Simple algorithms for target detection in FSR using local statistics

In this paper we consider two possible algorithms for the detection of marine targets using a Forward Scatter Radar (FSR) system. Contrary to traditional detection algorithms that process the input signal envelope, these algorithms perform the input signal phase. Before thresholding, the input signal phase is filtered and as a result the output signals are formed as values of local statistics of the signal phase calculated in the sliding window. In the case of large targets or fast moving targets, when the signal-to-noise ratio (SNR) is large enough, input signal phase samples are replaced by a local estimate of the signal phase variance calculated in the sliding window. In the case of small targets or slow moving targets when the SNR is relatively low, the input signal phase is filtered twice before thresholding. The first filter (Filter-1) removes the local average calculated in the sliding window from each sample of the input signal phase. The second filter (Filter-2) replaces the samples of the filtered signal by a local estimate of the variance calculated in the sliding window.