Remote Sensing Image Change Detection Based On Wavelet Transform And Fuzzy C-Means Clustering

Remote sensing image change detection is to identify changed information over time in the surface features by the two or multiple remote sensing images acquired over the same area at different data. With the increasing development of techniques of acquiring remote sensing images, change detection is more and more widely used in forest resources survey, agricultural resources surveys, environmental monitoring, natural disaster assessment, urban change analysis and so on.Although the research of remote sensing image change detection has made great progress, there have been a variety of change detection methods. But these methods need manual intervention with very low automation level. Therefore, how to achieve automatic detection of remote sensing images has become a hot of current research in this field. This paper studies some key technologies of unsupervised remote sensing image change detection, and has completed the following works:1. Introduce the theoretical knowledge and development status at home and abroad of remote sensing image change detection, summarize and review the existing detection methods, expound the processing flow and choice of remote sensing data for remote sensing image change detection, and preprocess remote sensing data by using radiometric correction.2. A change detection method of remote sensing images based on wavelet fusion and fuzzy C-means clustering (FCM) is introduced. The advantages and disadvantages of difference images acquired by differencing operator and log-ratio operator are taken into consideration, the method introduces discrete wavelet transform and uses the rules of mean and maximum absolute to construct new difference image by wavelet fusion the results of two difference images. The "hard" segmentation methods of traditional difference image make it’s difficult to select a threshold, fuzzy C-means clustering can achieve automatic segmentation of difference image and detect changed region.3. A change detection method of remote sensing images based on undecimated discrete wavelet transform (UDWT) and fuzzy local information C-means clustering (FLICM) is proposed. The difference image is decomposed by UDWT, the high-frequency wavelet coefficients of diagonal direction at each scale are removed, the rest of the multi-scale wavelet coefficients and difference image itself combine to obtain multi-scale feature vector. Then, multi-scale feature vector is clustering by using FLICM to determine changed region. The method consider the wavelet coefficients at different scales by constructing multi-scale vector, FLICM takes the spatial neighborhood information into account to improve the performance of change detection. The experiment results on real remote sensing data demonstrate the effectiveness of the method.