舉重槓鈴軌跡即時技術分析建構與研究

Realtime Weightlifting Barbell Trajectory Extraction And Performance Analysis

緒論：以往的舉重運動學分析常常需要使用到反光點及實驗室環境，不但成本高昂且無法即時呈現結果，本文將利用舉重運動及槓鈴軌跡移動之特性，設計視訊物件追蹤演算法，並以此演算法開發計算機輔助舉重訓練系統，自動為舉重影片繪製槓鈴軌跡，希望能達成即時回饋給教練並應用在比賽及訓練當中。方法：攝影機拍攝視訊的過程當中，同時擷取了被拍攝物件的時空域資訊。在本研究當中，我們利用空間域及時間域資訊提出一槓鈴自動追蹤演算法，自動化找尋槓鈴在畫面當中的座標，以此座標繪製槓鈴運動軌跡。除演算法之外，本文當中更利用此演算法實現計算機輔助舉重訓練系統，並利用此系統繪製軌跡，透過此軌跡，進一步分析選手的運動表現。結果：我們利用102學年度大專盃舉重賽為分析對象，視訊影像採用消費級數位相機或消費級數位攝影機於場邊攝影，從分析過程中發現自行開發的輔助訓練系統節省了93%的分析時間，並且，本系統產生之軌跡與市售運動學分析軟體所產生之軌跡一致，足以分析舉重不同階段之軌跡。結論：利用輔助訓練系統，我們可以透過槓鈴軌跡觀察選手的試舉過程是否有需要改進的地方或是可能因疲勞而導致受傷，本系統有低系統複雜性的特性，讓教練及選手在使用系統時能更簡單的使用。

Introduction: A realtime weightlifting automatic barbell trajectory extraction algorithm was proposed in this paper. We utilize the color and motion characteristics of the barbell to extract from the video sequence and further implement a computer-aided weightlifting training system by considering the barbell trajectory. Our aim was to create a new computer-aided training system which can utilize in a real competition and feedback analysis result to the coach in realtime. This system can effectively decrease the cost of the kinematic analysis which needs manual marker and lab environmental. Method: Spatial and temporal of the video sequence are both considering to find the coordinate of barbell in a frame and further connect the coordinates as the trajectory. The trajectory of the barbell was utilized in our computer-aided training system and evaluated the performance of the lifter. Result: 2013 Taiwan Weightlifting Universidad Game was utilized to verify our proposed algorithm. The video sequence was captured by consumer camcorders, the trajectory of barbell was then be extracted from these videos. Compare to traditional kinematic software, our proposed barbell trajectory extraction algorithm can effectively reduce the operation time more than 93% and gather the same trajectory as the traditional kinematic software. Conclusions: By observation the weightlifting barbell trajectory, we can evaluate the performance of the lifter. Our proposed algorithm effectively reduce the computational complexity. Automatically and effectively extracting the trajectory makes computer-aided weightlifting training system easily operated.