摘要
为了能够提高在一般环境下移动机器人定位的精度,文章采用栅格地图来描述环境;提出基于图像理解的移动机器人定位的概念,将栅格地图当成图像来处理,从而可以将图像配准中的优秀算法引入移动机器人定位中;对于栅格化的图像提取其Harris角点,同时提出两步匹配法寻找相邻2个时刻的2幅图像中匹配的Harris角点。首先采用里程计的数据对角点进行粗匹配,而细匹配采用的是基于随机抽样算法,能够有效地剔除粗匹配过程中误匹配的角点对,提高了定位算法的鲁棒性;最后用非线性最小二乘法估计机器人的位姿。通过实验可以看出在一般的环境中,基于图像理解的定位算法要优于基于线段特征的定位算法。
In order to improve the of localization precision of the robot in general environment, the grid map is used to describe the environment in this paper. The concept of locating mobile robot based on image understanding is presented when a grid map is processed as an image. So many excellent algorithms of image registration can be used in mobile robot localization. Then the Harris corner points are extracted from the image. And tow-step-matching algorithm is used to find the matching Harris corner points from two images of the k period and the k＋1 period. First the corner points are matched roughly by using the information from the odometry. Then by accurate matching algorithm based on random sample consensus, the mismatching points produced in rough matching can be effectively removed and the robustness of localization algorithm improved. At last the pose of the mobile robot can be estimated by the nonlinear least square method. It is proved by experiments that the algorithm based on image understanding is obviously more accurate than the algorithm based on line features.
出处
《合肥工业大学学报（自然科学版）》
CAS
CSCD
北大核心
2010年第4期510-513,620,共5页
Journal of Hefei University of Technology：Natural Science
基金
先进数控技术江苏省高校重点建设实验室基金资助项目(KXJ07127)
