Research On Underwater Targets Recognition Based On Principal Component Analysis And Neural Network

Nations around the world are now valuing the important strategic position very much.So how to recognize the target precisely and quickly become an important research direction.In the procedure of the recognizing,the feature extraction of the noise spectrogram is the key.However,after the feature extraction,the noise signal still have a problem of high dimension number.Meanwhile,in the research of the underwater targets recognizing,the neural network is applied in the classification and recognizing of the feature.So,finding a method dealing with the feature to make it easier to recognize is necessary.Principal components analysis is a kind of feature extraction technique point at data. Under the premise of persisting the main part of the original data,this method can effectively reduce the dimension of it After the dimension is reduced,the deal with the data is more fast and effective.To solve the problems above,this thesis present a method consist of PCA and BP neural network to recognize underwater targets.First,using PCA to extract the principal component.Then using the neural network to recognize the specific noise signal.Several researchs have been done in this thesis:(1)Study the noise theory of the underwater targets.On the base of the study,this thesis makes modeling and simulation of the signal;(2)Intruduce the method of principal components analysis into the recognizing of underwater targets.It helps reduce the dimension of the LOFAR spectrogram,and make it possible to do the identification.(3)Introducing a method based on PCA and BP neural network to do the identification.By means of simulation experiment, this thesis proves the advantage of this method.Results of the simulation experiment proves that this method helps the network to compute faster.This method also makes the recognizing more precise.