多分辨率配准点的ICP算法  

ICP Algorithm Based on Multi-resolution Registration Point

作　　者：王勇[1] 邹辉[1] 何养明 黎春 

机构地区：[1]重庆理工大学计算机科学与工程学院,重庆400054

出　　处：《小型微型计算机系统》2018年第3期406-410,共5页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61173184)资助;重庆市巴南区技术合作项目(2016TJ08)资助;重庆理工大学研究生创新基金项目(YCX2016226)资助

摘　　要：针对三维点云配准中精度低、速度慢的问题,提出一种多分辨率配准点的ICP算法.首先,该算法引入多分辨率概念,利用低分辨率匹配点对迅速完成配准,高分辨率匹配点对提高配准精度;然后引入匹配度概念并改进计算方法,从而寻找更合适的匹配点;同时根据计算需要来计算相应点的曲率信息,以进一步提升配准效率;最后使用四元数法迭代完成配准.实验表明了算法的有效性,在精度和速度上都有较大提升,且随点云规模增大,速度提升幅度有递增趋势.Aiming at the problem of low precision and slow speed in the 3D point cloud registration process, an ICP algorithm based on multi - resolution registration point is proposed. First of all, the concept of multi-resolution is introduced by this algorithm, the use of low-resolution matching point to quickly complete the registration, high-resolution matching points to improve the registration accura- cy;And then the concept of matching degree is introduced and the calculation method is improved, so as to find a more suitable matc- hing point of a registration point;At the same time, according to the calculation needs to calculate the curvature information of the rele- vant point to further enhance the registration efficiency ;Finally,the quaternion method is used to complete the registration. The effec- tiveness of the algorithm is proved by experiments, the accuracy and speed are greatly improved, and with the size of point cloud in- creases, the increase rate of speed tend to increase.

关 键 词：点云配准 多分辨率 曲率 四元数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]