基于虚拟现实技术的三维图像分层重构研究  

Research on Hierarchical Reconstruction of 3D Image Based on Virtual Reality Technology

作　　者：尤翔[1] YOU Xiang(College of Civil Engineering,Nanjing University of Technology,Jiangsu Nanjing 211816,China)

机构地区：[1]南京工业大学土木工程学院,江苏南京211816

出　　处：《计算机仿真》2020年第7期475-479,共5页Computer Simulation

摘　　要：针对传统三维图像分层重构方法存在特征点匹配性能较差的问题,提出一种基于虚拟现实技术的三维图像分层重构方法。利用基于深度图像渲染(Depth Image Based Rendering,DIBR)技术对三维图像实施深度预处理,包括对三维图像进行深度映射处理与深度平滑处理,减少三维图像的噪声与三维变换导致的图像空洞的数量与大小,利用固定块数法将三维图像分块,通过角点检测算法实现三维图像的角点检测。根据检测结果,通过灰度相关约束实现角点的初始匹配,基于基本矩阵,对初始匹配中存在的误匹配进行剔除,根据角点匹配结果,利用基于虚拟现实技术对三维图像进行三维分层,利用三角剖分算法实现三维图像的分层重构。实验结果表明,与传统方法相比,所提方法特征点匹配值额度高,表现出良好的特征点匹配性能,后续图片压缩质量高,为三维图像分层重构提供了新的路径。In traditional 3 D image hierarchical reconstruction methods,the matching performance of feature points is not ideal.Therefore,this paper presented a method of layered reconstruction for 3 D image based on virtual reality technology.Depth Image Based Rendering technology(DIBR)was used to implement depth preprocessing for 3 D image,including depth mapping and depth smoothing for 3 D image,so that the noise in 3 D image and the number and size of image holes caused by 3 D transformation were reduced.Moreover,the fixed block method was used to partition the 3 D image.The corner detection algorithm was used to detect the corner in 3 D image.According to the result,the initial matching of corner points was completed by gray correlation constraints.Based on the basic matrix,the mismatches in initial matching were eliminated.According to the matching results,the virtual reality technology was adopted to divide the layers of 3 D image.Finally,triangulation algorithm was used to achieve the layered reconstruction of 3 D image.In order to prove that the 3 D image layered reconstruction method based on virtual reality technology has strong feature point matching performance,the traditional 3 D image layered reconstruction method is compared with this method in feature point matching performance experiment.Simulation results show that the matching value of feature point of the proposed method is higher than the traditional method.The matching performance of feature points is good,and the quality of subsequent image compression is high.This method provides a new way for the layered reconstruction of 3 D image.

关 键 词：虚拟现实技术 三维图像 分层重构 深度映射处理 

分 类 号：TP751.1[自动化与计算机技术—检测技术与自动化装置]