System and method for providing tap-less, real-time visual search

Abstract

A system and method detects that an object within an image frame being captured via use of an imaging element associated with a computing device is an object of interest, tracks the object of interest within the image frame while determining if the object within the image frame remains the object of interest within the image frame for a predetermined amount of time, and, when the object within the image frame fails to remain the object of interest within the image frame for the predetermined amount of time causes the steps to be repeated. Otherwise, the system and method will automatically provide at least of part of the image frame to a cloud-based visual search process for the purpose of locating one or more matching products from within a product database for the object of interest with the located one or more matching products being returned to a customer as a product search result.

Description

BACKGROUND

In current apps that have visual search capabilities, when a customer is visually searching for a product a “tap to search” approach is used. In the “tap to search” approach, the customer must first point a camera, e.g., a camera of a mobile computing device such as a smart phone, a tablet computing device, etc., towards an object of interest. Once the object of interest is brought into focus, typically as seen on a display of the mobile computing device, the customer must then tap the display to capture a still image of the object of interest. The captured image is then processed, typically by being uploaded to a cloud server, whereupon the customer is presented with information about product(s) that are determined to be a match (or close match) to the object of interest. The information is generally presented to the customer using the display of the mobile computing device and the customer can interact with the information as presented to the customer to refine the information, perform further searching, to purchase product, etc.

SUMMARY

To address the deficiencies in the current apps that rely upon the “tap to search” approach, the following generally describes an example app that employs a “tap-less” approach. The example app will allow a customer to perform product searching/matching by simply pointing a camera, such as the camera of a mobile computing device, towards an object of interest. The example app thus allows a user to perform product matching in near real time by eliminating the need for the customer to further interact with (e.g., touch, tap, speak to, etc.) the mobile computing device.

More particularly, the following describes a system and method that detects that an object within an image frame being captured via use of an imaging element associated with a computing device is an object of interest, tracks the object of interest within the image frame while determining if the object within the image frame remains the object of interest within the image frame for a predetermined amount of time, and, when the object within the image frame fails to remain the object of interest within the image frame for the predetermined amount of time causes the steps to be repeated. Otherwise, the system and method will automatically provide at least of part of the image frame to a cloud-based visual search process for the purpose of locating one or more matching products from within a product database for the object of interest with the located one or more matching products being returned to a customer as a product search result.

To assist the customer in choosing the object of interest within a crowded scene, the example app may also convey visual cues to the customer. The visual cues, presented via use of the display of the mobile computing device, will implicitly guide the customer to bring the object of interest into focus. Once the object of interest is in focus and remains in focus for a preset duration, the example app will automatically trigger the product matching procedure without requiring any further interactions with the computing device on the part of the customer.

In a further example, the app will continuously record relevant data while the customer is stabilizing the camera towards the object of interest to provide for a best possible product matching experience.

A better understanding of the objects, advantages, features, properties, and relationships of the hereinafter described systems/methods will be obtained from the following detailed description and accompanying drawings which set forth illustrative embodiments and which are indicative of the various ways in which the principles of the described systems/methods may be employed.

BRIEF DESCRIPTION OF THE DRAWINGS

Example systems and method for providing tap-less, real-time visual search will be described hereinafter with reference to the attached drawings in which:

FIG. 1 illustrates an example computing device for use in capturing image frame information;

FIG. 2 illustrates an example method for implementing object detection and tracking and visual search using the captured image frame information;

FIG. 3 illustrates components of an example system/method for implementing object detection and tracking and visual search;

FIG. 4 illustrates an example of an object detection and tracking process as it is being performed on the computing device of FIG. 1;

FIGS. 5A-5F illustrate an additional example of an object detection and tracking and visual search process as it is being performed on the computing device of FIG. 1; and

FIG. 6 illustrates an example of an object detection and tracking and visual search process as it is being performed on the computing device of FIG. 1 to obtain search results for multiple objects within a crowded scene.

DETAILED DESCRIPTION

The following describes a new and innovative visual search product, e.g., an app or the like, having a “tap-less” capability.

In general, the “tap-less” capability is achieved by combining object detection and tracking techniques with visual search and scene understanding technologies.

Object detection is performed on-device in real time on image frames captured via use of a camera, data from object detection is presented in real time to the customer as visual cues for the prominent object being detected and tracked thus allowing the customer to choose the object of interest within a crowded scene, data from object detection is used for filtering out unnecessary information within the captured frame, and data from object detection is stored for later use as the input to the visual search process.

Object tracking is performed in real time in conjunction with object detection on the image frames captured via use of the camera. Data from object tracking, specifically the ID of the prominent object detected in the viewfinder frame, is used to present the customer with visual cues as to the data acquisition and to intuitively have the user stabilize the camera onto the object of interest.

Once the object of interest is in-focus, a visual search trigger algorithm will automatically cause product matching to be performed via use of a visual search engine that resides in the cloud. Multi-constrained optimization techniques are preferably used to choose the most-significant tracks in a given time-frame for triggering the cloud-based product matching process. Visual search is preferably performed in the cloud due to its algorithmically complex nature and the size of the products database. Thus, using the data captured during the object detection and tracking phase, the visual search engine will return to the customer one or more product matches for presentation to the customer via use of a computing device.

Turning now to FIG. 1, FIG. 1 illustrates, in block diagram form, an example computing device 100 usable with the subject app. Preferably, the computing device 100 is in the form of a mobile computing device. e.g., a smartphone, an electronic book reader, or tablet computer. However, it is to be understood that any device capable of receiving and processing input can be used in accordance with the various embodiments discussed herein. Thus, a computing device 100 can include desktop computers, notebook computers, electronic book readers personal data assistants, cellular phones, video gaming consoles or controllers, television set top boxes, and portable media players, among other devices so long as the device includes or is capable of being coupled to a movable image capturing element.

For use in connection with the visual search process, the computing device 100 has an associated display and one or more image capture elements 104. The display may be a touch screen, electronic ink (e-ink), organic light emitting diode (OLED), liquid crystal display (LCD), or the like element, operable to display information or image content to one or more customers or viewers of the computing device 100. Each image capture element 104 may be, for example, a camera, a charge-coupled device (CCD), a motion detection sensor, an infrared sensor, or other image capturing technology as needed for any particular purpose. As discussed, the computing device 100 can use the image frames (e.g., still or video) captured from the one or more image capturing devices 104 to capture data representative of an object of interest whereupon the captured image information can be analyzed to recognize the object of interest. Image capture can be performed using a single image, multiple images, periodic imaging, continuous image capturing, image streaming, etc. Further, the computing device 100 can include the ability to start and/or stop image capture, e.g., stop the visual search process, such as when receiving a command from a user, application, or other device.

As further shown in FIG. 1, the computing device 100 also includes one or more orientation-determining and/or position-determining elements 106 operable to provide information such as a position, direction, motion, and/or orientation of the computing device 100. These elements can include, for example, accelerometers, inertial sensors, electronic gyroscopes, and/or electronic compasses without limitation. Meanwhile, for communicating data to remotely located processing devices, the computing device 100 preferably includes at least one communication device 108, such as at least one wired or wireless component operable to communicate with one or more electronic devices, such as a cell tower, wireless access point (“WAP”), computer, or the like.

As yet further illustrated in FIG. 1, these and other components are coupled to a processing unit 112 which will execute instructions, including the instructions associated with the subject app, that can be stored in one or more memory devices 114. As will be apparent to one of ordinary skill in the art, the computing device 100 can include many types of memory, data storage, or computer-readable media, such as a first data storage for program instructions for execution by the processing limit(s) 112, the same or separate storage for images or data, a removable memory for sharing information with other devices, etc.

To provide power to the various components of the computing device 100, the computing device 100 also includes a power system 110. The power system 110 may be a battery operable to be recharged through conventional plug-in approaches, or through other approaches such as capacitive charging through proximity with a power mat or other such device.

In some embodiments the computing device 100 can include at least one additional input device 116 able to receive conventional input from a user. This input device 116 can include, for example, a push button, touch pad, touch screen, wheel, joystick, keyboard, mouse, keypad, or any other such device or element whereby a user can input a command to the device. These I/O devices could even be connected by a wireless infrared, Bluetooth, or other link in some embodiments. Some devices also can include a microphone or other audio capture element that accepts voice or other audio commands. As will be appreciated, the input device 116 can, among other things, be used to launch the app and to close the app as desired.

Turning to FIG. 2, in a preferred embodiment of the subject app the object detection 202 and tracking 204 processes are performed on-device as soon as the customer launches the app on the computing device 100. Once launched, the customer will point the imaging element(s) 104 towards the object of interest and, as the customer trains the camera on the scene that includes the object of interest, object detection and tracking will be performed on every frame presented to the user within the viewfinder that is caused to be displayed in the display 102 of the computing device 100. Object detection and tracking will be performed under real-time constraints, i.e., the process will consider the device processing power and frame processing may be skipped when necessary in order to achieve a real-time fluid experience.

While object detection may detect multiple objects within the viewfinder's frames, only the most prominent detected object shall be tracked and visually cued to the customer, allowing the customer to select the object of interest within a crowded scene by simply pointing the camera 104 towards that object and keeping the camera 104 focused on that object for a predetermined period of time 206 To assist the customer during this process, the viewfinder presented in the display 10, an example of which is illustrated in FIG. 4, may provide the customer with an indicia 402, such as a bounding box, that functions to emphasize the current focus of the camera 104, i.e., the current object of interest 404 within the scene, and a progress indicator 406 that indicates to the customer the amount of time the camera 104 has been focused on the object of interest 404 and, accordingly, the amount of time before the search process will be automatically triggered. Thus, as shown in FIG. 5A-5F, once the customer has trained the camera 104 on the object of interest 404 for a sufficient period of time, which is indicated to customer in this example by the filling of progress indicator 406 in the form of a progress bar, the system will automatically commence the process of matching the object of interest 404. If, however, the customer refocuses the camera 104 onto a different object of interest 404′ prior to the expiry of the measured time, the process will recommence. The viewfinder can further provide an indication to the customer when the search process has been automatically triggered, for example by changing/darkening the view of the scene as presented to the customer as shown in FIG. 5E as compared to FIG. 5D. It will also be appreciated that the example visual progress indicator 406 can also be associated with or alternatively implemented as an audible progress indicator.

As noted above, data from object detection is preferably presented in real-time to the customer in visual form, for example in the form of a bounding box 402 of the most prominent object 404 detected, overlaid on top of the captured image displayed in the viewfinder. This highlighting 402 of the object of interest 404 to the customer achieves two goals. First and foremost, highlighting 402 the object of interest 404 guides the customer into choosing the object of interest 404 from many objects within the field of view. Additionally, highlighting 402 the object of interest 404 guides the customer into bringing the object of interest 404 into a position of prominence in the field of view thus implicitly improving product matching by improving the captured object data used for product matching. Yet further, the prominent detected object's bounding box in this example—which defines an area of interest within the captured frame—may be used for filtering out unnecessary information (e.g., busy scenery or adjacent objects within the captured frame) from the captured image frame when performing the visual search process, thus improving product matching. Still further, data from object detection, specifically the prominent detected object bounds within the captured frames, may be used to crop the object image from the captured frame and these object images may be stored for optimally choosing the best data as input to the visual search process.

Data from object tracking, such as the ID of the prominent object being detected and tracked, is additionally used in connection with the progress indicator 406. For example, while the ID of the prominent object being detected and tracked remains unchanged over consecutive frames, the system may function to fill the progress bar in keeping with the embodiment illustrated in FIGS. 5A-5E. If, however, the value of the tracking ID changes then the progress indicator 406 will be reset to indicate to the customer that a new object has gained prominence and the device is now gathering data for that object. The use of object tracking in this manner will intuitively train the user into stabilizing the camera viewfinder onto the object of interest 404.

Once triggered, visual search is preferably performed in the cloud due to its algorithmically complex nature and the size of the products database. The input to the visual search is the data captured during the object detection phase, preferably after being subjected to a multi-constrained optimization technique that functions to choose the most-significant tracks in a given time-frame. In further embodiments the data may simply be an optimally chosen image of the prominently detected object.

As particularly illustrated in FIG. 5F, the output of the visual search may be product match IDs that can be thereafter translated into product metadata (product name, brand, images, price, availability etc.) and presented to the customer as purchasing options. For this purpose, the cloud-based visual search engine will have access to product data that is to be used during the matching process where that product data is further cross-referenced in one or more associated data repositories to the product metadata. The product metadata may then be provided to the computing device 100 of the customer for display whereupon the customer may interact with the product information to perform otherwise conventional e-commerce related actions, e.g., to place product into a shopping cart or list, to purchase product, etc.

In some instances, it may be desirable to pre-process the image information prior to the image information being provided to the visual search engine. A non-limiting example of a pre-processing technique is a cross-frames brightness correction technique that may be employed to enhance the object detection and tracking outcome. In addition, image stabilization techniques, such as the monitoring of the rotation vector as part of the exposed mobile OS motion sensors APIs, may be used to enhance the quality of the captured data during the object detection and tracking phase.

Turing now to FIG. 3, an example system/method that combines object detection and tracking techniques with visual search and scene understanding technologies to thereby provide a “tap-less” visual search capability is illustrated. As discussed above, when the subject app is launched, the system uses a real-time object detection component 302 to detect objects within a scene 300 that is being pointed to by a camera 104 of the computing device 100. The object detection component 302 can be implemented using, for example, GOOGLE's FIREBASE toolbox. The object detection component will provide output that identifies areas of possible interest in the frame, e.g., defines bounding boxes in the frame. While not required, the frames 300 can be provided to correction component 301 that functions to process the frames to reduce noise prior to the frames being provided to the real-time object detection component 302.

The output from the real-time object detection component 302 may then be provided to a bounding box/object locating component 304. The bounding box/object locating component 304 is intended to identify, via use of the data that is output by the real-time object detection component 302, the bounding-box with the highest confidence, i.e., identify the location of the object of interest within the frame. The output of the bounding box/object locating component 304, namely, the location within the image of the bounding-box surrounding the product of interest, is provided to the real-time tracking component 306. The real-time tracking component 306, in cooperation with the object location trajectory component 308, tracks the location of the bounding-box within the image to ensure that the camera is remaining focused on the same object through multiple frames/over time. These components may use a Kalman filter that functions to assign an ID to the object/bounding box location to assist in the location tracking procedure.

While the above described components are performing object detection and tracking, a time sampler component 310 is used to continuously capture the time a customer spends focusing on one object with the camera 104. In this example, the time sampler component 310 operates in conjunction with a motion detecting component 312 that uses data generated by the orientation/positioning element 106 of the mobile computing device 100 to track the motion of the mobile computing device 100 to determine if the customer is quickly shifting the focus from one object to another within the scene as described immediately below. It will also be appreciated that the output from the time sample component 310 may be used to update the progress indicator 406 as it is being presented in the viewfinder.

The data generated by the above components is provided to a multi-constraint optimization algorithm component 314 that functions to determine if visual search should be triggered or if processing should continue. More particularly, the multi-constraint optimization algorithm component 314 uses linear programming techniques to decide if the customer is interested in a given object, e.g., determines if the customer has kept the camera focused on the object for a predetermined amount of time. If the multi-constraint optimization algorithm component 314 determines that the customer is interested in the object in focus, the multi-constraint optimization algorithm component 314 will automatically trigger the visual search. If, however, the data indicates that the customer is not interested in the object in focus, e.g., the customer moves the computing device 100 prior to the expiry of the predetermined amount of time by an amount that changes the bounding box with the highest confidence/the ID of the object being tracked, the multi-constraint optimization algorithm component 314 will indicate to the system that the whole process must be reset 316, e.g., the system should reset the indicia 402, such as a bounding box, that functions to emphasize the current focus of the camera 104, and rest the progress indicator 406 that indicates to the customer the amount of time the camera 104 has been focused on the object of interest 404 within the viewfinder.

When the visual search process is automatically triggered, the image data is provided to the cloud-based, visual search engine 320. As further illustrated in FIG. 3, a region of interest (“ROI”) extraction component 322 will use the coordinates that define that area of interest within the image frame, e.g., the coordinates of the bounding box, to extract from the image frame the object of interest. A normalization component 324 may be used to normalize, at the pixel level, the extracted image information and an encoding component 326 may be used to encode to base 64 the normalized, extracted image information prior to the extracted image information being operated on by the image recognition component 326. As will be appreciated by those of skill in the art, the image recognition component 326 uses one or more algorithms to recognize the object of interest, e.g., to locate one or more products within a database of product information 332 that is/are an exact or close match to the object of interest. The located one or more products located within the database may then be ranked by a ranking component 330, for example based upon a confidence level, whereupon the located product information will be returned to the computing device 100 for presentation to the customer.

In view of foregoing, it will be appreciated that the described systems and methods for providing tap-less, real-time visual search provide, among other things, an improved shopping experience for customers by allowing a customer to find a product's replacement (usually an exact match replacement or near exact replacement) where the only user interaction needed is pointing a camera towards an object of interest. Furthermore, as seen by the sample screen images illustrated in FIG. 6, which depict a computing device 100 being used to execute multiple product searches and to display the corresponding search results, the subject system and method has the advantage of seamlessly providing information about plural objects within a crowded scene simply in response to a customer pointing a camera towards each of the objects in turn.

While various concepts have been described in detail, it will be appreciated by those skilled in the art that various modifications and alternatives to those concepts could be developed in light of the overall teachings of the disclosure. Further, while described in the context of functional modules and illustrated using block diagram format, it is to be understood that, unless otherwise stated to the contrary, one or more of the described functions and/or features may be integrated in a single physical device and/or a software module, or one or more functions and/or features may be implemented in separate physical devices or software modules. It will also be appreciated that a detailed discussion of the actual implementation of each module is not necessary for an enabling understanding of the invention. Rather, the actual implementation of such modules would be well within the routine skill of an engineer, given the disclosure herein of the attributes, functionality, and inter-relationship of the various functional modules in the system. Therefore, a person skilled in the art, applying ordinary skill, will be able to practice the invention set forth in the claims without undue experimentation. It will be additionally appreciated that the particular concepts disclosed are meant to be illustrative only and not limiting as to the scope of the invention which is to be given the full breadth of the appended claims and any equivalents thereof.

Claims

1. A non-transitory, computer-readable media having stored thereon instructions, the instructions, when executed by a computing device, cause the computing device to perform steps comprising: using data within an image frame being captured via use of an imaging element associated with the computing device to automatically detect that an object within the image frame is a most prominently detected object within the image frame; andin direct response to an expiration of a predetermined amount of time with the object continually being the most prominently detected object within the image frame automatically providing at least of part of the image frame to a cloud-based visual search process for the purpose of locating one or more matching products from within a product database for the object with the located one or more matching products being returned to a customer as a product search result.

2. The non-transitory, computer-readable media as recited in claim 1, wherein the instructions, when executed by the computing device, further cause the computing device to display in a display associated with the computing device the image frame being captured via use of the imaging element associated with the computing device, to manipulate the image frame to highlight the object within the image frame as it is being displayed in the display, and to provide an indication within the display of a passing of a measure of the predetermined amount of time.

3. The non-transitory, computer-readable media as recited in claim 2, wherein the image frame is manipulated to provide a bounding box around the object within the image frame as it is being displayed in the display.

4. The non-transitory, computer-readable media as recited in claim 3, wherein the instructions, when executed by the computing device, further cause the computing device to manipulate the image frame exclusive of a portion of the image frame within the bounding box in direct response to the expiration of the predetermined amount of time with the object continually being the most prominently detected object within the image frame.

5. The non-transitory, computer-readable media as recited in claim 3, wherein the image frame is manipulated exclusive of the portion of the image frame within the bounding box by being darkened relative to the portion of the image frame within the bounding box.

6. The non-transitory, computer-readable media as recited in claim 2, wherein the indication within the display of the passing of the measure of the predetermined amount of time comprises a displayed progress bar.

7. A computing device, comprising: a processor;an imaging element coupled to the processor, the imaging element having a field of view;a display element; andmemory including instructions that, when executed by the processor, enable the computing device to:capture image frame information by the camera, the image frame information including a representation of scene currently within a field of view of the camera, the image information being displayed on the display element of the computing device;use the image frame information to automatically detect that an object within the image frame information is a most prominently detected object within the image frame; andin direct response to an expiration of a predetermined amount of time with the object continually being the most prominently detected object within the image frame automatically providing at least of part of the image frame information to a cloud-based visual search process for the purpose of locating one or more matching products from within a product database for the object of interest with the located one or more matching products being returned to a customer as a product search result for display in the display element.

8. The computing device as recited in claim 7, wherein the instructions, when executed by the processor, further cause the computing device to manipulate the image frame information to highlight the object within the image frame information as it is being displayed in the display element and to provide an indication within the display element of a passing of a measure of the predetermined amount of time.

9. The computing device as recited in claim 8, wherein the image frame is manipulated to provide a bounding box around the object within the image frame information as it is being displayed in the display element.

10. The computing device as recited in claim 9, wherein the instructions, when executed by the processor, further cause the computing device to manipulate the image frame information exclusive of a portion of the image frame information within the bounding box in direct response to the expiration of the predetermined amount of time with the object continually being the most prominently detected object within the image frame.

11. The computing device as recited in claim 10, wherein the image frame information is manipulated exclusive of the portion of the image frame information within the bounding box by being darkened relative to the portion of the image frame information within the bounding box.

12. The computing device as recited in claim 8, wherein the indication within the display element of the passing of the measure of the predetermined amount of time comprises a displayed progress bar.

13. A method for performing visual search, comprising: using data within an image frame being captured via use of an imaging element associated with the computing device to automatically detect that an object within the image frame is a most prominently detected object within the image frame; andin direct response to an expiration of a predetermined amount of time with the object continually being the most prominently detected object within the image frame automatically providing at least of part of the image frame to a cloud-based visual search process for the purpose of locating one or more matching products from within a product database for the object with the located one or more matching products being returned to a customer as a product search result.

14. The method as recited in claim 13, further comprising displaying in a display associated with the computing device the image frame being captured via use of the imaging element associated with the computing device, manipulating the image frame to highlight the object within the image frame as it is being displayed in the display, and providing an indication within the display of a measure of a passing of the predetermined amount of time.

15. The method as recited in claim 14, wherein the image frame is manipulated to provide a bounding box around the object within the image frame as it is being displayed in the display.

16. The method as recited in claim 15, further comprising manipulating the image frame exclusive of a portion of the image frame within the bounding box in direct response to the expiration of the predetermined amount of time with the object continually being the most prominently detected object within the image frame.

17. The method as recited in claim 16, wherein the image frame is manipulated exclusive of the portion of the image frame within the bounding box by being darkened relative to the portion of the image frame within the bounding box.

18. The method as recited in claim 14, wherein the indication within the display of the passing of the measure of the predetermined amount of time comprises a displayed progress bar.