在現今的社會裡，人物追蹤與辨識的重要性越來越高，舉凡隨處可見的人物監視系統，老人的居家看護照顧等等都需要利用到人物辨識的技術。而在另一方面，對人物辨識的功能需求也越來越多。從過去只對固定的背景進行分析，到現今許多對動態背景的人物分割；從過去只作人物位置的判斷，到現在希望能對人物外型進行分割，進一步分析場景中的人物行為。對於動態背景的分析，難度在於背景的變化使得無法與前景人物做出區分；加上要對人物外型進行分割，更難以確保人物外型的正確性；而對這些問題提出解決方法的研究也較為缺乏。
在此本論文提出一個利用色彩與運動資訊，結合模糊理論，並利用分水嶺分割法簡化外型分割程序之人物外型追蹤方法。
本篇論文主要分為三個部份：運動物體偵測、人物外型偵測、人物外型追蹤。在人物與動態背景有相對運動情況下，本論文首先利用背景與人物運動方式的不同，區分出背景與前景的人物，藉此得到人物的概略位置。得到概略的位置之後，本論文利用分水嶺分割法將場景分為數個區域，簡化分割人物外型的複雜度。進一步建立人物與背景的色彩模型，並以色彩與運動方式兩個特徵，導入一個模糊系統來區分出背景與位於前景的人物，最後得到人物的外型。
而對一個影像序列來說，在偵測到人物外型之後，本論文將利用在人物外型邊緣上的色彩資訊，在前後影像建立人物外型的對應性，對人物外型進行追蹤，並利用運動方式的資訊，校正在人物外型追蹤時產生的錯誤。
最後由實驗結果顯示出，本論文對人物外型的追蹤，提供了一個可行的方法；且能夠在動態背景的場景下，維持一定的正確性。

In modern society, human tracking has become an important subject in computer vision research. It has been applied to various fields including human surveillance system and home care system. On the other hand, the subsequent process of personal identification or classification is also an interesting and demanding research topic nowadays. In the past, the tracking process is usually performed under fixed background. However, tracking under dynamic background is the essential problem to resolve when the camera moves during the human tracking process. In other words, we have segment the tracked subject when the background is varying. The human position can not be extracted from the subtracted background, instead, we have to estimate the foreground (subject) from the background by assuming that the motion of foreground is different from the one of background. Although the tracking under dynamic background is more difficult when reliable and stable results are requested, the researches on this topic are still under developing. Our thesis is also focused on this important topic. In the proposed system, we utilized the color and motion information under a framework of fuzzy system to track the human contour under dynamic background. The watershed segmentation is also applied to simplify the contour segmentation procedure.
There are mainly three parts which are motion object detection, human contour detection, and human contour tracking, in the proposed system. In order to have the related relationship of motion between dynamic background and human, we first classify the human from background based on the difference in motion distribution to obtain a rough position of human. After getting the human position, we apply watershed segmentation to divide the image into several regions. This operation can reduce the computation complexity when we segment the human contour. We also build two color models for human and background, and use a fuzzy system which uses color and motion features as inputs to classify the human areas from background. After all subject areas are classified, we can then obtain the human contour.
In a video sequence, after detecting the human contour, we track the human contour by using the color information around the detected contour to establish the connectivity in the subsequent images. Additionally, we also combine the motion information which is obtained in the previous operation to adjust the wrong tracking results.
Trough the experimental results, we have shown that the proposed method is a useful for human contour tracking, and it can maintain reasonable accuracy for tracking under dynamic background.

[1] J. Shin, S. Kim, S. Kang, Seong-Won Lee, J. Paik, B. Abidi, M. Abidi, “Optical flow-based real-time object tracking using non-prior training active feature mode,” ELSEVIER Real-Time Imaging, vol. 11, pp. 204-218, 2005.
[2] F. Jean, R. Bergevin, and A. B. Albu, “Body Tracking in Human Walk from Monocular Video Sequences,” Proceedings of the Second Canadian Conference on Computer and Robot Vision, 2005.
[3] T. Gevers, “Robust Segmentation and Tracking of Colored Objects in Video,” IEEE Transactions on Circuits and Systems for Video Technology, vol. 14, no. 6, June 2004.
[4] Yokoyama M., and Poggio T., “A Contour-Based Moving Object Detection and Tracking,” Visual Surveillance and Performance Evaluation of Tracking and Surveillance, 2005. 2nd Joint IEEE International Workshop on 15-16, pp. 271 – 276, Oct. 2005
[5] A. Koschan, S. Kang, J. K. Paik, B. Abidi, M. A. Abidi, “Color active shape models for tracking non-rigid objects,” Pattern Recognition Letters, vol. 24, pp. 1751-1765, 2003.
[6] J.L. Wang, B. L. Lo, and G. Z. Yang, “Ubiquitous Sensing for Posture/Behavior Analysis,” IEE Proceedings of the 2nd International Workshop on Body Sensor Networks (BSN 2005), pp. 112-115, April 2005.
[7] C. R. Wren, A. Azarbayejani, T. Darrell, A. P. Pentland, “Pfinder: Real-Time Tracking of the Human Body”, IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 19, no. 7, July 1997.
[8] T. Zhao, T. Nevatia, “Tracking Multiple Humans in Complex Situations,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 26, no. 9, September 2004.
[9] M. Zaki, M. Y. El Nahas and M. Youssef, “EMBOT: An enhanced motion-based object tracker,” The Journal of Systems and Software, vol. 69, no. 1-2, pp. 149-158, Jan. 2004.
[10] D. Murray and A. Basu, “Motion Tracking with an Active Camera,” IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 16, pp.449-459, May 1994.
[11] G. Rigoll, H.Breit and F. Wallhoff, “Robust tracking of persons in real-world scenarios using a statistical computer vision approach,” Image and Vision Computing, vol. 22, no. 7, pp. 571-582, July 2004.
[12] H. Breit, G. Rigoll, “Improved Person Tracking Using a Combined Pseudo-2D-HMM and Kalman Filter Approach with Automatic Background State Adaptation,” Proc. IEEE International Conference on Image Processing(ICIP), October 2001.
[13] D. Kelly, F. Boland, “Motion Modeling Selection in Tracking Humans,” Irish Signals and Systems Conference 2006, IET 28-30, pp.:363 – 368, June 2006.
[14] Bouguet J. , “Pyramidal implementation of the Lucas Kanade feature tracker description of the algorithms. OpenCV Documentation, Micro-Processor Research Labs, Intel Corporation, 1999.
[15] Richard O. Duda, Peter E. art David G. Stork, “Pattern Classification,” pp. 527, A Wiley-Interscience Publication, John Wiley & Sons, Inc. , 2000.
[16] L. G. Shapiro and G. C. Stockman, “Computer Vision,” Prentice Hall, 2001.
[17] Rafael C. Gonzalez, Richard E. Woods, “Digital Image Processing,” Prentice Hall, 2001.
[18] 楊英魁、孫宗瀛、鄭魁香等，”模糊控制理論與技術,” 全華圖書，1996.