The Research Of Change Detection In Remote Sensing Images Based On Fractional Fourier Transform And Gabor Wavelets

The change detection technique is one of most important applications for remote sensing image processing. Change detection of remote sensing images is a process of identifying changes occurred on the Earth surface by analyzing a pair of co-registered remote sensing images acquired on the same geographical area but at different time instances. It has been widely applied in many civilian and military fields, such as natural disaster assessment, ecological environment monitoring, and dynamic surveillance of battlefield, and so on.This thesis focuses on the research of change detection of remote sensing images based on fractional Fourier transform (FRFT) and Gabor wavelets, whose main content are stated as follows:In this paper, fractional Fourier transform, as an important branch of nonstationary signal processing theory, is first utilized in the change detection of remote sensing images for proposing a novel unsupervised change detection algorithm. Specifically, we perform the low-order FRFT on multi-temporal images acquired on the same geographical area but at different time instances, then calculate the difference image according to the type of remote sensing image. Next, principal component analysis (PCA) is adopted to yield the eigenvector corresponding to each pixel by considering its spatial context information. Further, to postulate change detection as a binary classification issue, the change map is obtained by classifying eigenvectors into two classes:changed and unchanged with K-means. Finally, experimental results, reported on both optical and synthetic aperture radar (SAR) images, verify the validity of the proposed algorithm.Secondly, another change detection algorithm of multi-temporal remote sensing images is proposed based on Gabor wavelets and two-stage clustering, for which the difference image, obtained from multi-temporal images acquired on the same geographical area at two different time instances, is decomposed by Gabor wavelets to extract multi-scale and multi-direction feature vector for each pixel. Meanwhile, two-stage clustering scheme with Fuzzy C-Means (FCM) is introduced to improve the accuracy of change detection. Compared with the existing algorithms, this proposed algorithm can achieve better detection performance.Furthermore, we present a novel technique based on Gabor wavelets and PCA for unsupervised change detection of multi-temporal remote sensing images. Similarly, the multi-scale and multi-direction Gabor feature vectors are extracted from the different image by implementing Gabor wavelets transform. But to reduce the computational complexity, PCA is applied for dimensionality reduction of Gabor feature vectors. And then the change map is generated by clustering the feature vectors with the K-means and FCM algorithms, respectively, whose performance is compared. Experimental results, reported on both optical and SAR images, verify the validity of the proposed algorithm.