Unsupervised Change Detection of Remote Sensing Images Based on SURF and SVM

Abstract

Change detection is one of the most exciting application of remote sensing image interpretation and processing. In this paper, we propose a novel approach for unsupervised change detection by integrating Speeded Up Robust Features (SURF) key points and Supporting Vector Machine (SVM) classifier. The approach starts by extracting SURF key points from both images and matches them using RANdom SAmple Consensus (RANSAC) algorithm. The matched key points are then viewed as training samples for unchanged class; on the other hand, those for changed class are selected from the remaining SURF key points based on Gaussian mixture model (GMM). Subsequently, training samples are utilized for training an SVM classifier. Finally, the classifier is used to segment the difference image into changed and unchanged classes. To demonstrate the effect of our approach, we compare it with the other four state-of-the-art change detection methods over two datasets, meanwhile extensive quantitative and qualitative analysis of the change detection results confirms the effectiveness of the proposed approach, showing its capability to consistently produce promising results on all the datasets without any priori assumptions.