Synthetic Aperture Radar

Recent developments in high resolution SAR imaging has made possible the use of sophisticated machine vision algorithms on these images. Radar is an all time/all weather sensor, thus it can be used in situations (cloud cover, night time, etc.) where visible light sensors are practically useless. The poor resolution in Radar images compared to EO images have prevented sophisticated recognition schemes to be implemented for Radar Images until recently.

SAR provides a way of collecting and processing data such that the resulting image has 1 foot by 1 foot resolution. Moreover, the development of the Polarimetric Whitening Filter (PWF) by Lincoln Laboratory has provided a way of reducing speckle in SAR images without trading off resolution.

Our main interest is in modelling and detecting buildings. Unlike EO imagery, the full geometric structure of an object is not present in SAR imagery. The three prominent features of the SAR image of a building are:

- Bright line(s), resulting from the wall-ground dihedral(s) facing the radar.

- Shadow regions where there is no energy return.

- Backscatter from small structures on the roof of a building.

The detection of buildings in SAR images is of interest because of various reasons. It represents an interesting problem in Statistics/Signal Processing/Computer Vision. It also bears military significance for cutting down on false alarms in urban areas and for succesfully detecting targets close to buildings.

Back to research topics.