現今社會科技不斷發展及變遷之下，自動化所帶來的便利與安全性廣受一般廠商的青睞也慢慢被一般社會大眾所肯定，本文將自動化的概念應用到自走車上來完成車輛導引任務，並利用視覺系統作為感測器將影像資訊傳送至主電腦系統再藉由無線傳輸的方式對自走車進行控制且達成停車任務。 在本文中，我們以視覺為基礎對自走車執行倒車入庫任務；首先，為了拍攝整個實驗環境，採用DFK21AF04工業攝影機搭配5-50mm/F1.3鏡頭架設於實驗環境上方，並將影像透過影像擷取卡傳至電腦端；有了環境影像之後，利用影像處理來獲得自走車位置及自走車與水平軸之角度，影像處理如影像校正(image calibration)、色彩空間轉換(color space conversion)、顏色辨識(color Recognition)、連通物件(connected component)、感興趣區域選取(region of interest)及歐蘇法(Otsu’s method)；得到所需的影像資訊後，採用模糊控制器來完成停車任務，並利用自走車位置與自走車方位來當作控制器之輸入；最後，電腦模擬及實際實驗結果來證明此系統是能夠順利對自走車完成倒車入庫任務。 In this thesis, we perform a vision-based garage-parking fuzzy control system for the car-like mobile robot. Firstly, in order to capture the images we set the DFK21AF04 CCD camera and 5-50mm/F1.3 lens above the experiment place. The image is captured by CCD camera and the computer video capture card. Secondly, we process the images to obtain the horizontal location of the mobile robot and the angle between horizontal axis and mobile robot. The image processes includes image calibration, color space conversion, color recognition, connected component, region of interest (ROI) and Otsu’s method. After the image processes, we use the location and orientation of the mobile robot as input variables of the controller. Then two kinds of fuzzy controllers are designed for this system. Lastly, by computer simulation, the control design is proved which can complete the garage-parking mission successfully. Furthermore, the real-time experiments show this control system is feasible and effective.
