A building dense point cloud rapid reconstruction method based on PMVS

Abstract

The invention discloses a building dense point cloud rapid reconstruction method based on PMVS. In an application scene of 3D reconstruction of buildings, the method uses the quasi-dense point cloud to replace the self-built sparse point cloud of a PMVS algorithm as a diffusion seed point. The main flow of the method of the invention comprises the following steps: first, an initial spatial triangular mesh model of the building is constructed, spatial interpolation diffusion, patch optimization and filtering, triangular mesh information updating and other operations are carried out to obtain accurate quasi-dense point cloud, and the quasi-dense point cloud is used to replace the self-built sparse point cloud of the PMVS algorithm as a diffusion seed point, and thus acceleration of the reconstruction process of the PMVS algorithm is realized. The method can effectively improve the operation efficiency of the PMVS algorithm. The method plays a certain filtering effect on the point cloud noise brought by natural landscapes and other non-building main body point cloud noise.

Description

technical field [0001] The invention relates to the field of image processing, in particular to a PMVS-based fast reconstruction method of dense point clouds of buildings. Background technique [0002] The PMVS (patch-based multi-view stereo) algorithm proposed by Furukawa et al. is one of the mainstream dense point cloud reconstruction algorithms with excellent performance. The basic process of the algorithm is: use Harris and DoG operators to detect the feature points of the input image set, use feature matching and triangulation reconstruction to obtain sparse seed point clouds, use these seed points to iteratively perform steps such as patch diffusion, optimization, and filtering, and finally A dense set of patches with normal information that tightly covers the surface of the object to be reconstructed is reconstructed. The advantages of this algorithm are: (1) It does not need any initial information such as convex hull, bounding box, etc.; (2) The algorithm has a wid...

AI Technical Summary

Problems solved by technology

[0003] Although the performance of the PMVS algorithm is excellent, there are also some shortcomings, one of which is the low operating efficiency of the algorithm. When dealing with high-resolution image data and large-scale scene reconstruction, the operating efficiency of the PMVS algorithm will be significantly reduced, resulting in PMVS Algorithms cannot meet the requirements of rapid reconstruction of large-scale urban scenes in smart cities

[0004] One of the reasons for the low operating efficiency of the PMVS algorithm is that the PMVS algorithm uses weak constraints to perform feature matching on the image to reconstruct the sparse point cloud of the object to be reconstructed. The weak constraints of feature matching make the process not have global optimality, making The reconstructed sparse point cloud is full of noise

Using these noise-rich sparse point clouds as seed points for subsequent patch diffusion optimization operations will cause more meaningless diffusion, and the patch diffusion and optimization process is the main time-consuming process in the PMVS algorithm. To a certain extent, the running time of PMVS algorithm is increased

[0005] Aiming at the efficiency of the PMVS algorithm, many scholars have proposed different improvement schemes, but there is no precedent for improving the efficiency of the PMVS algorithm from the perspective of improving the seed point density and accuracy by using the geometric characteristics of buildings.

The existing PMVS algorithm is not efficient and cannot meet the requirements of rapid reconstruction of large-scale urban scenes in smart cities

Images