Non-parametric detector in non-homogeneous clutter environments with knowledge-aided permutation test

In this study, a new kind of knowledge-aided (KA) permutation test (PT) detector is proposed for the problem of performance deterioration of the PT in complex non-homogeneous clutter environments. The proposed detector consists of two parts, a data selector and a conventional PT. The data selector uses the geographic information system (GIS) to select the reference cells which have similar clutter characteristics with the cell under test. Samples from the reference cells selected by the data selector are used to determine the detection threshold of the PT. Because the PT is a non-parametric method, the proposed detector can maintain a constant false alarm rate without knowing the clutter's distribution type, which is very important in complex clutter environments. Considering the computation complexity of the PT, the authors also provide an implementation algorithm for the PT called ‘valid strategy method’, which is more efficient than the direct calculation method. In order to prove the excellent performance of the KA-PT detector, this detector is compared with some conventional detectors in terms of the false-alarm number and detection probability with the real data collected by IPIX radar. The results show that incorporating the non-parametric method with the prior GIS can significantly improve detectors’ performance in practice.