基于概率匹配的栅格地图拼接算法  

Mosaic Algorithm of Grid Map Based on Probability Matching

作　　者：郝奇 陈智君 郑亮 陈双 曹雏清 HAO Qi;CHEN Zhi-jun;ZHENG Liang;CHEN Shuang;CAO Chu-qing(Wuhu Institute of Robot Industry Technology,Harbin Institute of Technology,Wuhu 241000,China;Wuhu HIT Robot Technology Research Institute Co.,Ltd.,Wuhu 241000,China)

机构地区：[1]哈尔滨工业大学芜湖机器人产业技术研究院,芜湖241000 [2]芜湖哈特机器人产业技术研究院有限公司,芜湖241000

出　　处：《组合机床与自动化加工技术》2022年第5期83-87,共5页Modular Machine Tool & Automatic Manufacturing Technique

摘　　要：栅格地图拼接是栅格地图扩展和多地图合并成大地图的一项关键技术,现有拼接方法主要是基于图像的特征提取和匹配,应用在彩色图像中的效果较好,但在栅格地图中却存在着匹配可靠性差和精度不高的问题。针对现有不足,提出了一个基于自适应分支定界匹配和双向概率残差优化匹配的栅格地图拼接算法。首先,从源地图和目标地图中提取源点云和目标点云;其次,将目标点云和源地图进行分支定界匹配,获取目标点云在源地图中的初始位姿;然后,将目标点云通过初始位姿投影到源地图,构建概率残差的第一个约束,再将源点云通过初始位姿的逆变换投影到目标地图,构建概率残差的第二个约束;之后,构建两个约束的代价函数,通过非线性优化得到目标地图在源地图中的最优位姿;最后,通过占据概率融合得到拼接后的栅格地图。结果表明,该算法匹配效果好、迭代收敛速度快、匹配精度高,满足工业场景的可靠性和精度要求,对于工业场景下的地图构建和AGV导航具有重要意义。Grid map mosaic is a key technology for grid map expansion and multi map merging into large map.The existing mosaic methods are mainly based on image feature extraction and matching,which has a good effect in color image,but there are some problems in grid map,such as poor matching reliability and low accuracy.Aiming at the existing shortcomings,this study proposes a grid map mosaic algorithm based on adaptive branch and bound matching and bidirectional probability residual optimization matching.Firstly,the source point cloud and target point cloud are extracted from the source map and target map;Secondly,the initial pose of the target point cloud in the source map is obtained by branch and bound matching between the target point cloud and the source map;Then,the target point cloud is projected to the source map through the initial pose to construct the first constraint of probability residual,and then the source point cloud is projected to the target map through the inverse transformation of initial pose to construct the second constraint of probability residual;Then,the cost function of two constraints is constructed,and the optimal pose of the target map in the source map is obtained through nonlinear optimization;Finally,the spliced grid map is obtained by occupancy probability fusion.The results show that the algorithm has good matching effect,fast iterative convergence speed and high matching accuracy,and meets the reliability and accuracy requirements of industrial scene.It is of great significance for map construction and AGV navigation in industrial scene.

关 键 词：地图拼接 概率地图 概率融合 非线性优化 

分 类 号：TH164[机械工程—机械制造及自动化] TG659[金属学及工艺—金属切削加工及机床]