Method of track data association based on multi-scale convolution and improved LSTM
To address the issue of low accuracy in intelligent trajectory association for radar data,this paper proposes a trajectory association method that integrates multi-scale convolution(CNN)and improved long short-term memory networks(LSTM).Firstly,multi-scale CNN is utilized to extract spatial features across multiple dimensions,avoiding the limited field of view problem caused by fixed-size convolution kernels and enabling the network to capture higher-dimensional features.Subsequently,the features are fed into the improved LSTM network to capture temporal features,and the enhanced unit structure takes into account the correlation of adjacent moments in trajectory data,effectively suppressing the impact of noise and errors.Finally,the results from simulation experiments demonstrate that this method effectively improves trajectory association accuracy.
convolutional neural networklong short-term memory neural networktrack association
万方数据
维普
