Based On Five-dimensional Point Cloud Data Reconstruction Research

With the development of technologies such as computer vision,computer graphics,optics,,and the increasing demands for object models in games animation,reverse engineering,and industrial inspection,the core part,3D reconstruction technology,is getting more and more attention.Cloud reconstruction could scan the surface of a real object based on three-dimensional scanning device and rebuilds its virtual three-dimensional model in a computer.Moreover,the scanning device acquires three-dimensional points cloud data which is a set of three-dimensional points that record the geometry and spatial distribution of objects in the same spatial coordinate system.In addition,there is no topological structure among these data points.The three-dimensional point cloud reconstruction could restore the surface of the object as quickly and optimally as possible.The author conducted in-depth research on the key technologies of 3D points cloud data reconstruction such as point cloud simplification,point cloud smoothing,grid reconstruction,hole repair,and innovatively achieved five-dimensional point cloud reconstruction based on innovative 3D point cloud reconstruction with time dimension and RGB color dimension.Firstly,in order to improve the efficiency of grid reconstruction,we adopt a concise method which is based on the gravity center method.The central idea is to create a minimum cubic grid that can surround all the point clouds,then divide the cube into multiple small cubes,and save all of points' gravity in the cube serves as new sampling points bring about the streamline of the point cloud.To further improve the quality of grid reconstruction,the author creatively presented an improved moving least squares smoothing filtering method which calculate the distance from the sampled point to the fitted surface after building a fitted surface instead of migrating all the sampling points into the fitting surface as other ordinary MLS methods.If the distance of sample points is large,it would be regarded as large noise and be deleted directly.Experimental results demonstrate that the algorithm can further improve the point cloud refinement rate and reconstruction efficiency without affecting the reconstruction.Then,we took advantage of a region-based growth grid reconstruction algorithm which constructs seed triangles and then continue to expand outward with the border of the initial triangle following certain rules,until a complete surface is completed.It was confirmed that this method can complete the point cloud surface reconstruction with high quality and efficiency.This method searches the neighboring region for the sampling points on the boundary side,then constructs the fitting surface by moving least squares method and upsampling,triangulates the upsampled points,finally,repair holes effectively.This method could retrieve the internal and external parameters of the depth camera and the color camera through the joint calibration of the depth camera and the color camera,and resolve the registration of the depth data and the color data by coordinate transformation.Furthermore,color point cloud data could be acquired in real time by continuous shooting by the camera.The topological structure is constructed and the reconstructed object surface is mapped based on and spatial coordinates of the color point cloud data and the RGB information.It shows that this method can achieve better matching of depth and color data and obtain high-quality color point cloud data.Finally,this paper designs and implements a five-dimensional reconstruction system that can process color point cloud data in real time based on the proposed algorithm and standards,and further validates the feasibility and efficiency of the proposed algorithm.