Underwater target material classification method based on multi-domain feature extraction

To solve the problem of the classification of underwater target materials,an underwater target material classification method based on multi-domain feature extraction was proposed. By combining four features of the target echo signal including the auto regressive(AR)coefficients in the time domain,the cepstrum domain feature,the spectral peak and the frequency in the time-frequency joint domain,the classification of four underwater target materials such as metal,rock,plastic and rubber was achieved. The measured data from the anechoic tank were used to verify the effectiveness of the method. The results show that for the simulated data of four materials,the classification accuracy of the proposed method is higher than 80%,and the classification performance using multi-domain features is obviously better than that using a single feature. For the measured data of plastic and metal,the classification accuracy of the proposed method is more than 80% when the signal-to-reverberation ratio is no less than 0 dB. This method is robust to the differences in geometric characteristics of the target such as the size or shape. © 2024 Southeast University. All rights reserved.