The Research On Underwater Acoustic Ship Target Recognition Method Based On Deep Learning

Ship target recognition is one of the key contents of underwater acoustics research,which is of great significance for ocean exploration and modern naval warfare.Using the radiated noise of ships is an important means of ship target recognition.Because the mechanism of ship radiated noise and marine environment are complex,a recognition method with stable performance and high accuracy is needed.In recent years,artificial intelligence(AI),represented by machine learning(ML)and deep learning(DL),has developed rapidly,which provides new ideas and development direction for ship target recognition technology.According to the generation principle of ship radiated noise,this paper extracts the characteristics of the simulated ship radiated noise signal.Support vector machine(SVM),knear neighbor(KNN),back propagation algorithm(BP),radial basis function(RBF)and extreme learning machine(ELM)classifiers are studied and analyzed.It paves the way for the application of machine learning method combined with target feature extraction scheme in underwater target recognition.Traditional machine learning underwater acoustic target recognition methods need a lot of professional knowledge when extracting features.Therefore,this paper uses convolutional neural network(CNN),which can automatically extract features and accurately classify,to recognize underwater acoustic targets.In the study of convolutional neural network,its structure and working mechanism are analyzed.A deep convolution neural network available for one-dimensional sequence signal input is designed to extract the features of the target signal,and compared with the manual feature extraction,which verifies the effectiveness of the feature extraction of the original ship noise signal through deep learning.Finally,this paper adopts a deep learning underwater acoustic target recognition model based on Dense Net to recognize the measured ship radiated noise signal,and uses different super parameters to analyze the recognition performance of the model.Different from the traditional recognition model,this model does not use the time-frequency feature image,but directly uses the original audio signal in the time domain as the network input data.The experimental results show that the deep learning underwater acoustic target recognition model based on Dense Net achieves an overall accuracy of 94.39% at 10 d B signal-to-noise ratio,and is better than four traditional machine learning technologies and three other existing deep convolution neural network models.