首页 | 本学科首页   官方微博 | 高级检索  
     

任意视角的多视图立体匹配系统
引用本文:郭禾,傅新元,陈锋,王宇新,李寒.任意视角的多视图立体匹配系统[J].计算机工程与设计,2007,28(12):2873-2877.
作者姓名:郭禾  傅新元  陈锋  王宇新  李寒
作者单位:大连理工大学,计算机科学与技术系,辽宁,大连,116024
摘    要:为了得到高精度深度图,通过特征点提取与匹配、计算基础矩阵F、引导互匹配、捆集调整等一系列技术,求出每幅视图在同一空间坐标系下的高精度摄像机矩阵P.为任意角度的多个视图建立可扩展的主框架,不需要图像矫正.基于图像坐标和自动计算出的视差范围建立三维网络,为每条边分配适当的权值,把多视图立体匹配问题转化为网络最大流问题.算法保证了高准确率和平滑性,实验结果表明此方法适用于三维模型重建.

关 键 词:SIFT特征点  基础矩阵  摄像机矩阵  可扩展主框架  网络最大流  任意视角  多视图  立体  匹配系统  system  stereo  matching  三维模型重建  方法  结果  实验  平滑性  准确率  算法  最大流问题  三维网络  转化  匹配问题  权值  分配  范围
文章编号:1000-7024(2007)12-2873-05
修稿时间:2006-05-29

Integrated multi-view stereo matching system with arbitrary viewpoints
GUO He,FU Xin-yuan,CHEN Feng,WANG Yu-xin,LI Han.Integrated multi-view stereo matching system with arbitrary viewpoints[J].Computer Engineering and Design,2007,28(12):2873-2877.
Authors:GUO He  FU Xin-yuan  CHEN Feng  WANG Yu-xin  LI Han
Affiliation:Department of Computer Science and Technology, Dalian University of Technology, Dalian 116024, China
Abstract:The algorithm is decomposed into a number of successive steps to obtain precise depth map. Feature extraction and matching, computing F matrix, guided complementary matching, bundle adjustment etc. are first applied on an image sequence. Then the precise camera matrix for each image is extracted. The algorithm builds extendable main frame on the images with arbitrary viewpoints, constructs 3D network based on image coordinates and automatically computed disparity range, and assigns cost value for each edge. The multiview stereo matching problem is finally translated into the maximum flow problem. The algorithm ensures to high precision and smoothness in 3D scene reconstruction, and it is well performed in the experiment.
Keywords:SIFT feature  fundamental matrix  camera matrix  extendable main frame  maximum flow
本文献已被 CNKI 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号