Study On Sea Clutter Model And Small Target Detection In Sea Clutter Background

Sea surface target detection is an important branch of radar target detection and the basis and premise of radar system design.It has important research significance in military and civil fields.Clutter is a key factor that restricts the performance of radar target detection and tracking.And the research on the characteristics of clutter is always a research hotspot in the field of radar.With the improvement of radar resolution,the traditional detection algorithm cannot meet the requirements of active equipment.Therefore,this paper focuses on the multi-dimensional characteristics of sea clutter and target as well as their application in sea surface target detection.The main research work of this paper is as follows:1.Parameters of the different clutter models affect the performance of target detection in high resolution heavy tailed sea clutter directly.In order to solve the problem that the detection performance of the sea clutter with Pareto distribution is reduced due to the limited application range of the existing estimators,a parameter estimation method based on logarithmic cumulants is proposed.The first-order and second-order logarithmic cumulants of Pareto distribution are calculated by Mellin transform and complete beta function property.The traditional moment estimation method and mixed moment estimation method are combined effectively.It can not only estimate the shape parameters in all defined domains,but also is not limited by the number of observation times.Monte Carlo simulation results show that the proposed method can reduce the relative bias of estimation.The difference is improved by at least one order of magnitude.At the same time,through logarithmic transformation,the traditional unit average CFAR Detector in the background of Gaussian distribution is extended to the background of Pareto distribution clutter.And the distributed target CFAR detection method under the background of Pareto distribution is obtained.The influence of the parameters of clutter on the CFAR detection method is verified by simulation experiments.2.In view of the poor performance of fractal theory when applied to sea surface target detection in short observation time,a sea surface target detection method based on non-extensive entropy of Doppler spectrum is proposed by extending entropy in Doppler domain.According to the difference of aggregation between sea clutter and target Doppler spectrum,the non-extensive entropy which can reflect the nonlinear characteristics of clutter is derived from Shannon entropy,which can only reflect the macroscopic value of echo.Secondly,the relationship between non-extensive entropy and fractal dimension is given.At the same time,the selection of non-extensive parameters to describe multi-fractal is analyzed.Compared with the Doppler spectrum of pure clutter unit echo,the Doppler spectrum of target unit echo has obvious aggregation and multi-fractal property,while the target does not have multi-fractal property.A method of small target detection on sea surface based on Doppler spectrum non-extensive entropy is proposed.Compared with the existing multi-fractal frequency domain Hurst index method and Shannon entropy method,the detection probability of the proposed method can be increased by about 10% in the case of short observation time.3.Aiming at the problem that the target detection method only extracts the polarization features from a single angle and cannot make full use of the polarization echo information,this paper proposes a sea surface target detection method based on polarization joint features from the perspective of polarization coherent decomposition and polarization incoherent decomposition,which combines polarization features with pattern recognition.This method extracts the polarization characteristics which represent the scattering difference between the target and the clutter from different angles:(1)since the scattering mechanism of sea clutter is random scattering with low grazing angle and the specific structure of the target reduces the randomness of the scattering in the echo,the polarization covariance matrix of the sea clutter and the target is decomposed by Cloude polarization to extract the mathematical expectation of polarization entropy and anti-entropy.Secondly,the scattering components of sea clutter and target echo are different.And the normalized coefficients of spherical scatterers,dihedral scatterers and spiral scatterers are much larger than those of clutter due to the artificial structure of targets.Therefore,the three scattering components are extracted by Krogager polarization feature decomposition of polarization scattering matrix of sea clutter and target.The normalized coefficients of scattering components reflect the structural difference of scattering mechanism between clutter and target echo.Based on the extracted polarization features,One-Class Support Vector Machine(OCSVM)is used to distinguish the target from the clutter.Compared with the existing polarization detection methods,the polarization information of the proposed method is more abundant and comprehensive,thus improving the detection performance.The experimental results of IPIX data under different sea conditions show the effectiveness of this method.4.In view of the limited applicability and flexibility of the fixed pattern classifier in feature detection,which cannot adapt to the complex and changeable sea clutter detection scene,a multi-dimensional feature detection method based on Localized One-Class Support Vector Machine(LOCSVM)is proposed.In the feature extraction stage,the features reflecting the characteristics of small targets and clutter are extracted from three aspects: 1)from the polarization scattering angle of the target,the relative power of the spherical and dihedral scattering components is extracted based on the polarization scattering matrix.2)From the Doppler and fractal dimensions,the average non-extensive entropy is extracted from the full polarization echo channel to reflect the Doppler and fractal characteristics of the unit to be measured.In the detection phase,by introducing k-means clustering into OCSVM,a new class of classifier LOCSVM is designed as detector.By integrating multiple classifiers,LOCSVM can simulate the complex nonlinear classification interface.In the complex detection background,when the target and clutter samples appear cross aliasing,it can realize the non-linearity of the target and clutter classification interface,so as to improve the detection performance.The experimental results of IPIX data show that the improved classifier can enhance the target detection performance by adjusting the number of clusters in different sea conditions.