Multi-frame joint detection for passive radar based on multi-layer perceptron

For the problem of target detection of the low-altitude, slow-speed and small target (Lss-target) under complex clutter, a multi-layer perceptron (MLP) detector is developed in this paper. The detector combines the multi-frame range-Doppler (RD) spectrum as the observation space, and takes n (usually n=3) units along the range dimensional sliding window on each RD spectrum to construct the observation vector. To evaluate the performance of the detector, an unmanned aerial vehicle (UAV) detection experiment is carried out. LSS-targets such as UAVs usually appear near the zero-doppler region on the RD spectrum, clutter on which has a great impact on target detection. Experimental results show that the proposed MLP detector can approximate the likelihood ratio (LR) detector well in the Gaussian clutter environment. Meanwhile, it has better detection performance than the mismatched LR detector when the analytical solution for clutter distribution model is difficult to be obtained.