棒球是目前國內熱門運動之一，一場棒球比賽，投手往往是擔任比賽勝敗的關鍵角色，而一個好投手需要可以精準的控制自己的球路，但是投手投出至捕手的好球帶位置是由主審來決定的，因此判別一個球為好壞球容易影響到一場比賽的勝負。目前大聯盟的K-Zone系統具備著分析投手好壞球的能力，這類系統越來越受到棒球愛好者的矚目，但是由於系統昂貴無法普及化，所以本文嘗試著實作此類的系統，以低成本的理念下，實作出一套能立即的重建棒球三維軌跡並且具備判斷好壞球能力的系統。

為了減少成本，本論文使用便宜的家用手持攝影機配合影像擷取卡，利用簡單容易製作的棋盤格特徵校正版，透過兩台電腦做攝影機校正，兩台電腦則是透過網路連接交換資料。偵測移動球體在捕手附近容易被干擾，所以本文透過卡爾曼濾波器來幫助預測球體落點，藉著雙攝影機抓取到的影像對應點重建球體三維軌跡並且將影像擷取後等相關資訊儲存在系統建置的資料庫上，使用者可於日後觀看幫助運動分析。

實驗結果顯示，在短距離空間場景測試下，假設虛擬好球帶的實驗中，其理論邊長與實際邊長誤差分別為0.31%、1.57%和0.51%，其值已達到相當精確的程度；在程式執行效能方面，一次軌跡重建與資料庫建立所需之執行時間最多僅需500ms的時間，可在下一球投出之前完成所有程式流程，達到 Soft Real-Time。

Baseball is one of the most popular sports in our country. A pitcher plays the major role in a baseball game. A good pitcher has to control his pitch precisely in order to be successful in the game. But the strike zone is up to the umpire's call. A strike or a ball called in a critical situation pitch can dramatically affect the result of a game. In MLB, they use K-Zone or similar system to analyze whether the pitched ball is a strike or a ball. These systems are attracting people who love baseball. But the system is too expensive to be available for simple training purpose. We try to implement such a system with the low cost in mind. Our system can reconstruct baseball three dimension trajectory in real time and judge automatically the pitched ball a strike or a ball.
For reducing cost, we use inexpensive home-use cameras witch feed two image capture cards into two laptops. We use easily produced checkerboard pattern to calibrate the camera’s parameters. Two laptops connect by each other via Ethernet interface to exchange data. The moving objects near catcher are prone to be interfered, so we use Kalman filter to help the system estimate moving trajectory of baseball. The baseballs are detected in corresponding image points by two cameras and are used reconstruct baseball three dimension trajectory. We store the video information to database in the system. User can review the historical video and related information to help ball location analysis in the future.
The experiment result shows that in the scenario of short distance the errors between the measured values of strike zone border and actual values of border are 0.31%, 1.57% and 0.51% respectively. This shows that our system is quite precise in short distance usage. The system performs the trajectory reconstruction and establishment of database requires 500ms. It can complete the program flow before the next pitch so that it can be used in live broadcast situation once we complete better GUI interface.

[1] A. Gueziec, “Tracking pitches for broadcast television,” Journal of Computer, Vol.35, pp. 38-43 (2002)
[2] L. Susman, “Calibration of a six-port reflectometer using projective geometry concepts,” Journal of Electronics Letters, Vol.20, pp. 9 (1984)
[3] 呂傑棋，「3D視覺校正軟體之研製」，碩士論文，中華大學，新竹 (1997)。
[4] 馬維章，「以多部攝影機實現埲球偵測與追蹤系統」，碩士論文，國立台灣科技大學，台北 (2008) 。
[5] 龔雨軒，「互動式投影遊戲之視覺平台發展」，碩士論文，國立台灣科技大學，台北 (2008) 。
[6] G. Bradski and A. Kaehler, Learning OpenCV: Computer Vision with the OpenCV Library, O, Reilly, pp. 370-374 (2008)
[7] G. Bradski and A. Kaehler, Learning OpenCV: Computer Vision with the OpenCV Library, O, Reilly, pp. 378-381 (2008)
[8] M. Kinsner, D. Capson and A. Spencer, “Scale-Space Feature Detection for Close Range Camera Calibration,” IEEE Canadian Conference on Electrical and Computer Engineering, pp. 1464-1467 (2007)
[9] R. D. Sampson, A. E. Peterson and E. P. Lozowski, “Photogrammetric calibration of a consumer grade flat-bed scanner,” IEEE Canadian Conference on Electrical and Computer Engineering, Vol.2, pp. 622-626 (1999)
[10] R. I. Hartley, “An algorithm for self calibration from several views,” IEEE Conference on Computer Vision and Pattern Recognition, pp. 908-912 (1994)
[11] W. Guanghui, Q. M. Wu and W. Zhang, “Camera Self-Calibration and Three Dimensional Reconstruction under Quasi-Perspective Projection,” IEEE Canadian Conference on Computer and Robot Vision, pp. 129-136 (2008)
[12] S. Florczyk, Robot Vision: Video-based Indoor Exploration with Autonomous and Mobile Robots, Wiley-VCH, pp. 113-122 (2005)
[13] Q. T. Luong and O. Faugeras, “Self-calibration of a moving camera from point correspondences and fundamental matrices,” Journal of Computer Vision, Vol.4, pp. 880-883 (1997)
[14] 繆紹綱，數位影像處理，培生出版社，台北，第392-395 頁、726-730 頁 (2009)。
[15] M. Sonka, V. Hlavac and R. Boyle, Image Processing, Analysis, and Machine Vision, PWS Publishing, pp. 68-102 (1999)
[16] J. R. Parker, Algorithms for Image Processing and Computer Vision, NY, John Wiley & Sons (1997)
[17] 劉翁昌，「複雜環境下之即時人臉偵測與辨識系統」，碩士論文，國立台灣科技大學，台北 (2009)。
[18] R. Yan, X. Song and S. Yan, “Moving object detection based on an improved gaussian mixture background model,” Journal of Computing, Communication, Control, and Management, Vol.1, pp. 12-15 (2009)
[19] S. Ju, X. Chen and G. Xu, “An Improved Mixture Gaussian Models to Detect Moving Object Under Real-Time Complex Background,” International Conference on Cyberworlds, pp. 730-734 (2008)
[20] 劉冠廷，「基於運動歷史影像實現行人追蹤與計數系統」，碩士論文，國立台灣科技大學，台北 (2010)。
[21] R. E. Kalman, “A new approach to linear filtering and prediction problems,”Journal of Basic Engineering, pp. 35–45. Retrieved (2008)
[22] G. Bradski and A. Kaehler, Learning OpenCV: Computer Vision with the OpenCV Library, O, Reilly, pp. 348-363 (2008)
[23] P. Shrestha, M. Barbieri, H. Weda and D. Sekulovski, “Synchronization of Multiple Camera Videos Using Audio-Visual Features,” IEEE Transactions on Multimedia, Vol.12, pp. 79-92 (2010)
[24] D. N. Brito, “Synchronizing Video Cameras with Non-Overlapping Fields of View,” Journal of Computer Graphics and Image Processing, pp. 37-44 (2008)
[25] 鄭子凡，「以多組CMOS感測器實現高速攝影系統之SOPC硬體架構，碩士論文，國立台灣科技大學，台北 (2007)。
[26] Jean-Yves Bouguet (2010),Camera Calibration Toolbox for Matlab
Available：http://www.vision.caltech.edu/bouguetj/calib_doc/htmls/example.html
[27] K. Kim, T. H. Chalidabhongse, D. Harwood, and L. Davis, “Background Modeling and Subtraction by Codebook Construction,” Proc. IEEE International Conference on Image Processing, vol. 5, pp. 3061-3064, (2004)
[28] P. KaewTrakulPong, and R. Bowden, “An Improved Adaptive Background Mixture Model for Real-time Tracking with Shadow Detection,” Proc. 2nd European Workshop on Advanced Video Based Surveillance Systems, AVBS01, VIDEO BASED SURVEILLANCE SYSTEMS: Computer Vision and Distributed Processing, (2001)
[29] H. S. Chen, H. T. Chen, W. J. Tsai, S. Y. Lee and J. Y. Yu, “Pitch-by-Pitch Extraction from Single View Baseball Video Sequences,” IEEE International Conference on Multimedia and Expo, pp. 1423-1426 (2007)
[30] J. Smart and K. Hock, Cross-Platform GUI Programming with wxWidgets, Prentice Hall (2005)
[31] P. K. Janert, Gnuplot in Action: Understanding Data with Graphs, Manning Publications (2009)
[32] G. Bourgeniere, P. Moulon, C. Rosenberger and W. Smari, “On the determination of 3D trajectory of moving targets by stereovision,” International Conference on Computers, Communications, & Signal Processing with Special Track on Biomedical Engineering, 1st, pp. 38-44 (2005)
[33] H. Fujiyoshi, S. Shimizu, T. Nishi, Y. Nagasaka and T. Takahashi, “Fast 3D position measurement with two unsynchronized cameras,” IEEE International Symposium on Computational Intelligence in Robotics and Automation, Vol.3, pp. 1239-1244 (2003)
[34] 鐘國亮，影像處理與電腦視覺，東華書局，台北 (2004)。
[35] W. Niblack, An Introduction to Digital Image Processing, Prentice-Hall, New Jersey (1986)
[36] E. Angel, Interactive Computer Graphics: A Top-Down Approach Using OpenGL, Addison Wesley (2008)
[37] R. C. Gonzalez, Digital Image Processing, New Jersey, Prentice Hall (2002)