基于CART决策树的星载GNSS-R海冰检测方法

针对星载GNSS-R海冰检测中阈值选取的问题,基于DDM观测数据提取4个敏感特征参数,构建特征参数组合,采用CART决策树方法,通过与实际观测数据训练,构建了GNSS-R海冰检测决策树模型,利用北极附近海域的TDS-1卫星观测数据进行海冰检测计算,通过与实际观测数据比对验证,结果表明检测成功率为98.86%,初步说明了该算法的有效性。该算法能够解决星载GNSS-R海冰检测中阈值确定问题,可为业务化应用奠定基础。

Sea Ice Detection Using Spaceborne GNSS-R Data by CART Decision Tree

To solve the problem of threshold selection in GNSS-R sea ice detection,four sensitive feature parameters are extracted based on DDM observation data,and the combination of characteristic parameters is constructed.The GNSS-R sea ice detection decision tree model is constructed by CART decision tree method,training by actual observation data.Sea ice detection is carried out by using TDS-1 satellite observation data in the sea area near the Arctic,according to the verification,the results show that the detection success rate is 98.86%,which preliminarily shows the effectiveness of the algorithm.The algorithm studied in this paper can solve the problem of threshold determination in GNSS-R sea ice detection,which can lay the foundation for operational application.