Research On DOA Estimation Technology Of 3D Forward-Looking Sonar Based On Sparse Array

Under the promotion of the national strategy of developing and transforming the marine science technology to an innovation-led type,the research and development of the home-grown sonar is underway.Forward-Looking Sonar(FLS)can actively detect target of interest within the field of view,and has become key equipment to solve the target detection and positioning for small marine platforms.According to the current development trend,the actual demand of FLS,and the support of Natural Science Foundation of China and key scientific research project,this paper mainly discusses the development of three-dimensional(3D)FLS prototype from three aspects: theoretical analysis of sparse array design and typical Direction-of-Arrival(DOA)estimation technology,sparse DOA estimation technology for point source target and phase-difference DOA estimation technology for surface target.First,theoretical basis of sparse array design and DOA estimation technology is explored.Compared with the uniform linear array,the sparse arrays have several advantages such as larger array aperture,better resolution capacity,and higher estimation accuracy by reasonable placement of sensors through certain rules,which do not increase the additional sensor cost.DOA estimation technology of sparse array is carried out from the following three aspects.In the first aspect,the typical coprime array and nested array geometry are introduced,and corresponding evaluation criteria in DOA estimation are quantitatively analyzed according to the sparse array design theory.In the second aspect,we propose a Cubic Nested Array(CNA)and derive its array geometry.The estimation performance of the proposed CNA is compared with other sparse arrays through the analysis of Cramér-Rao bound(CRB),and the simulation results show that the proposed CNA has higher DOA estimation accuracy.In the third aspect,we analyze the drawbacks of existing difference co-array based DOA estimation methods in terms of estimation accuracy from the Monte-Carlo simulation.Second,DOA estimation technique of point source target based on sparse reconstruction technology is investigated.The research is related to two aspects.Firstly,for high-precision DOA estimation requirements in sparse arrays,the sparse reconstruction technique is applied to point target positioning.Nevertheless,it is often unrealistic for existing sparse methods to assume that the signal DOA lies exactly on a predefined grid because of the continuous angle space.In this paper,we propose a root sparse Bayesian learning(RSBL)algorithm for sparse array,in which the hyperparameters and grid point in the dictionary can be updated as the iteration process progresses,so that the grid point gradually approach the true echo direction.Furthermore,we propose a root Sparse Bayesian Learning with Support Selection(RSBL-SS)algorithm that adopts a Constant False Alarm Rate(CFAR)rule to enforce the sparsity.The simulation results show that proposed method has a remarkable performance,such as good estimation accuracy close to the CRB,and improved convergence performance.Secondly,considering the problem that high correlation of dictionary column is unfavorable to distinguish two objects,we propose a Structured-optimization Sensing Matrix Design with Adaptive Stepsize(SSMDAS)algorithm for high-resolution DOA estimation requirements in sparse arrays,which reduces correlation of dictionary column and promote convergence performance.Both the simulation and pool experiment results verify the superiority of the proposed method.Third,DOA estimation technology of surface target based on phase difference is studied.Considering the problem that conventional DOA estimation methods of point source fail on surface target since it neglects the geometric features of surface target and cannot satisfy the point source assumption.To address this problem,we propose a phase-difference DOA estimation method on surface targets which contains following four aspects.Firstly,to address the problem of spatial continuity of surface terrain echo strips,the beamforming method is adopted for water column imaging and spatial filtering,resulting that there is only one echo in each beam angle at the same time.Secondly,aiming at the interference of water scattering echoes and multi-path sound propagation,we propose a method combining the outlier labeling with re-detection method to detect terrain echo stably.Thirdly,the surface terrain echo has complex acoustic scattering superposition,resulting in many abnormal jump points which can be effectively filtered out by applying nonlinear filter processing.Fourth,due to the sparse array design of FLS,there is a problem of phase unwrapping ambiguity.Based on the ambiguity eliminate technology of caprice principle,we propose a phase-difference DOA estimation method without ambiguity problem.Finally,sound field simulations by the Field II tool and pool experiment are implemented to verify that the proposed method can achieve accurate reconstruction of seabed topography.At the end,the FLS prototype is developed on the basis of the above key technologies.At the beginning,the structure of the FLS system is determined according to its key technical indicators.Then we design the embedded software,propose a multi-channel matched filter with large pipeline and small pipeline mode,and implement the real-time parameter calculation module of SOPC system based on NIOS II.After implementing the system joint debugging,the effect of the prototype in performing the forward-looking collision avoidance and seabed topographic mapping tasks are verified by the tests of underwater target detection and the surface target detection carried out in the lake.The results verify that the developed FLS could be used for collision avoidance and surface topography tasks,and the measurement results are accurate and reliable.