Detection of abandoned and vanished objects

• US 8,422,791 B2
• Filed: 11/23/2009
• Issued: 04/16/2013
• Est. Priority Date: 11/24/2008
• Status: Active Grant

• Current Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Original Assignee: Canon Kabushiki Kaisha (Canon Inc.)
• Inventors: Liu, Qiang, McLeish, David Grant, Springett, Jarrad Michael
• Primary Examiner(s): STREGE, JOHN B
• Application Number:
  US12/623,856
• Publication Number:
  US 20100128930A1
• Time in Patent Office: 1,240 Days
• Field of Search: 382/103, 382/199, 382/206, 382/224, 382/235, 382/266, 340/568.1, 348/169
• US Class Current: 382/199
• CPC Class Codes: G06V 10/46 Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features colour feature extraction G06V10/56
• CPC Class Codes: G06V 20/52 Surveillance or monitoring of activities, e.g. for recognising suspicious objects recognising microscopic objects G06V20/69

Abstract

Disclosed herein are a method and system for classifying a detected region of change of a video frame as one of an abandoned object event and an object removal event, wherein a plurality of boundary blocks define a boundary of said region of change. For each one of a set of said boundary blocks (510), the method determines a predicted edge characteristic (520) and an observed edge characteristic (530) for said boundary block. The method then determines an individual block score (540) for said boundary block, based on said predicted edge characteristic (520) for said boundary block and said observed edge characteristic (530) for said boundary block. Once all of the set of boundary blocks have been processed, the method determines a global score (560) for said region of change, based on said individual block scores of said boundary blocks. The method then classifies the region of change (570) as an abandoned object event or an object removal event, based on how the overall score relates to a threshold.

Claims

1. A computer-implementable method for classifying a detected region of change of a video frame as one of an abandoned object event and an object removal event, wherein a plurality of boundary blocks define a boundary of the region of change, said method comprising the steps of:

(a) for each boundary block of a set of the boundary blocks;

(i) determining an observed edge characteristic of the boundary block, the observed edge characteristic including an observed edge orientation;

(ii) determining a predicted edge characteristic for the boundary block based on a configuration of the boundary block and other boundary blocks adjacent to the boundary block, the predicted edge characteristic including a predicted edge orientation; and

(iii) determining an individual block score for the boundary block, based at least upon a correspondence between the predicted edge orientation for the boundary block and the observed edge characteristic orientation for the boundary block;

(b) determining a total score for the region of change, based on the individual block scores of the boundary blocks; and

(c) classifying the region of change as the one of an abandoned object event and an object removal event based on the total score.

2. The method according to claim 1, wherein the region of change is defined by a set of spatially adjacent elements that differ from a background model of a scene in the video frame, and

wherein the plurality of boundary blocks comprise the outer boundary of those elements.

3. The method according to either one of claims 1 and 2, wherein each boundary block is a Discrete Cosine Transform (DCT) block of a JPEG image.

4. The method according to claim 1, wherein the boundary blocks used in steps (b) and (c) are from the same video frame.

5. The method according to claim 1, wherein the configuration is determined based on a relative position of at least one neighbouring boundary block for each boundary block.

6. The method according to claim 1, wherein the predicted edge characteristic is determined based on user input.

7. The method according to claim 1, wherein the predicted edge characteristic is determined based on a geometrical attribute of the boundary.

8. The method according to claim 3, wherein the observed edge characteristic of each one of the set of boundary blocks is determined based at least on DCT coefficients of the boundary block.

9. The method according to claim 1, wherein determining the total score for the region of change includes computing a sum of the individual block scores and normalizing the sum by the number of boundary blocks in the set of boundary blocks.

10. The method according to claim 1, wherein the set of boundary blocks includes all of the boundary blocks that define the boundary of the region of change.

11. The method according to claim 1, wherein the set of boundary blocks is a subset of the boundary blocks that define the boundary of the region of change.

14. The method according to claim 1, wherein the total score is compared to a predetermined threshold to classify the region of change as one of an abandoned object event and an object removal event, the predetermined threshold being determined by one of:

(a) averaging an edge strength of the region of change at initialization;

(b) averaging an edge strength of a relevant background model; and

(c) receiving the predetermined threshold from a user.

15. The method according to claim 14, wherein a total score above the predetermined threshold indicates an abandoned object, and a total score below the predetermined threshold indicates a removed object.

12. A non-transitory tangible computer-readable storage medium having recorded thereon a computer program for directing a processor to execute a method for classifying a detected region of change of a video frame as one of an abandoned object event and an object removal event, wherein a plurality of boundary blocks define a boundary of the region of change, said computer program comprising:

(a) code for processing each boundary block of a set of the boundary blocks to;

(i) determine a predicted edge characteristic for the boundary block based on a configuration of the boundary block and other boundary blocks adjacent to the boundary block, the predicted edge characteristic including a predicted edge orientation;

(ii) determine an observed edge characteristic for the boundary block, the observed edge characteristic including an observed edge orientation; and

(iii) determine an individual block score for the boundary block, based on at least a correspondence between the predicted edge orientation for the boundary block and the observed edge orientation for the boundary block;

(b) code for determining a global score for the region of change, based on the individual block scores of the boundary blocks; and

(c) code for classifying the region of change as the one of an abandoned object event and an object removal event, by comparing the global score with a predetermined threshold.

13. A camera system for classifying a detected region of change of a video frame as one of an abandoned object event and an object removal event, wherein a plurality of boundary blocks define a boundary of the region of change, said camera system comprising:

a lens system;

a camera module coupled to said lens system to capture the video frame;

a storage device for storing a computer program; and

a processor for executing the program, the program comprising;

(a) code for processing each one of a set of the boundary blocks to;

(i) determine a predicted edge characteristic for the boundary block based on a configuration of the boundary block and other boundary blocks adjacent to the boundary block, the predicted edge characteristic including a predicted edge orientation;

(ii) determine an observed edge characteristic for the boundary block, the observed edge characteristic including an observed edge orientation; and

(iii) determine an individual block score for the boundary block, based on at least a correspondence between the predicted edge orientation for the boundary block and the observed edge orientation for the boundary block;

(b) code for determining a global score for the region of change, based on the individual block scores of the boundary blocks; and

(c) code for classifying the region of change as the one of an abandoned object event and an object removal event, by comparing the global score with a predetermined threshold.