High precision indoor model contour extraction algorithm based on geometric information

The use of terrestrial laser is increasing in the field of architecture, especially in the design and construction of indoor space, and the point clouds generated by laser scanning is increasingly needed for assistance. Traditional 3D reconstruction method is to use RGB-D camera or LiDAR scanner to obtain depth map or point clouds data, and then use algorithm to get Mesh models. The indoor model obtained by traditional methods has some problems, such as low precision, serious cavity, etc. In order to obtain the 3D model with high precision and high simulation, manual modeling must be used to cooperate, which consumes a lot of manpower and material resources, which is an urgent problem to be solved. Among them, indoor space reconstruction is a hot topic at present, but due to the disorder of indoor space and a lot of occlusion, Thereby increasing the difficulty of indoor space reconstruction. In this paper proposes a high-precision indoor model contours extraction algorithm based on geometric information of point clouds, which can obtain the spatial layout information of the model. Use plane segmentation and corners detection in three-dimensional space to solve the problem of poor reconstruction effect caused by indoor building occlusion and point cloud model inclination is reasonably. The information of points, lines and planes in 3D indoor space can be extracted, and realize efficient reconstruction of indoor space. Finally, evaluated this method has accuracy in static laser scanning model and simulation laser data set is evaluated, and prove this method is workable in various scenes.