Research On Ship Detection In High Resolution SAR Images Of Complex Scene

Ship detection provides basic information for marine rescue,ship traffic management,fishery supervision and so on.As a non-contact,long-distance detection technology,remote sensing can provide a large range of ocean data,and then obtain the ship information.In particular,Synthetic Aperture Radar(SAR)system is not affected by cloud,fog,rain and snow,and can observe the sea in all weather and at all the time,which has become an important technical means for ship detection at sea.Therefore,it is of great significance to carry out the research of SAR image ship detection.Ship targets in SAR image are easily affected by marine environment,SAR system parameters and ship material structure,and show different characteristics in images.There may be a large number of false targets in the image scene,which seriously affects the ship detection performance.Weak ships are easily submerged in SAR image sea clutter because of the small Radar Cross Section.In addition,due to SAR special imaging mechanism,C-band SAR image usually has serious azimuth ambiguities.In order to improve the performance of ship detection in the above scenes,this paper makes the following exploration on ship detection in SAR image.Aiming at the scene with many false targets,a ship detection optimization method based on multi-feature weighting is proposed.The marker-based watershed algorithm is employed to remove land from SAR amplitude image.Log-normal distribution is selected to participate in the Constant False Alarm Rate(CFAR)detection to obtain candidate targets in the above result.The length to width ratio,the target area and the contrast ratio of the candidate targets are extracted.Then,a variance coefficient method is proposed to distribute the weight of the three features,to adapt to the proportion of each feature in the ship weight decision.The confidence levels are calculated by combining the normalized feature vectors of the candidate targets with the feature weight.By determining the best confidence level,false targets among the candidate targets are removed to optimize ship detection results.The two-parameter CFAR algorithm and the Kapur,Sahoo and Wong(KSW)two-threshold algorithm are used to compare with the proposed method.Results show that the proposed method can adapt to the complex scene with many false targets and improve the ship detection performance.Aiming at the scene with weak ships and azimuth ambiguities,a ship detection method based on the scattering gradient vector is proposed.Based on the scattering vector of Polarimetric Synthetic Aperture Radar(Pol SAR)image,the scattering gradient vector is used to improve the contrast between the ship and sea,and then the Shannon entropy is extracted.Through the difference of Shannon entropy between ship and sea,the best Shannon entropy threshold is selected to carry out ship detection,so as to improve the detection performance of weak ships.Finally,the azimuth ambiguities are removed by the direction angles,the displacement distances and the Shannon entropy between them and the corresponding true ships.The proposed method is compared with the HH channel method,the Improved Polarimetric Notch Filter(IPNF)method and the Generalized Optimization of the Polarimetric Contrast Enhancement(GOPCE)method.Results show that the proposed method can effectively detect weak ship,remove the azimuth ambiguities,and its detection performance is well.The paper has 23 diagrams,9 tables and 61 references.