本論文使用視覺感測器輔助機器人進行巡航任務，研究的議題包括攝影機校準方法的應用、單攝影機同時定位與建圖、與多攝影機系統輔助巡航等。應用攝影機校準方法進行攝影機內部參數與外部參數的校準，以及多攝影機之間位移向量與旋轉矩陣的求算；單攝影機搭配反深度參數化方法進行影像特徵初始化，以達到無延遲的狀態初始化程序；使用環境中多組的視覺感測器，改善巡航中機器人狀態估測的準確度。 A robot navigation system using multiple vision sensors is developed in this thesis. Research topics include the application of camera calibration method, the development of visual monocular simultaneous localization and mapping (MonoSLAM), as well as the robot navigation using multiple cameras. The camera calibration method is applied to calibrate the intrinsic and extrinsic parameters of each camera. It is also employed to determine the translation vector and rotation matrix between every two cameras. The inverse depth parameterization method initializes the landmark state of MonoSLAM and provides an un-delay state initialization procedure. Multiple cameras in the environment are used in this research to improve the accuracy of robot state estimation during navigation.
