Statistical Classifier of Radar Returns

Abstract

For the last years many algorithms for radar return identification based on wideband or imaging radar have been developed. However, because of its all-weather performance, microwave radar is still the most reliable sensor for surveillance.

This paper considers the problem of radar return identification using microwave radar. Radar return is classified into several different classes, namely ground, weather, birds and aircraft, using Stehwien-Haykin classifier based on features derived from maximum entropy spectral analysis. To evaluate the practicality and effectiveness of this classifier its performance is tested by Monte Carlo simulation. Clutter samples, are generated by passing a white noise sequence through linear digital filter. The power spectral density is reasonably represented by second or third order Batterworth. The results show that the classifier correctly identifies the radar return with mean error rate less than 4 %.

Partially supported by Center of Excellence BIS21 grant ICA1-2000-70016 and Bulgarian National Foundation for Scientific Investigations grant No I-902/99.