Research On Active Sonar Target Classification Based On The Union Of Multi-order FrFT Domain Features

The development of sonar technology has attracted much attention due to the particularity of its environment and its significance in military and civilian fields.The classification of active sonar target is an important part of sonar technology.The basic principle is to use the acoustic signal emitted by the active sonar transmitting terminal,and return the target echo signal after the interaction between the transmitting signal and the target.The target echo signal received by the receiving terminal is processed and analyzed to classify the target.Due to underwater reverberation,serious noise interference,weak target echo signal and high similarity of target,the target classification performance of active sonar is seriously affected.In the face of this problem,aiming at the problem of serious interference and high similarity of target echo,this paper uses four types of similar target echo signals with reverberation and noise interference in real environment extracted from the actual environment of Moganshan underwater acoustic test site,and adopts the idea of multi-order feature union for fractional Fourier domain features to carry out active sonar target classification research.The specific research contents are as follows :Firstly,the theory and simulation of multi-order fractional Fourier domain features are studied.According to the concept and characteristics of fractional Fourier transform(FrFT),the theoretical model of multi-order FrFT domain characteristics of active sonar target echo signal is derived.The echo signal of the active sonar target under different signal-to-reverberation ratios is obtained by using the highlight echo model and the reverberation signal based on the unit scattering model,and the multi-order FrFT domain feature extraction of the signal is carried out.The energy aggregation and reverberation suppression performance of the optimal order FrFT domain feature are verified,and the target resolution ability of different order features is also verified.Secondly,for the classification problem under the condition of multi-order feature union and low signal-to-noise ratio,a Fisher discriminant dictionary learning(FDDL)classification method based on multi-order feature union is proposed.The order is randomly selected and the weight is set by setting the threshold of FrFT domain amplitude feature to realize the union of multi-order features.Based on sparse representation classification,the sparse coefficients are limited by Fisher discriminant method,and a Fisher discriminant dictionary(FDDL)classification method based on multi-order feature union is proposed.The performance of the method is verified by lake measured data.The experimental results show that the Fisher discriminant dictionary learning(FDDL)classification method based on multi-order feature association contains more feature information than the optimal order domain feature from the point of view of feature extraction,which is more beneficial to classification.From the perspective of classifiers,Fisher discriminant dictionary learning(FDDL)classifier uses Fisher discriminant to limit the sparse coefficient,so the classification effect of FDDL classifier is better than other traditional classifiers.Thirdly,in order to make full use of the characteristics of each order of the active sonar echo signal,an active sonar target classification method for multi-order FrFT domain distribution images is proposed.Through the multi-order FrFT domain feature theoretical model,the FrFT domain feature distribution images of each order are obtained.Combining image processing method with image entropy theory,an active sonar target classification method for multi-order FrFT domain distribution images is proposed.The performance of the method is verified by lake measured data.The experimental results show that the image features of multi-order FrFT domain distribution have certain advantages for underwater similar target classification under the condition of high signal-mixing ratio.Then,in order to obtain more informative and more separable order features,a multi-order FrFT domain feature union method based on information entropy weighting is proposed.Based on the theoretical model of information entropy and multi-order FrFT domain features,the information entropy of multi-order FrFT domain features is calculated by using the range averaging method.Through the order selection and weight setting method based on information entropy,the order of information content higher than the optimal order is obtained,and the union of multi-order features is realized.The performance of the method is verified by lake measured data.The experimental results show that the multi-order FrFT domain joint feature based on information entropy weights has a great advantage in the classification of similar underwater targets under different signal-mixing ratios,and the classification accuracy is stable and the classification performance is excellent among different classifiers.Finally,the summary and outlook are made.