Monitoring the urban environment with multitemporal SAR data

The TerraSAR-X images, characterized by resolutions up to one meter and an eleven days revisit time, provide new opportunities in various applications that range from environmental monitoring, to rapid emergency response, to the detection of infringement of local building regulations. This paper investigates the monitoring of land use and the novelty introduced by it is the classification of the single pixel through the exploitation of information contained in its amplitude's time series rather than taking advantage of the high spatial resolution for texture estimations. More specifically, this work covers the identification of the scattering behavior in time of three classes of urban targets: buildings, roads and squares and trees. The performance of the proposed algorithm has been tested on a data-set of 126 TerraSAR-X images of the Italian city of Milan. The proposed approach is based on the separation between the metropolitan area and surrounding countryside based on interferometric coherence, the subsequent pixel based classification on the urban pixels and the monitoring of land changes with a Bayesian step detector. The sinusoidal scattering in time of some buildings and the stealth nature of some skyscrapers have also been discussed. This three-step classification and monitoring algorithm has shown to provide a satisfactory classification performance and has great potential to benefit public institutions in the detection of urban sprawl and soil sealing.