应用改进ICP算法的点云配准  

Point cloud registration based on improved ICP algorithm

作　　者：杨小青[1] 杨秋翔[1] 杨剑[1] 郑晓璐[1] 

机构地区：[1]中北大学计算机与控制工程学院,山西太原030051

出　　处：《计算机工程与设计》2015年第9期2457-2461,共5页Computer Engineering and Design

基　　金：总装预研基金项目(9140A17020113BQ04226)

摘　　要：针对三维重建领域点云数据配准过程中存在的配准效率低和精度误差大问题,提出一种改进ICP算法。应用盒子结构划分点云数据,对每一独立单元盒提取特征点构建三角形,根据相似原理,选用与其相似度最大的三角形各顶点作为初始对应点对,引入支持度的概念,给出评价准则,使当前待确定点对的成立最大支持其余已确认匹配点对,保证对应点查找的正确性。同理,以现有三角形各边为基准,继续构建新的三角形,完成所有匹配点对的查找。实验结果表明,改进算法较传统ICP算法有大幅提高,缩短配准误差至0.03%,配准效率提高了65%以上,优势明显。For the problem of low registration efficiency and high accuracy error existing in the three-dimensional reconstruction point cloud data registration process, an improved ICP algorithm was proposed. Box structure was applied to divide and sort the point cloud data, for each separate box, feature points were extracted to construct triangles, according to the similar principle, each vertex of the greatest similarity triangle was chosen as the initial corresponding point. To ensure the finding correctness of the corresponding point, the concept of support degree was introduced, the evaluation criteria was given to guarantee the estab- lishment of the current point to be ascertained giving the best support for the remaining matched points. Similarly, every side of the existing triangle was chosen as a reference and new triangle was built until all closest points were found. Experimental results indicate that the improved method significantly enhances the performances compare with traditional ICP algorithm, the registra- tion error is shortened to 0. 03% and the registration efficiency is increased by more than 65%, its advantages are obvious.

关 键 词：点云配准 ICP算法 盒子结构 相似原理 支持度 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]