【摘要】 针对大尺寸立体视觉测量中存在较大视差和透镜畸变等因素导致极线约束匹配率低的问题,提出了将图像校正与随机采样算法相结合的立体视觉图像匹配方法.对立体图像对进行线性校正,建立初始匹配点集;采用随机采样算法估计图像对之间的基本矩阵,恢复原始图像对之间的对极几何约束关系,剔除初始匹配点集中未匹配和误匹配的特征点,从而获得精确的匹配点集.该方法已应用于合成孔径雷达(SAR)大型可展开微波天线网面的实际测量,匹配率高于96%.更多还原

【Abstract】 An image matching method of stereo vision through combining image correction and random sampling algorithm is proposed.Accurate matched point sets are obtained in the process of performing linear correction to a stereo image,constructing original matching point sets according to the correcting result, estimating the fundamental matrix between image pair by using of random sampling algorithm,restoring the epipolar geometry restriction relationship between the original image pair,and picking out the unmatched and error matched feature points in the original matched pint sets.This method has been applied in the measurement of the surface of large-scale expandable microwave antenna of synthesized aperture radar(SAR). It is indicated by the measurement result that the matching ratio of the method is greater than 96%.更多还原