Abstract: Synthetic Aperture Radar (SAR) produces images of terrain targets by synthetically processing the returned signals and simulate them into a few hundred or even thousands meters antenna in order to increase the azimuth resolution through the same pulse compression technique, while using the movement of the installation platform of the radar and shining beams to the target from different time and from different positions. Currently, SAR imaging technique has become very crucial to obtain all kinds of information in many applications. For example, SAR imagery system is installed extensively in spy planes and satellites. Therefore, the vast amounts of data will be gathered via SAR every day, and it is very necessary to process them fast and accurately obtain the important information. This article developed a SAR image comparison method based on multi-scales wavelets transform by means of feature matching. This method will identify the Region of Interest (ROI) automatically and quickly for the SAR images of same region collected at different time. The experimental data indicates that this method can perform accurate SAR image comparison and ROI identification which has huge practical value.

Key words: Synthetic aperture radar, wavelet transform, multi-scales image feature matching, image comparison

摘要： 合成孔径雷达(SAR)就是利用雷达安装平台的运动，在不同的时刻从不同的位置照射目标，并使用信号处理技术将多次照射的回波仿真成尺寸长达数百米、甚至数千米长的天线来提高分辨率，从而对地面目标成像。目前，SAR图像技术应用已经成为一种极为重要的获取情报信息的手段。高空侦察飞行器和卫星上都装备有SAR成像系统，因此，每天都可获得巨大的图像数据。而快速、准确地从SAR图像中提取出重要信息已经成为一种迫切需要。本文提出一种基于小波变换在多尺度条件下用特征匹配进行SAR图像比对的方法，针对不同时间所采集的同一区域的SAR图像，自动快速地确定感兴趣的区域（ROI）并作出标识。实验结果表明，该方法能准确地进行SAR图像比对和ROI标识，具有实际应用价值。

关键词: 合成孔径雷达(SAR), 小波变换, 多尺度图像特征匹配, 图像比对