運用數位攝影機及全面最佳化之人體模型開發動作擷取系統

Abstract

人體動作分析已廣泛應用於運動醫學的探討、人因工程的研究、機器人的研發、以及電腦動畫的發展等各領域。目前已發展出各式各樣的人體運動學量測設備，其中光學式量測系統的使用最為頻繁，已發展出許多應用光電子技術的商業化光電子影像系統。其優點為具有極高的準確性，但其最大的缺點為價錢昂貴且維修花費較高。隨著數位影像技術的進步，一般數位攝影機的普及化，運用數位攝影機為基礎的人體運動學量測系統也因應而生，其主要的優點為價格較低，不受限於實驗環境，然其準確性則仍需再加強。 本研究目的在於發展一套使用數位攝影機的人體動作擷取軟體系統，並加入全面性皮膚移動誤差最佳化以提高準確性。使用兩台數位攝影機，從受試者側面拍攝步行動作，同時使用商業化動作擷取系統進行拍攝。經由視窗介面，將實驗拍攝的影像資料輸入本系統，影像處理得到標記點平面影像座標、再建立三維空間座標，配合下肢模型以及全面性最佳化方法校正標記點座標誤差，得到最佳化後的三維空間座標資料。並利用所開發的軟體，比較本系統與一般商業化動作擷取系統所獲得的空間座標誤差及關節角度。 研究結果顯示，使用一般數位攝影機所得之結果，經過全面最佳化方法校正之後，所得之標記點空間座標與關節角度，誤差有明顯的改善，與目前市面上商業化動作擷取設備準確度相當接近。

The motion analysis has been widely applied to various domains, such as research of sports medicine, ergonomics, the development of robot, and the computer animation . For the need of the motion analysis measurement, various kinds of human motion capture device have been developed. The optics type system has been most frequently adopted and, has been also applied into the development of commercialized photoelectron image system. The advantage of this kind of system is its extremely high accuracy; however, the biggest shortcoming is its high cost and its higher service expenditure for the repaired cost. Along with digital cameras being available, human motion capture system which utilizes digital cameras in accordance to be developed. The main advantage is its lower price, but its accuracy has to be improved. The purposes of this research are to develop a human motion capture software system using digital cameras in applying a global optimization method (GOM) to increase its accuracy. Two digital cameras were used to capture subject’s gait on level walking. And commercialized motion capture system were used simultaneously. Afterwards by using graphical user interface, users can easily load images captured by digital cameras to this system. After image processing, the three-dimensional coordinates reconstruction using lower extremity model and the global optimization method to reduce the artifact of markers three-dimensional motion data were obtained. These data were thus compared to those data obtained from present commercialized motion capture system. The results showed that the errors of the coordinates of the skin markers and joint angles significantly decreased after correction with GOM.