中低分辨率合成孔径雷达影像多纹理特征的Otsu变化检测

Otsu Change Detection of Low and Moderate Resolution Synthetic Aperture Radar Image by Using Multi-Texture Features

利用中低分辨率合成孔径雷达(SAR)影像,通过灰度共生矩阵提取不同纹理窗口大小的纹理特征来构造差异影像,并结合Otsu阈值分割方法来获取变化图像。实验结果表明,当检测地物单一、变化较明显的区域时,通过选用均值纹理特征并结合相应纹理窗口,中低分辨率SAR影像能够满足变化检测精度的要求。

Low and moderate resolution synthetic aperture radar （SAR） images are used to extract texture features with different texture window sizes by the gray-level co-occurrence matrix and construct the difference images. Otsu threshold segmentation method is combined to obtain the change image. The experimental results show that, as for the detection of areas with simple surface features and obvious changes, the low and moderate resolution SAR images can meet the high detection accuracy demand by using the mean-value texture feature in combination with the corresponding texture window.