Robust monoparametric multiradar CFAR detection against non-Gaussian spiky clutter

The problem of decentralised constant false alarm rate (CFAR) detection in a distributed multiradar system is considered. First, the way in which an incorrect assumption of the clutter model affects the false alarm rate (FAR) of state-of-the-art decentralised multiradar systems is investigated, revealing current flaws of distributed detection techniques developed under the Gaussian assumption. Results show that the FAR of current distributed architectures degrades intolerably when non-Gaussian spiky clutter is present. On the other hand, state-of-the-art biparametric CFAR algorithms for non-Gaussian clutter are computationally demanding and lossy, so the question that arises is: are they really necessary? It is shown that the answer is ‘no’, or rather, ‘not always’. A new distributed detection strategy with simple monoparametric multipulse CFAR detectors is proposed, that, in spite of the single parameter estimation, allows one to obtain a false alarm probability at the fusion centre robust against changes of the degree of clutter spikiness. The robustness is obtained by the joint action of multisensorial integration and local temporal integration, in conjunction with a specific choice of the local monoparametric estimator. After a robustness analysis of the algorithm the detection performance of the multiradar network is determined and a comparison with a decentralised system employing biparametric estimation algorithms presented.