Classification of UAV-to-Ground Targets Based on Micro-Doppler Fractal Features Using IEEMD and GA-BP Neural Network

Abstract

In modern wars, the unmanned aerial vehicles (UAVs) have become the main means of local high-precision strike. The ground tracked vehicle and the ground wheeled vehicle are widely used to assist soldiers in ground operations because the vehicles can carry a variety of weapons while still move quickly. Therefore, in order to realize the strike from the UAV, it is of great significance to classify ground targets correctly. In this paper, a classification method based on micro-Doppler signatures and the improved Ensemble Empirical Mode Decomposition (IEEMD) is proposed. At first, models are built to describe the micro-Doppler characteristics of ground targets. Measured data of corresponding targets is analyzed. Secondly, principle of IEEMD is given and comparison of IEEMD and Ensemble Empirical Mode Decomposition (EEMD) is also made to prove the superiority of IEEMD in accuracy and calculation. Thirdly, three micro-Doppler fractal features are extracted from different Intrinsic Mode Functions (IMFs) obtained by IEEMD. Combined with Genetic algorithm-back propagation (GA-BP) neural network, accurate classification of ground targets is realized. Last but not least, classification experiments in different cases are carried out to indicate the effectiveness of proposed method. Comparison with current algorithms under various signal-to-noise ratios (SNRs) demonstrates that method in this paper has higher accuracy and better anti-noise performance.