本研究比較兩種機器人的定位演算法：過濾器型式(filter-based)與視覺里程計(visual odometry, VO)。感測器則使用雙眼視覺系統，以手持的方式在環境中行進感測任務。過濾器型式的定位方法使用擴張型卡爾曼過濾器進行同時定位與建圖，再取用其定位部分的功能；視覺里程計的定位方法則是以為位置求解問題為基礎，搭配隨機取樣一致程序(RANSAC)，求解機器人的定位問題。論文的研究議題包括：進行雙眼視覺感測器的校準與量測；實現過濾器型式與視覺里程計等兩種機器人定位演算法；比較兩種演算法在未知環境中執行定位任務之優缺點。 This study investigates the performance of two robot localization algorithms based on filter method and visual odometry (VO), respectively. A handhold binocular vision system is used as the sensing device which is free-moving in the environment. The filter-based localization algorithm uses the extended Kalman filter to simultaneously implement the tasks of localization and mapping, and then utilizes the localization function. The VO-based localization algorithm uses random sample consensus (RANSAC) to solve the location determination problem. This study investigates the issues include: calibration and measurement of binocular vision system, implementation of filter-based and visual odometer based robot localization algorithms, discussion of the advantages and disadvantages of these two localization algorithms using in an unknown environment.
