Determination of object presence and motion state

• US 9,594,963 B2
• Filed: 09/29/2016
• Issued: 03/14/2017
• Est. Priority Date: 08/21/2012
• Status: Active Grant

• Current Assignee: Kyndryl Incorporated
• Original Assignee: International Business Machines Corporation
• Inventors: Bobbitt, Russell P., Feris, Rogerio S., Zhai, Yun
• Primary Examiner(s): Park, Chan
• Assistant Examiner(s): SCHWARTZ, RAPHAEL M
• Application Number:
  US15/279,524
• Publication Number:
  US 20170017845A1
• Time in Patent Office: 166 Days
• Field of Search: None
• US Class Current: 1/1
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06V 20/41 Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items segmenting video sequences G06V20/49
• CPC Class Codes: G06V 20/49 Segmenting video sequences, i.e. computational techniques such as parsing or cutting the sequence, low-level clustering or determining units such as shots or scenes
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

Foreground feature data and motion feature data is determined for frames of video data acquired from an object region of interest. The frames are labeled as “object present” if the determined foreground feature data value meets a threshold value, else as “object absent; and as “motion present” if the motion feature data meets a motion threshold, else as “static.” The labels are used to classify segments of the video data comprising groups of consecutive video frames, namely as within a “no object present” segment for groups with “object absent” and “static” labels; within a “object present and in transition” segment for groups “object present” and “motion present” labels; and within a “object present and stopped” segment for groups with “object present” and “static” labels. The presence or motion state of an object at a time of inquiry is thereby determined from the respective segment classification.

Claims

1. A computer-implemented method for applying computer vision techniques to automatically detect and classify the presence or absence of an object within a region of interest, the method comprising executing on a processor the steps of:

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, wherein 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, exchanging the object present and the object absent labels that are assigned to each of the video frames;

classifying segments of the input video data comprising consecutive video frames that are each assigned the object absent label and the static label as “

no object present”

segments;

classifying segments of the input video data comprising consecutive video frames that are each assigned the object present label and the motion present label as “

object present and in transition”

segments; and

classifying segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label as “

object present and stopped”

segments.

2. The method of claim 1, further comprising:

integrating computer-readable program code into a computer system comprising the processor, a computer readable memory and a computer readable hardware storage device, wherein the computer readable program code is embodied on the computer readable hardware storage device and comprises instructions for execution by the processor via the computer readable memory that cause the processor to perform the steps of determining for the auto-correction time period the distribution of the object present and the object absent labels, exchanging the assigned object present and object absent labels in response to determining that the distribution of the object present and object absent labels does not meet the expected distribution for the auto-correction time period, classifying the segments of the input video data comprising consecutive video frames that are each assigned the object absent label and the static label as the “

no object present”

segments, classifying the segments of the input video data comprising consecutive video frames that are each assigned the object present label and the motion present label as the “

object present and in transition”

segments, and classifying the segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label as the “

object present and stopped”

segments.

3. The method of claim 1, further comprising:

accumulating appearance features extracted from each of the frames into histograms for each of the classified segments of the input video data;

comparing the accumulated histogram for a current one of the classified segments to the histograms of each of a plurality of segments that are previous in time to the current segment, wherein each of the compared plurality of segments occurred within a comparison time period of the current motion state segment, and wherein the comparison time period is selected to encompass each of the “

object present and stopped”

segment, the “

object present and in transition”

segment and the “

no object present”

segment under anticipated or historical object movement timings for the region of interest; and

changing a current classification of the current segment to the “

object present and stopped”

classification, the “

object present and in transition”

classification or the “

no object present”

classification of a one of the compared plurality of segments that has an accumulated histogram that best matches the accumulated histogram for the current segment.

4. The method of claim 3, further comprising:

selecting the auto-correction time period to encompass a plurality of object stopping events over a time frame longer than an anticipated or dynamically observed time that an object is stopped within the region of interest for one stopping event.

5. The method of claim 4, wherein the extracted, accumulated features are color features, and the accumulated histograms are color histograms.

6. The method of claim 4, wherein the step of accumulating the appearance features extracted from each the frames into the histograms for each of the classified segments of the input video data further comprises:

constructing the accumulated appearance histogram after elapse of a latency delay time period from a beginning of the classified segment.

7. The method of claim 6, wherein the latency delay time period ranges from one to two seconds.

8. The method of claim 6, wherein the object is a train car, further comprising:

defining the region of interest to comprise a train track area within an image scene of a stream of the video image data, and to exclude a pedestrian platform within the image scene that is adjacent to and visually distinguishable from the object track area.

9. An article of manufacture, comprising:

a computer readable hardware storage device having computer readable program code embodied therewith, the computer readable program code comprising instructions for execution by a computer processing unit that 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, wherein 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, exchange the object present and the object absent labels that are assigned to each of the video frames;

classify segments of the input video data comprising consecutive video frames that are each assigned the object absent label and the static label as “

no object present”

segments;

classify segments of the input video data comprising consecutive video frames that are each assigned the object present label and the motion present label as “

object present and in transition”

segments; and

classify segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label as “

object present and stopped”

segments.

10. The article of manufacture of claim 9, wherein the computer readable program code instructions for execution by the computer processing unit, further cause the computer processing unit to:

accumulate appearance features extracted from each of the frames into histograms for each of the classified segments of the input video data;

compare the accumulated histogram for a current one of the classified segments to the histograms of each of a plurality of segments that are previous in time to the current segment, wherein each of the compared plurality of segments occurred within a comparison time period of the current motion state segment, and wherein the comparison time period is selected to encompass each of the “

object present and stopped”

segment, the “

object present and in transition”

segment and the “

no object present”

segment under anticipated or historical object movement timings for the region of interest; and

change a current classification of the current segment to the “

object present and stopped”

classification, the “

object present and in transition”

classification or the “

no object present”

classification of a one of the compared plurality of segments that has an accumulated histogram that best matches the accumulated histogram for the current segment.

11. The article of manufacture of claim 10, wherein the computer readable program code instructions for execution by the computer processing unit, further cause the computer processing unit to select the auto-correction time period to encompass a plurality of object stopping events over a time frame longer than an anticipated or dynamically observed time that an object is stopped within the region of interest for one stopping event.

12. The article of manufacture of claim 11, wherein the computer readable program code instructions for execution by the computer processing unit, further cause the computer processing unit to accumulate the appearance features extracted from each the frames into the histograms for each of the classified segments of the input video data further by constructing the accumulated appearance histogram after elapse of a latency delay time period from a beginning of the classified segment.

13. The article of manufacture of claim 12, wherein the object is a train car, and wherein the computer readable program code instructions for execution by the computer processing unit, further cause the computer processing unit to define the region of interest to comprise a train track area within an image scene of a stream of the video image data, and to exclude a pedestrian platform within the image scene that is adjacent to and visually distinguishable from the object track area.

14. A system, comprising:

a processing unit;

a computer readable memory in communication with the processing unit; and

a computer-readable storage medium in communication with the processing unit;

wherein the processing unit executes program instructions stored on the computer-readable storage medium via the computer readable memory and thereby;

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, wherein 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, exchanges the object present and the object absent labels that are assigned to each of the video frames;

classifies segments of the input video data comprising consecutive video frames that are each assigned the object absent label and the static label as “

no object present”

segments;

classifies segments of the input video data comprising consecutive video frames that are each assigned the object present label and the motion present label as “

object present and in transition”

segments; and

classifies segments of the input video data comprising consecutive video frames that are each assigned the object present label and the static label as “

object present and stopped”

segments.

15. The system of claim 14, wherein the processing unit executes the program instructions stored on the computer-readable storage medium via the computer readable memory, and thereby further:

accumulates appearance features extracted from each of the frames into histograms for each of the classified segments of the input video data;

compares the accumulated histogram for a current one of the classified segments to the histograms of each of a plurality of segments that are previous in time to the current segment, wherein each of the compared plurality of segments occurred within a comparison time period of the current motion state segment, and wherein the comparison time period is selected to encompass each of the “

object present and stopped”

segment, the “

object present and in transition”

segment and the “

no object present”

segment under anticipated or historical object movement timings for the region of interest; and

changes a current classification of the current segment to the “

object present and stopped”

classification, the “

object present and in transition”

classification or the “

no object present”

classification of a one of the compared plurality of segments that has an accumulated histogram that best matches the accumulated histogram for the current segment.

16. The system of claim 15, wherein the extracted, accumulated features are color features;

wherein the accumulated histograms are color histograms;

wherein the processing unit executes the program instructions stored on the computer-readable storage medium via the computer readable memory, and thereby accumulates the appearance features extracted from each of the frames into the histograms for each of the classified segments of the input video data by constructing the accumulated appearance histogram after elapse of a latency delay time period from a beginning of the classified segment; and

wherein the latency delay time period ranges from one to two seconds.

17. The system of claim 16, wherein the object is a train car, and wherein the processing unit executes the program instructions stored on the computer-readable storage medium via the computer readable memory, and thereby defines the region of interest to comprise a train track area within an image scene of a stream of the video image data, and to exclude a pedestrian platform within the image scene that is adjacent to and visually distinguishable from the object track area.