Research On The Feature-based SAR Image Automatic Registration Of SAR Images

Remote Sensing image registration is to register data from different time or different sensor platform to one pixel system. Due to different resolution, time and imaging mechanism for different sensors, remote sensing image is inevitably moving, rotating and zooming which lead to unable fusion directly. All of these decide that image registration is necessary. Synthetic Aperture Radar (SAR) can work all day, all-weather, and can go through the trees to get high resolution SAR image. High resolution SAR image is not only using widely in army application, but also in civil application. More and more research on SAR image is deep.SAR image registration is a necessary step of SAR image processing. It is essential for the fusion between SAR image and optical image, and SAR image detection and so on.This thesis does work based on these states. Firstly, chapter 2 introduces the principle knowledge about remote sensing image registration. And then, chapter 3 has a detailed description on the present condition and methods of SAR image registration. After comparing these algorithms and analyzing them, some conclusions can be obtained. At the same time, experiments on Mutual Information and Alignment Metric methods show their application in SAR image registration. Based on these theories, chapter 4 proposes a coarse-to-fine feature-based method for SAR image registration. The segmentation process and the match process have considered the speckle noise of SAR image. The experiment shows it applicable and effective, no matter from the time consuming and precision. At last, it concludes the above content and gives an expectation for SAR image registration.