Device and method for point cloud optimization

Abstract

A method for point cloud optimization includes point cloud data preprocessing steps, point cloud sharp feature recovering steps and point cloud sharp feature enhancing steps, wherein the point cloud data preprocessing steps include large-scale random scanning point cloud simplifying, noise reducing, exterior point removing, homogenization, normal vector calculating and space structure division preprocessing, the point cloud sharp feature recovering steps include further projection and normal vector calculating and point cloud sharp feature recovering, and the point cloud sharp feature enhancing steps include point cloud sharp feature up-sampling and point cloud sharp feature enhancing. By enhancing sharp features of random scanning point cloud, a three-dimensional model finally achieved is up to the requirement of practical application. Meanwhile, optimized point cloud is more concise and neat, so that the existing art can be more suitable for adjusting parameters and capable of improving efficiency. In other words, under the condition that existing point cloud three-dimensional reconstruction processes are not changed, by optimizing raw materials of point cloud, degree of automation, production efficiency and quality of products are improved.

Description

【Technical Field】 [0001] The invention relates to an image processing technology, in particular to design a method and device for optimizing large-scale scattered point clouds. 【Background technique】 [0002] In today's production applications, a major bottleneck in the development of 3D technologies such as computer-aided design, reverse engineering, virtual reality, 3D animation and games is that there is still no convenient way to quickly obtain 3D models stored by computers. In recent years, 3D laser scanners have been widely used due to their advantages in easily and flexibly obtaining 3D surface data of real objects. However, in the world, how to directly and quickly obtain a practical point cloud model directly from the scanned point cloud data is still an unsolved problem. [0003] The so-called point cloud model generally refers to a collection of three-dimensional coordinate points on the surface of the object that are calculated by the three-dimensional scanning device ...

AI Technical Summary

Problems solved by technology

This method can better get a smooth and continuous fitting surface, but the biggest disadvantage is that it assumes that the surface of the object is smooth everywhere, but it cannot maintain the sharp features of the object.

[0005] Second, the existing 3D point cloud reconstruction methods are heavily dependent on the normal vector information of the point cloud

But there is still no good way to solve this problem, especially the normal vector of the points at the sharp features of the point cloud model is difficult to estimate

[0006] like figure 1 As shown, it is the 3D reconstruction of a point cloud of a square foam box. The input is the 3D point cloud data directly obtained by the existing 3D scanning equipment. Points (outlier) and holes, and the data scale is large and unevenly distributed

Images