基于激光扫描和SFM的非同步点云三维重构方法

发布时间：2018-07-01 08:32

本文选题：D激光扫描 + SFM　；参考：《仪器仪表学报》2016年05期

【摘要】：室外场景具有测量数据量大、扫描数据易重叠及建筑物表面信息复杂等特点,单靠激光扫描方法能够获得场景精确的深度信息,但缺乏颜色和纹理信息,利用从运动中恢复结构(SFM)方法可获得丰富的彩色信息,但重构精度不高,若将两种设备固定进行在线实时同步测量,易受到测量环境和系统制约不易实现。针对此问题,提出了一种基于激光扫描和SFM结合的非同步点云数据融合的三维重构方法。首先,提出利用手动选择控制点进行7自由度初始配准,再利用迭代最近点(ICP)算法对初始配准结果进行精确配准,最后利用最近点搜索算法将分布在经基于面片的多视图立体视觉(PMVS)算法优化后的SFM数据中的颜色信息与激光扫描的点云坐标进行融合。实验结果和数据分析显示,本文的方法能有效地将激光扫描与SFM点云数据进行融合,实现了室外大场景的三维彩色重构。
[Abstract]:Outdoor scene is characterized by large amount of measurement data, easy overlap of scanning data and complex information on building surface. Laser scanning method alone can obtain accurate depth information of scene, but lack of color and texture information. The method of recovering structure from motion (SFM) can obtain rich color information, but the reconstruction accuracy is not high. If two kinds of equipment are fixed for on-line real-time synchronous measurement, it is easy to be restricted by measurement environment and system. To solve this problem, a 3D reconstruction method based on the combination of laser scanning and SFM is proposed. First of all, it is proposed to use manual selection of control points for 7 degrees of freedom initial registration, and then iterative nearest point (ICP) algorithm is used to accurately match the initial registration results. Finally, the nearest point search algorithm is used to fuse the color information distributed in the optimized SFM data based on multi-view stereo vision (PMVS) algorithm and the point cloud coordinates of laser scanning. Experimental results and data analysis show that the method proposed in this paper can effectively fuse laser scanning with SFM point cloud data and realize 3D color reconstruction of outdoor large scenes.
【作者单位】：上海电力学院自动化工程学院;上海大学机电工程与自动化学院;德国维尔茨堡大学计算机科学学院;
【基金】：国家自然科学基金(61171145) 上海市自然科学基金(16ZR1413400) 上海电力学院人才引进基金(K2015-016)项目资助
【分类号】：TN249

【相似文献】