In this thesis, different active contour models are applied to dealing with the motion detection and region tracking problems. The information of the moving objects is obtained and classified based on the “active contour without edges” model.
Some previous works about contour modeling is introduced in Chapter 2. In this thesis, contour models are all implemented based on the level set theory.
When the background is static, the motion region can be obtained by subtracting the background. The absolute difference data can be classified by the “active contour without edges” model. In order to locate the moving objects in the center of the image, the camera must change its pan and tilt angles. As the camera moves, the background subtraction can no longer be used. In this thesis, we adopt the region tracking model to identify moving objects.
The region tracking model is used to find in the current image the region which is similar to the region defined in the previous frame. The original region tracking model has some problems when the color of the moving object is very similar to that of the surrounding clutter. According to the level surface constructed from the previous frame, a reliability weight is introduced to suppress erroneous estimations.
This is because it is reasonable to assume that the unknown level surface in the new frame should be very similar to that in the previous frame.
The statistical property is also taken into consideration in this thesis. A method that can estimate the prior probability is proposed for the maximum a posteriori (MAP) estimation. The probability model can strengthen the information of the inside region and eliminate the regions which should be outside.
A background modeling method is proposed and the static region is accumulated.
The motion region is extracted by collecting the inter-frame difference results. A simple model is proposed to update the level surface.
A surveillance system is built based on background modeling, motion detection, and region tracking. The system restarts the background modeling if the background model is not perfect or if the area of the tracked region varies too dramatically.
Because the region tracking process does not need the background information, the camera can be moved to always locate the moving objects in the center of the image.
In order to reduce the computational cost, a low-resolution level surface is used in this thesis. That is, the input image data is smoothed and down sampled before being processed. This results in problems in thin regions, such as the neck and limbs of a person. This is because the magnitude of the level surface in these thin regions is too small. To deal with this problem, the input image can be interpolated to increase the areas of thin regions. Then, we may use an associated high-resolution level surface to accomplish the region tracking task. The evolution will become more stable if the maximum magnitude of the level surface becomes larger inside the objects.
References
[1] Kass, M., Witkin, A., and Terzopoulos, D. “Snakes: Active contour models.”
International Journal of Computer Vision, 1:321–331. 1988.
[2] V. Caselles, R. Kimmel, and G. Sapiro, “Geodesic Active Contours,”
International Journal of Computer Vision 22(1), 61-79, 1997
[3] Dubrovin, B.A., Fomenko, A.T., and Novikov, S.P. “Modern Geometry — Methods and Applications I,” Springer-Verlag: New York.1984.
[4] Gilles Aubert, Pierre Kornprobst. “Mathematical Problems in Image Processing:
Partial Differential Equations and the Calculus of Variations,” Springer 2001.
[5] Tony F. Chan, Luminita A. Vese, “Active Contours Without Edges,” IEEE Transactions on Imgae Processing, Vol. 10, No. 2, pp.266-277, February 2001.
[6] Song Chun Zhu and Alan Yuille, “Region Competition: Unifying Snakes, Region Growing, and Bayes/MDL for Multiband Image Segmentation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 18, No. 9, pp.884-900, September 1996.
[7] Nikos Paragios and Rachid Deriche, “Coupled Geodesic Active Regions for Imgae Segmentation,” European Conference in Computer Vision, Nov. 1999.
[8] Abdol-Reza Mansouri, “Region Tracking via Level Set PDEs without Motion Computation,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, No.7, 947-961, July 2002.
[9] Alper Yilmaz, Xin Li, Mubarak Shah, ”Contour-Based Object Tracking with Occlusion Handling in Video Acquired Using Mobile Cameras,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 26, No. 11, 1531-1536, November 2004.
[10] Alper Yilmaz, Xin Li, Mubarak Shah, “Object Contour Tracking Using Level Sets,” Proceedings of Asian Conf. on Computer Vision (ACCV), South Korea, 2004.
[11] Osher, S. and Sethian, J.A., “Fronts Propagating with Curvature Dependent Speed: Algorithms Based on Hamilton-Jacobi Formulations,” Journal of Computational Physics, 79, pp.12-49, (1988)
[12] Stanley Osher, Ronald Fedkiw, “Level Set Methods and Dynamic Implicit
Surfaces,” Sprigner, 2003.
[13] Sethian, James Albert., “Level Set Methods and Fast Marching Methods, Evolving Interfaces in Computational Geometry, Fluid Mechanics, Computer Vision, and Materials Science,” 2nd ed., Cambridge University Press, 1999. New York.
[14] M.P. Wand and M. Jones, ”Kernel Smoothing,” Chapman and Hall, 1995
[15] Yizong Cheng, “Mean Shift, Mode Seeking, and Clustering,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.17, No. 8, pp.790-799, August 1995.
[16] Dorin Comaniciu, Peter Meer, “Mean Shift: A Robust Approach toward Feature Space Analysis,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 24, No. 5, pp.1-18, May 2002.
[17] Dorin Comaniciu, Visvanathan Ramesh, Peter Meer, “Kernel-Based Object Tracking,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol.25, No.5, pp.564-577, May 2003.
[18] Ahmed Elgammal, Ramani Duraiswami, David Harwood, Larry S. Davis,
“Background and Foreground Modeling Using Nonparametric Kernel Density Estimation for Visual Surveillance,” Proceedings of the IEEE, Vol. 90, No.7, 1151-1163, July 2002.
[19] R. T. Collins, A. J. Lipton, T. Kanade, H. Fujiyoshi, D. Duggins, Y. Tsin, D.
Tolliver, N. Enomoto, O. Hasegawa, P. Burt, and .Wixson, “A system for video surveillance and monitoring,” Carnegie Mellon Univ., Pittsburgh, PA, Tech.
Rep., MU-RI-TR-00-12, 2000.
[20] Weiming Hu, Tieniu Tan, Liang Wang, Steve Maybank, “A Survey on Visual Surveillance of Object Motion and Behaviors,” IEEE Transactions on Systems, Man, and Cybernetics ─ Part C: Applications and Reviews, Vol.34, No.3, pp.334-351, August 2004.
[21] Ying-li Tian, Arun Hampapur, “Robust Salient Motion Detection with Complex Background for Real-time Video Surveillance,” IEEE Computer Society Workshop on Motion and Video Computing, Breckenridge, Colorado, January 5 and 6, 2005
[22] Micheal Isard and Andrew Blake, “CONDENSATION ─ Conditional Density Propagation for Visual Tracking,” International Journal of Computer Vision
29(1), pp.5-28, 1998.
[23] Nikos Paragios, Rachid Deriche, “Geodesic active Contours and Level Sets for the Detection and Tracking of Moving Objects,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, No. 3, pp.266-280, March 2000.
[24] Nikos K. Paragios, “Geodesic Active Regions and Level Set Methods:
Contributions and Applications in Artificial Vision,” PhD Thesis in University of Nice Sophia Antipolis School of Computer Engineering, January 2000.
[25] Rosario Feghali, Amar Mitiche, “Spatiotemporal Motion Boundary Detection and Motion Boundary Velocity Estimation for Tracking Moving Objects With a Moving Camera: A Level Sets PDEs Approach With Concurrent Camera Motion Compensation,” IEEE Transactions on Image Processing, Vol. 13, No.
11, pp.1473-1490, November 2004.
[26] Roman Goldenberg, Ron Kimmel, Ehud Rivlin, Michael Rudzsky, “Fast Geodesic Active Contours,” IEEE Transactions on Image Processing, Vol. 10, No. 10, pp.1467-1475, October 2001.
[27] Yonggang Shi, William Clem Karl, “A Fast Level Set without Solving PDEs,”
Acoustics, Speech, and Signal Processing, 2005. Proceedings. (ICASSP '05).
IEEE International Conference on, Vol. 2, pp.97- 100, March 2005.
[28] Yonggang Shi, William Clem Karl, “Real-time Tracking Using Level Sets,”
International Conference on Computer Vision and Pattern Recognition (CVPR) June 2005.
作者簡歷
蔡孟修，西元1982 年 6 月 11 日出生於台中縣。西元 2000 年畢業於國立台中第
二高級中學，西元2004 年畢業於國立交通大學電機與控制工程學系。之後進入
國立交通大學電子研究所攻讀碩士學位。碩士論文題目為「基於等位函數法之運
動物體偵測與追蹤」。
聯絡地址：台中市西屯區大聖街455 號
電子信箱：[email protected]