Research On Active Sonar Target Classification Based On Separability Discrimination

Sonar technology is of decisive importance in the military that maintains the country’s lasting stability and the people’s livelihood that guarantees the people’s happiness and wellbeing.In the military field,the improvement in the accuracy and speed of detecting targets has made military combat decision-making more real-time and stable;in the field of people’s livelihood,improved understanding of the marine environment has made it possible to adequately exploit sea resources.Active sonar target classification is the most important thing in the development of modern sonar technology,its principle is to emit hydroacoustic signals through sonar equipment and the signal is reflected back after encountering the target with the echo containing the target characteristic information,and then the echo signal is classified and processed.However,due to the serious reverberation of the marine development,and the weak and similar target,which seriously affect the classification performance of the active sonar equipment,the classification method that can solve the similar target in the reverberation environment is particularly important.Based on this,this paper studies the classification of similar underwater targets based on reverberation environment.The specific research works are as follows:Firstly,the application of sparse representation classification in active sonar target classification is studied.Based on the good anti-noise of the sparse representation and the excellent classification performance of the sparse representation classification algorithm,this paper applies the sparse representation classification algorithm to the active sonar target classification.Using the echo signals of underwater similar targets in reverberation environment,the experimental comparison of the sparse representation classification algorithm is carried out from two dimensions of robustness of classification and comparison with classical methods.Thus,the excellent performance of sparse representation-based classification in active sonar target classification is verified.Secondly,based on the good anti-reverberation performance of the Fractional Fourier Transform,by extracting the time-frequency domain features of the target signal through the Fractional Fourier Transform,a multi-order fractional Fourier domain feature fusion extraction method of active sonar target is proposed.The experiment uses the echo signals of underwater similar targets in reverberation environment,then compares the time domain signal combined with the sparse representation classification algorithm and the single-order Fractional Fourier domain feature combined with sparse representation classification algorithm to verify that the proposed active sonar target feature extraction method greatly improves the final classification effect indirectly.Finally,because the limited sparse coefficients of the sparse representation classification will increase the possibility of active sonar target classification,by combining the idea of dictionary learning,this paper proposes an active sonar target classification based on LC-KSVD algorithm and FDDL algorithm.These two methods limit the sparse coefficients in the sparse representation classification process,so that the finally learning dictionary is also recognizable.Combined with the multi-order fractional Fourier domain feature fusion extraction method of active sonar target,the algorithm is applied to the echo signals of underwater similar targets in reverberation environment,and is compared with classical classification,sparse representation-based classification and dictionary learning classification.Experimental results show that the two algorithms have excellent performance in similar target classification in reverberation environment.