摘要
提出了基于非下采样Shearlet和几何结构的遥感图像无监督变化检测新算法。首先将两幅SAR图像相减取绝对值得到差异图像,然后利用基于非下采样Shearlet自适应贝叶斯阈值去噪算法对差异图像进行去噪处理来减少噪声的影响。最后根据差异图像的局部几何特征和邻域信息构造跨特征矢量,再利用模糊C-means聚类算法对跨特征矢量聚类,聚类的结果为变化类和未变化类即最终的变化检测结果。实验证明:该算法对噪声的抗噪性能平稳而且有效,可以得到较好的检测结果。
A novel unsupervised change detection algorithm of remote sensing images based on Nonsubsampled shearlet and Geometrical structure is proposed.Firstly,the difference image is composed of the absolute value of the difference of two remote sensing images.Then denoising algorithm based on Nonsubsampled shearlet adaptive Bayesian threshold is used to deal with the difference image to reduce the influence of noise.Finally,local geometric features and neighborhood information of the difference image are used to construct the cross-feature vector,and then the cross-feature vector is clustered by Fuzzy C-means clustering algorithm.The results of clustering is change class and no change class,which are the final change detection results.Experiments show that Anti-noise performance of the proposed algorithm is steady and effective and the proposed algorithm can get a better change detection results.
出处
《遥感技术与应用》
CSCD
北大核心
2014年第3期482-488,共7页
Remote Sensing Technology and Application
基金
教育部促进与美大地区科研合作与高层次人才培养项目(20101595)
