Single MLP-CFAR for a radar Doppler processor based on the ML criterion. Validation on real data

Mitigation of the Effects of Unknown Sea Clutter Statistics by Using Radial Basis Function Network

“…Network and support vector machine were trained with generated target range profiles and generally, support vector machine showed higher recognition rates under low signal to clutter ratios. In [24], detection of target with unknown Doppler shift, embedded in sea clutter, was improved (in comparison to conventional Doppler-preprocessed CFAR), by application of MLP neural network, trained with generated clutter patterns. Application of MLP neural network reduced clutter power variation and thus maintained fixed threshold value applied to conventional non-coherent CFAR detector.…”

“…In contrast to methods described in preceding paragraph, in this paper we use online learning of the clutter statistics. Similar to [15], the proposed method is not restricted to clutter statistics and does not require offline training or prior clutter model as in [19][20][21] or clutter samples as in [18,[22][23][24]. In a more wider context of application of the proposed detection scheme in target tracking process, this is due to application of external association mechanism that separates set of samples into target echo and set of sea clutter echos, like in [12], where Viterbi based association scheme acts as target from clutter separator.…”