Improving Change Detection Using Geometrical Features

While pixel-based or object-based change detection approaches have been proposed in technical literature, change detection techniques based on feature extraction and comparison are less numerous, due to the reliability the extraction. This paper discusses how to exploit aggregated information obtained from geometrical features to obtain hints to changed areas and to unchanged areas as well. This approach not only allows to detect changes, but also to improve existing change detection results by reinforcing change areas and rejecting portions of the scene where geometrical features do not appear to have undergone any change. Results exploiting COSMO/Skymed data in Haiti show that the approach is valuable, and has great potential for improving change detection results obtained by more standard per-pixel approaches.