一种基于相位一致性相关的多源遥感影像配准方法  

A New Multi-source Remote Sensing Image Registration Method Based on Correlation of Phase Congruency

作　　者：范登科[1] 潘励[1] 叶沅鑫[1] 

机构地区：[1]武汉大学遥感信息工程学院,武汉430079

出　　处：《铁道标准设计》2012年第2期118-123,共6页Railway Standard Design

基　　金：国家973项目支持"高分辨率遥感数据精处理和空间信息智能转化的理论与方法"(2012CB719905)

摘　　要：针对几何校正过程中多源遥感影像同名点匹配率低的问题,提出一种基于相位一致性相关的遥感影像配准方法。该方法首先使用多尺度Harris提取出不受高斯平滑影响、位置稳定的角点,而后以金字塔分层映射为搜索策略,在参考影像上预测可能包含同名点的子区域,通过计算该子区域与待匹配点领域的相位一致性,同时引入相关系数作为相似性度量获取同名点对,最终实现不同光谱影像间的配准。实验表明,该方法的误匹配率较低,精度和稳定性高于传统原始影像灰度相关及同类配准方法,适用于不同传感器或不同光谱通道间影像的匹配。This paper provides a new image registration method based on the correlation of phase congruency to solve the problem that the matching ratio of the tie points of multi-source remote sensing images is rather low during geometric correction course.In this method,the corners of stable positions without influence of Gaussian smoothing are firstly filtered by multi-scale Harris,and then layered pyramid structure is utilized as the searching strategy to predict the subregion of reference image in which the tie point maybe locates.So,by calculating the phase congruency between this subregion and that of point to be matched,and by introducing the correlation coefficient as the similarity measurement to obtain the tie point-pairs,the image registration between different spectrums can be achieved.Experiment results showed that this method got low mismatching ratio but high accuracy and high stability in contrast to the traditional grayscale correlation method of original image and other similar methods,and could be suitable for the image matching between different sensors or different spectral channels.

关 键 词：影像配准 多尺度卷积 相位一致性 相关系数 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]