基于词包模型的高分辨率SAR图像变化检测与分析

Change Detection and Analysis of High Resolution Synthetic Aperture Radar Images Based on Bag-of-words Model

该文面向高分辨率SAR图像解译中的变化检测问题,针对其研究现状与难点,重点解决高分辨率SAR图像变化检测中的语义信息缺失问题,提出一种基于词包模型的变化检测与分析的方法。该方法利用词包模型,对两个时相的图像做词包表征,将视觉直方图的差作为变化向量进行分析。由于变化向量包含有语义信息,因此可通过对其分析,结合像素级变化结果,实现对变化区域的语义分析及感兴趣变化类型检测。经实验验证,该框架对高分SAR影像变化语义分析具有应用前景。

This paper discusses the change detection in high-resolution SAR image interpretation. Referring to the unfavorable elements in the change detection and the status quo, this paper focuses on resolving the semantic information deficiency problem in SAR image change detection. A method named change detection base on Bag of Words Model （BoWM） is proposed. By using the BoWM, two visual histograms of two different temporal images are obtained, and the histogram difference, which contains semantic information, is defined as the change vector. By analyzing the change vector and combining it with the statistical change detection method, the semantic analysis and interest change-type detection of the change area can be obtained. Experiments show that the proposed method may be applicable to the semantic analysis of the change area in high-resolution SAR images.