Composited Classification Research Based On Multi-resolution SAR Images Of Sea Ice

With the increasing of marine production and activities,the influence of sea ice on people is also increasing.Something serious even damage to people's property or life safety.With the advantage of all-day and all-weather characteristic,Synthetic Aperture Radar(SAR)is also rapidly developed and widely used in recent years.So the research on the classification of sea ice by SAR image is increasing.However,there are some unavoidable problems to classify sea ice with single resolution SAR image.The pixel interpretation ability of high resolution image is strong,but its coverage range is small,while the coverage of low resolution data is very wide,but it can't show the target's detail features.Therefore,it is of great significance to combine the advantages of both in order to overcome their disadvantages.In this paper,the sea ice composite classification based on multi-resolution SAR images is studied.The main work and research results are as follows:Firstly,a texture likelihood feature,which can represent the correspondence between high and low resolution images,is designed.The research direction of this paper is to guide the sea ice classification on low-resolution image with the classification results on high-resolution image.Therefore,it is very important to establish the relationship model between high and low resolution images.The research direction of this paper is to guide the sea ice classification on low-resolution image with the classification results on high-resolution image.Therefore,it is very important to establish the relationship model between high and low resolution images.Secondly,an algorithm combining texture likelihood feature with Naive Bayesian classifier is proposed.In this paper,three methods are respectively used to realize sea ice composited classification.First,the eigenvalue of mean is used as input of Naive Bayesian classifier to classify sea ice.Second,the texture likelihood feature is used to classify sea ice.Third,the texture likelihood feature is used as input of Naive Bayesian classifier to classify sea ice.The experiment shows that the last method has the best classification effect.Finally,a voting classification approach is designed according to weight.In this paper,multi-groups of ROI are used to carry out the sea ice composite classification with Naive Bayesian classifier based on texture likelihood feature.The Kappa coefficients of all classification results are normalized,and the result obtained is regarded as the weight of each group.Then the final classification results are obtained by voting.The experimental results show that the results obtained by voting classification approach are more accurate than those obtained from single group of ROI.