當在一個未知的環境當中如何使移動式機器人,能夠快速的自我定位,瞭解本身的所在,為一個熱門的話題,本文利用全景式攝影機為感測器結合擴展是卡曼濾波器, 完成同步定位與環境地圖建立(Simultaneous Localization and Mapping, SLAM)演算法,讓機器人在移動探索環境的同時,能夠建立起環境的特徵地圖,並且定位出自身的位置。全景式攝影機擁有360 度的視野,能夠獲得更多的環境資訊,在於對特徵點的追蹤亦擁有較長的追蹤時間,以及較豐富的特徵點,加強了SLAM系統的穩健度。利用哈裡斯角點偵測配合尺度不變特徵轉換(SIFT)演算法的特徵描述,做為影像特徵匹配。利用影像參考點建立,讓機器人可以藉由匹配參考點資訊來達到定位,最後控制機器人於長廊中來回行走60 公尺之距離,並建立起環境地圖,證明演算法之可行性。
When a mobile robot in an unknown environment, how to be located ,and knowing the position as soon as possible had becoming a hot issue. In this thesis achieves using Panoramic Camera combining Extend Kalman Filter to complete Simultaneous Localization and Mapping, SLAM. It can be used when a robot exploring an environment, building a feature, and mapping finding the location at same time. Panoramic Camera has a 360 degrees of view, not only can capture more environment information but also can have a longer time in tracing the features. This also makes the SLAM system more steadily. In here the feature matching basis is Harris corner detection with scale-invariant feature transform (SIFT) description. By comparing established reference points with reference images information the robot can be located in the environment. The experimental results show that the localization algorithm can help robot to walk in the environment and build the feature map.