使用電腦視覺技術在複雜房間房境內作自動車導航

Abstract

本論文提出以電腦視覺技術，判斷路面狀況，作為自動車在複雜房間 環境內導航的一種方法。本方法是利用事先學習所得到的路面資訊，來判 斷何者是路面，何者是障礙物，並進而運用走在路面中間的觀念讓自動車 航行於房間內。我們提出一種模型學習的方法，用以建立自動車航行的路 徑以及學習停車環境，以便讓自動車航行並停車到任何你所指定的地方。 我們也提出一套產生路徑的方法。我們也提出一套讓自動車停車到固定位 置的方法。最後，我們製作一輛實際的自動車作測試，並達到自動車在複 雜房間環境內作平順且安全航行的目標。經過多次成功的實驗，我們證明 了此套方法的可行性。 An integrated approach to autonomous land vehicle (ALV) guidance by judgingthe conditions of routes to generate appropriate paths using computer visiontechniques in complex room environments is proposed. Judgement of the existenceof a route or a collision is based on the information of route surfaces which islearned in advance. A method of guiding the ALV to the central path of a routeis posposed to make the ALV navigate successfully in a complex room environment.Techniques of model learning are proposed to extract the route features, recordthe information about turnings, and learn the park location environments.Techniques of path generation are proposed to generate paths automatically.Furthermore, three strategies for the ALV to park in desired locations are alsoproposed. Finally, a real ALV was used as a test-bed, and smooth and safenavigation sessions were achieved. Lots of successful experiments confirm thefeasibility of the proposed approach.