利用综合差异图和按块分类的SAR图像变化检测

发布时间：2018-07-10 05:23

  本文选题：变化检测 + 差异图像　； 参考：《遥感信息》2017年04期

【摘要】：针对基于差异图像分类的SAR图像变化检测方法中差异图像保持完整变化区域难的特点,以及按像元分类容易受噪声的干扰,提出了一种基于综合差异图像和按块k均值聚类法的SAR图像变化检测方法。首先,分别通过差值法和对数比值法得到2幅不同时相同一地理位置的SAR图像差异图像,为了使差异图像更加平滑和保持边缘信息,分别对这2种差异图像进行均值滤波和中值滤波。然后,通过简单的线性结合得到最终的差异图像,随后将差异图像分成若干个大小为h×h且不重叠的块,通过主成分分析提取每个块的特征向量,再利用k均值聚类法将特征向量空间分成2类。最后,根据最近邻法将差异图像分为变化区域和未变化区域。实验结果表明,该方法不仅能有效地检测出变化区域,还在一定程度上降低了虚警。
[Abstract]:In order to solve the problem that it is difficult for the difference image to maintain a complete region of change in SAR image change detection method based on differential image classification, the pixel classification is vulnerable to noise interference. A new method of SAR image change detection based on synthetic differential image and block k-means clustering is proposed. Firstly, two different SAR images with the same geographic position are obtained by difference method and logarithmic ratio method, respectively, in order to smooth the difference images and keep the edge information. The two kinds of differential images are filtered by mean value and median filter respectively. Then, the final differential image is obtained by a simple linear combination. Then, the differential image is divided into several blocks of h 脳 h size and no overlap, and the feature vectors of each block are extracted by principal component analysis (PCA). Then the k-means clustering method is used to divide the eigenvector space into two categories. Finally, according to the nearest neighbor method, the difference image can be divided into variable region and unchanged region. The experimental results show that this method can not only detect the region of change effectively, but also reduce the false alarm to a certain extent.
【作者单位】： 中国科学院遥感与数字地球研究所;中国科学院大学;
【分类号】：TN957.52
，

本文编号：2112222