基于Skinner-Ransac的移动机器人单目视觉SLAM  

Mobile Robots Mono-SLAM Based on Skinner-Ransac

作　　者：阮晓钢[1] 王丹阳[1] 魏若岩[1,2] 李笑漪 

机构地区：[1]北京工业大学电子信息与控制工程学院,北京100124 [2]河北经贸大学信息技术学院,石家庄050000

出　　处：《北京工业大学学报》2016年第9期1281-1285,共5页Journal of Beijing University of Technology

基　　金：国家"973"计划资助项目(2012CB720000);国家自然科学基金资助项目(61375086);北京市教育委员会科技计划重点项目(KZ201210005001)

摘　　要：针对移动机器人单目视觉同时定位与地图创建(simultaneous location and mapping,SLAM)数据关联问题,提出一种Ransac抽样算法.该算法(Skinner-Ransac)基于Skinner操作条件反射原理,结合扩展卡尔曼滤波(extended Kalman filter,EKF)运动模型获得的先验信息,对图像匹配点抽样集合中的每个匹配点对赋予权值,并根据判断函数对其权值进行更新.最后,针对数据关联的效率要求提出新的迭代终止条件,以公开图像数据集作为图像采集样本.实验结果表明:Skinner-Ransac算法高效可靠,SLAM的位姿估计结果可以达到移动机器人自主导航的需求.A method of sample consenus algorithm based on Skinner probabilistic automata( SkinnerRansac) was proposed to solve the data association problem in monocular vision simultaneous location and mapping( SLAM) for mobile robots. Combined with the priori information of extended kalman filter( EKF) motion model,to assign weight for each sample in the set of image matching points samples,and update the weight for each sample based on the current sample results,a new iteration terminal condition was put forward for the lack of the priori knowledge. A set of open image data were taken as test samples.Results show that Skinner-Ransac algorithm is efficient and reliable, and SLAM's pose estimation accuracy can be achieved for the need of mobile robot autonomous navigation.

关 键 词：Skinner-Ransac算法 移动机器人 单目视觉同时定位与地图创建 数据关联 

分 类 号：TP242.6[自动化与计算机技术—检测技术与自动化装置]