摘要
提出了一种适用于未标定图像三维重建的立体匹配算法。该算法首先引入限制因子来消除Harris角点聚簇的现象,使用高斯曲面拟合内插使Harris角点达到亚像素级;接着采用特征点的Sift特征描述符进行初始匹配,利用随机抽样算法估计基础矩阵的同时剔除误匹配点对;最后在估计的基础矩阵的引导下进行双向匹配。实验证明,该算法能够很好地恢复物体的结构,是一种有效的用于未标定图像三维重建的立体匹配算法。
This paper proposed a stereo matching algorithm for 3D reconstruction based on uncalibrated images. Firstly,this algorithm refered limiting factor to eliminate the clustering phenomenon of Harris corners,and used Gaussian quadrics fitting to let Harris corners reach sub-pixel level. Then,constructed the Sift feature descriptor around the feature point for initial match, used RANSAC algorithm to estimate fundamental matrix,at the same time,excluded outlier points. Finally,proposed bidirectional matching to find unmatching points by using fundamental matrix. The experimental results show that object structure can be well reconstructed by using this algorithm,and it is an effective stereo matching algorithm for 3D reconstruction based on uncalibrated images.
出处
《计算机应用研究》
CSCD
北大核心
2010年第10期3964-3967,共4页
Application Research of Computers
基金
国家自然科学基金资助项目(60873094)
国家"863"计划资助项目(2008AA01Z301)
西北大学研究生自主创新基金资助项目(09YZZ65)
关键词
立体匹配
未标定图像
三维重建
限制因子
亚像素
双向匹配
stereo matching
uncalibrated images
3D reconstruction
limiting factor
sub-pixel
bidirectional matching
