Multiple Sensor Track Fusion Algorithm Based on LSTM Network

Aiming at the problem of low accuracy of multisource sensor fusion caused by the difficult acquisition of prior information and weak compatibility of fusion data when tracking missiles in air combat, a track fusion algorithm based on the LSTM method is proposed. Since the track data is a kind of time-series information, the LSTM network is used to learn the relevant prior knowledge of the front and rear sequences, combined with the good nonlinear processing ability of deep learning, the network can learn the track information that is closer to the actual battlefield environment and realize track fusion. Experiments show that the effect of this method has significantly less error than the traditional method when there is common process noise, and compared with other neural networks, this method also shows a better fusion effect, indicating that the algorithm can effectively avoid the estimation error caused by the public a priori estimation and is feasible.