Change Detection In Synthetic Aperture Radar Images Based On Data Fusion On Ration Images And The Extraction Of The Homogeneous Regions

Change detection belongs to the field of image processing, which is generally referred as a process to determine the change of the state of an object or the change in a phenomenon according to many observations in different times”. With the development of remote sensing technology, the Synthetic Aperture Radar(SAR) images become the most widely used images in the change detection problem. The SAR images have the incomparable advantages of having the ability to all-weather working and to penetrate to the ground and not being affected by the atmosphere, climate and other random factors. Until now, many scholars at home and abroad have done a lot of researches in the change detection problem and the precision of the change detection results have been improved. The unsupervised change detection algorithms are the most widely used change detection algorithms. The main steps of the unsupervised change detection algorithms include the preprocessing of the SAR images, the construction of the difference image and the analysis of the change detection map. This paper focuses on the study of the construction of the difference image in change detection problem. Here, we propose two methods, both of which are based on data fusion, to do the construction of the difference image in change detection problem. The details are described as follows:1. The method of the construction of the difference image in the change detection problem is proposed, which is based on data information. This method directly fuses two difference images produced by mean-ratio operator and log-ratio operator, separately. The chosen fused rule is numeric and very simple, but it can effectively utilize the complement information in two ratio images and is helpful for the next processes in the change detection problem. For the fused rule being numerical and simple, the change detection result is not so excellent without some extra processes. The experiments validate our judgment, namely the method can only produce the difference image of which the performance is close to the better difference image produce by either of the two ratio operators. Thus, we consider adding an extra processing step to excavate the abundant information contained in this fused difference image produced by the fusion operator, producing the final change detection difference image, so as to obtain more excellent change detection results. So we propose a second change detection method, which is also base on data information, namely the second method, the one based on data fusion and the extraction of the homogeneous regions for change detection in SAR images.2. The second method is also to construct the difference image in change detection problem, which is based on data fusion and the extraction of the homogeneous regions. The second method can improve the change detection results produced by the first method, which is achieved by doing a multi-scale analysis and process of the difference image produced by the first method. And the second method is validated by the experiments with more precise change detection results. The second method is not only applied to the widely used SAR images, but also to the SAR images at the region of the Yellow River Estuary. The dataset has its own specificity for the huge difference of the spark level between the images taken in different times and most widely used change detection methods are hard to deal with the datasets. Compared with the first method, the second one add another process to extract the homogeneous areas in the change-detection difference images.