基于NSCT与FCM的多时相遥感图像的变化检测

Based on NSCT Combination with FCM Multitemporal Remote Sensing Image Change Detection

提出了一种基于非下采样Contourlet变换和模糊C均值聚类相结合的方法。该方法首先对两时相遥感图像进行相减运算得到差异图像。再对差异图像进行NSCT多尺度分解得到子带图像,将各子带图像与差异图像本身构成特征向量。最后通过使用模糊C均值聚类算法对多尺度特征向量进行分类得到最终的变化检测结果(变化和非变化类)。该算法不受变化类和非变化类统计分布的限制,不需要先验知识,适用性强。对真实遥感数据集进行研究,实验结果表明本文方法可以得到较好的检测效果;将本文算法与传统方法相比,该方法具有更好的检测精确度和抗噪性能。

we propose a method which is based on nonsubsampled contourlet transform （NSCT） combination with fuzzy C-means clustering algorithm in this paper. By subtracting the two remote sensing images, we can get difference image at first. Then the method can decompose the difference image with NSCT into subband images, the feature vector space is formed with the subband images and the difference image. Finally, by using Fuzzy C-means clustering algorithm for multi-scale feature vectors to get the final classification change detection results （change and unchanged）. The algorithm is not limited by the statistical distribution of change class and non-change class. It doesn’t require prior knowledge and is great applicability. Experimenting on real remote sensing data shows that it can get better detection effect, compare with the traditional methods, it has better detection accuracy and anti-noise per-formance.