电子学报 ›› 2021, Vol. 49 ›› Issue (3): 619-624.DOI: 10.12263/DZXB.20191217

• 学术论文 • 上一篇    下一篇

一种基于离散属性BNT的双偏振气象雷达降水粒子分类方法

李海1, 尚金雷1, 孙婷逸1, 冯青1, 庄子波2   

  1. 1. 中国民航大学天津市智能信号与图像处理重点实验室, 天津 300300;
    2. 中国民航大学飞行技术学院, 天津 300300
  • 收稿日期:2019-10-28 修回日期:2020-05-30 出版日期:2021-03-25
    • 通讯作者:
    • 李海
    • 作者简介:
    • 尚金雷 男,1993年生,山东威海人,硕士研究生,主要研究方向为机器学习、机载气象雷达信号处理.E-mail:18660039930@163.com;孙婷逸 女,1995年出生,河北张家口人,主要研究方向为机载气象雷达信号处理.E-mail:812901035@qq.com;冯青 女,1980年生,陕西蒲城人,讲师,硕士,主要研究方向为气象雷达信号处理、阵列信号处理.E-mail:fengqing_tj@163.com;庄子波 男,1980年生,山东潍坊人,副教授,硕士,主要研究方向为航空气象.E-mail:zbzhuang@cauc.edu.cn
    • 基金资助:
    • 民机项目 (No.MJ-2018-S-28); 航空基金项目 (No.20182067008); 国家自然科学基金项目 (No.U1433202); 中央高校基本科研业务费项目 (No.3122018D008); 中国民航大学蓝天教学名师培养经费资助课题; 天津市自然基金重点项目 (No.20JCZDJC00490)

A BNT Hydrometeor Classification Algorithm for Dual-Polarization Radar

LI Hai1, SHANG Jin-lei1, SUN Ting-yi1, FENG Qing1, ZHUANG Zi-bo2   

  1. 1. Tianjin Key Lab for Advanced Signal Processing, Civil Aviation University of China, Tianjin 300300, China;
    2. Flight Technology College, Civil Aviation University of China, Tianjin 300300, China
  • Received:2019-10-28 Revised:2020-05-30 Online:2021-03-25 Published:2021-03-25
    • Corresponding author:
    • LI Hai
    • Supported by:
    • Civil Aircraft Project (No.MJ-2018-S-28);  Aeronautical Science Foundation of China (No.20182067008); National Natural Science Foundation of China (No.U1433202); Program of Fundamental Research Funds for the Central Universities (No.3122018D008); Supported by Teaching Master Training Foundation of Civil Aviation University of China; Key Project of Tianjin Nature Foundation (No.20JCZDJC00490)

摘要: 针对传统降水粒子分类算法存在的过度依赖专家经验和模型预设误差问题,本文提出了一种基于离散属性贝叶斯网络(Bayesian NeTwork,BNT)的双偏振气象雷达降水粒子分类(Hydrometeor Classification,HC)方法.首先对双偏振气象雷达获取的偏振参量取值进行离散化处理生成离散化标准,并根据离散化标准制作训练数据集合;然后使用训练数据集合对贝叶斯网络进行结构学习学得贝叶斯网络结构,以及参数学习学得与贝叶斯网络结构匹配的条件概率表;最后加入附加信息计算出每种降水粒子类先验概率,与贝叶斯网络结构和条件概率表共同组成贝叶斯网络分类器.训练好的贝叶斯网络分类器根据最大后验概率准则完成对测试数据的降水粒子分类,与模糊逻辑算法对比评价结果.实验证明:该方法能有效区分不同的降水粒子得到准确的降水粒子分类结果.

 

关键词: 双偏振雷达, 降水粒子分类, 贝叶斯网络, 贝叶斯分类, 不确定知识, 离散化, BNT

Abstract: The over-reliance on expert experience and model preset errors in traditional precipitation particle classification algorithms are discussed. This paper proposes a dual-polarization hydrometeor classification (HC) method based on discrete attribute Bayesian NeTwork (BNT). Firstly, the value of polarization parameters obtained by the dual-polarization meteorological radar is discretized to generate a discretization standard, and the training data set is made according to the discretization standard. Then the training data set is used to learn the structure of the Bayesian network and the conditional probability table matching the structure of the Bayesian network. At last, additional information is added to calculate the prior probability of each precipitation particle class, and the Bayesian network classifier is composed of Bayesian network structure and conditional probability table. The trained Bayesian network classifier classifies the precipitation particles according to the maximum posterior probability criterion and compares the evaluation results with the fuzzy logic algorithm. Experiments show that this method can effectively distinguish different precipitation particles.

 

Key words: dual polarization radar, hydrometeor classification, Bayesian networks, Bayesian classification, uncertain knowledge, discretization, BNT(Bayesian NeTwork )

中图分类号: