Research On Fine Registration Of Multidate SAR Images Of Varieties Of Characteristcs

Because synthetic aperture radar can work all day and night no matter if theweather is good. In recent years, the use of SAR images for disaster monitoring, surfacefeatures explained has been paid more and more attention. Before using INSARtechnology, DINSAR technology and change detection in disaster detection, theregistration of two images should to be done firstly. The main content of this paper isSAR image accurate registration used a variety of features.Harris operator is a classic algorithm for the fine registration of images; it ischaracterized with relatively stable feature. The same geographical features and can bequickly found among SAR images of different phase. Harris algorithm also hasdisadvantage, it is easily affected by noise as well as the terrain, which will lead to thefailure of the matching. In this paper, the corner points of the same geographical featurewhich are mismatched are excluded at the last. It can be proved that the algorithm caneliminate most of the false matching points, and the registration accuracy is greatlyimproved.The algorithm based on SIFT (Scale Invariant Feature) is an algorithm, which hasrelatively stable characteristic. Since the algorithm can the find the features in differentscale space, which are presence in the multiple scale space. The feature is made by the128-dimensional vector, which is based on the pro-domain information. The algorithmcan position the feature which is stable in different scale in different SAR images ofdifferent phase. By the distance between the feature vectors, the feature points of theSAR image are matched. As is the same in the Harris method, there are manymismatching points which are due to the impact of factors such as noise, phase andtopography or error in the match, the same method is used to weed out the falsematching points in this article.