臺灣博碩士論文加值系統

傳統監控系統大致分為兩種型態，一為以固定式攝影機對一特定區域進行監控，二為以可動式PTZ相機對特定目標進行監控。固定式攝影機常常會因為監控廣大區域而使得目標物過小致使身分不易辨認，而可動式PTZ相機需要手動控制難免有疏漏情況產生。本篇論文提出一智慧型雙攝影機之多目標物特寫追蹤系統，以場景攝影機建立背景機率模型(Background probability model )偵測各目標物位置及利用卡曼濾波器(Kalman filter)分別做追蹤，再配合雙攝影機之間的座標對應矩陣，使可動式PTZ攝影機能隨時針對目標物之特定部位(如上半身或臉部)進行特寫鏡頭追蹤。如此，我們便可以場景相機對多目標物做追蹤，並隨時以PTZ相機對特定目標物特寫並追蹤，隨時監控目標物之臉部表情或做行為分析。本系統以降解析度方式配合卡曼濾波器(Kalman filter)增快運算速率，在場景相機前景偵測部分每張frame處理時間約為0.073秒，而PTZ參數計算加上相機轉動處理時間約為0.064秒，總和每張frame0.137秒的處理時間，因此雙相機特寫追蹤每秒約可處理6~7張frame，以達到「即時」追蹤之目的。

Generally, video surveillance systems could be categorized into the one with only fixed cameras and the one with active PTZ cameras. The target tracking using the fixed camera often acquires a vague target image if the target is distant from the camera. On the contrary, target tracking using the active PTZ cameras may obtain a clear target image but it needs the manual control. To improve the abovementioned shortcomings, this paper proposes a cooperative dual-camera scheme that can track the multiple targets with a pantoscopic camera and track the close-up scene of the corresponding target simultaneously with an active PTZ camera. The target detection and tracking are developed by using the multiresolution-based spatial-temporal MOG model and the Kalman filter respectively. The cooperation between the pantoscopic and PTZ cameras is constructed by using the correspondence transformation matrices. Once the area of interesting in the tracked target is determined, its closed-up scene is also tracked. The area of interesting may be the region of face for recognizing the face expressions or the region of half-length body for recognizing the hand gestures. The time for target detection is about 0.073 second per frame and the time for the cooperative control of PTZ camera for the close-up scene tracking is 0.064 second per frame. Therefore, the close-up scene tracking consumes about 0.137 second per frame.

第一章 緒論
第二章 目標物追蹤
第三章 場景相機與PTZ相機之座標對應
第四章 實驗結果
第五章 結論

[1]Rosen, Jeffrey, “A cautionary tale for a new age of surveillance, ” The New York Times Magazine, Oct. 2001.
[2]C. Wren, A. Azarbayejani, T. Darrell, and A. Pentland, “Pfinder: real-time tracking of the human body,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 19, No. 7, pp. 780-785, July 1997.
[3]I. Haritaoglu, D. Harwood, L. S. Davis, ” W4：who? when? where? what? a real-time system for detecting and tracking people,” Third Face and Gesture Recognition Conference, pp. 222-227, Apr. 1998
[4]C. Anderson, Peter Burt, G. van der Wal, “Change detection and tracking using pyramid transformation techniques,” In Proceedings of SPIE - Intelligent Robots and Computer Vision, Vol. 579, pp. 72–78, 1985.
[5]J. Barron, D. Fleet, S. Beauchemin, “Performance of optical flow techniques,” International Journal of Computer Vision, Vol. 12, No. 1, pp. 42-77, 1994.
[6]C. C. Lien and S. C. Hsu, “The target tracking using the spatial-temporal probability model,” IEEE International Workshop on Nonlinear Signal and Image Processing, pp. 393-398, May 18-20 2005, Sapporo, Japan.
[7]A. Heyden and K. Astrom, “Euclidean reconstruction from constant intrinsic parameters,” In Proc.s 13th International Conference on Pattern Recognition, pp. 339–343, 1996.
[8]李昆霖, 基本矩陣之最佳化與歐基里德重建, 交通大學資訊科學研究所碩士論文, 2002.
[9]J. H. Han, “Contour matching using epipolar geometry,” IEEE Transactions on Pattern Analysis and Machine Intelligence, Vol. 22, No. 4, pp. 358-370, 2000.
[10]鄭智明, 雙攝影機校正研究, 中華大學資訊工程研究所碩士論文 ,2004.
[11]Y. Ren, C. S. Chua and Y. K. Ho, “Motion detection with nonstationary background,” Machine Vision and Applications, Vol. 13, No. 5-6, pp. 332-343, Mar. 2003.
[12]D. Salmond, “Target tracking: introduction and kalman tracking filters,” IEE Target Tracking: Algorithms and Applications, Vol. 2, pp. 1/1-1/16, 2001.
[13]G. Xu, and Z. Zhang, “Epipolar geometry in stereo, motion and object recognition,”.