基于射线投影的多视场深度像融合  

Integration of Multiple Range Images Based on Ray Casting

作　　者：刘晓利[1] 彭翔[2] 李阿蒙[2] 高鹏东[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,天津300072 [2]深圳大学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10期1286-1291,1297,共7页Journal of Computer-Aided Design & Computer Graphics

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

摘　　要：首先确定匹配好深度像的最小轴对齐包围盒.通过盒面,沿着x,y,z3个轴方向投射3个均匀分布的射线组,计算得到与所有深度像的交点,并通过Dexel结构来存储射线组与交点.加入距离和法向夹角2个判断依据来去掉重叠点,从而将多个视场深度像融合为完整的、无冗余的三维模型.实验结果表明,该方法误差小、速度快,而且简单有效.An axis-aligned bounding box for all the registered range images is set up first. Then three sets of uniform-distributed rays are cast and passed through the faces of the box along three orthogonal axes of x,y, z, respectively. The intersections between the rays and the range images are then calculated and stored in Dexels data structure. The overlapped intersections are identified by one specific restriction based on their locations and normals. After removing the overlapped points, a complete non-redundant threedimension model is obtained. The experimental results show that the proposed algorithm is accurate, and efficient.

关 键 词：深度像匹配 深度像融合 射线投影 Dexet结构 轴对齐包围盒 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]