本論文利用微型飛行器(Micro Aerial Vehicle, MAV)來進行室內環境之自動導航。本文主要以單攝影機作為感測器,利用攝影機所得影像進行影像處理,而透過影像處理求得資訊偵測週遭環境,並用此進行飛行控制。本篇論文將利用光流法及消失點執行室內走廊導航任務,透過光流法找到局部導航點,而透過消失點找到全域導航點。局部導航點可以閃避動態或靜態障礙物,全域導航點則可以找到走廊中心位置。由於光流法進行導航任務會有重灑特徵點時,光流資訊無法估測之問題,本論文將提出雙程序光流法來克服此問題。本論文將局部導航點及全域導航點相結合,來進行微型飛行器之穿越走廊飛行規劃。由於室內導航任務中,常會因偵測不出牆面進而造成導航失敗,所以本文提出透過影像之光流方向統計分析,進行牆面偵測,來執行閃避牆面動作,進而規劃出漫遊飛行策略。接著,本文設計視覺伺服控制器來達成微型飛行器之各種飛行導航任務目標。最後實驗證實,雙程序光流法結合消失點的導航方法及牆面偵測的方法,可以有效的應用於室內環境。
This paper used MAV (Micro Aerial Vehicle) to navigate in the indoor environment. From the images captured by the monocular camera on the MAV, the information of the MAV's surroundings can be observed and utilized to navigate the MAV. Here, the optical flow and vanishing point are estimated and integrated to assist the navigation through the corridor. The optical flow obtained by the pyramid Lacus-Kanade algorithm is utilized to construct the local navigation point for avoiding the static of moving obstacles, and the vanishing point is estimated to be the global navigation point for indicating the trend of the corridor. Since the feature points tracked by the Lacus-Kanade algorithm must be re-sprinkled every several image frames for keeping the tracking correctness, the information of the optical flow cannot be employed as local navigation point at the moment of re-sprinkling feature points. The pyramid Lacus-Kanade algorithm with dual thread is proposed to overcome this problem. On the other hand, the wondering navigation in the room space is presented with the wall detection method which analyzes the distribution in direction of the optical flow. The visual servoing law is also developed here to accomplish the task designed by the corridor-passing or room-wondering navigation. The experimental results show the efficiency of our approaches for MAV automatic flying in indoor environment.