中文摘要：

针对SfM重建点云的曲面建模问题,提出一种改进的区域增长网格化算法。定义k近邻影响域提高拓扑稳定性,引入二叉排序树高效地组织候选三角片,采用无向环搜索策略完成孔洞的检测,最终获得完整的三角网格面。实验结果表明,该算法相比于Possion曲面重建,在获得高的重建精度的同时显著提高了计算效率,有助于提升3D曲面重建与模型表现的性能。

英文摘要：

This paper proposed an improved region growing based triangulation algorithm for surface modeling problem from point clouds reconstructed by SfM.Defined a k-nearest neighbor influence region to improve the topological stability.It orga-nized candidate triangles efficiently by binary sort tree and accomplished holes detection by a searching strategy using undirected loop.Finally,achieved a complete triangular mesh.Experimental results show that,compared to Possion surface reconstruction,the algorithm can significantly improve the computational efficiency and acquire a high reconstructed accuracy,which helps to improve the performance of 3D surface reconstruction and model rendering.