Radar HRRP Target Recognition Based on Hybrid Quantum Neural Networks

This article proposes a hybrid quantum model called convolutional neural network-long short-term memory-quantum circuit (CNN-LSTM-QC) for binary classification tasks of high-resolution range profile (HRRP) targets. The CNN captures spatial information from the HRRP sequence data, while the LSTM grasps and processes long-term dependence relationships between the range cells within an HRRP sample. In addition, the model incorporates a QC for feature extraction and classification. By leveraging quantum entanglement, it captures correlated information of the data, enabling efficient feature extraction even in scenarios with low signal-to-noise ratio (SNR) and incomplete data. Moreover, to tackle the multiclassification tasks, the QC-CNN-recurrent neural network (QC-CNN-RNN) model is proposed, employing CNN-LSTM-QC as a feature extractor, with its output serving as input to the CNN-bidirectional-RNN (CNN-Bi-RNN) model. Experimental results show that the proposed model surpasses others in binary classification tasks, achieving an average recognition rate of 89% at $ 0\text{-dB}$ SNR and 80.14% when the training set is limited to the $0^\circ$ to $90^\circ$ azimuth range. In multiclassification tasks, the QC-CNN-RNN model improves classification accuracy by 5.09% at $ 0\text{-dB}$ SNR compared to the CNN-Bi-RNN model.