Studies On Change Detection Methods For Multitemporal Remote Sensing Images

The change detection of remote sensing images is a kind of technique of identifying change information by the two images acquired over the same area at different date. With the increasing 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 environmental monitoring, land use/land cover, agricultural surveys, urban change analysis, military surveillance, hit effectiveness assessment and so on.Three change detection methods of single-band multi-temporal remote sensing images are proposed in this paper.(1) A change detection method of remote sensing images based on non-parametric density estimation and weighted Markov random fields (MRF) is proposed. Kernel density estimation is used to estimate the class conditional probability density of differencing image, and is combined with weighted MRF to classify the differencing image. Finally, the change detection map is generated. The effectiveness of proposed method is verified by the experiments of the simulated data sets and real remote sensing data sets.(2) A change detection method of remote sensing images based on geostatistics texture and reliable scale fusion is proposed. Many mobile windows are used for computing the value of pseudo cross variogram function and constructing differencing images. These differencing images form multi-scale differencing image sequences with the differencing image which is constructed by difference method. The reliable scale fusion is applied to multi-scale differencing image sequences. Finally the change detection map is generated. The effectiveness of proposed method is verified by the experiments of the simulated data sets and real remote sensing data sets.(3) A change detection method of remote sensing images based on nonlinear anisotropic diffusion and fuzzy enhancement is proposed. The differencing image which is constructed by difference method is processed by nonlinear anisotropic diffusion filter operation and fuzzy enhancement operation. The final result of change detection map is produced by utilizing Otsu threshold and two-dimensional fast Otsu threshold for the enhanced differencing image. The effectiveness of proposed method is verified by the experiments of real remote sensing data sets.This paper is supported by National Natural Science Foundation of China (No.60672126 and No.60970066) and National High Technology Research and Development Program of China (863 Program) (No.2007AA12Z223 and No. 2009AA12Z210).