向量场一致性约束的对立颜色特征匹配

Opponent Color Feature Matching Constrained by Vector Field Consistency

针对Color SIFT类特征提取算法进行特征匹配时误匹配点对多的问题,引入向量场一致性约束改进了Opponent SIFT的匹配算法。首先,在对立颜色空间的3个通道中采用SIFT算子进行特征描述,构建384维特征向量,增强颜色对比效果;然后,利用最近邻距离比率进行粗匹配,降低整体匹配失效的风险;最后,采用向量场一致性约束提纯,规避RANSAC算法因局部最优导致正确点对丢失的缺陷。实验结果表明:在处理图像综合变化时,本文算法较SIFT结合RANSAC的算法,提高了匹配精度,鲁棒性更好。

According to a large number of mismatched point pairs in image feature matching with the Color SIFT extracting algorithms,this paper introduces vector field consistency constraint to improve an Opponent SIFT matching algorithm.First,SIFT operator is used to describe local feature in 3 channels of opponent color space and 384 dimensional feature vectors are constructed to enhance the color contrast effect.Then,the nearest neighbor distance ratio is utilized to make coarse matching and reduce the overall failure risk of matching.Finally,vector field consistency constraint is used to purify and evade the defect of the correct point lost due to local optimization caused by the RANSAC algorithm.Experiment results show that when dealing with comprehensive image changes,the proposed algorithm improves the matching accuracy and robustness compared with the SIFT with RANSAC algorithm.