A Real-Time System for Contour-Based Object Detection and Tracking

Masayuki Yokoyama & Tomaso Poggio

The Problem:

We consider the problem of motion analysis, particularly detection and tracking of moving objects.  We suppose to use inputs from a single and static camera. (However, vibrations must be taken into account.) It has been a difficult problem to detect moving objects precisely because of following factors.

• illuminance changes
• small motions of background (e.g. leaves and edges of trees.)
• occlusion
• interference of objects

Motivation:

For designing an optimal circuitry for computer vision, tasks of low level vision cannot be ignored.  Proposing and designing an applicable motion analysis system using both low and high level tasks helps us developing LSIs for computer vision systems.

Previous Work:

In case of using a static camera, background estimation is a popular method for detecting foreground objects.  There are many researches of background estimation [1],[2],[3],[4].  However, it usually takes several seconds for background model estimation because the speed of illumination changes and small movement in the background are very slow.

Approach:

Our method is based on using lines computed by a gradient-based optical flow and an edge detector.  While it is known among researchers that gradient-based optical flow and edges are well matched for accurate computation of velocity, not much attention is paid to creating systems for detecting and tracking objects using this feature.  In our method, extracted edges by using optical flow and the edge detector are restored as lines, and background lines of the previous frame are subtracted.  Contours of objects are obtained by using snakes to clustered lines.  Detected objects are tracked, and each tracked object has a state for handling occlusion and interference.

Difficulty:

An accurate motion analysis without using expensive resources (e.g. large occupation of processor tasks and memory capacity) is required.  So how to decrease calculation time is one of the most difficult problem, especially in case of real-time analysis.  Such a problem of cost and performance often become a trade-off between them.

Impact:

We suppose that the platform hardwares for our research do not have high-speed processors and much memory capacity, so it is our hope that our research promotes development of a practical motion analysis system. Mobile platform is our ideal target, whose clock frequency and memory size are rigidly restricted for its low power consumption.

Acknowledgments:

This report describes research done at the Center for Biological & Computational Learning, which is in the McGovern Institute for Brain Research at MIT, as well as in the Dept. of Brain & Cognitive Sciences, and which is affiliated with the Computer Sciences & Artificial Intelligence Laboratory (CSAIL).  

This research was sponsored by grants from: Office of Naval Research (DARPA) Contract No. MDA972-04-1-0037, Office of Naval Research (DARPA) Contract No. N00014-02-1-0915, National Science Foundation (ITR/SYS) Contract No. IIS-0112991, National Science Foundation (ITR) Contract No. IIS-0209289, National Science Foundation-NIH (CRCNS) Contract No. EIA-0218693, National Science Foundation-NIH (CRCNS) Contract No. EIA-0218506, and National Institutes of Health (Conte) Contract No. 1 P20 MH66239-01A1.  

Additional support was provided by: Central Research Institute of Electric Power Industry (CRIEPI), Daimler-Chrysler AG, Compaq/Digital Equipment Corporation, Eastman Kodak Company, Honda R&D Co., Ltd., Industrial Technology Research Institute (ITRI), Komatsu Ltd., Eugene McDermott Foundation, Merrill-Lynch, NEC Fund, Oxygen, Siemens Corporate Research, Inc., Sony, Sumitomo Metal Industries, and Toyota Motor Corporation.

References:

[1] A. Elgammal, D. Harwood, and L. Davis. Non-parametric model for background subtraction. In Proc. European Conf. Computer Vision, volume II, pages 751?767, May 2000.

[2] L. Li,W. Huang, I. Y. H. Gu, and Q. Tian. Foreground object detection in changing background based on color co-occurrence statistics. In Proc. IEEE Workshop on Applications of Computer Vision, 2002.

[3] A. Mittal and N. Paragios. Motion-based background subtraction using adaptive kernel density estimation. In Proc. IEEE Conf. Computer Vision and Pattern Recognition, volume 2, pages 302?309, 2004.

[4] C. Stauffer and W. E. L. Grimson. Adaptive background mixture models for real-time tracking. In Proc. IEEE Conf. Computer Vision and Pattern Recognition, 1999.