Robust adaptive detection of optical patterns in low resolution airborne radar images

Abstract

Deals with the conception and design methodology of detectors for the location of prominent objects in airborne radar imagery. In its first half, several detection configurations are compared (linear, squared, logarithmic and binary detectors) working on impulsive clutter environments (Weibull model). The election of one detector over the others is based on its average behaviour over the whole range of clutter impulsiveness characteristics expected for the type of environment. The main conclusion is that logarithmic and binary detectors outperform the others. An adaptive threshold mechanism is incorporated into the detectors. Its aim is to maintain their CFAR property against variations of clutter impulsiveness. This scheme is based on the determination of the CFAR detection threshold factor as a function of the PFA estimated at a lower threshold (pseudothreshold).<>