Unknown target recognition method for radar target recognition

Abstract

The invention proposes an unknown target recognition method for radar target recognition. The method comprises a recognition stage of an unknown target. According to the recognition stage of the unknown target, an MSTFT-WVD algorithm is adopted to obtain the micro-Doppler characteristic time-frequency graph of the unknown target; and the micro-Doppler characteristic time-frequency graph of the unknown target is inputted into a deep learning model which is trained in a second stage, so that the category information of the unknown target is obtained. With the method of the invention adopted, a higher recognition accuracy rate can be obtained.

Description

technical field [0001] The invention relates to the technical field of radar target recognition, in particular to an unknown target recognition method for radar target recognition. Background technique [0002] The different tasks performed by different targets on the battlefield determine the difference in their threat levels, so the identification of battlefield targets is of great significance. At present, the vehicle-mounted radar of a certain type of radar reconnaissance vehicle installed by our army's armored troops is a pulse Doppler radar. The identification of the target is usually determined by humans. After the target is found, the scout listens to the target through the headset. Echo the audio signal to identify the target. This method has high requirements on the professional quality of scouts, especially for scouts who have not undergone long-term special training, it is more prone to misjudgment or misjudgment, the correct recognition rate of the target is ex...

AI Technical Summary

Problems solved by technology

[0007] (1) When using the time-frequency analysis method for micro-Doppler feature extraction, the results obtained by the linear time-frequency analysis method usually have the problem of low time-frequency resolution, such as STFT, Gabor transform and other methods; quadratic time-frequency analysis The results obtained by the method usually have the problem of cross-item interference, such as Wigner-Ville Distribution (WVD), Pseudo Wigner-Ville Distribution (PWVD), etc.; Problems with too high complexity, such as Reassigned Smoothed Pseudo Wigner-VilleDistribution (RSPWVD), etc.

[0008] (2) In the target recognition method based on micro-Doppler features, many current classification algorithms are shallow structures, the application range is narrow, and the complex function cannot be well represented when the sample size is small. Complex recognition problems have limited processing capabilities, such as Fisher linear classifiers, Bayesian classifiers, SVM, etc., so there is room for improvement in improving the learning ability of classifiers

Therefore, it is difficult to complete the recognition of ground moving targets, especially the high-accuracy and real-time recognition

Images