Abstract: Image registration and object recognition algorithm based on local feature match often fails when aligning multispectral images. Aiming at such shortcoming, this paper presents a registration algorithm based on local feature match by using clustering. It builds optimized local feature descriptors according to the characteristic of multispectral images and obtains the correct set of corresponding points by using the statistical information of the relative main orientation. Experimental results confirm that the correct rate and registration accuracy are substantially improved by the algorithm.

Key words: multispectral image, local feature, feature clustering, registration, match

摘要： 针对基于局部特征匹配的图像配准和识别算法对于多谱段图像配准性能较差的缺点，提出一种基于局部特征聚类匹配的图像配准方法，该方法针对多谱图像的特点构建优化的局部特征提取算子，根据特征点对的相对主方向统计信息对初始匹配点集重新聚类，求出正确匹配的点对子集以实现配准。实验结果表明，该算法较大地提高了多谱图像配准的正确率和精度。

关键词: 多谱图像, 局部特征, 特征聚类, 配准, 匹配

CLC Number: 

• TN911.73