Key Technologies For The Fusion Of Point Cloud And Image And Its Application In Hole Repairing

With the rapid development and progress of technologies such as computers and three-dimensional measurement,reverse engineering technology is widely used in many fields such as mechanical design and manufacturing,mold repair,and virtual reality.In the process of three-dimensional measurement of the sample,due to the damage of the sample to be tested or the influence of factors such as measurement equipment and measurement methods,the obtained point cloud data may have holes.The existence of point cloud holes will directly affect the quality of the reconstructed model,so it is necessary to repair the holes in the point cloud data.This article focuses on the problem of repairing holes in 3D point cloud data,focusing on the use of the fusion of the image of the sampled sample in the 3D point cloud data with the hole area and the point cloud data to repair the point cloud hole,the main work is as follows:(1)When the hole area of the measured object is taken as a supplementary image,the difference of imaging angle interval of image sequence will affect the effect of the point cloud generated by the image.In order to solve this problem,a method of image sequence compensation is proposed to determine the optimal range of imaging angle interval.This method uses the way of rotating platform to get the image.After determining the best imaging angle interval range of the supplementary image,it can provide a method for the image supplementary part in the process of point cloud hole repair,improve the efficiency of acquiring supplementary images and the quality of the image point cloud generated by them.(2)The density of the image point cloud and the density of the measured point cloud are often quite different,which is not conducive to the registration of the two.In view of the inconsistency between the two density,a density control method of point cloud generated by sequence image is proposed.This method is based on the structure from motion.Firstly,it determines the increasing or decreasing direction of the point cloud density of the image,and then divides the two-dimensional feature point area of the first image into uniform grids.By adjusting the density of the two-dimensional feature points in the cell,the point cloud density of the three-dimensional image can be controlled.After adjusting the density of image point cloud,it can reduce the difference between the density of image point cloud and that of measurement point cloud,which is conducive to the registration of the two.(3)In the registration of image point cloud and measurement point cloud,due to the different ways of obtaining point cloud data,the two groups of point cloud data often have the problem of inconsistent scaling scale.Aiming at the problem that the registration accuracy of two sets of point cloud data is not high due to the inconsistency of point cloud scale in the registration process,an improved ICP registration method is proposed.In this method,the scale parameter is introduced into the registration process,the best scale coefficient is found through iteration,the point cloud data is processed to reduce the dimension to achieve the coarse registration,and then the k-d tree algorithm is used to accelerate the selection of corresponding points to achieve the fine registration,complete point cloud hole repair.