An Unsupervised Change Detection Approach for Remote Sensing Image Using SURF and Gabor Wavelet Features

In this paper, we propose a novel approach for unsupervised change detection by integrating Speeded up robust features (SURF) key points and Gabor wavelet features. First, SURF key points and Gabor wavelet features are extracted from two temporal remote sensing images respectively. Then, the difference image is generated based on Gabor wavelet features, in which the adjacency between pixels is incorporated to obtain a local relationship. Subsequently, the SURF key points are matched by using Random Sample Consensus (RANSAC) algorithm. The matched key points are viewed as training samples for unchanged class and those for changed class are selected from the remaining SURF key points based on Gaussian mixture model (GMM). Finally, training samples are utilized for training a support vector machine (SVM) classier which is used to segment the difference image into changed and unchanged classes. Experiments employing LandSat images demonstrate the effectiveness of the proposed approach.