基于自标定的两视图三维重构研究  

Research on 3D Reconstruction of Two Views Based on Self-calibration

作　　者：吴文欢[1] 陈占伟[1] 

机构地区：[1]周口师范学院计算机科学与技术学院,河南周口466001

出　　处：《科技广场》2012年第7期86-88,共3页Science Mosaic

基　　金：国家自然科学基金(No.60973096);航空科学基金(NO.2010ZC56007);周口师范学院青年科研基金资助项目(zknuqn201127B)

摘　　要：相机标定是三维重构中的关键步骤,本文对在自标定下的两视图三维几何模型恢复进行研究。先采用SURF算法得到图像匹配点对,在此基础上,利用RANSAC鲁棒估计基础矩阵;然后根据Kruppa方程所推导出的二次方程自标定出相机内参数,进一步地,通过分解本质矩阵求解出相机外参数;最后,根据三角化法求得空间三维坐标值。实验结果证实,该方法具有效性和正确性,并且能够真实地再现物体的三维模型。Since camera calibration is a critical step in the three-dimensional reconstruction,the 3D geometric model recovery of two views based on self-calibration is studied in this paper.Firstly,image matching points are obtained by using SURF algorithm,and then fundamental matrix is robustly estimated with RANSAC.Secondly,the intrinsic parameters are calculated according to a quadratic equation which is educed the Kruppa equation;moreover,the external parameters are also solved through SVD decomposition of the essential matrix.Finally,spatial three-dimensional coordinate values are easily obtained by the triangulation method.The experimental results validate the validity and correctness of this method,which can factually represent 3D model of objects.

关 键 词：SURF算法 基础矩阵 本质矩阵 自标定 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]