Sar Image And The Panchromatic Image And High Precision Registration And Change Detection Technology Research

The rapid development of science and technology has brought a lot of information, but the technology of information processing can not meet the requirement of a large amount of information. Image, as a one of the main carrier, contains a lot of information from different times in different regions. More and more information can be obtained from image processing, and precise image alignment guarantees adequate extraction of information characteristics.Because of the difference of imaging mechanism and pixel manifestations between SAR and panchromatic image, and the inherent image speckle noise of SAR image, it is very difficult to align the two kinds of images. Traditional alignment methods such as mutual information registration and cross-related registration can not show the desired effect of SAR image and panchromatic image registration. In response to these issues, this paper designed two filtering algorithms for SAR image noise suppression to reduce the interference of the SAR image noise. And then introduce a new registration measure cross-cumulative residual entropy.Cross-cumulative residual entropy can be applied to both discrete and continuous environment, and it has better effect on noise rejection. Use Cross-cumulative residual entropy to measure the similarity of SAR and panchromatic image, extract the ratio gradient feature, and then compare these two results to complete SAR and optical image registration. The experiment result show the method can solve image registration with larger angle rotation, it has higher registration accuracy and better performance than other usual methods.Based the requirement of remote image interpretation, the identification method of building region is discussed in this paper. The paper proposed the idea of building area recognition and detection using knowledge base. The knowledge base technology can combine the surface features of image and the features property rules to identify the target area from the image. Because the remote sensing image data have large capacity information, the entire image change detection will be time and memory consuming. In this paper, the interested region were detected from different time image by the knowledge base technology, then change detection was done on the target areas. The experiment results the method is feasible.