Multi-temporal Remote Sensing Image Change Detection Based On Pixel-Level

Change Detection based on remote sensing images obtained at different time is a technology which analyzes and identifies the change characteristics and process of the earth’s surface quantitatively. As a key technology, remote sensing image change detection has a very wide range of applications such as resource survey, environmental monitoring, basic geographic database update, battlefield situation analysis, objective dynamic monitoring, evaluation of attack effect and so on, and it is an important research direction of remote sensing technology.So far, lots of researchers have proposed many methods on change detection. However, with the study goes deep continuously, the researchers come to realize that the methods of change detection differ in many ways due to different application requirements, The universality of these algorithms is poor, Till now there is not a algorithm whose performance is good in all ways. Meanwhile ,the direct result of the image change detection method is the estimate to the possibility of each pixel change, and then distinguish each pixel through a given threshold. However, the threshold selection is very difficult at actual application, it is difficult to distinguish all detection results use a fixed threshold. And some algorithms is very sensitive to the threshold change, the mild change of threshold may lead to the sharply change of detection results indicator.Basic on above, taking national defense pre-research project "Multi-source XXXXXX Fusion Technology and Demonstration System" as the background, and with the support of a large number of optical remote sensing image data, the paper made a deep study and discussion combining the method of theoretical analysis and experimental verification to the remote sensing image change detection based on pixel-level. To sum up, the main work as follows:1, following the same steps, it reviewed and summarized comprehensively the previous studies, and summarized the existing change detection methods;2, based on the summarization previously, it described systemically the mathematical model and the general process flow of remote sensing image change detection, and selected six kinds of typical pixel-level change detection algorithms which based on different statistical assumptions model to validate using experiment;3, according to experiment tests results, it made theoretical analysis and evaluation of performance indicators to the six kinds of algorithms, and proposed a fusion change detection methods based on D-S evidence theory in order to achieve complementary advantages between different algorithms, and to minimize factors constraint such as light and noise and threshold selection difficult, to improve detection accuracy, reliability and robustness; 4, a multi-temporal remote sensing image change detection system is build, to realize human-computer interaction, and expected to apply at the port and airport target change analysis in the future.