Study Of Ship Detection And Discrimination Technology In Synthetic Aperture Radar Imagery

As the development of the image technology in SAR (Synthetic Aperture Radar), itis possible to achieve the ship detection and surveillance, when ship target detection isbrought forward.Ship detection and surveillance is an important task for the oceanapplication in SAR imagery.It has important significance both in civilian and militarydomain to detect the ship target from the SAR images.The paper has fully analysed thefeature in SAR images under the background of engineering application, which basedon the theory of image understanding. As well it has summarized the actuality of shipdetection and ship wake, and discussed the critical workflow and technology in theantumatic system of ship detection full-scaled. Thus the article emphasized on theaccurate statistics model, the CFAR ship detection and target discrimination.The mainwork in the paper is summed up as follows:1.The paper analysed the statistic characteristic of sea clutter in SAR imagesfirstly.Based on the randomicity and non-stability of sea clutter in SAR images, thepaper has usedG0distribution in sea clutter statistics modeling.Then evaluate theapplicability ofG0distribution to sea clutter in theory and the results of modelingexperience. It illustratesG0distribution models both the homogeneous andheterogeneous sea clutter well.2.For the ship detection in medium and high resolution SAR imagery, a fastmethod for ship detection from SAR imagery based on pixel excision inG0distributedsea clutter is proposed in this paper. The method excises clutter pixels with a thresholdthat determined by the appearance frequency first, then reduces resolution by thesampling procedure and carries through automatic ship detection finally. Detectionresults on data illustrate the method can reduce the number of false alarm and improvethe calculation efficiency remarkably.3.In the target discriminnation, the paper has choosen the frame of feature selectionand extracted the feature based on the analysis of the investigation and the ship targetfeature.Two adaptive genetic algotithms has been contrasted in the feature selection, theexperiment results illustrate LAGA is better than AGA in suppressing “premature” andcan prevent falling into a local optimum and enhance the race of the convergence.Thepaper has choosen the support vecor machine as the classifier and given thediscrimination method by the results of feature selection. Discriminnation results ofclassify and ship target discriminnation on real data illustrate the method is available.