Methods Of Ship HRRP Estimation And Masked Small Target Detection In High-resolution Radars

Generally,maritime radars are designed to have high range resolution or dual-mode operation,where low resolution is for detection and high resolution for discrimination.Due to directly proportional to the area of spatial resolution cell,low power level of sea clutter in high-resolution maritime radars makes it possible to find small and low-observable targets.High-resolution maritime radars often transmit pulses of large time-bandwidth product such as wide linear/nonlinear frequency modulated pulses and achieve high range resolution by the pulse compression technique.However,while obtaining high range resolution,it also brings unavoidable side-effects: strong non-Gaussianity of sea clutter and range sidelobe effect.The strong non-Gaussianity of high-resolution sea clutter partly counteracts the benefit from lower power level of sea clutter in detection.The range sidelobe effect means that range sidelobes of pulse compression are mistaken as real targets and mask small targets located within the range sidelobe interval.The range sidelobe effect is an inherent property of the pulse compression radars and will inevitably occur.The difference lies in the extent of the range sidelobe effect that depends on the signal-to-clutter ratio(SCR)and the range sidelobe level of the pulse compression.The non-Gaussianity of sea clutter and the range sidelobe effect make it extremely complicated to detect sea-surface small targets masked by the range sidelobes of large objects.Besides target detection,high-resolution maritime radars also undertake the tasks of ship discrimination and classification.Ship length is one of salient features for ship discrimination and classification and it can be estimated from the estimate of ship radial size and its heading at tracking mode.However,an overestimation of ship radial size often occurs due to the existence of the range sidelobe effect.Therefore,in this dissertation,the estimation methods of ship radial sizes and high-resolution range profiles(HRRPs)and detection method of masked small targets in high-resolution maritime radars are investigated to improve the performance of target detection and classification under complex environments of non-Gaussian sea clutter and range sidelobe effect.The main work and contributions are as follows,1.For the problem of the constant false alarm rate(CFAR)of non-coherent detectors under sea clutter of the generalized Pareto distribution(GPD),the cell-average(CA)and orderstatistic(OS)non-coherent detectors are analyzed,the false alarm probability formulas of the two non-coherent detectors are derived,and the two detectors are CFAR with respect to the power level of sea clutter.However,it is found that the two non-coherent detectors do not have CFAR with respect to the non-Gaussianity and speckle covariance matrix of sea clutter.In order to ensure the CFAR property in the overall scene,the block-whitening before noncoherent integration using estimated speckle covariance matrix is proposed.The noncoherent detectors with the block-whitening are CFAR with the power of sea clutter and the speckle covariance matrix.In this way,the detectors can obtain the CFAR property in the overall scene when the decision threshold matching the shape parameter of sea clutter is used,which is verified by the experimental results.2.In view of the overestimation of ship radial size caused by the range sidelobe effect in high-resolution maritime surveillance radar,a new method for ship radial size estimation is proposed.The method consists of detection and localization of ships,decision of severe range sidelobe effect,recovery of sparse high-resolution range profiles(HRRPs)of ships,and radial size computation.When the severe range sidelobe effect occurs,the length of the range interval covered by the detected strong returns of a ship is much larger than its radial size.In this case,a ship radial size overestimate occurs as traditional methods are used to estimate ship radial sizes by finding the borders of strong returns.An approach to solve the problem of severe overestimation of ship radial sizes is to find the borders of ship HRRPs.Therefore,the ship radial size estimation boils down to the sparse recovery of ship HRRPs in non-Gaussian sea clutter.A sparse recovery method using linear programming(LP)is proposed to recover sparse HRRPs of ships in non-Gaussian sea clutter.And then the ship radial size is estimated by finding the borders of the recovered HRRP.Experimental results using simulated and measured data show that the proposed method attains more accurate radial size estimates of ships than the traditional methods.3.For the estimation of ship HRRPs,a sparse recovery via iterative minimization(SRIM)method is proposed to estimate ship HRRPs in non-Gaussian sea clutter.It is always an important problem to recover sparse signals from noisy observations and has been extensively investigated.However,existing sparse recovery methods focus on signals rather than background interference,which is often assumed to be Gaussian noise.When these methods are used to estimate ship HRRPs in non-Gaussian sea clutter,there will inevitably suffer from a performance loss due to the mismatch of the background model.The SRIM method adapts the non-Gaussianity of the interference by the compound Gaussian model(CGM)with inverse Gamma texture and characterizes ship HRRPs by the bi-parametric generalized Gaussian distributions(GGD)(0<p(?)1,p is the sparsity parameter,which determines the sparsity of ship HRRPs)model.In the SRIM method,the optimal sparsity parameter of the GGD model is searched by the minimal criterion of the KolmogorovSmirnov distance(KSD)of the residue and the interference model.The SRIM method is verified by using the simulated and measured radar data and the experimental results show that it obtains better performance of ship HRRP estimation compared to the existing methods.4.In high-resolution maritime radars,sea-surface small targets masked by range sidelobes of large objects such as ships are difficult to be detected.For this problem,an effective method is proposed to detect masked small targets.It consists of the first cycle of target detection,localization of large objects,high-precision reconstruction of radar returns of large objects,removal of radar returns of large objects,and detection of masked small targets.Besides small targets of interest,there exist many uninterested large objects such as big ships,reefs,and offshore platforms for oil production on the sea surface.Large objects have too strong radar returns and their range sidelobes often mask a wide range interval and small targets within the interval are difficult to be detected.For detection of masked small targets,high-precision reconstruction of radar returns of large objects is crucial because tiny reconstruction errors are enough to mask weak returns of small targets.An algorithm using the LP and oversampled HRRP model is presented for high-precision reconstruction of radar returns of large objects.Experimental results show that the proposed masked small detection method is effective unless a small target falls into the range interval occupied by a large objects and has a comparable radial velocity with the large object.