基于编码结构光的物体表面点云获取算法  

Algorithm to obtain point-cloud data of object surface based on coded structured light

作　　者：梁东魁[1] 

机构地区：[1]燕山大学信息科学与工程学院,河北秦皇岛066004

出　　处：《计算机工程与设计》2010年第12期2905-2908,共4页Computer Engineering and Design

摘　　要：为了解决基于结构光的三维重建中摄像机观测条纹与投影仪模式之间的匹配问题,提高点云数据的精度,采用DeBruijn序列对结构光进行编码,基于全局优化思想对条纹边界进行最优邻域匹配(BN-Matching),并针对结构光图像的特点和用途提出了边界检测算法。该编码方法结合时分多路复用与空间邻域两种编码的思想,对物体表面的连续性和单调性不作限制。实验结果表明,边界检测及匹配算法有较强的鲁棒性,获得的点云数据能够满足后续三维表面重建的要求。In the three-dimensional reconstruction process based on structured light,for resolving the matching problem between the camera observation pattern and projection pattern,and improving the precision of point-cloud data,encoded the structured light using De Bruijn sequence.The color-stripe-borderline correspondence is solved with the BN-Matching algorithm based on the global optimal,and an edge detection algorithm is proposed aiming at the specialty and purpose of structured-light images.The coded strategy integrates the ideas of time-multiplexing code and spatial neighborhood code,and does not restrict the continuity and monotony of object surface.The experiments show the edge detection and BN-Matching are robust,and the point-cloud data is fine to reconstruct the 3D surface.

关 键 词：时分多路复用编码 空间邻域编码 DeBruijn序列 边界检测 匹配 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]