根據複合式地圖的移動機器人視覺定位技術

Visual Localization for Mobile Robots Based on Composite Map

黃雅軒(Yea-Shuan Huang) ； 歐志鴻(Zhi-Hung Ou) ； 游鴻修(Hung-Hsiu Yu)

《前瞻科技與管理》 3卷2期 (2013/11) Pp. 139-161

https://doi.org/10.6193/JATM.2013.0302.09

摘要

本篇論文提出了一種移動機器人的視覺定位方法，首先利用雷射機與攝影機來建構複合式地圖，而在之後機器人進行定位時，只需要單顆攝影機就能完成機器人的定位工作。由於雷射機價格相當昂貴，如果每個移動機器人都需配載一台雷射機，這樣高成本的系統是不容易被普及使用的。所以在本論文中，我們提出了一種新穎的複合式地圖建構方法，利用雷射機的深度資訊與攝影機的影像特徵進行融合，來產生具有定位功能的複合式地圖，而在進行機器人定位時，利用機器人平台上的攝影機，將當下所拍攝影像中的特徵點擷取出來，並與先前所建構的複合式地圖進行比對，則可準確地估測出機器人當時位於地圖上的所在位置，來達到機器人定位目的。經實際測試後所得到的實驗數據，本論文所提出的複合式地圖之視覺定位方法在多種情況下的定位誤差值，其總平均位移誤差量為0.168公尺，總平均角度偏移量為3.911度。故由實驗數據可知，本論文除了能大幅降低定位系統的硬體成本之外，也有著相當不錯的定位效果。

關鍵字

自主式移動機器人 ； 複合式定位地圖建構 ； 視覺定位 ； 特徵點比對 ； 相機校正

並列摘要

In this paper, we propose a novel mobile robot visual localization method which mainly contains two processing stages: map construction and position localization. In the map construction stage, this method uses both a laser range finder and a camera mounted on a robot to construct a composite map. The laser range finder can provide depth information and the camera can provide distinct features of salient points from images. In the position localization stage, a robot system detects salient points from a current processing image by just using a camera, computes features of the detected salient points, matches them with the features recorded in the previously constructed composite map, and finally decides the robot positions. By this way, a robot can be located its position effectively without expensive laser range finder so that such system can be generally accepted to different applications for its low cost. With the proposed method, several experiments have been performed, the average displacement is 0.168 meters, and the average angle error is 3.911. From these experimental data, it shows that our method not only can reduce the hardware cost of a localization system, but also can achieve high localization accuracy.

並列關鍵字

autonomous mobile robot ； composite map construction ； visual localization ； feature point matching ； camera calibration

參考文獻

Bouguet JY, n.d., “Camera Calibration Toolbox for Matlab,” (accessed June 17, 2003).

Brown, M.,Lowe, D.(2002).Invariant Features from Feature Point Groups.BMVC 2002: 13th British Machine Vision Conference.(BMVC 2002: 13th British Machine Vision Conference).:

Burgard, W.,Fox, D.,Hennig, D.,Schmidt, T.(1996).Position Tracking with Position Probability Grids.First Euromicro Workshop on Advanced Mobile Robot.(First Euromicro Workshop on Advanced Mobile Robot).:

Chen, C.H.,Chan, Y.P.(2008).SIFT-Based Monocluar SLAM with Inverse Depth Parameterization for Robot Localization.IEEE Workshop on Advanced Robotics and Its Social Impacts.(IEEE Workshop on Advanced Robotics and Its Social Impacts).:

Davison, A.J.,Reid, I.D.,Molton, N.D.,Stasse, O.(2007).MonoSLAM: Real-Time Single Camera SLAM.IEEE Transaction on Pattern Analysis and Machine Intelligence.29,1052-1067.