ICP框架下基于表面间平均体积测度的深度像配准  

Range Image Registration with ICP Frame Using Surface Mean Inter-Space Measure

作　　者：高鹏东[1] 彭翔[1] 李阿蒙[2] 刘晓利[1] 

机构地区：[1]天津大学精密测试技术与仪器国家重点实验室 [2]深圳大学光电子学研究所教育部光电子器件与系统重点实验室,深圳518060

出　　处：《计算机辅助设计与图形学学报》2007年第6期719-724,共6页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(60275012);广东省普通高校自然科学研究重点项目(04Z010);广东省自然科学基金(031804);深圳市科技计划项目(200341).

摘　　要：提出一种利用深度像重叠区域间的空间体积作为误差度量的精确配准算法.通过寻找2幅深度像重叠区域内的有效三角形对,并将这些三角形对所夹的三维空间作为误差测度来指导深度像的配准;然后将对应三角形的质心作为对应点对,估计出新的空间位置转换关系.实验结果表明:该算法具有较高的配准精度,收敛速度快并且具有一定的抗噪声能力.A novel approach is presented for precise registration of range images with a new error metric, based on the inter-space between overlap regions. The proposed method aligns range images using the surface mean inter-space measure （SMISM）. This measure takes on the mean 3D inter-space associated with each valid triangle-pair in the overlap region as the error metric to guide the range image registration. And then the centroids of corresponding triangles are employed as point-pairs to calculate the new transformation. Experimental results demonstrate that the presented method is insensitive to noises and has high precision as well as a nice performance on convergence.

关 键 词：表面间平均体积测度 深度像配准 误差测度 三角形对 

分 类 号：TP391.72[自动化与计算机技术—计算机应用技术]