Research Of Saliency Detection And Classification For SAR Images

SAR(Synthetic Aperture Radar)has high resolution,all-weather,all-time characteristics.It has been widely used in many field such as environment,resources and military.SAR image classification is an important application of SAR image processing.But due to the existence of speckle noise in SAR image,the processing of the SAR image is very difficult,optical image processing method is not suitable for SAR image.This paper is based on the research of saliency detection method,combine with super pixel method in-depth study on the classification of high resolution SAR images,the main contents include:1.Active learning SAR image classification based on the saliency.First,the classification method constructs saliency detection,the method is applied to SAR image saliency feature extraction,and image is divided into saliency region and non-saliency region.Then SAR image is classified through active learning support vector machine after active learning sample selection.This method greatly improve the efficiency of the support vector machine(SVM)and reduce the number of samples of the artificial mark.Finally,the effectiveness of the method is verified by experiment.2.SAR image classification method based on sparse representation and super pixel.The method firstly uses the SLIC super pixel segmentation method to divide SAR image into super pixels.Secondly,features of super pixels are extracted based on the co-occurrence matrix;then Features are dealt with sparse representation.Finally,SAR image is classified based on sparse representation method,and through the experiment,the classification of the methods in this paper have achieved good results.