基于聚类评分的暴雨/冰雹分类模型

针对暴雨和冰雹2种强对流天气的区分问题,研究了主成分分析联合线性鉴别分析对雷达图像中提取的冰雹及暴雨特征降维和去相关的作用,设计了基于聚类评分的暴雨/冰雹分类模型,采用K均值聚类评分的办法确定具有高分类性的主成分,并以此主成分设计分类器对暴雨和冰雹进行区分.结果表明:主成分分析联合线性鉴别分析进行特征处理能够在降维的同时保留大部分分类性信息,基于聚类评分的分类模型有利于提高冰雹识别的命中率并降低其误警率,且对一般类型公共数据分类问题有效.

Heavy Rain/Hail Classification Model Based on Cluster Scoring

For the discrimination of two severe convective weathers, heavy rain and hail, we investigated the effectiveness of combination of principal component analysis （PCA） and linear discriminant analysis method on the dimension reduction and feature decorrelation, and a classification model based on K-means cluster scoring was devised. This classification model is able to determine the principal component with high classification performance and distinguish heavy rain and hail based on this principal component design classifier. The experimental results show that most of the classification information is remained by principal component analysis and linear discriminant analysis in the dimension reduction. Moreover, classification model based on cluster scoring can improve the probability of detection （POD） and reduce the false alarm rate （FAR） of hail identification. This model has similar effectiveness for general data classification problem as well.