Linear system based, qualitative independent motion detection from compressed mpeg surveillance video

• US 8,164,629 B1
• Filed: 09/29/2009
• Issued: 04/24/2012
• Est. Priority Date: 02/12/2003
• Status: Active Grant

• Current Assignee: The Research Foundation for The State University of New York (State University of New York)
• Original Assignee: The Research Foundation for The State University of New York (State University of New York)
• Inventors: Zhang, Zhongfei
• Primary Examiner(s): Rao, Andy S.
• Assistant Examiner(s): HALLENBECK-HUBER, JEREMIAH CHARLES
• Application Number:
  US12/569,828
• Time in Patent Office: 938 Days
• Field of Search: 348/142, 348/144, 348/152, 348/155
• US Class Current: 348/155
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/30232 Surveillance
• CPC Class Codes: G06T 7/20 Analysis of motion motion estimation for coding, decoding, compressing or decompressing digital video signals H04N19/43, H04N19/51
• CPC Class Codes: H04N 19/48 using compressed domain processing techniques other than decoding, e.g. modification of transform coefficients, variable length coding [VLC] data or run-length data motion estimation in a transform domain H04N19/547; processing of decoded motion vectors H04N19/513
• CPC Class Codes: H04N 19/543 using regions
• CPC Class Codes: H04N 19/61 in combination with predictive coding

Abstract

The present invention features a qualitative method to detect independent motion revealed in successive frames of a compressed surveillance MPEG video stream using linear system consistency analysis without decompression of the stream, identifying the segments containing independent motion in a real-time or faster manner, for the retrieval of these segments. The linear system is constructed using the macroblocks of MPEG compressed video frames. The normal flow value of the macroblock is obtained by taking the dot product between the macroblock gradient vector, computed by averaging the four block gradient vectors, and the motion vector of this macroblock. The normal flow value is filtered for inclusion in the linear system, and the statistic of the matrices of the resulting linear system is determined, filtered to screen out false negatives and outliers, and used to determine the presence or absence of independent motion.

Claims

1. A method for detection of independent motion of a plurality of objects presented in a time sequence of images acquired from different perspectives, comprising:

a) processing the plurality of images with an automated processor to generate a linear system representing the plurality of objects in the form of a matrix D having at least three dimensions stored in a memory;

b) analyzing the stored matrix D with an automated processor to qualitatively determine the consistency of said linear system based on a relationship of respective rank of the matrix D an augmented matrix DB, substantially without quantitatively solving said linear system consistency being represented as a decision; and

c) outputting the consistency decision, representing whether or not there is independent motion between the plurality of objects represented in the time sequence of images.

2. The method according to claim 1, wherein:

Dξ

=b;

3. The method according to claim 1, wherein a change in camera position with respect to objects represented in the time sequence of images does not render the linear system inconsistent, while a motion of one abject with respect to another object renders the linear system inconsistent.

4. The method according to claim 1, wherein said analyzing comprises performing a consistency analysis of said linear system.

5. The method according to claim 1, wherein processing and said analyzing are performed in at least real time.

6. The method according to claim 1, wherein the time sequence of images are in an Motion Picture Experts Group 2 (ISO 13818-2) digital video format.

7. The method according to claim 2, wherein sequence of images is compressed in a digital video format which comprises blocks comprising arrays of image pixels, and macroblocks comprising arrays of the blocks, further comprising estimating a motion parameter from a transformation at least one of said macroblocks over time, without rendering the compressed digital video data as a raster image.

8. The method according to claim 2, further comprising estimating said motion parameters using a normal flow computation.

9. A method of detecting independent motion in a stream of digitally compressed video data, the steps comprising:

a) selecting a pair of video frames from a stream of digitally compressed video data representing a three dimensional scene;

b) automatically building a linear system representing the three-dimensional scene for said pair of video frames using at least a three-dimensional to two-dimensional affine transform in which parallel lines remain parallel after affine transformation, and a change in camera position with respect to objects represented in the pair of video frames does not render said linear system inconsistent;

c) automatically analyzing said linear system represented in a matrix D to qualitatively determine the consistency thereof, substantially without quantitatively solving said linear system, based on an equivalence of a respective rank Rank(D) of the matrix D and rank Rank(Db) of an augmented matrix Db, the consistency being represented as a decision; and

d) storing the decision in a memory,

wherein;

Dξ

=b,

10. The method according to claim 9, wherein Rank is defined such that a matrix A is of rank r if it contains at least one square submatrix of order r with a non-zero determinant, while the determinant of any square submatrix having (r+1) or higher order, possibly contained in A, is zero, the equivalence of the of respective rank Rank(D) of the matrix D and rank Rank(Db) of an augmented matrix Db being determined by applying a Singular Value Decomposition to both matrix D and augmented matrix Db, defining

11. The method according to claim 9, wherein a change in camera position of a camera acquiring the video frames of the digitally compressed video data, with respect to objects represented in the time sequence of images does not render the linear system inconsistent, while a motion of one abject with respect to another object renders the linear system inconsistent.

12. The method according to claim 9, wherein automatically building a linear system and automatically analyzing said linear system are performed faster than real time.

13. The method according to claim 9, wherein the digitally compressed video data is in an Motion Picture Experts Group 2 (ISO 13818-2) digital video format.

14. The method according to claim 9, wherein digitally compressed video data is comprises blocks comprising arrays of image pixels, and macroblocks comprising arrays of the blocks, further comprising estimating a motion parameter from a transformation at least one of said macroblocks between respective ones of the video frames.

15. The method according to claim 9, further comprising estimating said motion parameter using a normal flow computation.

16. The method according to claim 9, wherein the consistency of the linear system is determined substantially without spatial segmentation in the image domain.

17. An apparatus for detection of independent motion of a plurality of objects presented in a time sequence of images acquired from different perspectives, comprising:

a) a processor operating on a data flow in at least in real time, to produce a sequence of outputs corresponding to a current image and a prior image, configured to;

receive and process the current image in conjunction with the previously received prior image, to generate a linear system representing the plurality of objects represented in the images in the form of a matrix D having at least three dimensions stored in a memory;

analyze the stored matrix D to determine the qualitative consistency of the linear system based on a relationship of respective Rank of the matrix D and the Rank of an augmented matrix Db, substantially without quantitatively solving the linear system and

output the consistency decision after determination, representing whether or not there is independent motion between the plurality of objects represented in the time sequence of images.

18. The apparatus according to claim 17, wherein:

Dξ

=b;

19. The apparatus according to claim 18, wherein time sequence of images is compressed in a digital video format which comprises blocks comprising arrays of image pixels, and macroblocks comprising arrays of the blocks, further comprising estimating a motion parameter of the affine transform from a movement of at least one of said macroblocks over time, without rendering the compressed digital video data as a raster image, using a normal flow computation.

20. The apparatus according to claim 18, wherein Rank is defined such that a matrix A is of rank r if it contains at least one square submatrix of order r with a non-zero determinant, while the determinant of any square submatrix having (r+1) or higher order, possibly contained in A, is zero, the equivalence of the of respective rank Rank(D) of the matrix D and rank Rank(Db) of an augmented matrix Db being determined by applying a Singular Value Decomposition to both matrix D and augmented matrix Db, defining