Study On Thinned Semi-circular Array Based On Multi-beam Imaging Sonar

With the continuous advancement of the national marine strategy,scientific and technological investment in marine resource exploration and national maritime rights and interests has been increasing.Multi-beam imaging sonar,as an important device for underwater detection,has been widely used.However,in order to obtain high-resolution images,multi-beam imaging sonars are often equipped with large number of transducers,resulting in high system hardware costs,complex subsequent processing circuits,and large system power consumption.Thinned array techniques have been extensively studied to solve these problems.At present,researches of thinned array on single beam case are meature.The sparseness of the array in the case of multi-beam is limited by more constraints,which is a high-dimensional synergistic consistency problem and needs further study.In this thesis,we focus on the sparseness of semicircular arrays in the case of multiple beams.The main research contents and achievements of this thesis are as follows:(1)The thinned array design of multi-beam semi-circular array based on niche invasive weed optimization(NIWO)and convex hybrid algorithm is studied.The hybrid algorithm uses the positions of transducers as the optimization variable of the invasive weed algorithm,and performs population reproduction,spatial diffusion,niche learning and survival of the fittest to pick up the thinned array elements.Because the thinned array will cause the rise of peak side lobe level(PSLL)of the beam,then,the convex optimization theory is used.A lower PSLL is chosen as target function of convex optimization,and multiple sets of weighting coefficients for multi-beams are solved simultaneously.The results of numerical simulation and measured data show that the proposed algorithm can effectively obtain thinned array layout and multiple sets of weighting coefficients that meet the imaging requirements.The algorithm results in high array sparseness rate and narrow beams,but the algorithm takes a lot of time.(2)The thinned array design of multi-beam semi-circular array based on multiple focal undetermined system slover(M-FOCUSS)and convex optimization hybrid algorithm is studied.Given the reference beams and full array structure,the problem of seeking the minimum number of effective array elements and multiple sets of weighting coefficients is a highly nonlinear optimization problem,which can be transformed into the reconstruction problem of sparse signals in the theory of compressed sensing.IOn the analogy of multi-measure vector problem in compressed sensing,the thinned array model is built,which is solved by M-FOCUSS to determine the sparse array structure.In order to form multiple beams with low side lobe,a convex optimization model of excitation amplitude and phase is established.The numerical simulation and the measured data verify that the algorithm can obtain a thinned array structure and multiple sets of weighting coefficients that meet the requirements of imaging.Compared with the invasive weed intelligent algorithm,the proposed algorithm greatly reduces the computational time,reduces the PSLL of obtained beams,but the algorithm needs to specify the reference beams and the optimized sparse rate of the thinning elements decreases.(3)The thinned array design of multi-beam semi-circular array based on multiple linear search and fast iterative shrinkage threshold(FISTA)hybrid algorithm is studied.Firstly,the similarity between thinned array synthesis problem and least square regression problem is analyzed,and a thinned array model of single beam is constructed.Then multi-task learning model is introduced to extend the thinned array problem of single beam to multi-beam case.In this model,mixed L1/L2 norm of weighting matrix is added in objective function to ensure a block sparse array structure.This multi-objective optimization problem is solved by linear search and fast iterative shrinkage threshold method.The algorithm solves both the array structure and the multiple sets of weighting coefficients of each array element simultaneously to avoid the mismatch.The results of numerical simulation and measured data show that the algorithm can obtain a compromise between sparse rate,beam sidelobe peak level and computational consumption,but the algorithm relies on the sampled reference beams.In summary,the three algorithms in this thesis have their own advantages and disadvantages,and thus have different application scenarios.The multi-beam thinned array design based on NIWO and convex optimization results in good optimization performance and high array sparse rate,but its calculation buren is heavy.The multi-beam thinned array design based on M-fOCUSS and convex optimization can obtaine low and narrow beams,but the array sparse rate is slightly lower.The multi-beam thinned array design based on linear search and FISAT method,the optimized array has high sparseness rate and its side lobe level is relatively low,but the optimization leads to a serious mainlobe widening phenomenon.The latter two methods can lead to a faster converge speed than the first algorithm,but both require reference beams.