Research On Radar Ship Target Detection At Different Grazing Angles

Radar ship target detection plays an important role in maritime traffic management,fishery management,maritime security and anti-smuggling.Sea state,radar parameters and observation geometry are the main factors affecting the detection of targets,among which grazing angle is one of the important factors.The radar works at high altitude platforms(eg.satellite SAR)or low altitude platforms(eg.airborne or shore-based radars),hence the former works at medium-high grazing angle,while the latter works at low grazing angle limited by the radar height.In addition,radar observes sea surface in down-looking mode,and sea clutter is strong and rapid change,resulting in target detection more difficult.Therefore,it is of vital importance to analyze the sea clutter characteristics to radar target detection.According to extensive literature review,the existing problems of target detection are mainly as follows:(1)The existing sea clutter models were all proposed under specific conditions.The applicability of sea clutter models under different grazing angle,azimuth angle and other parameters needs to be evaluated,so as to select suitable models for target detection.In addition,the existing goodness-of-fit evaluation methods can only get one overall error value,which leads to inaccurate evaluation results,or it’s computationally complex and takes a long time to get the results.(2)For radar detection at medium-high grazing angles,it lacks comprehensive evaluations and interpretations of polarization features.Furthermore,the polarization characteristics are also affected by sea state and other conditions,which makes the detection of small targets or high sea state targets more difficult.(3)For radar detection at low grazing angles,the problems of false alarm and missing detection exist.On the one hand,the increase of false alarms(eg.sea spikes)interfere with target detection;on the other hand,slow targets are easy to be submerged in sea clutter owing to the Doppler broadening.The full utilization of polarization and interframe correlation information is an effective way to improve the detection accuracy.Therefore,the characteristics of radar echo at different grazing angle was analyzed first,then the information of multi-polarization and interframe correlation was utilized to present algorithm for target detection,especially for slow/high sea state targets.The main research contents and results are summarized as follows:Firstly,the applicability of sea clutter models at different grazing angles was studied,and a method was proposed to describe the amplitude characteristics of sea clutter.As we know,the interaction mechanism between electromagnetic wave and sea surface is different for low and medium-high grazing angles.For medium-high grazing angle,the sea surface echo is dominated by Bragg scattering,while for low grazing angles,the shielding and sea spike effects are enhanced,hence the characteristics of sea clutter at different grazing angles may be different.The applicability of sea clutter models at different grazing angles was studied.(1)At low and medium-high grazing angle,six sea clutter models were evaluated by five methods(eg.MSE and K-L divergence).The suitable models at different grazing angles were obtained,which can provide reference for the selection of sea clutter models in ship detection.(2)Comparisons about the applicability of sea clutter models at different grazing angles was analyzed,which was the first time to compare the sea clutter characteristic of different radar systems(Airborne wide area surveillance radar and satellite SAR).(3)A new method was proposed to characterize the amplitude of sea clutter,which is composed of maximum value,statistical peakiness,leading-edge width and trailing-edge width.The proposed method can be used not only to describe the characteristic difference of sea clutter amplitude under different grazing conditions,but also to evaluate the goodness-of-fit of sea clutter model.Especially when the overall goodness of fit error of the two models is similar and the optimal distribution model cannot be obtained,the proposed method has significant advantages,which was simple and efficient.Compared with MSE,K-L divergence,Bhattacharyya distance and threshold error methods,the proposed method can evaluate the goodness of fit of the sea clutter model at the peak and the trailing point simultaneously;compared with the graphical goodness of fit,the computational complexity of the proposed method was low,and repeated Monteraro experiment was not required.Secondly,for ship detection at medium-high grazing angles,a new polarization feature and a new polarization multi-feature fusion method were proposed:(1)Euclidian distance and mutual information analysis were used to comprehensively evaluate the detection capability of full-polarization,compact-polarization and dual-polarization features,and features with best detection capability were obained.(2)In order to propose a polarization feature that is independent of sea surface roughness,combined with X-Bragg scattering model,the sea surface roughness parameter was introduced and a polarization scattering model was derived to interpret the features with good detection capacity.Then a new feature-polarization phase factor was proposed,which has a strong sea-surface contrast and was independent of sea surface roughness.Theoretical and experimental results show that the detection method based on the polarization phase factor was robust and can be applied to different sea conditions,and its detection performance was better than the circular polarization ratio,roundness,relative phase and K-CFAR detectors.(3)A polarization multi-feature fusion method was proposed for ship target detection.Polarization features with good capability were used to construct multi-dimensional polarization feature vectors for the design of target detection algorithm.To reduce the redundancy between features and reduce the computational complexity,a feature dimension reduction method was introduced to extract new polarization features,and then multi-feature dimension reduction GP-PNF was designed using the new features.The proposed method makes full use of the full-polarization,compact-polarization and dual-polarization information.Simulation experiments and measured data verified the effectiveness of the polarization multi-feature fusion detection method,which was better than K-CFAR and GP-PNF methods.Finally,a low rank and sparse representation based ship target detection method was proposed at low grazing angle.The differences between sea clutter and ship target in range-Doppler sequences were analyzed,which shows the low rank of sea clutter and sparsity of ship target respectively.Hence the sea clutter was modeled as background with low rank characteristics,and the radar target was modeled as foreground with sparse characteristics(represented by Markov random field).The separation of ship targets and sea clutter was realized by the optimization solution,and the sea clutter was suppressed to the maximum extent under the premise of keeping the target power unchanged.Considering the motion of the sea surface,some sea clutter may be mistaken for foreground targets.For this problem,a sea clutter discriminator was constructed by defining the background and foreground frequency to further remove false alarm targets(sea clutter)in the foreground,so as to realize the detection of ship targets.Simulation experiments and measured data verified the effectiveness of the proposed method,which was better than GΓD-CFAR and PCP methods.In addition,combining with the inertial navigation information of radar,the tracks of the ship targets were obtained.