融合SFM先验信息的城市三维线段重建算法  

Urban 3D line segments reconstruction with SFM prior information

作　　者：张瑞[1] 颜青松 曲英杰 邓非[1] 蔡亚锋 ZHANG Rui;YAN Qingsong;QU Yingjie;DENG Fei;CAI Yafeng(Wuhan University,Wuhan 430079,China;Wuhan TianJiHang Information Technologies Inc.,Ltd.,Wuhan 430079,China)

机构地区：[1]武汉大学,武汉430079 [2]武汉天际航信息科技股份有限公司,武汉430079

出　　处：《测绘科学》2021年第8期55-62,119,共9页Science of Surveying and Mapping

基　　金：四川省科技计划资助项目(2019YFG0460)。

摘　　要：针对城市场景中大量存在的边线结构特征,该文提出了一种融合SFM先验信息的城市三维线段重建算法。该文算法首先在影像金字塔中构建精简的二维线段集;然后将相机位姿及稀疏点等数据作为先验信息,引入极线约束、方向约束和深度约束构建匹配线对;最后根据匹配线对重建出三维线段集。实验结果表明,该文提出的三维线段重建算法,在保持场景完整性的同时,能够有效地减少线段误匹配和重建结果冗余,缩短重建时间,提高三维线段重建结果精度。Aiming at the large number of edge structure features in urban areas, this paper proposed an urban three-dimensional line segments reconstruction algorithm that merged structure from motion(SFM) prior information. The algorithm in this paper firstly constructed a compact two-dimensional line segment set in the image pyramid, then took the camera pose and sparse point data as prior information, introduced the epipolar constraints, direction constraint and depth constraint to construct the matching line pairs. Finally, the 3 D line segment set was reconstructed according to the matching line pairs. Experimental results showed that the 3 D line segments reconstruction algorithm proposed in this paper could effectively reduce line segments mismatch and redundancy of reconstruction results, shorten the reconstruction time, and improve the accuracy of 3 D line segments reconstruction results while maintaining the integrity of the scene.

关 键 词：三维线段重建 LSD SFM 线段匹配 

分 类 号：P237[天文地球—摄影测量与遥感]