Online Dictionary Learning Techniques for Sea Clutter Suppression
Maritime airborne surveillance radars operating at high grazing angles experience significant sea clutter returns with a time and range-varying Doppler spectra, thus making the detection of small targets extremely difficult. Many target detection schemes assume models for the amplitude distribution that require estimation of specific parameters. These may not be accurate in all situations and hence reduce the detection performance. In this paper, we propose a new sea clutter suppression approach based on a dictionary learning strategy. The success of this data-driven approach is that it does not rely on statistical modeling of the clutter, but rather on the quality of the training data used to form the dictionary. Our analysis looks at the learning performance between batch and online strategies and then uses the obtained dictionaries to assess the clutter suppression capabilities in terms of improvement in the signal to interference ratio and the detection performance using a Monte Carlo analysis. To demonstrate these algorithms, data from the Australian Ingara L-band radar system has been used.