Research On Resolution Enhancement Technology Of Multi-beam Forward-looking Sonar Image

High-resolution sonar images can be more clearly to show the targets information of underwater scene,which are important basis of target detection,location,identification and tracking,and play important roles in national defense and people's livelihood.In high-resolution sonar,range resolution can be improved easily by increasing the system bandwidth and azimuthal resolution can be improved by increasing array aperture size.However,increasing the aperture size will significantly increase the number of channels,power consumption,and size of the system.Based on the multi-beam forward-looking sonar developed by the laboratory,some resolution enhancement algorithms were conducted in this paper.The main contributions of this paper are listed as follows:(1)A time-varying focused multi-beam forming algorithm based on distortion evaluating is designed and implemented.According to requirement of high resolution imaging for complicated and changeable underwater scene,focused points of time structure are combined with the spatial structure of semicircular array to obtain accurate phase information of all focused points through near-field time varying focus model.Some necessary focused planes and speed level of sound are selected by a kind of distortion evaluation model,and near-field focusing is implemented by piecewise approximation method.The optimal beams are designed by using second-order cone programming,which balances various performance indicators of pattern.Finally,the algorithm is implemented on FPGA,and effectiveness and practicability of the algorithm is validated by imaging tests.(2)A monopulse forward-looking imaging algorithm is proposed because the resolution orientation of the method of beam scanning imaging is limited by the main lobe width of beam.The method adds monopulse angle measurement into imaging processing,which achieves precise localization of strong scattering point within the beam and makes certain characteristics targets more clearly.As a result,azimuth imaging resolution of the sonar system and image quality is improved.In addition,the algorithm does not involve complex data processing operations,which is conducive to engineering.(3)A super resolution algorithm in azimuth for scanning beams based on convolution inversion is proposed to solve the problem that sonar image is clearly blurred because of scanning beam convolution effect.Considering imaging data is actually the convolution result of target scattering information and beam pattern,cyclic convolution model in azimuth is established.Bayesian maximum likelihood criterion is selected to regularize the super resolution inverse problem,and approximation iterative solutions are chosen to recover the scene scattering distribution information and achieve scanning beam super resolution in azimuth.(4)A super-resolution reconstruction algorithm for sonar image based on sparse representation is proposed.Considering the importance of the sonar image contour feature,the improved edge protection interpolation method are used to get the initial estimates of high resolution,which keep edge and detail information unchanged well.On the assumption that high and low resolution image space have similar manifold,high and low resolution dictionary are trained by K-SVD algorithm,which corrects dictionary atoms update steps so that convergence rate of training and effectiveness of dictionary are greatly improved.In addition,principal component analysis algorithm is used to extract the main composition of the training image block,which dramatically reduces computation of dictionary training.The simulation results show that the proposed method can obtain good reconstruction effect,and has obvious advantages in visual effect and objective index.