条纹边界编码的动态3D重建及流水建模  

Dynamic 3D Reconstruction and Pipelined 3D Modeling Based on Stripe Boundary Encoding

作　　者：郭继平[1] 彭翔[1] 刘晓利[1] 李阿蒙[1] 殷永凯[1] 

机构地区：[1]深圳大学光电工程学院教育部光电子器件与系统重点实验室,广东深圳518060

出　　处：《光学学报》2010年第10期2884-2890,共7页Acta Optica Sinica

基　　金：国家自然科学基金(60775021;60907005);深圳市科技计划项目(200734);深圳大学科研项目(000018);中国博士后科学基金(200902334)资助课题

摘　　要：提出一种基于彩色条纹边界编码的动态三维重建及流水式三维建模的方法,通过定位隐边界,消除隐边界对边界跟踪的影响,减少错误数据,提高扫描效率,实现了对运动物体深度数据的动态获取。进而,提出一种基于标志点匹配的流水式三维建模方案并给出建模过程中的视点规划方法,由此提高了匹配精度和匹配效率,以最少的单面数据完成了建模过程。实验证实了此方法不仅可以更快更精确地重建出单面深度数据,而且可以自动、快速完成深度数据间的匹配,得到更高的匹配精度和效率,实现了流水式三维建模。An approach for dynamic 3D shape acquisition of moving object and pipelined 3D modeling is proposed.Based on a brief review of previously reported methods a modified sensor prototype to realize the proposed approach is designed,which relies on color stripe boundary encoding.By locating the ghost boundary and removing its influence on boundary tracking,it is possible to reduce the erroneous range data and improve the scanning efficiency.Moreover,an efficient method for pipelined 3D modeling based on the registration using markers is proposed.An interactive view planning method in the modeling is also presented,leading to an improved accuracy and efficiency for registration and resulting in a modeling process with minimum range images.Experiment results show that the proposed approach can not only reconstruct the range image of single viewpoint accurately,but also automatically register the range images from different viewpoints,achieving 3D modeling in a pipeline mode.

关 键 词：图像处理 三维测量 流水式三维建模 条纹边界编码 视点规划 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]