| 内容提要: |
With the successful launch of HY-2B, the new satellite will be dual-star networked with HY-2A, which provides a new data source for arctic research. The Level 1b products of HY-2A/B are used for analysis. Comparison and analysis of two different retracking datasets of SMLE and OCOG were carried out. Waveform analysis shows the feasibility of the SMLE retracking algorithm on the sea ice surface. Using AGC and PP of Ku/C band as the extraction parameters to extract the sea ice, three different classification methods (Threshold, K-Nearest Neighbor, Libsvm) were applied to As the extraction parameters of sea ice and sea water have a large number of identical values, the noise is easy to appear with the low accuracy (10%) of the threshold method. The KNN have highly accurate, all above 92%. The accuracy of KNN is related to the value of K. The Libsvm method got a good result (higher than 88% of sea ice, higher than 74% of the sea water). HY-2B acquires a better extraction result than HY-2A. As a change of the value of K, great variation of accuracy was occurred using the Ku band. However, the result of the C band changed slightly. Using the Libsvm method, the C band has a better distinguish capacity than the Ku band in the Arctic sea ice. |