A non-rigid body 3D reconstruction method and system based on multiple depth maps

Abstract

The present invention proposes a non-rigid body three-dimensional reconstruction method based on multiple depth maps, which includes the following steps: photographing depth maps of non-rigid bodies at different angles and different postures to obtain multiple depth maps; converting each depth map into a set of three-dimensional point cloud, and obtain multiple matching point pairs between multiple sets of three-dimensional point clouds; perform position transformation on each matching point, and obtain the transformation parameters corresponding to each matching point after position transformation; splice all transformation parameters , and establish an energy function based on the splicing results; solve the energy function to reconstruct a non-rigid three-dimensional model based on the solution results. According to the method of the present invention, the required input information is simple and easy to collect, and a high-precision complete three-dimensional model can be obtained. This method is accurate and robust in solving problems, eliminates the impact of cumulative errors, has fast running speed, and has broad application prospects. The present invention also proposes a non-rigid three-dimensional reconstruction system based on multi-depth maps.

Description

technical field [0001] The invention relates to the technical field of computer vision, in particular to a non-rigid three-dimensional reconstruction method and system based on multi-depth images. Background technique [0002] 3D reconstruction is an important problem in the field of computer vision. High-quality 3D models have important application value in various fields such as film and television entertainment, cultural relics protection, and mechanical processing. However, the acquisition of high-quality 3D models is usually achieved by expensive laser scanners. Although the accuracy is guaranteed, there are also two disadvantages: First, the laser scanner requires the object to be scanned to be absolutely still during the scanning process. It will lead to obvious errors in the scanning results; second, laser scanners are expensive, and it is difficult to popularize them to expand their civilian value. Contents of the invention [0003] The present invention aims to...

AI Technical Summary

Problems solved by technology

However, the acquisition of high-quality 3D models is usually achieved by expensive laser scanners. Although the accuracy is guaranteed, there are also two disadvantages: First, the laser scanner requires the object to be scanned to be absolutely still during the scanning process. It will lead to obvious errors in the scanning results; second, laser scanners are expensive and difficult to popularize to expand civilian value

Images