運動物體追蹤的程序分為運動物的偵測，以及後續運動軌跡的預測。本論文考慮一個從天花板下懸的網球，受重力作錐型單擺運動，運動追蹤控制器的目的是，經由裝設在地面上的數位相機取像，要讓同在地面上由馬達驅動的雷射筆同步將光束隨時指向擺動中的網球。針對此問題，本論文提出一以影像處理為基礎的定位追蹤平台，其控制迴圈內包括：取像、偵測、預測以及輸出等模組。錐型擺球體在數位相機所取像之後，先利用背景相減法等前置處理，取得目前的球體中心位置的估測值。之後，我們採取擴展型卡爾曼濾波器的方法估測下一瞬間運動物體所處的新位置。最後，為了隨時調整雷射筆在下一瞬間指向前一步驟裏所預測新的球體位置，我們推算如何由影像球體中心點，直接推估驅動雷射筆的馬達，在二維角度的方向上所需脈衝數指令的對應關係。本研究的實驗顯示我們的方法在控制雷射筆光指向運動中球體的控制目的上，明顯優於不預測、或是只用簡單一次差分估測下瞬間位置的方法。本論文的主要貢獻是在DSP內嵌式系統平台上，設計出能即時反應 (每秒處理17張影像)，並且能搭配實體輸出週邊的運動物體追蹤控制器。此技術可應用至智慧建築內的保全監測，並進一步推展至更高速移動物的追蹤。

Position tracking of a moving object involves the detection of the moving object at the present moment, and the prediction of it at the next. Applications of moving object tracking include surveillance systems at the lower end or missile intercept systems at the higher end. To investigate the technicalities involved, a conical pendulum, hung from a ceiling and tipped with a tennis ball, is to be targeted real-time by a laser mean light from a pointer mounted on a motor of two degree of freedom. This paper proposes an image-based moving object tracking system, whose control loop consists of the four modules: acquisition, detection, prediction, and output. After an image is taken from a digital camera, the following three steps are taken: Center of the ball is first processed by background subtracting method, position of the center at the next moment is then predicted by extended Kalman filter, and in the wake of this prediction, the angle discrepancy of the laser pointer needed to follow the light beam into the new predicted position is mapped into the pulses required at each of the two dimensions by the driving motor. Experiments have been conducted to show that the proposed approach performs much better than the otherwise tracking solutions when either no prediction or simple first order prediction approach are employed. The main contribution of this study is the development of an image-based position tracking system for moving objects. Built on DSP embedded platform, this position tracking system is capable of calculating drive commands output to targeting peripherals in order to follow the moving target with a real-time processing capability (17 images per second). The technology and platform developed in this study can also be applied to surveillance systems, and adjusted with higher end components to track higher speed moving objects.

[1] K. Rajgopal and N. Pappa, “State Estimation of a Nonlinear System by Neural Extended Kalman Filter,” Proceedings of Annual India Conference, pp. 1-6, Sept. 2006.
[2] A.E. Nordsjo, “A Constrained Extended Kalman Filter for Target Tracking,” Proceedings of IEEE Conference on Radar, pp. 123-127, Apr. 2004.
[3] F.E. Alsaqre and B. Yuan, “Multiple moving objects tracking for video surveillance systems,” Proceedings of ICSP Conference on Signal Processing, vol. 2, pp. 1301-1305, 2004.
[4] 宋炫慶，<基於粒子濾波技術的多個移動物體之即時視覺偵測與追蹤>，台灣科技大學資訊工程所，2003。
[5] M.S. Lee, “Detecting People in Cluttered Indoor Scenes,” Proceedings of IEEE Conference on Computer Vision and Pattern Recognition, vol. 1, no. 2, pp. 804-809, Jun. 2000.
[6] A. Talukder and L. Matthies, “Real-Time Detection of Moving Objects from Moving Vehicles Using Dense Stereo and Optical Flow,” Proceedings of IEEE/RSJ International Conference on Intelligent Robots and Systems, vol. 4, pp. 3718-3725, Oct. 2004.
[7] N. Wanopoulos, N. Vasanthavada and R.L. Baker, “Design of an Image Edge Detection Filter Using the Sobel Operator,” IEEE Journal of Solid-State Circuits, vol. 23, no. 2, pp. 358-367, Apr. 1988.
[8] O. Javed, K. Shafique, and M. Shah, “A Hierarchical Approach to Robust Background Subtraction Using Color and Gradient Information,” Proceeding of Workshop on Motion Video Computing, pp. 22-27, Dec. 2002.
[9] I. Haritaoglu, D. Harwood and L.S. Davis, “Hydra: Multiple people detection and tracking using silhouettes,” Proceedings of International Conference on Image Analysis and Processing, pp. 280-285, Sept. 1999.
[10] M. Amiri, H.R. Rabiee, F. Behazin and M. Khansari, “A new wavelet domain block matching algorithm for real-time object tracking,” Proceedings of ICIP Conference on Image Processing, vol. 3, pp. 961-964, Sept. 2003.
[11] Z.J. Yu, S.L. Dong, J.M. Wei, T. Xing and H.T. Liu, “Neural Network Aided Unscented Kalman Filter for Maneuvering Target Tracking in Distributed Acoustic Sensor Networks,” ICCTA International Conference on Computing: Theory and Applications, pp. 245-249, Mar. 2007.
[12] R.E. Kalman, “A New Approach to Linear Filtering and Prediction Problems,” Transactions of the ASME-Journal of Basic Engineering, vol. 82, pp. 35-45, 1960.
[13] X. Xu and B. Li, “Adaptive Rao–Blackwellized Particle Filter and Its Evaluation for Tracking in Surveillance,” IEEE Transactions on Image Processing, vol. 16, no. 3, pp. 838-849, Mar. 2007.
[14] M.P. Horton, “Real-time identification of missile aerodynamics using a linearized Kalman filter aided by an artificial neural network,” IEE Proceedings of Control Theory and Applications, vol. 144, no. 4, pp. 299-308, Jul. 1997.
[15] K.A. Kramer and S.C. Stubberud, “Tracking of Multiple Target Types with a Single Neural Extended Kalman Filter,” IEEE Conference on Intelligent Systems, pp. 463-468, Sept. 2006.
[16] 盧怡仁、蔡偉和，《數位信號處理平台在嵌入式系統的應用》，文魁資訊，2006。
[17] http://www.ti.com/
[18] http://en.wikipedia.org/wiki/Pendulum
[19] 鍾國亮，《影像處理與電腦視覺》，東華書局，2006。
[20] 蔡仁翔，<基於色彩關聯之即時多目標移動物體追蹤系統>，台灣科技大學自動化及控制工程所，2005。
[21] http://en.wikipedia.org/wiki/Kalman_filter
[22] 吳秉勳，<以分散式廣義卡曼濾波器為基礎之行動機器人多樣感測器資料融合>，義守大學機械與自動化工程所，2006。
[23] http://www.cs.unc.edu/~tracker/media/pdf/SIGGRAPH2001_CoursePack_08.pdf
[24] 蕭德琪，<使用平均移動法及卡式濾波器實現即時物體影像追蹤系統>，交通大學電機與控制工程所，2003。
[25] 陳昱仁，<機器人視覺追蹤之研究>，中正大學電機工程所，1996。