Research On Multi-band Full Polarimetric SAR Sea Ice Classification

As a serious kinds of meteorological disasters, sea ice troubles many countries and regions. It has caused countless casualties and property losses. How to accurate real-time monitoring of sea ice, and then on the basis of the collected data to provide strong support, is becoming the problem we need to solve. Satellite remote sensing technology has the advantages, like getting data in a short cycle,getting data from a wide boundary and so on. So it can meet our above requirement. So this research has adopted the polarization SAR image analysis methods to classify sea ice.This study adopted the classification algorithm based on the spectrum characteristic of sea ice, sea ice supervised classification algorithm based on texture feature and texture feature of different polarization modes to classify the sea ice.The general idea of classification algorithm based on spectral characteristics of sea ice is: Getting multiple wavelengths of gray image; Calculating the average pixel spectrum curve in the district of coexist as a reference of the class for knowledge base; Choosing the biggest difference square pixels in the district finally take the maximum coexist for the knowledge base. Repeat the steps above we can get the two knowledge base. So as to achieve the purpose of the classification of sea ice. But after the above algorithm of sea ice are classified, the result is not satisfactory. Then this study puts forward the sea ice supervised classification algorithm based on texture feature and texture feature of different polarization modes to classify the sea ice.The last method obtain a more satisfactory result.This study using multi-band polarization SAR image to classify sea ice is a preliminary research, future research can also add other features fusion, to improve the precision of the classification results.