, e.g., with the encryption provided by the cellular channels 107/108, using SSL, or similar communication channel encryption, the messaging between the two end-point devices can further be end-to-end encrypted. For example, the two devices can use the cloud messaging to do a Diffie-Hellman key exchange and use each other's key to encrypt the payloads of the messages sent via the cloud messaging system. Alternatively, as described above, the QR code may include a public key for the client device 101/106 that the mobile device can use to encrypt its own public key and include it in the pairing information with the paring request. Having each other's public keys, the two devices can then encrypt the payload of all the messages they exchange from that point on. In an alternative embodiment, the QR code can include a “shared secret” as described above with reference to
As those in the art will understand, a number of variations may be made in the disclosed embodiments, all without departing from the scope of the invention, which is defined solely by the appended claims. It should be noted that although the features and elements are described in particular combinations, each feature or element can be used alone without the other features and elements or in various combinations with or without other features and elements. The methods or flow charts provided may be implemented in a computer program, software, or firmware tangibly embodied in a computer-readable storage medium for execution by a general-purpose computer or a processor.
Examples of computer-readable storage mediums include a read only memory (ROM), a random-access memory (RAM), a register, cache memory, semiconductor memory devices, magnetic media such as internal hard disks and removable disks, magneto-optical media, and optical media such as CD-ROM disks.
Suitable processors include, by way of example, a general-purpose processor, a special purpose processor, a conventional processor, a digital signal processor (DSP), a plurality of microprocessors, one or more microprocessors in association with a DSP core, a controller, a microcontroller, Application Specific Integrated Circuits (ASICs), Field Programmable Gate Arrays (FPGAs) circuits, any other type of integrated circuit (IC), and/or a state machine.
One or more processors in association with software in a computer-based system may be used to implement methods of video data collection, cloud-based data collection and analysis of event-based data, generating event-based video clips and high-priority event notifications, sharing event-based video, verifying authenticity of event-based video data files, and setting up client devices according to various embodiments, as well as data models for capturing metadata associated with a given video data object or file or for capturing metadata associated with a given event-based video clip according to various embodiments, all of which improves the operation of the processor and its interactions with other components of a computer-based system. The camera devices according to various embodiments may be used in conjunction with modules, implemented in hardware and/or software, such as a cameras, a video camera module, a videophone, a speakerphone, a vibration device, a speaker, a microphone, a television transceiver, a hands free headset, a keyboard, a Bluetooth module, a frequency modulated (FM) radio unit, a liquid crystal display (LCD) display unit, an organic light-emitting diode (OLED) display unit, a digital music player, a media player, a video game player module, an Internet browser, and/or any wireless local area network (WLAN) module, or the like.
This application is a continuation-in-part of and claims priority to U.S. patent application Ser. No. 16/323,201, entitled “Video-Based Data Collection, Image Capture and Analysis Configuration,” filed Feb. 4, 2019, which is a national stage application of PCT Patent Application No. PCT/US17/50991, filed Sep. 11, 2017, which claims the benefit of U.S. Provisional Application No. 62/412,764, filed Oct. 25, 2016, the contents of which are hereby incorporated by reference in their entirety.
| Filing Document | Filing Date | Country | Kind |
|---|---|---|---|
| PCT/US19/34437 | 5/29/2019 | WO | 00 |
