双目立体测量中多视角深度图同时融合算法  

An Approach for Simultaneous Integration of Range Images in Binocular Stereo Vision

作　　者：韦虎[1] 张丽艳[1] 张辉[1] 

机构地区：[1]南京航空航天大学江苏省精密与微细制造技术重点实验室,南京210016

出　　处：《计算机辅助设计与图形学学报》2008年第11期1446-1451,1458,共7页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家自然科学基金(50475041);江苏省自然科学基金(BK2006192);教育部新世纪优秀人才支持计划.

摘　　要：提出一种将多次测量得到的深度图数据融合生成完整单一的网格曲面来描述被测物体表面的融合算法.依次将每幅深度图作为基准图,在基准图的基础上定义多幅深度图重叠对应关系;然后合并重叠数据,保留基准图网格;最后将相邻的基准图通过网格缝合连接成单一无缝的曲面.该算法能同时对多幅深度图进行融合,通过对重叠区域的优化,使得最终融合结果更精确地逼近被测物体表面的真实数据.融合实验结果验证了该算法的有效性.A new integration algorithm is proposed for combining multi-view range images into a single mesh surface to completely describe the object measured. By the algorithm, each range image is in turn treated as a benchmark image. Based on the benchmark images the overlapped correspondence relationship of multi-view is defined. Then the overlapped points are merged and the mesh on the benchmark image is kept. Finally a uniform waterproof surface is generated by stitching the gaps between the benchmark images. The algorithm can make the final integrated surface closer to the actual measured object by simultaneously integrating multi-view range images and by optimizing the overlapped regions. The integration experiments demonstrate the validity and efficiency of the approach.

关 键 词：双目立体测量 多视角深度图 同时融合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]