Chaotic CFAR detectors for detection of land and sea point targets

Abstract

This paper presents a new formulation for chaotic detectors which is based on a combination of statistical detectors and chaotic predictors. The chaotic predictors are user to estimate the clutter (i.e., modulation component), while the statistical detectors are used at the output of the squared residual error of the chaotic predictor. Here, the background clutter is this error component and any man-made point target that may be present. The residual error consists of the residual modulation component, the speckle and additive thermal noise. The proposed detector has been used for detecting man-made point targets using a wide range of radar data. Detection of small man-made targets in radar or infrared clutter is an important area of interest for many applications such as ocean surveillance, search and rescue, remote sensing, mine detection, etc. It has been shown that infrared and radar clutter exhibit chaotic rather than purely random behaviour. From the chaotic point of view, a neural network predictor has been developed using a generalized regression neural network (GRNN).