Ocean clutter suppression using one-class SVM

This paper presents a new algorithm based on pattern classification technique to suppress ocean clutter of high frequency ground wave radar (HFGWR). We first decompose radar returns as weighted sum of several chirp-like signals and then extract chirp rates and initial frequencies of these chirps to construct the feature space. The support region of ocean clutter in this feature space is calculated by one-class support vector machine (OCSVM). New data that falls inside this support region are labelled as ocean clutter and are discarded. We validate our method by conducting experiments on real radar data. The result shows that the selected features can discriminate between ocean clutter and target echo effectively and our classifier can suppress ocean clutter successfully by defining a fine support region of ocean clutter in that feature space