Hybrid analysis for SAR change detection based on time series data

This paper presents the Hybrid Analysis for Synthetic Aperture Radar (HASAR) framework for time series change detection (TSCD). The HASAR framework utilizes spatiotemporal observations to identify abrupt changes in SAR data stacks. Full resolution is used and no filtering processes are involved. Moreover, no thresholds or prior knowledge of the changes are required. The HASAR framework includes two steps: 1) extraction of spatiotemporal observations and 2) automatic change detection. First, it extracts the time series of statistically homogeneous pixels (SHP) to acquire the spatial temporal observations. Second, it performs the automatic change detection on the spatiotemporal observations to target significant changes in time. This hybrid analysis integrates pixel-/object-based processing techniques with spatial/temporal observations. It preserves details while mitigating the speckle effect. The temporal observations also help justify the significance of the changes. In our experiment, we apply this technique on 30 ERS-2 images. The results show that the HASAR system performs well and possesses the characteristics of generalization. A vast amount of change analysis can be carried out using the same framework. Therefore, the proposed method reveals high potentials for continuous Earth monitoring.