Deep Learning Identification of Unmanned Aerial Vehicles with Hardware Implementation

In this paper, a deep learning identification of Unmanned Aerial Vehicles (UAVs) targets based on a two-stage loss function is proposed and implemented on Xilinx ZCU104 board. Through modeling and simulation, we obtain the scattering fields of UAVs, then it is converted into time-frequency image data set, and the custom two-stage loss function is used in the training process to obtain the optimal model. Finally., the model is transferred to the hardware platform to realize the classification process of the target by convolution neural network(CNN).