Change Detection Approach Using Fuzzy Labeled Co-Occurrence Matrix on Multi-Temporal TerraSAR-X Images

In this paper, an automatic unsupervised change detection approach specifically for the analysis of single-channel single-polarization high-resolution SAR images is proposed. The temporal change feature images, namely positive change and negative change, are derived from the fuzzy labeled co-occurrence matrix. Then, the multi-level Otsu method is applied on the temporal change feature images to classify the changed and unchanged class, and the change detection map is obtained finally. Experiments were carried on the multi-temporal TerraSAR-X images. Quantitative classification results confirm the effectiveness of the proposed approach.