基于双目光束法平差的机器人定位与地形拼接  

Binocular bundle adjustment based localization and terrain stitching for robot

作　　者：徐进[1] 沈敏一[1] 杨力[1] 王炜强[1] 刘济林[1] 

机构地区：[1]浙江大学信息与电子工程学系,浙江杭州310027

出　　处：《浙江大学学报（工学版）》2011年第7期1141-1146,共6页Journal of Zhejiang University：Engineering Science

基　　金：国家自然科学基金资助项目(60534070);国家自然科学基金重大研究计划资助项目(90820306)

摘　　要：为了实现自主移动机器人完成复杂智能任务,如路径规划和避障等,针对移动机器人在复杂未知环境中精确定位并对周围环境进行致密地形构建问题,提出一种基于双目光束法平差的机器人定位与致密地形拼接算法.利用安装在移动机器人上的立体相机获取图像序列,跟踪前后帧图像序列中的对应特征点,基于双目光束法平差优化(BBA)精确估计机器人位置姿态.对左右相机采集图像对进行立体匹配获取致密三维地形信息,结合定位时获取的旋转平移姿态,实现了地形的构建与拼接.实验结果表明,该算法具有较好的实时性和鲁棒性.In order for robot to achieve future complex intelligent tasks such as path planning and obstacle avoidance,a binocular bundle adjustment based localization and stitching method was proposed to solve the problem of precise localization of the mobile robot and dense terrain mapping of unknown environment.The corresponding features in the inter-frames were tracked from the image sequences captured by the stereo vision system mounted on the mobile vehicle.Then the system estimated the location and pose of robot by using binocular bundle adjustment（BBA） method.Dense three-dimensional terrain reconstruction information was obtained by stereo matching of images captured by stereo camera.Then terrain reconstruction and stitching were realized incorporated with the pose information estimated by localization.Experimental results demonstrated the effectiveness and robustness of the system.

关 键 词：立体视觉 运动估计 SIFT RANSAC 地形拼接 双目光束法平差(BBA) 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]