Studies On Change Detection Technique For Multitemporal Remote Sensing Images

With the rapidly development of techniques and tools of acquiring remote sensing images and mass accumulation of remote sensing images data, change detection is more and more widely used in resource and environmental monitoring, geography space information updating, agricultural surveys, urban change analysis, military surveillance, hit effectiveness assessment and so on. The problem is usually divided into three stages: data selection and preprocessing, use the appropriate detection algorithm to obtain change info, postprocessing and detection performance evaluation. The key techno- logies are change detection technique and discrimination technology. Therefore, it is significant to research further the change detection technique and target discrimination technology.The main research work in the dissertation is as follows:(1)Since the tendency of multi-source data and integration process in the future applications, a novel change detection method based on template which is made by high resolution optical image is studied systematically. Firstly, high resolution optical images are used to make binary templates. Secondly, the template and the two images to be detected do registration separately, on this basis of that, we obtain the region of interesting by calculating the spatial transformation of the template and the image sequence. Thirdly, we adopt the segmentation sea area images through weighted fusion of template and different temporal images, then use Constant False Alarm Rate (CFAR) algorithm to extract the changed targets. Finally, target features are extracted to perform change analysis.(2)Since the tendency of more and more intelligent methods in the future applications, a novel change detection technique for multi-temporal SAR images using nonsubsampled contourlet transform has been proposed. Firstly, we use the ratioing map which is obtained by taking ten times the logarithm in base ten of the ratio of the intensities. Secondly, we get the sub-images through the multi-scale decomposition of ratioing map, and use the scale correlation image denoising for speckle reduction. Thirdly, use the improved CFAR detector to detect changes in every sub-image. Finally, the result map is derived according to an adaptive fusion algorithm.(3)Since the tendency of integration knowledge in the future applications, focusing on the discrimination technology in the postprocess of change detection, a novel ship target discrimination method with the combination of ordinary target discrimination technique and knowledge based target discrimination technique has been proposed. Firstly, extract the shape features of length, width, length-width ratio, area, bow shape, stern shape, shape complexity, outline moment sequence etc, combine with the knowledge information of mean CFAR, peak CFAR, target location, formation and so on. Secondly, select the optimal features, and then do a comprehensive process. finally, we obtain the discrimination results. Through the complementary process of the knowledge and features, the technique is more efficient to reduce false alarms.The proposed methods and algorithms are applied to both real and synthetic multitemporal remote sensing images and the experiment results are satisfied. Some of our methods have also been applied in the relevant projects.