Multi-View Three-dimensional Reconstruction Feature Point Detection Matching And Point Cloud Area Clipping Of Improvement

Multi-view-based 3D reconstruction is a hot research topic in the field of computer vision.It captures multiple views of a target scene from different perspectives and obtains three-dimensional geometric and texture information of the target object from the image.The target information is reconstructed using these spatial information 3D scene model.Through specific research on multi-view 3D reconstruction algorithms,combined with the characteristics of ancient scenes with large scenes,complex details,and diverse structure types,it is found that there are many places in the 3D reconstruction system for ancient buildings based on multi-views that need to be studied and improved.The first is to improve the feature point detection and matching algorithm in the three-bit reconstruction technique,and then proposes an automatic clipping algorithm for point cloud regions based on the camera coordinate edges.,and summarized as follows: 1.The feature point detection and matching capabilities for multi-view based 3D reconstruction are poor,and the number of misidentified feature points is more.An adaptive scaling algorithm based on Bicubic algorithm combined with Harris algorithm to improve the Asift algorithm is proposed.The improved algorithm BH-Asift(Bicubic Harris-Asift)is applied to the multi-view 3D reconstruction system based on PMVS.The improved feature point detection and matching of the algorithm is proposed.Stronger ability to produce the final 3D model is more realistic.Aiming at the problems of multiple point cloud and redundant model information generated by multi-view 3D reconstruction algorithm,this paper proposes a point cloud region clipping algorithm based on Alpha Shapes algorithm to detect the edge of 2D plane coordinate point set.The point set is obtained by the camera projecting three-dimensional coordinates onto the plane.The Alpha Shapes algorithm detects the edge points of the scattered point set,and connects adjacent edge points to form an area.The point cloud is generated according to the three-dimensional reconstruction algorithm.The coordinate position on the plane clears the point cloud outside the area.