視覺伺服控制的概念就是視覺系統偵測與可動機構系統控制技術的結合,本文建立了一套具有雙CCD攝影機系統,摸擬人類的眼睛,其中包含了一個控制馬達,可上下轉動,以及一個有5個自由度的機械手臂系統,其第一軸採用中空旋轉平台設計,仿人類的腰部功能,負責調整機器手臂的工作面向。而手臂部份則使用4顆步進馬達設計成右手型式。利用影像處裡、立體視覺對環境進行偵測及影像識別,藉此引導可動機構系統的動作,在真實環境可識別特定顏色物體及偵測移動目標物,並由末端的夾爪,精準的抓取空間中的目標物。本文以靜態及動態的追蹤實驗驗證立體視覺演算的正確性,再經由動態追蹤後的抓取目標物完成整個視覺伺服系統的結合。本文並利用曲線擬合中的線性回歸分析修正演算後的誤差,計算出更加接近目標的正確座標。本文運用了影像處理中的色彩空間轉換、多重閥值顏色篩選、影像二值化、影像重整、質心計算及立體視覺演算。手臂部份運用的理論則為正逆向運動學,控制部份則使用運動控制卡,影像擷取卡,控制軟體則使用Matlab。最後成功完成本文的實驗結果。
The concept of visual servoing is combining visual system with movable mechanism technology. In this paper, a dual CCD camera system is established for simulating human’s eyes, which contains a control motor for rotating up and down, as well as a 5 degrees of freedom robotic arm system, which adopted the first hollow shaft rotating platform design, imitation human lumbar function, responsible for adjusting the robot’s work-oriented. The arm portion is designed to use four stepper motors like right arm of human. The modified stereovision algorithm is on the environment detection and image recognition, thereby guiding the movable robotic arm system operation in the real environment, to get the designated color object and identifies specific moving target by the end-effector. In this paper, static and dynamic tracking experiments are designed to verify the correctness of stereovision algorithm, and then getting target after tracking to complete the entire visual servo system. This paper also use curve fitting linear regression analysis to calculate the corrected errors, and let us close to the correct coordinates of the target. This paper used the image processing including color space conversion, multi- threshold screening, binary image processing, centroid calculation, and modified stereo vision algorithms. Theoretical part of the robotic arm is positive and inverse kinematics, and using motion control card, video capture card, the control software is Matlab. Finally, we got successful results in completion of our experiments.