System and method for robust motion detection

• US 9,245,187 B1
• Filed: 07/07/2014
• Issued: 01/26/2016
• Est. Priority Date: 07/07/2014
• Status: Active Grant

• Current Assignee: Geo Semiconductor Incorporated
• Original Assignee: Geo Semiconductor Incorporated
• Inventors: Varghese, Gijesh
• Primary Examiner(s): BHATNAGAR, ANAND P
• Application Number:
  US14/324,593
• Publication Number:
  US 20160004912A1
• Time in Patent Office: 568 Days
• US Class Current: 1/1
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10032 Satellite or aerial image; Remote sensing
• CPC Class Codes: G06T 7/194 involving foreground-background segmentation
• CPC Class Codes: G06T 7/215 Motion-based segmentation
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/50 by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
• CPC Class Codes: G06V 20/40 in video content extracting overlay text G06V20/62; video retrieval G06F16/70; processing of video elementary streams in video servers H04N21/234; processing of video elementary streams in video clients H04N21/44
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69
• CPC Class Codes: H04N 19/136 Incoming video signal characteristics or properties
• CPC Class Codes: H04N 19/87 involving scene cut or scene change detection in combination with video compression

Abstract

Method and system for detecting objects of interest in a camera monitored area are disclosed. Statistical analysis of block feature data, particularly Sobel edge and spatial high frequency responses is used to model the background of the scene and to segregate foreground objects from the background. This technique provides a robust motion detection scheme prone to catching genuine motions and immune against false alarms.

Claims

1. A method for detecting motion in a sequence of video frames captured from a scene, each frame comprising a plurality of pixels grouped in a plurality of image blocks, said method comprising:

(a) receiving pixel data and block feature data for each of the plurality of blocks of a current frame and a previous frame, the block feature data being at least one of Sobel edge and spatial high frequency response values for each pixel averaged over the block;

(b) classifying the blocks as one of background, strong foreground, and weak foreground based on temporal profile of the block feature data;

(c) producing an initial list of rectangles that enclose a plurality of connected foreground block, wherein each rectangle is assigned with a strength score and a frame by frame tracking count;

(d) identifying the rectangles as one of;

(i) new, (ii) persistent and (iii) recurring based on their strength score and tracking count;

(e) validating the new rectangles by comparing their constituent block data with that of corresponding collocated blocks from the previous frame; and

(f) producing a final list of rectangles comprising validated new, recurring and persistent rectangles.

2. The method of claim 1, wherein distribution of the at least one feature data, monitored for a period of time, is represented in a histogram.

3. The method of claim 2, wherein a block is classified as background if the normalized distance of the block feature value from the mean of the histogram is smaller than a first threshold, or as foreground if larger.

4. The method of claim 3, wherein the strength score for the rectangle is calculated as total number of strong foreground blocks within the rectangle.

5. The method of claim 2, wherein a foreground block is classified as strong if the feature value has a deviation from a short term running average larger than a second threshold, or weak if smaller.

6. The method of claim 1, where background of the scene is modeled by mapping the background blocks from a plurality of frames.

7. The method of claim 6, wherein the background model is used for inter-frame encoding of the sequence of video frames.

8. The method of claim 7, wherein encoding bit-rate is lowered by skipping frames when no activity is detected in the scene.

9. The method of claim 1, wherein the spatial high frequency and the Sobel edge features data are collected and processed simultaneously.

10. The method of claim 9, wherein a first background map produced from the Sobel edge data is combined with a second background map produced from the spatial high frequency data to produce a final background map.

11. The method of claim 9, wherein a block is classified as foreground if at least one of the features classifies the block as a foreground block.

12. The method of claim 1, wherein the tracking count is incremented when an overlap with a rectangle from the initial list of rectangles in the previous frame is found.

13. The method of claim 1, wherein frame to frame relative changes in size and shift in position of the rectangles are used to validate a match.

14. The method of claim 1, wherein a rectangle is validated as a new rectangle if the following measures are above given thresholds:

(i) the Sum of Absolute Differences (SAD) between the pixel data of the current rectangle and its collocated rectangle from the previous frame;

(ii) L2 norm between histograms of the two rectangles; and

(iii) foreground strength of the rectangle is above a strength threshold.

15. The method of claim 1, wherein an alarm is issued upon validation of the new rectangles.

16. The method of claim 1, wherein validation of the new rectangles triggers at least one of recording, transmission and content analysis of the frames.

17. An electronic processor comprising:

(a) means for receiving data from a sequence of video frames captured from a scene, wherein each frame comprises a plurality of pixels grouped in a plurality of image blocks;

(b) means for receiving block feature data for each block of a current frame and a previous frame, the block feature data being at least one of Sobel edge and spatial high frequency response values of each pixel averaged over the block;

wherein the processor is configured to detect motion in the scene by;

(i) classifying the blocks as one of background, strong foreground, and weak foreground based on temporal profile of the block feature data;

(ii) producing an initial list of rectangles that enclose a plurality of connected foreground block, wherein each rectangle is assigned with a strength score and a frame by frame tracking count;

(iii) identifying the rectangles as one of new, persistent and recurring based on their strength score and tracking count;

(iv) validating the new rectangles by comparing their constituent block data with that of corresponding collocated blocks from the previous frame; and

(v) producing a final list of rectangles comprising validated new, recurring and persistent rectangles.

18. The processor of claim 17, wherein said processor is configured to model background of the scene by mapping the background blocks from a plurality of frames.

19. The processor of claim 18, wherein said processor is configured to implement the background model in inter-frame encoding of the sequence of video frames.

20. A system for detecting motion in a scene, said system comprising:

(a) at least one stationary positioned capture device to capture a sequence of video frames from the scene, wherein each frame comprises a plurality of pixels grouped in a plurality of image blocks;

(b) an input interface configured for receiving pixel data and block feature data for each block of a current frame and a previous frame, the block feature data being at least one of Sobel edge and spatial high frequency response values of each pixel averaged over the block; and

(c) a processing unit configured for;

(i) classifying the blocks as one of background, strong foreground, and weak foreground based on temporal profile of the block feature data;

(ii) producing an initial list of rectangles that enclose a plurality of connected foreground block, wherein each rectangle is assigned with a strength score and a frame by frame tracking count;

(iii) identifying the rectangles as one of new, persistent and recurring based on their strength score and tracking count;

(iv) validating the new rectangles by comparing their constituent block data with that of corresponding collocated blocks from the previous frame; and

(v) producing a final list of rectangles comprising validated new, recurring and persistent rectangles.