3D Dense Point Cloud Processing And Application Based On Building Sequence Image

The building's three-dimensional dense point cloud can be fine to express the structure of the building,which has wide application prospects.Resolving the problem that building three-dimensional reconstruction of the low level of automation,this paper is devoted to the study of dense point cloud generation and application technology based on building sequence images.The main research contents are as follows:(1)An adaptive SIFT feature point detection algorithm based on local region statistical information is studied.By calculating the difference between the mean value of the gray value of the local area and the gray value of the building,the SIFT parameter is adjusted adaptively to solve the problem that the feature points detected in the local area of building are too few.Experiments show that the improved feature point detection algorithm for subsequent 3D reconstruction will reduce the occurrence of noise outside the building,and it is better in the local area.(2)The algorithm of coordinate calibration in message processing and three-dimensional reconstruction is studied,and the data reception of Haixinda total station is processed by message detection algorithm.According to the existing packet loss in the process of message receiving and message of the next group of pollution problems,read the content of the message at the same time detection message header,so as to obtain accurate position information which is used to coordinate transformation in 3D reconstruction,and get the real coordinates of 3D reconstruction.Experiments show that the message processing and coordinate transformation algorithm proposed in this paper can obtain the exact real coordinates of the building.(3)This paper studies a clutter removal algorithm based on the clustering of members and the geometric characteristics of buildings.Based on the analysis of the spatial structure of the miscellaneous points,two algorithms are proposed to remove the miscellaneous points away from the buildings and the geometric characteristics of the buildings.Experiments show that the proposed algorithm can eliminate the building noise(4)This paper studies the normal vector inconsistency of the dense point cloud to extract the edge of the building,obtain the basic contour of the building,project it onto the longitudinal plane of the building,and calculate the building's floor area on the projection plane.Experiments show that the algorithm proposed in this paper estimates that the area of the building and the actual measurement area error within 5%.