本論文主要設計一人體動作辨識系統，經由影像處理的方式辨識及紀錄人類的動作，在紀錄的同時並以人型機器人及3D模型做出即時的模仿。此外，人形機器人在模仿的同時能夠經由加速度感測器達到自主平衡的功能。機器人以攝影機當作眼睛，藉由影像處理能在複雜環境的背景中，正確辨識人體身上的特殊標記，並以標記之相對位置，辨識並記憶人體動作流程。依標記點的座標位置建立一人形之多連桿模型，再利用資料庫的標記運動軌跡，讓機器人手舞足蹈，實現模仿人類的動作。模仿的動作包含了各種變化之雙手動作和下半身動作。論文的目標為人體動作的辨識、利用3D人形模型及機器人即時模仿人類的動作。由研究成果顯示，本論文提出之方法確實能夠藉由人體上標記點的位置，並即時計算出人類運動的姿勢，再透過3D人形模型與實際機器人同時展示出人類的運動動作。 此外，機器人能夠即時模仿人類動作且保持平衡是相當困難的工作，為了防止機器人在模仿人類動作的過程中因為傾斜角度過大而傾倒，所以本論文利用加速度感測器的回授設計了模糊控制器來保持機器人的平衡。The goal of this thesis is to design a human motion imitation system which can imitate human motions by recognizing the color marks pasted on the human body under any complex background. By the proposed system, the human motions are shown in real time in the monitor and the humanoid robot can imitate those motions simultaneously. By image processing, the system can calculate the joint angles of the human motion and take them into the 3D human model to imitate human motions. The imitated human motions include some motions such as “raise hand,” “kick forward,” “kick sideward,” and “Stand by one foot” etc. The experimental results show that the human motions can be recognized by image processing. At the same time, the human motions are imitated by a 3D human model and a real humanoid robot simultaneously. The construction of 3D human model and the interface between the human and robot are presented too. Moreover, it is difficult for a humanoid robot to imitate human motions and keep his balance simultaneously. In order to avoid the robot falling down when he imitates human motions, we design a fuzzy controller to guarantee the stability of the robot by using a G-sensor.
