用電腦視覺及模糊輪向調整技術以沿路資訊作自動車室外導航

Abstract

本論文提出一套用於戶外自動車導航的技術。我們選擇道路上的明顯路標（例如各種路線和燈柱邊緣）作為導航的特徵，同時將燈柱位置和路況以人工測量的方式建立成簡單的模式資訊，用來作為自動車定位的依據以輔助導航。本研究利用一個控制自動車沿線導航的方法，可以控制自動車在任何畫有路線的道路上行駛，並且藉著模式資訊來預測燈柱出有在影像中的時機，快速地抓出燈柱邊緣而正確地作自動車定位。在影像處理方面，我們以即時分析所攝得影像亮度的方法，來克服戶外光線變化的影響；同時我們使用一個改進過的最小平方誤差法來求得較為精確的路線方程式。此外，我們還提出一個以車子偏離路線的距離和車子偏斜角度為參數、利用模糊集合理論解決自動車輪向控制的方法。另外，因為室外環境的高低起伏而使得車b子五不易定速航行，本文提出一個穩定車速的方法來達到平穩航行的效果。最後實際應用本論文所提的方法在一輛自動車的導引上，多次的實驗證實本系統可行。

An approach to the guidance of autonotnous land vehicles (ALV's) for outdoor navigation by road information using comater vision and fuzzy wheel adjustment techniques is proposed. Several types of landmarks or features on roads, such as different kinds of path lines and lampposts, are usilized for vision-based navigation. Lamppost positions and road conditions are extracted as the model information for navigation guidance. Using a line following method, the ALV can be guided on roads with path lines on them. By model information, the time that lampposts appear in the image can be predicted. The edges of lampposts are extracted to locate the ALV correctly and quickly. A dynamic image thresholding method is used in real time to solve the problem caused by the sunlight change A modified least square-error line approximation method is employed to extract path lines. Fuzzy set theory is applied to determine the turn angle for wheel adjustment. In addition, a speed adjustment approach is used to keep the ALV at stable speeds on fluctuating roads. Lots of successful navigation experiments using a real ALV confirm the effectiveness of the proposed methods.