研究目的為發展藉由立體視覺伺服控制擲球的機械手臂系統,機械手臂具有4個自由度,其末端為4隻手指機構的機械手掌,還有一組由雙攝影機構成的立體視覺系統。首先,立體視覺系統擷取環境影像,經過色彩空間轉換、顏色篩選、影像二值化、特徵萃取後代入質量中心演算法後,經由立體視覺演算法獲得目標點三維空間座標位置;其次,將物體放置在視覺範圍與手臂工作空間的重疊範圍內,並經由逆向運動學計算出機械手臂各軸所需的旋轉角度後手臂到達物體位置,控制機械手指將物體抓取至手掌上,並使手臂回到原始位置;預先在空間中選定數點目標點建立參數網狀格點,針對每點目標點,經過數次的擲球動作後,紀錄機械手臂各軸馬達的參數。因此,針對空間中任意目標點,使用雲形線擬合內插法估算機械手臂各軸馬達的參數,使機械手臂能擲球擊中目標。最後,根據實驗結果顯示,在距離機械手臂1.6公尺遠的平面上,機械手臂系統擲準的平均誤差約為5公分。
Purpose of study is that using visual servoing on pitching robotic arm system in this paper. A four-degree of freedom manipulators set with a four finger robotic hand. Furthermore, a stereo vision system has been made by two CCD cameras. Frist, the visual system can find the target point in three-dimension by using color conversion, color filter, feature extraction, binary images and centroid calculation. Second, choosing some known points to find arm parameter, motor speed and angles. Finally, the parameter function could be fitted by spline. According to experiment result shows that the system could pitch a target that is 1.6 meter far away and only 5 centimeter difference.