Research On Key Technologies Of Passive Detection For Underwater Small-scale Platform

Detecting underwater targets based on sonars mounted on unmanned underwater vehicle?UUV?,which is a small-scale platform,is hidden,flexible,and highly efficient.However,in the practical working scenarios of passive sonar on small-scale UUV,some nonideal factors,including the limit of array aperture and system energy,the lack of prior information of targets,the mismatch of array model and the motion of platform,will cause negative effects on sonar signal processing.This paper focuses on the problems that small-scale UUV passive sonar may face in practice and conducts research on some key technologies of passive sonar detection for UUV based on the characteristics of radiated noise of targets.Influenced by the steady-state adaptive weight noise,the performance of the conventional adaptive line enhancer?ALE?will deteriorate.To solve this,this paper proposes the sparsity-driven ALEs.First,motivated by the sparsity of the tonals in the frequency domain,the frequency-domain sparsity of the optimum adaptive weights of ALE is derived and analyzed.Based on this,the time-domain adaptation of ALE is transferred into the frequency domain.Then,three sparsity penalties,L₁-norm,log-sum function,and L₀-norm are incorporated into the cost function of the frequency-domain adaptation,which yield three sparsity-driven ALEs,L₁-ALE,RL₁-ALE,and L₀-ALE.The sparsity-driven ALEs can reconstruct the frequency-domain sparsity of the steady-state adaptive weights of ALE and reduce the steady-state weight noise.Thus,a higher signal-to-noise ratio?SNR?gain can be obtained.By the proposed technique,the ability of UUV passive sonar to detect the radiated tonals can be improved.Finally,the effectivenesses of the sparsity-driven ALEs are verified by simulation and real data processing.The computational complexities of the sparsity-driven ALEs are relatively high,which contradicts the energy-saving requirements of UUV sonar systems.To solve this,this paper proposes two fast implementation techniques.First,motivated by the characteristic of the frequency-domain transformation of the sparsity-driven ALEs,this paper proposes the fast implementation technique based on the sliding discrete Fourier transform?SDFT?.This technique replaces the fast Fourier transform?FFT?with the efficient SDFT to perform the frequency-domain transformation.It can reduce the computational complexity effectively on the premise that there is no performance deterioration.Furthermore,motivated by the characteristic that the sparsity-driven ALEs perform the adaptation in the frequency domain,the fast implementation technique based on the frequency-domain fast block processing is proposed.This technique introduces the idea of parallel fast block processing in the schematic of sparsity-driven ALEs.It can reduce the computational complexity significantly on the premise that the performance deterioration is very slight.Finally,the effectivenesses of the two fast implementation techniques are verified by simulation and real data processing.When UUV passive sonar detects the signal of continuous spectrum,the performance of adaptive beamformer will be affected by the lack of prior information of targets and the mismatch of array model.To solve this,this paper proposes a robust adaptive beamformer based on the two-step steering vector estimation.Through our study,it is found that if the actual steering vector is projected obliquely into a constructed subspace,it can be presented by the sum of two independent components.Then,based on the geometrical characteristics of the two components,two different optimization problems are designed to estimate them respectively.In this way,the actual steering vector can be estimated precisely.Using the estimated actual steering vector,the weights of proposed robust adaptive beamformer is thus obtained.The proposed method is very robust against the large direction-of-arrival?DOA?mismatch,array element amplitude-phase error and array shape distortion.The output signal-to-interference-plus-noise ratio?SINR?can be well enhanced.The effectiveness of the proposed method is verified by simulation and real data processing.In UUV motion case,the relative directions between the targets and the receving array are time-varying rapidly.Thus,a large number of feasible array data snapshots is not available.As a result,the performance of the subspace-based DOA estimation methods will deteriorate.To solve this,this paper proposes a DOA estimation method based on the spatial rotation and focusing.First,by decomposing the UUV motion into the translation and heading variation,the array signal in UUV motion case is modelled and the effect on the covariance matrix estimation is analyzed.Based on this,the spatial rotation matrix is designed with the real-time heading angle provided by the navigation system.This matrix can compensate the relative variations of the DOAs of targets in the snapshots,which is caused by the UUV heading variations.In this way,the DOAs of targets are focused in the spatial domain.A large number of successive snapshots can then be used to estimate the covariance matrix precisely.By applying the covariance matrix in the subspace-based methods,the DOAs of weak far-field targets with high resolutions can be estimated in UUV motion case.The simulation results verify that the effectiveness of the proposed method.The accuracy and resolution of this method is also compared and analyzed.