本論文提出一套新的基於Homography之視覺定位方法,以獲得目前載具相對於目標位置與姿態的關係,並實際應用於視覺路徑之導航。其中,視覺路徑為多個目標位置與姿態所構成,亦為環境資料庫。將事先於定點獲取的影像定義為目標位置;攝影機的視野方向定義為目標姿態。於視覺系統採用單一固定式攝影機;載具的控制方法採用一種簡易的直覺式控制。並且針對載具的初始位置為未知之情況,提出一套有效的環境資料庫建立與搜尋技術。最後,以實際的實驗結果驗證此系統架構。
The thesis presents a new homography-based localization approach which obtains autonomously the relation between the current position of vehicle and the target. The approach is applied to path navigation in an unknown environment for the first time. The visual path is consisted by the target positions (is called the environment database) and the target position is defined by an image taken previously in that position. The vision system consists of a fixed camera mounted on the vehicle and the control law is based on the instinctive, which achieves turn left, turn right or nothing. Efficient database building and searching approaches are presented to solve the unknown environment for the first time. Real experiment results are presented to show the performance of the framework.