System and method for object tracking and activity analysis

• US 20080130948A1
• Filed: 09/13/2006
• Published: 06/05/2008
• Est. Priority Date: 09/13/2005
• Status: Active Grant

• Current Assignee: Verification Corporation
• Original Assignee: Verification Corporation
• Inventors: Ozer, Ibrahim Burak
• Granted Patent: US 7,526,102 B2
• Time in Patent Office: Days
• US Class Current: 382/103
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• CPC Class Codes: G06V 10/42 Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
• 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: G06V 40/161 Detection; Localisation; Normalisation

Abstract

Embodiments of the invention relate to methods and systems for providing real-time video surveillance of crowded environments. The method consists of several object detection and tracking processes that may be selected automatically to track individual objects or group of objects based on the resolution and occlusion levels in the input videos. Possible objects of interest (OOI) may be human, animals, cars etc. The invention may be used for tracking people in crowded environments or cars in heavy traffic conditions.

Claims

1. A method for object tracking in a surveillance area, the method comprising the steps of:

eliminating the background and identifying foreground regions using color models in compressed and uncompressed domains;

identifying one or more objects of interest (OOI) in low resolution, high occlusion levels;

selecting an appropriate detection and tracking algorithm based on an estimated congestion level and an estimated occlusion level;

detecting a group OOI;

tracking the group OOI in low resolution, high occlusion levels;

analyzing the surveillance area in low resolution, high occlusion levels; and

identifying one or more individual OOIs.

2. The method of claim 1, wherein the step of eliminating the background and identifying foreground regions using color models in compressed and uncompressed domains comprises the steps of:

adapting a background model using high-level temporal semantics;

adapting the background model using high-level scene semantics;

analyzing the surveillance area using a combination of high-level temporal abd high-level scene semantics; and

improving the robustness of the background model with temporal statistics.

3. The method of claim 1, wherein the step of identifying one or more objects of interest (OOI) in low resolution, high occlusion levels comprises the steps of:

extracting regions of interest that are divided into blocks based on a calibration of a camera used, and an analysis of sizes of objects in the area;

computing templates for low and high frequency components of the blocks;

training to match the computed templates to previously stored labeled templates; and

comparing the computed templates to the previously stored labeled templates to identify OOIs.

4. The method of claim 1, wherein the step of selecting an appropriate detection and tracking algorithm based on an estimated congestion level and an estimated occlusion level comprises selecting one or more of group object tracking algorithms and individual object tracking algorithms.

5. The method of claim 1, wherein the step of tracking the group OOI in low resolution, high occlusion levels comprises the steps of:

corresponding flow blocks to areas marked as potential OOI; and

computing confidence levels for group movements based on motion parameters;

6. The method of claim 1, wherein the step of analyzing the surveillance area in low resolution, high occlusion levels comprises the steps of:

estimating a congestion level;

estimating a flow of a group OOI; and

detecting possible events based on the estimated congestion level and flow.

7. The method of claim 1, wherein the step of identifying one or more individual OOIs comprises the steps of:

matching image blocks to templates of OOI by combining weighted shape, color or motion patterns of image blocks obtaining one or more data elements;

changing the weights of patterns by using the confidence levels of the patterns;

forming a list of OOI detected in the surveillance area to match to different foreground regions;

determining a pattern of motion of the OOI; and

determining high-level semantics of the surveillance area.