Radar Target Detection via Deep Learning

This work proposes a Deep Learning based technique for radar target detection to replace standard radar signal processing techniques. The proposed technique comprises a Convolutional Neural Network (CNN) with the range-Doppler ambiguity function serving as its input. A radar simulator is developed to generate range-Doppler data for a single target with an isotropic antenna. The method is compared with the classical Cell-Averaging Constant False Alarm Rate (CA-CFAR) detector, and it is demonstrated that the proposed method outperforms CFAR approximately by a factor of 2 terms of probability of detection (Pd) and by a factor of 100 in terms of probability of false alarm (PFA), at a minor cost of computational complexity.