Research Of Target Highlight Feature Extraction,recognition And Classification In Reverberation Background

The target highlight can effectively reflect the physical information such as the shape and material of the target,and has stable attributes.Therefore,extracting the characteristics of underwater target highlight is still a powerful tool for active sonar detection and recognition.However,when there is reverberation interference in the target echo,especially in the case of low signal to mixing ratio,the target bright spot feature is often seriously "polluted",which significantly reduces the target detection and recognition performance of active sonar.In view of the above problems,the research work of this paper is as follows:(1)Aiming at the problem of target highlight feature extraction in reverberation background,this paper uses the difference of energy distribution between reverberation and target echo in time and frequency domain,and adopts effective filtering technology to suppress reverberation interference in time and frequency domain,so as to better extract target highlight feature.There are four anti reverberation methods in this paper,which are based on two different principles.In the first principle,the energy distribution of reverberation in time-frequency plane is low rank,while the signal is sparse,so the two are separated to achieve the purpose of anti reverberation.According to this principle,the methods include robust principal component analysis,random projection and dynamic mode decomposition;the other principle uses the energy intensity difference and energy distribution of signal and reverberation in time-frequency domain First,the lower energy part of each time point in the time-frequency plane is filtered,and then the distribution of LFM signal on the time-frequency plane is a straight line,while the reverberation distribution is chaotic.The time-frequency plane is rotated at an appropriate angle,and then the projection filtering operation is carried out to further filter the reverberation interference,so as to achieve a more ideal anti reverberation effect,Based on this principle,the time-frequency energy filtering method is proposed.After the anti reverberation processing in time-frequency domain,Hough transform is used to detect the straight line of the filtered time-frequency plane,and then the bright spot features of the target echo are extracted.Through the simulation and experimental analysis,it is shown that the proposed anti reverberation method can effectively suppress the reverberation interference,improve the effect of highlight feature extraction,and is more conducive to the subsequent target classification and recognition.(2)The echo signal of underwater target is more complex,and it is difficult to obtain data samples,which leads to the problem that the number of existing underwater target samples is too small.In this paper,the artificial bee colony algorithm optimized support vector machine is used to classify and recognize underwater targets.This algorithm can not only get better recognition results in the case of small number of samples,but also solve the problem of traditional support vector machine It can’t adjust parameters adaptively for different application conditions.The simulation and experimental data show that,compared with the recognition rate without anti reverberation processing,the recognition rate after processing by the four anti reverberation methods in this paper has been effectively improved,which verifies the effectiveness of the four anti reverberation methods in this paper,thus providing a new method for target bright point feature extraction and recognition classification in reverberation background.