Automatic Detection of New Building Construction from Sentinel-1 Multi-temporal Imagery

An automatic processing chain for detecting a new building construction solely from multitemporal satellite synthetic aperture radar data was developed and introduced in the paper. Although it was developed while utilising Sentinel-1 imagery, it can be adapted to other satellite radar systems. The solution was developed to be used mainly in protected critical infrastructure zones to identify prohibited constructions. In order to eliminate detection of changes happening on natural surfaces due to common agricultural works or seasonal changes in vegetation, the algorithm mainly uses a combination of a backscatter change statistical evaluation and coherence change evaluation. The solution was first tested in a selected area, including urban and natural environments. Although mean success in detecting a new building construction from a single track was only 36%, processing data from more tracks significantly increased the probability of a new building detection, especially of the larger ones.