一种基于双目立体视觉的多视角三维重建方法

A Multiple-view 3D Reconstruction Approach Based on Binocular Stereo Vision

论文提出了一种基于角点与直线特征联合的三阶段立体匹配算法,能较好解决视差不连续区域和遮挡区域的误匹配问题,将局部坐标系中得到的数据点进行三维数据融合统一到一个坐标系中,基于OpenGL开发平台采用局部三维Delaunay的网格构造算法构建出物体的全景三维空间信息。实验证明该算法具有良好的鲁棒性,能够对自然物体进行精确三维重建并实现多视角显示。

In this research,an approach of three-phase 3Dcorresponding algorism based on Harris and line feature points is proposed,which can solve the false matching problem caused by inconsistent or shadowed areas.Through coordinate transformation between views,3Ddata of scene form different views integrated into a uniform reference coordinate frame.OpenGL programming using Delaunay triangulation and texture mapping is adopted in the 3Ddisplays process.Experiment demonstrates the robustness of this corresponding method,in which the 3D model of the targeted object can be displayed form all views.