Land Cover Change Detection for Fully Polarimetric SAR Images

Land cover change detection is an important application of remote sensing data. Polarimetric Synthetic Aperture Radar (PolSAR) image offers increased potential to detect and monitor changes in land cover over multitemporal images. This paper explores the use of unsupervised Normalize Difference Ratio (NDR) as the change index (CI) followed by supervised thresholding to extract binary mask for changed and unchanged areas. The datasets used for this experiment are UAVSAR fully polarimetric L-band images acquired on two different dates over Hayward, California, USA. The proposed change detection technique is compared with standard supervised techniques. The result indicates that the proposed NDR technique has a higher detection rate than the traditional Differencing techniques and PCCD techniques.