Maritime Radar Target Detection Using Time Frequency Analysis

Abstract

Small target detection in sea clutter remains a challenging problem for radar operators as the backscatter from the sea-surface is complex, involving both time and range varying Doppler spectra with strong breaking waves which can last for seconds and resemble targets. The goal of this thesis is to investigate two different time frequency wavelet transforms to filter the sea clutter and improve target detection performance. The first technique looks at an application of stationary wavelet transforms (SWT) to improve target detection. The SWT decomposes a signal into different components (or sub-bands) which contain different characteristics of the interference (clutter + noise) and target. A method of selecting the sub-band with the most information about the target is then presented using an ‘entropy’ based metric. To validate the SWT detection scheme, real radar data recorded from both an airborne and a ground based radar systems are analysed. A Monte-Carlo simulation using a cell averaging constant false alarm rate detector is implemented to demonstrate and quantify the improvement of the new scheme against unfiltered data. The second technique utilises a sparse signal separation method known as basis pursuit denoising (BPD). Two main factors contribute to the quality of the separation between the target and sea-clutter: choice of dictionary that promotes sparsity, and the regularisation (or penalty) parameter in the BPD formulation. In this implementation, a tuned Q-factor wavelet transform (TQWT) is used for the dictionary with parameters chosen to match the desired target velocity. An adaptive method is then developed to improve the separation of targets from sea-clutter based on a smoothed estimate of the sea clutter standard deviation across range. A new detection scheme is then developed and the detection improvement is demonstrated using a Monte-Carlo simulation.