结合全局信息描述子的局部特征匹配算法

A SIFT matching algorithm based on global context DAISY descriptor

针对影像间存在相似性场景时容易出现误匹配的问题,该文提出了局部DAISY描述子与全局纹理信息相结合的算法。详细论述了DAISY描述子和全局纹理的生成过程。该算法首先提取SIFT特征;然后计算特征点的DAISY描述子与全局描述子,获得结合全局信息的DAISY描述子;最后采用卡方统计量作为度量函数进行特征描述子的匹配。实验结果表明,该文算法能得到较好的匹配结果。但是该算法在大角度旋转情况下误匹配率较大。

For the mismatching problem of images which have similar scenes, this paper presented a feature matching algorithm based on partial DAISY descriptor and global texture information. The generation process of DAISY descriptor and global texture were discussed. Firstly, SIFT feature was extracted; then, the DAISY descriptors and global feature point descriptor were calculated to obtain the DAISY de- scriptors combined with global information; finally, the Z2 statistic was used as a measure of function to match feature descriptors. Experimental results showed that the proposed algorithm could get better matc- hing results; meanwhile, the mismatch rate decreased with large rotation angle.