4D change detection based on persistent scatterer interferometry

We propose a novel time series analysis based on persistent scatterer interferometry (PSI) to detect spatial big changes (3D) such as construction along with their occurrence times (1D). PSI detects and analyses persistent scatterer (PS) points, which are characterized by strong and coherent signals throughout time-series SAR images and usually form building-shaped patterns in urban areas. Hence, potential PS points that disappear or emerge at a specific date because of big changes are discarded. We define such points as big change (BC) points. In our approach, pixels with high temporal coherences are first detected as PS points by a standard PSI processing. We introduce change index sequence for each pixel, which are computed from its temporal coherences in different image subsets defined by time-series break dates, to quantify its probabilities of being BC points at different dates. The change indices of the pixels are used to design an automatic thresholding method to extract BC points. Afterwards, the disappearing or emerging date of each BC point is detected from the break dates based on temporal variation in its change index sequence. The simulation test proves the overall, producer's and user's accuracies better than 99%. In the real data test, the patterns of the disappearing and emerging buildings are successfully recognized in Berlin, Germany along with the occurrence dates.