针对高分辨率遥感影像城市区域地类复杂的特点,提出了一种融合光谱和几何特征的多时相高分辨率遥感变化检测方法。在获取光谱和几何差异影像的基础上,采用模糊分类的方法,进行变化类和非变化类区分,建立各类的隶属度图像。采取基于模糊集理论的融合算法,对各种检测方法获取的隶属度图像进行有效融合,以减小部分区域的模糊度,提高变化类和非变化类的检测精度。采用城区多时相高分辨率影像为实验数据,进行各种变化检测方法的对比与分析,实验结果表明,融合光谱和几何特征的方法所获取的检测结果与光谱、几何差异影像法相比,具有检测精度高、漏检率低的特点。

In this paper, an automatic change detection method based on fusion of spectral and geometrical features for multi-temporal high-resolution remote sensing image is proposed. Firstly, as to the characteristic of complicated classes in urban area, spectral and geometrical difference images are developed using multi-temporal images. Secondly, with difference image as input, the membership images of changed and unchanged classes are acquired using
fuzzy classification method. Thirdly, a fusion model based on fuzzy logic theory is used to combine all kinds of membership images in order to reduce the fuzziness and to distinguish the changed and unchanged classes of special areas efficiently. Finally, change detection result is obtained using threshold segmentation algorithm and accuracy evaluation is given. Multi-temporal high-resolution images of urban area are taken as experimental data, and the experimental results prove that the change detection result fusing spectral and geometrical features has higher detection precision and lower undetected ratio compared with those from spectral or geometrical difference image methods.