Detection of Magnetic Dipole Target Signals by Using Convolution Neural Network

An intelligent method for detection of magnetic dipole target signals (MDTS) by using the convolution neural network (CNN) is proposed. First, the contaminated MDTS are intercepted and converted into 2D images through the Wigner-Ville time-frequency transforms. Then, a set of images are selected for training of the CNN until a satisfying level. Last, all images are tested to identify which images contain MDTS. Both simulation and experiment results show that the proposed method possess very high detection possibility, which, under the same signal-noise rate (SNR), outperform the conventional signal energy-based constant false alarm rate (CRAR) algorithm.