Target recognition method for harmonic radar based on CNN

Nowadays, the target recognition of harmonic radar mainly adopts specific frequency points as the target characteristics, and the difficulty lies in the weak adaptability to the external actual environment, and false alarms and false negatives often occur in high-intensity clutter environments, making it difficult to effectively detect the target. In this paper, the harmonic radar echo signal dataset is established by near-field radar equipment, the harmonic radar echo data is processed by CNN, and the neural network parameters are iteratively trained by the Adam optimization algorithm to achieve effective recognition of PN junction and metal junction targets. Experiments show that the accuracy of the harmonic radar target recognition method based on CNN is higher than that of the traditional Constant False Alarm Detection (CFAR) technology and the LSTM harmonic radar target recognition method, and it also reduces the occurrence of false alarms.