The 3D Reconstruction Of Modern Beams And Columns Of Indoor Scene

In recent years,indoor 3D reconstruction technology is widely used in building restoration,smart city and intelligent robot navigation,and the related reconstruction technology is more and more mature.Whether it is the reconstruction of building surface,the detection of doors and windows,or the refined surface reconstruction of interior furniture,or even intelligent classification,there are a lot of relevant research.The beams and columns region is easily ignored because of its low proportion of point cloud in the whole point cloud,but it is also an important research object in architecture project.For this reason,we focus on the three-dimensional reconstruction of beams and columns.In the indoor scene,when the plane of beam,column,wall,ceiling and floor are combined in pairs,most of them are orthogonal to each other.Therefore the important point cloud areas can be extracted in pair by orthogonality.On the top view image of these important areas,the beams and columns area can be distinguished well.According to the geometric characteristics of the beams and columns area,the detection scheme can be developed for them respectively.In the three-dimensional reconstruction of beams and columns,the commonly used PLANE model can’t extract the plane of small area,the information of beams and columns is easy to be lost in the process of reconstruction.In view of this deficiency and the orthogonality of beams and columns and adjacent planes,the extraction algorithm of pairwise orthogonal point cloud(POP)model is proposed.The algorithm can not only extract the small plane of beams and columns,but also has high efficiency,and can be used to remove furniture and other debris.After the pre-process,the simplified point cloud is mapped to the two-dimensional image,and the beams can be extracted by using the scalability,connectivity and pixel terms of candidate rectangles in the beams region.Based on the symmetry,square characteristic and pixel term of candidate C-shapes in column region,the column can be extracted.The experimental results show that the method proposed in this paper can effectively extract the location information of beams and columns of three-dimensional indoor scene and reconstruct it.