Research And Applications Of 3D Face Digitization

As an important symbol of human identity,human face is of great significance in medical plastic surgery,video making,virtual reality,and other fields.The density and accuracy of point cloud are key factors related to 3D digital quality.Therefore,there are two key issues in 3D face digitization.One is gathering point cloud with high-density and high-precision.Another is point cloud registration with high effectiveness.Aiming at these two problems,this paper designs two systems: binocular structure light measurement system and multi Kinect measurement system.These corresponding algorithms are studied and refined,and the validity of the method is verified through experiments.The main work of this article is as follows:(1)The binocular vision measurement can't obtain the depth information of area which is with not obvious or no texture and has low point cloud density.The structural light measurement system has complex calibration.Due to these two problems,a binocular structured light measurement system and a special color structure optical template was designed to collect high density and high precision point cloud.The full projector pixel space was utilized to extract feature points quickly and accurately based on a pixel unit.It greatly increases the density of feature points and increases the density of point cloud.A feature point matching algorithm based on surface structure optical geometry is proposed,which has low computational complexity,high efficiency and high correct match rate.(2)Subject to the equipment perspective and the size of the measured object size restrictions,a single measurement equipment can't complete the full face three-dimensional digitization precisely.So,this article designed a multi-Kinect measurement system.Firstly,three Kinect sensors worked together to capture point clouds from three sides of the face.These points cloud were registered to obtain the complete face point cloud.Then,the bilateral filtering algorithm was refined to remove the noise of point cloud,and the point cloud registration area was simplified.Finally,this article introduced point feature histogram and the refined nearest point iterative algorithm to improve the iterative speed and efficiency of the ICP algorithm.The full point cloud was complete in the simplified point cloud registration area.(3)In order to verify the validity of the algorithms proposed in this paper,two 3D digitization experiments on face sculpture and real face were designed.Meanwhile,those related algorithms were compared with the algorithms proposed in this paper.In this paper,the root mean square error after point cloud registration was 0.0854 mm with the time of 18.47 s.This paper completed the full face 3D digitization fast and accurately with low cost.The algorithm performance has good superiority.