Underwater target identification method based on multi-deep learning model joint judgment system

Abstract

The invention provides an underwater target identification method based on a multi-deep learning model joint judgment system, which comprises the following steps of: firstly, generating a one-dimensional sound signal spectrum and a two-dimensional time-frequency spectrogram as deep learning processing objects from a frequency domain and a time-frequency domain according to the characteristics of underwater target radiation noise data, and secondly, constructing a stacked noise reduction auto-encoder and a one-dimensional convolutional neural network model for processing according to the one-dimensional sound signal spectrum, outputting various target recognition confidence coefficients, constructing a two-dimensional convolutional neural network model for processing according to the two-dimensional sound signal time-frequency spectrogram, and outputting confidence coefficients; performing weighted fusion judgment on a multi-model confidence result, and optimizing a weighting coefficient of an output result of each model based on a genetic algorithm; and finally, achieving unknown target noise data identification based on the model and the criterion. According to the method, deep mining is carried out on multi-dimensional numerical features based on deep learning, advantage complementation of different-dimensional separable numerical features is realized, and improvement of target recognition robustness is facilitated.

Description

technical field [0001] The invention belongs to the technical field of underwater target identification and artificial intelligence, and mainly relates to an underwater target identification method based on a joint decision system of multiple deep learning models. Background technique [0002] The target radiation noise identification in water mainly uses the target radiation noise received by the sonar and other sensor information to identify the target type, and provide target feature information to provide a basis for the sonar operator to make comprehensive decisions. With the large-scale application of precision-guided weapons, the dependence on target recognition ability has become more and more prominent. At the same time, the development of sonar detection technology has greatly increased the number of targets found, which has also brought more challenges to target recognition. [0003] Traditional sonar target recognition mainly realizes classification by extractin...

AI Technical Summary

Problems solved by technology

For underwater acoustic target recognition, many scholars have also carried out research on deep learning application methods, but generally the models used are relatively single

Images