Object-based Building Facades Recognition And Reconstruction From Mobile Laser Scanning Point Clouds

This paper is guided by an object-based point cloud analysis theory, to carry out mobile laser scanning (MLS) point cloud data processing and information extraction, and the studies mainly around the building facade recognition and building facade reconstruction themes. The main contents include the following four points:(1) MLS point cloud segmentation and mergence.In this paper, the normal vector and distance constraints point cloud segmentation method which has robustness is applied. But the excessive segmentation phenomenon is existed. For this reason, this paper use kd-tree to determine approaching relationships, use the normal vector angle and distance between facets as the criterion of mergence.(2) Prior knowledge-based building facades recognition and extractionThe common features of point cloud classification and recognition include:size, shape, position, direction, color, topology relationships and so on. In view of the building facade is generally lager than the other targets, perpendicular to the horizontal plane and significantly higher than the surrounding surface, this paper use the characteristics of the angle between facet normal vector and horizontal vector, facet area, absolute elevation and elevation changes to extract building facades.(3) Outline extraction of MLS point cloud building facadeBecause of the exist of vast facade point in MLS point cloud, the traditional outline extraction method based on triangular network of ALS lose effectiveness. In this paper, improved the traditional method to apply to MLS point cloud.(4) The correction of building facades topology relationshipsSince the inaccuracy of building facades information with PCA method, this paper uses the prior knowledge of building facade perpendicular to the horizontal plane and the perpendicular and parallel relationship between building facades, detects building facade topological relationships based on normal vector of facet and correct the topological relationship.The effort of object-based method of MLS point cloud building facades recognition and reconstruction can provide an excellent results with high robustness. And it can be applied to processing and analysis of point cloud data with complex scenes region. So it provides an effective method to identify and reconstruct building facades.