Improved radar detection using evolutionary optimised filter

Matched filter concepts are extensively used in radar applications to maximise target detection. Usually, the receiver filter is matched to the transmitted signal, which is optimum for a point target return. For extended targets it is possible to do better by matching with the returned target signal. However, in most cases, the returned target signal is not known a priori. The authors propose a method to approximate the filter matched to the signal returned from an unknown extended target. A genetic algorithm is used to optimise filter output signal-to-noise ratio on a pulse-to-pulse basis. The gain, relative to the result from a filter matched to the transmitted signal, is dependent on the actual backscatter profile of the target. Tests on three different target profiles, derived from real high-resolution radar data, verify the performance of the method. The signal-to-noise ratio threshold for a given detection rate is found to be lowered by 2-3-dB on these target signatures, where each profile consists of 20 backscatterers. Furthermore, the steady increase in available computational resources would most likely make real-time applications of the novel evolutionary radar filter possible in the near future.