Patent Application
20170017845
Embodiments of the present invention relate to applying computer vision techniques to automatically determine the presence or absence of objects, as well as their motion states, within stations or other specific locations.
Reliably detecting the presence or absence of an object, such as a train in railway interchange environments such as station platforms, and on tracks in train and subway stations is desirable for railway system management. Automated video systems may be used to monitor pluralities of such railway interchange environments from a remote, centralized point, enabling efficiencies in centralizing and providing comprehensive and contextual train traffic management. Such systems may be useful in alerting train system managers to collision and pedestrian hazards posed by moving trains, for example by providing advance warnings to train operators or other parties remote from a station stop that another train is unexpectedly occupying or entering a station platform area, early enough to enable avoidance measures. Alarms from such systems may notify a train engineer to begin stopping a train early enough, so that its inertia may be safely overcome prior to entry of a problematic area.
However, video monitoring systems suffer from a variety of limitations. For example, the capabilities of such systems may be limited by reliance on human perception to both review video feeds of pluralities of different station areas, and make the necessary determinations to spot and abate problems in a timely fashion. The number of personnel available to watch video footage from vast camera arrays is generally limited by budgetary and other resource limitations, as is the ability of any one human monitor to monitor and perceive a threat in multiple, simultaneous video feeds. The process of watching video streams is resource consuming, suffers from high costs of employing security personnel, and efficiency in such systems to detect events of interest is also limited by the constraints of human comprehension.
The field of intelligent visual surveillance seeks to address this problem by applying computer vision techniques to video stream inputs to automatically detect the presence or absence of trains at given track locations. However, the efficacy of such systems in real-world conditions is limited. Accurately determining the presence or absence of a train at a given train platform area may be challenging, in one aspect due to a large variability in amounts of reflect light generated by changing weather conditions (sunny, cloudy, nighttime, transient moving cloud shadows, etc.). A wide variety of reflected or occluded lighting profiles must be processed that may be caused by different train cars, the numbers of train cars, speeds or movement relative to the video cameras, and the different sizes, shapes and reflective behaviors of the train car element surfaces. Strong visual textures are generally observed in rail track areas, and even a minor change in lighting may cause incorrect foreground classification in video analytic systems. Thus, high rates of false positive detections, or low rates of accuracy in detecting true events, generally limit the usefulness and trustworthiness of such systems.
In one embodiment of the present invention, a method for applying computer vision techniques to automatically detect and classify the presence or absence of an object within a region of interest includes determining for an auto-correction time period a distribution of object present and object absent labels that are assigned one each to each of a plurality of the video segments of video image scene data acquired from a region of interest. The video segments are each assigned the object present label in response to determining that foreground feature data values meet a background subtraction threshold value indicative of a presence of a foreground object having a specified size; the object absent label in response to determining that the foreground feature data values do not meet the background subtraction threshold value; a motion present label in response to determining that motion feature data values for the frames differ from motion feature data values of previous frames by a motion threshold value that is selected as indicative of a change in visual features consistent with movement of the object within the frame data; and a static label in response to determining that the motion feature data values for the frames do not differ from the motion feature data values of the previous frames by the motion threshold value. In response to determining that the distribution of the object present and the object absent labels does not meet an expected distribution for the auto-correction time period, aspects exchange the object present and the object absent labels that are assigned to each of the video frames. Accordingly, segments of the input video data comprising consecutive video frames assigned the object absent label and the static label are classified as “no object present” segments; segments comprising consecutive video frames that are each assigned the object present label and the motion present label are classified as “object present and in transition” segments; and segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label are classified as “object present and stopped” segments.
In another embodiment, an article of manufacture has a tangible computer-readable storage medium with computer readable program code embodied therewith, the computer readable program code comprising instructions that, when executed by a computer processing unit, cause the computer processing unit to determine for an auto-correction time period a distribution of object present and object absent labels that are assigned one each to each of a plurality of the video segments of video image scene data acquired from a region of interest. The video segments are each assigned the object present label in response to determining that foreground feature data values meet a background subtraction threshold value indicative of a presence of a foreground object having a specified size; the object absent label in response to determining that the foreground feature data values do not meet the background subtraction threshold value; a motion present label in response to determining that motion feature data values for the frames differ from motion feature data values of previous frames by a motion threshold value that is selected as indicative of a change in visual features consistent with movement of the object within the frame data; and a static label in response to determining that the motion feature data values for the frames do not differ from the motion feature data values of the previous frames by the motion threshold value. In response to determining that the distribution of the object present and the object absent labels does not meet an expected distribution for the auto-correction time period, aspects exchange the object present and the object absent labels that are assigned to each of the video frames. Accordingly, segments of the input video data comprising consecutive video frames assigned the object absent label and the static label are classified as “no object present” segments; segments comprising consecutive video frames that are each assigned the object present label and the motion present label are classified as “object present and in transition” segments; and segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label are classified as “object present and stopped” segments.
In another embodiment, a system has a processing unit, computer readable memory and a tangible computer-readable storage medium with program instructions, wherein the processing unit, when executing the stored program instructions, determines for an auto-correction time period a distribution of object present and object absent labels that are assigned one each to each of a plurality of the video segments of video image scene data acquired from a region of interest. The video segments are each assigned the object present label in response to determining that foreground feature data values meet a background subtraction threshold value indicative of a presence of a foreground object having a specified size; the object absent label in response to determining that the foreground feature data values do not meet the background subtraction threshold value; a motion present label in response to determining that motion feature data values for the frames differ from motion feature data values of previous frames by a motion threshold value that is selected as indicative of a change in visual features consistent with movement of the object within the frame data; and a static label in response to determining that the motion feature data values for the frames do not differ from the motion feature data values of the previous frames by the motion threshold value. In response to determining that the distribution of the object present and the object absent labels does not meet an expected distribution for the auto-correction time period, aspects exchange the object present and the object absent labels that are assigned to each of the video frames. Accordingly, segments of the input video data comprising consecutive video frames assigned the object absent label and the static label are classified as “no object present” segments; segments comprising consecutive video frames that are each assigned the object present label and the motion present label are classified as “object present and in transition” segments; and segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label are classified as “object present and stopped” segments.
These and other features of this invention will be more readily understood from the following detailed description of the various aspects of the invention taken in conjunction with the accompanying drawings in which:
The drawings are not necessarily to scale. The drawings are merely schematic representations, not intended to portray specific parameters of the invention. The drawings are intended to depict only typical embodiments of the invention, and therefore should not be considered as limiting the scope of the invention. In the drawings, like numbering represents like elements.
As will be appreciated by one skilled in the art, aspects of the present invention may be embodied as a system, method or computer program product. Accordingly, aspects of the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment (including firmware, resident software, micro-code, etc.) or an embodiment combining software and hardware aspects that may all generally be referred to herein as a “circuit,” “module” or “system.” Furthermore, aspects of the present invention may take the form of a computer program product embodied in one or more computer readable medium(s) having computer readable program code embodied thereon.
Any combination of one or more computer readable medium(s) may be utilized. The computer readable medium may be a computer readable signal medium or a computer readable storage medium. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the context of this document, a computer readable storage medium may be any tangible medium that can contain or store a program for use by or in connection with an instruction execution system, apparatus, or device.
A computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in a baseband or as part of a carrier wave. Such a propagated signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device.
Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including, but not limited to, wireless, wire line, optical fiber cable, RF, etc., or any suitable combination of the foregoing.
Computer program code for carrying out operations for aspects of the present invention may be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the “C” programming language or similar programming languages. The program code may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet Service Provider).
Aspects of the present invention are described below with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer readable medium that can direct a computer, other programmable data processing apparatus, or other devices to function in a particular manner, such that the instructions stored in the computer readable medium produce an article of manufacture including instructions which implement the function/act specified in the flowchart and/or block diagram block or blocks.
The computer program instructions may also be loaded onto a computer, other programmable data processing apparatus, or other devices to cause a series of operational steps to be performed on the computer, other programmable apparatus or other devices to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide processes for implementing the functions/acts specified in the flowchart and/or block diagram block or blocks.
Intelligent visual surveillance comprehends applying computer vision techniques to automatically detect specific events in video stream data, including detecting and distinguishing static from moving objects. Foreground analysis-based background modeling may be used for direct object detection and tracking processes, wherein a train is detected whenever a large portion of foreground object is detected in a given monitoring region. However, railway environments present a number of challenges to automated video monitoring implementations, and the performance of background modeling techniques may be unreliable in detecting whether a train is present or not.
More particularly, effectively using video analytics to detect train presence status at platforms tracks or in train and subway stations or other railway environments generally requires a very low false alarm rate for event triggering; else the usefulness of the system is diminished. For example, a maximum of five false positives per analysis task per camera per day may be required in some applications. This poses a very challenging situation for tasks that are directly related to activities around train platform and other areas where trains frequently pass by and cause interferences, such as rail crossing detection, entering tunnel detection, alerting of people close to a platform edge or of people falling onto track, etc. Generally, a majority of such false alarms are caused by passing trains due to their large size, reflecting lighting, speed, etc. Therefore, being able to accurately determine train's presence is very important in false alarm rate control
Background analysis may not be robust enough in difficult lighting environments. Outdoor locations are subject to wide varieties of lighting changes due to weather and sunlight variability. Such problems are more prominent in rail track areas, where strong textures are observed and a minor change in lighting is likely to cause an incorrect foreground classification. While some systems implement interest point analysis methods to overcome certain deficiencies in foreground analysis method results, interest point analysis lacks reliability in low lighting environments where no texture can be seen and, consequently, no interest points can be found.
Referring again to
With respect to the BGS process at 104, rather than detecting moving objects as is typically the focus in the prior art, the ROI 202 and the BGS process 104 are selected to detect and distinguish train car foreground objects that do not belong to background environment settings. At 108 the process compares the BGS values determined for each of the video frames at 104 to a BGS threshold value that is selected as indicative of the presence of a foreground object having the size of a train car, and labels each frame accordingly as “train present” or “train absent.” If the determined BGS value meets (is equal to or greater than) this BGS threshold, it may be inferred that a train car-sized object is present in the image and the video frame is accordingly labeled as “train present.” Otherwise, if the determined BGS value is less than the threshold, this indicates that any detected foreground object is too small to be a train car (for example, a pedestrian or other smaller object, or a transient shadow or other lighting event), and the process may infer that a train is not present within the frame and label it as “train absent.”
Motion features are captured by the frame-differencing feature process at 106 via computing ratios of differences in motion features accumulated over time between consecutive frames of the video input 101. Frame differencing ratios are determined for each frame as a ration of motion feature data determined for the frame to a value of motion feature data determined for an immediately previous frame of the plurality of frames. At 110 the Frame-Diff ratio value determined for each of the video frames is compared to a motion threshold, a ratio value selected as indicative of a change in visual features consistent with the movement of a train car within the frame relative to a previous frame. If the determined Frame-Diff ratio value meets (is equal to or greater than) the motion threshold, it may be inferred that objects in the image are in motion and the video frame is labeled as “motion present”; otherwise, if less than the motion threshold, it may be inferred that objects in the image are not moving, static, and the video frame is labeled as “static”.
At 112 the input video 101 is segmented into one of three presence and motion state groups of consecutive video frames that are commonly labeled at 108 and 110 according to the following:
(i) each group of consecutive frames that are labeled at 108 as “train absent,” and labeled at 110 as “static,” are classified into a “no train present” segment of the video input, signifying that no train is at the station ROI 202;
(ii) each group of consecutive frames that are labeled at 108 as “train present,” and labeled at 110 as “motion present,” are classified into a “train present and in transition” segment of the video input; signifying that a train is at the station and within the ROI 202, generally pulling into or pulling out of the station; and
(iii) each group of consecutive frames that are labeled at 108 as “train present,” and labeled at 110 as “static” are classified into a “train present and stopped” segment of the video input; signifying that a train is at the station and stopped within the ROI 202, generally for passenger egress and ingress.
The present embodiment also provides processes for backward matching of the train presence and motion classifications of the video segments. More particularly, appearance features are extracted from the frames and accumulated over time for each of the motion state segments at 114, from a first frame in each segment up to the point in time at which the classification and labeling decision is made at 112 to transition to a next, different segment. In the present embodiment, color features are extracted from the monitoring ROI and accumulated in a color histogram, though other features may be extracted and accumulated via other methodology in other embodiments. In some embodiments, a latency delay time is specified at the beginning of each motion state segment, so that the accumulated appearance can be constructed with a latency that stops the system from responding to lighting changes due to cloud movements and other transient situations that might otherwise cause a false determination of the presence of a train. Examples of the delay time period include one and two second periods, and still others may be practiced.
At 116 the accumulated histogram for a current video segment is compared against the histograms of each of a plurality of previous (precedent) segments that each occurred within a comparison time period of the current segment, and a best-matching one of the previous motion state segments is identified, for example the one having a highest matching score . The comparison time period is selected to encompass each of the three different motion segment classifications determined at 112 under anticipated or historical train movement timings for the ROI 202. For example, if the ROI 202 is of a train station wherein trains stop for no longer than one minute to allow for passenger egress and ingress, then one exemplary comparison time period is five minutes, which should capture each of the three motion states associated with a stopped train. Other comparison time periods may be specified, in response to schedules or dynamic observations of actual train movements.
If the best matching histogram is sufficiently close to the histogram of the current segment, for example has a matching score that meets a threshold score at 118, then at 120 the current segment classification is updated to the classification of the previous segment that is identified as having the highest matching score, which may thereby change the classification initially determined for the current segment at 112 if said previous segment has a different classification.
The present embodiment also provides for an auto-correction process. At 122 the distribution of “train present” and “train absent” labels of the video segments is determined within an auto-correction time period, which is selected to encompass a plurality of train stopping events over a time frame longer than an anticipated (for example, scheduled) or actual (dynamically observed) time that a train is stopped within the ROI for one stopping event. This determination may occur periodically. If at 123 the distribution of the “train present” and “train absent” labels does not meet an expected distribution over the auto-correction time period, this is indicative of an error in the original labeling process at 112, and accordingly at 124 the “train present” or “train absent” labels of the video frames are exchanged and the segmentation classifications accordingly re-determined at 112 for the video input 101. Thus, at 126 the presence or absence of a train within the ROI of the video scene at a time of inquiry is determined from the “train present” or “train absent” label of the video segment classification comprehending the time of inquiry.
The lower graph plotline 304 illustrates another case wherein the train enters the ROI at 314 then stops at the platform 206 and remains static long enough for the Frame-Diff feature values to drop back to its initial low or zero value at 320 (as there should be no differences between consecutive frames of the video input while the train is static), while continuing to impose a large BGS area value in the ROI at 322. Ideally the process of
However, when the train stops at the platform for a long time problems may arise with the BGS values determined at 104. BGS modeling methods are generally based on online updating mechanisms wherein more recent observations gradually take over the models corresponding to the background. Therefore, if a train stops for a long enough period of time it may be learned into and become part of the background in a process is known as “healing”. For example, referring now to
This error is corrected by the backward motion state matching process at 114, 116, 118 and 120 of
However, comparing the accumulated histogram 432 of segment N 422 (at 116,
In a similar fashion, comparison of the accumulated histogram 439 of the segment 429 erroneously labeled as “S” to the accumulated histograms 430, 432, 434, 436 and 438 of the respective preceding video segments 420, 422, 424, 436 and 428 finds that the histogram 436 of the “N” (“no train present”) classification 426 labeled is the highest matching one, and since the match is close enough to meet the threshold score (at 118,
Another problem may occur in the automated processes of
More particularly, each ROI should have an expected or observed distribution of “train present” relative to “train absent” labels. In one example, the loading of trains on the track in the ROI over a long time period comprising a plurality of discrete, historical train stops as the station may indicate that the majority of time there should be no train present on the track: for example, a train may only be scheduled or otherwise expected to stop once, for at most a five minute period, during any given hour on any given day. Therefore, reviewing the labels applied at 112 as updated/corrected through the backward motion process effected at 120 should find that “No train” labels should be in the majority over a long time period, for example over any auto-correction time period of at least 30 minutes. If they are not, then it may be inferred that the “presence”/“absence” decisions are incorrectly reversed (due to the presence of a static train at the initiation of the BGS process) and that they should be reversed back along all the history records, and the classification accordingly re-determined at 112. Therefore, if the system gets into a sub-optimal state as a result of starting when a train is at the platform, or a train stopped for an excessive time due to exception, the process of
In one embodiment, the extracted color features are accumulated in the color histogram at 114 with a fixed weighting factor that is selected to bias or more heavily weight the most recent observations, and thereby enable the most recent observations to dominate the accumulated model. In one example, new observations are multiplied by a weighting factor of 0.1, and all accumulated observations in the histogram are multiplied by 0.9, and the factored products of the new and old observations are then combined to generate new, updated accumulated values for the histogram, thereby progressively biasing the accumulated histogram to reflect a most recent observation with more weight (represent 10% of the accumulated data) over the older observations as the observations accumulate beyond nine total observations, wherein each of the older observations becomes a smaller percentage of the aggregate 90% allocated to all previous observations.
Referring now to
Embodiments of the present invention may also perform process steps of the invention on a subscription, advertising, and/or fee basis. That is, a service provider could offer to integrate computer-readable program code into the computer system 522 to enable the computer system 522 to perform computer vision technique analysis of image data for automatically detecting and classifying the presence or absence and motion state of a train within a railway track area region of interest as described above with respect to
The terminology used herein is for describing particular embodiments only and is not intended to be limiting of the invention. As used herein, the singular forms “a”, “an” and “the” are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms “comprises” and/or “comprising” when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof. Certain examples and elements described in the present specification, including in the claims and as illustrated in the Figures, may be distinguished or otherwise identified from others by unique adjectives (e.g. a “first” element distinguished from another “second” or “third” of a plurality of elements, a “primary” distinguished from a “secondary” one or “another” item, etc.) Such identifying adjectives are generally used to reduce confusion or uncertainty, and are not to be construed to limit the claims to any specific illustrated element or embodiment, or to imply any precedence, ordering or ranking of any claim elements, limitations or process steps.
The corresponding structures, materials, acts, and equivalents of all means or step plus function elements in the claims below are intended to include any structure, material, or act for performing the function in combination with other claimed elements as specifically claimed. The description of the present invention has been presented for purposes of illustration and description, but is not intended to be exhaustive or limited to the invention in the form disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the invention. The embodiment was chosen and described in order to best explain the principles of the invention and the practical application, and to enable others of ordinary skill in the art to understand the invention for various embodiments with various modifications as are suited to the particular use contemplated.
| Number | Date | Country | |
|---|---|---|---|
| Parent | 14711871 | May 2015 | US |
| Child | 15279524 | US | |
| Parent | 13590269 | Aug 2012 | US |
| Child | 14711871 | US |
