Interpretation And Classification Algorithm For Multi-polarization SAR Image Of Sea Ice In Liaodong Bay

Sea ice is the major disaster of Liaodong Bay in winter; the implementation of sea ice monitoring using remote sensing method has significant importance to the safety of marine transportation and marine projects. Multi-polarization SAR has the ability of working at all-weather situation and all-day, it also can provide rich information about the type of sea ice. It is an ideal data source for sea ice business monitoring system. However, by now, the major means of monitoring the sea ice in internal are ground based radar and optical remote sensing, but research for characteristics of sea ice of Liaodong Bay on multi-polarization SAR imagery is still lack. In order to grasp the characteristics of sea ice and ice-type characteristics in Liaodong Bay, this paper carried out an investigation in Liaodong Bay, accessed many kinds of information of sea ice in this area, include GPS location, salinity, ice temperature, surface roughness and thickness and so on. At the same time, the synchronous multi-polarization SAR and high-resolution optical remote sensing image of this area were accessed. Based on the measured data and remote sensing images, this paper interpreted the sea ice on the multi-polarization SAR image, enriched the priori knowledge of the characteristics of sea ice on the multi-polarization SAR images.To adapt the requirement of automatic identification and classification of sea ice in multi-polarization SAR imagery, this paper analyzed and compared the texture features of different types of sea ice on the multi-polarization SAR image. Based on the result of quantitative comparison, some texture features are screened out. At the same time, this paper described a method of selecting parameters of Gray Level Co-occurrence Matrix used when texture analyzing.This paper designed a sea ice multi-polarization SAR imagery classification method based on Agglomerative Hierarchical Clustering algorithm, using texture features screened out. Experiment result shows that, our method can get the result of close to visual interpretation. This paper is just a preliminary research of sea ice classification using multi-polarization SAR imagery; much more depth research will be done using more remote sensing images and experiment data.