基于修正小波变换插值-TAN的雷达降水粒子分类

对于双偏振天气雷达在获取数据分辨率较低情况下的降水粒子分类问题, 提出一种双偏振体制天气雷达下基于修正小波变换插值-树扩展朴素贝叶斯(tree augmented naive Bayesian, TAN)的降水粒子分类方法。首先，需要在原雷达偏振参量数据分辨率比较低的情况下进行修正小波变换插值处理，以获得较高分辨率的雷达偏振参量数据。然后，基于互信息理论利用离散化的高分辨率偏振参量数据进行TAN网络结构和参数训练，得到可用于降水粒子分类的TAN网络。最后, 将高分辨率的雷达偏振参量数据带入到最终获得的TAN网络中以实现降水粒子分类。对实测数据的处理结果表明, 对于低分辨率雷达偏振参量数据, 所提方法可以取得较好的降水粒子分类结果。

Radar hydrometeor classification based on modified wavelet transform interpolation-TAN

For the hydrometeor classification problem of dual-polarization weather radar with low data resolution, a hydrometeor classification method based on modified wavelet transform interpolation tree augmented naive Bayesian (TAN) is proposed. Firstly, it is necessary to carry out modified wavelet transform interpolation when the resolution of the original radar polarization parameter data is relatively low, so as to obtain high resolution radar polarization parameter data. Then, based on the mutual information theory, the discrete high resolution polarization parameter data are used to train the TAN network structure and parameters, and the TAN network which can be used for hydrometeor classification is obtained. Finally, the high resolution radar polarization parameter data are brought into the final TAN network to realize the hydrometeor classification. The processing results of measured data show that the proposed method can achieve better classification results of hydrometeor for low resolution radar polarization parameter data.