基于相关面特征的多子区异谱图像匹配算法

为实现用于异谱图像之间的精确匹配,提出一种基于相关面特征的多子区关系约束匹配算法。首先在实时图上选择出边缘信息丰富的区域作为待匹配子区,然后进行相关匹配,分析相关面特征找到符合形态特征的多个个局部最大值位置,将各子区位置关系作为约束条件对得到的局部最大值位置进行聚类,对匹配结果进行综合和可信度判断。试验表明,该方法可有效提高多子区相关匹配方法的性能,具有更高的匹配概率和更好的图像尺度和旋转适应性。

Multiple Subimage Matching Algorithm Based on Correlation Plane Feature

In order to match heterogenous images, a multiple subimage matching algorithm based on correlation coefficients feature was proposed. Firstly the multiple subimages were exteacted from the real image according to a predefined selection method, They were regarded as template and correlated with the reference image. Secondly the correlation coefficient features were computed and used to find multiple matching candidate positions corresponding to features’ top multiple local maxium . The multiple matching positions were integrated with the clustering method based on the spatial relation among the subimage. The confidence of the final matching result was given. The result shows the proposed method can improve matching performance of multiple subimages matching, including higher matching rate and better adability to scale and rotation.