使用視訊資料庫及動作辨識方法之棒球暗號輔助教學系統

A Baseball Sign Training System using the Video Database and Hands Movement Recognition

李宗人(Tsung-Jen Li)

指導教授 ： 張厥煒

國立臺北科技大學/電資學院/資訊工程系研究所/碩士(2012年)

https://doi.org/10.6841/NTUT.2012.00618

摘要

棒球是一種競賽性質的體育活動。在棒球比賽中，教練為贏取比賽所做的一切措施，發展至今日，成為了各樣種類的棒球戰術。教練利用棒球上的身體語言，將戰術傳達給選手，為了避免被識破，不能永遠都只用一套暗號型態。因此選手們平時必須練習自己球隊所設計的暗號型態，以免在比賽中執行錯誤的戰術。
本論文以棒球場上三壘指導教練下達的進攻戰術暗號為目標，建立一個棒球暗號動作的視訊資料庫系統。對每一個視訊資料作解讀後編輯出一組動作序列，搭配選擇設定好的戰術及人體部位對應表，隨機抽取視訊資料產生題庫，達到暗號練習的目的，讓選手能透過此系統掌握暗號的要領。
在對視訊資料作編輯時，由於手動編輯耗費人力，因此除了利用視窗化介面手動編輯外，系統利用影像動作辨識，判斷在各個時間點上，手部觸摸到的人體部位。首先透過人臉偵測以及人體比例，定義出各個人體部位的範圍。再以基於輪廓尋找的MHI運動檢測，結合膚色資訊，找出運動的手部位置，並分別對左右手進行個別追蹤。利用找出的手部位置，計算其出現在各個人體部位範圍的時間長度，作為判斷是否觸摸到部位的依據。實驗結果在判斷上有良好的效果，也利用影像辨識技術讓視訊的編輯上更為省時省力。

關鍵字

運動資訊 ； 視訊資料庫 ； MHI運動偵測

並列摘要

Baseball is a sport with competitive property. In baseball games, measures that implemented by coaches in order to win the contest, have been developed into varied baseball tactics. Coaches pass specific signals to their players through body language, aware of being recognized, signals and tactics matching should be changed after a few games. Thus, avoid misunderstanding the message and executing wrong tactics, players should practice the matching models of their own team often.
This paper has developed a computer-assisted learning system with video databases for baseball signs, which are mainly passed to players by their third base coach. By recognizing hands position, each video in database has its’ own motion queue. When using the system, users can choose which sign module being practice, and the system will choose some videos as exercises automatically. This system will help baseball players to recognize the coach’s movement more precise.
Because editing data by manual windows-based interface consume manpower resources, we add another Automatically generated function to help us to edit it. By using the method of motion recognition, the time code is found in the function that a coach touched the body parts we predefined in the video. First, we define the body parts scope in an image by using face detection and human proportion. Then it find the position of coach’s hands by using MHI motion detection and skin color information, and track the hands separately. Finally, we determine whether the coach touched the body parts we predefined by the length of time, and it automatically generate a motion queue to help us to edit it.

並列關鍵字

sport information ； video database ； MHI motion detection

參考文獻

[3] 陳明達、陳進發、陳鵬仁、黃建松，排球裁判執法判決電腦輔助評估系統之研發，體育學報 中華民國體育學會，第41卷第2期，頁83~96(2008.6).

[4] Aaron F. Bobick, James W. Davis, "The Recognition of Human Movement Using Temporal Templates", IEEE Transactions On Pattern Analysis and Machine Intelligence, Vol. 23, No. 3, March 2001.

[5] Daniel Weinland, Remi Ronfard, Edmond Boyer, "Free Viewpoint Action Recognition using Motion History Volumes", Computer Vision and Image Understanding, Volume 104, Issues 2-3, November 2006, pp.249-257

[6] S. Piva, L. Comes, M. Asadi, C. S. Regazzoni, “Grouped-People Splitting Based on Face Detection and Body Proportion Constraints”, Proceedings of the IEEE International Conferenc on Video and Signal Based Surveillance, 2006.

參考文獻