基于二维粒子谱仪的固态降水粒子自动分类研究-雪花和霰

Auto-classification of Solid Precipitation Particles based on A 2DVD into Snowflake and Graupel

固态降水粒子进行准确而细致的分类对许多大气过程及天气雷达的应用是十分重要的。使用二维光学粒子谱仪(2DVD)对单个降水粒子进行测量,并基于测得的粒子微物理参数及特性提供降水过程中--分钟单位时间间隔内主要降水粒子类型的估测,对固态降水粒子进行自动分类。为实现自动分类任务,考虑将该工作与常用的机器学习分类算法相结合,应用朴素贝叶斯,支撑向量机(SVM),决策树三种监督学习算法对单位时间间隔内的粒子分类。文中将降水粒子归类为雪花和筱两种主要类型,并结合人工检测进行结果验证,最终利用独立的数据集进一步验证,证明分类算法的准确性。

Civing an accurate and detailed classification of solid precipitation particles is of a paramount importance to most of the atmospheric processes and the application of weather radar.This paper aims to use two-dimensional optical disdrometer(hereinafter referred to as 2DVD)to measure the precipitation of a single particle,and based on its micro-physical parameters and characteristics of precipitation in the course of a minute of the main types of precipitation part-cles in the unit Interval estimation,this paper classifies the solid precipitation particles.In order to actualize the auto-matic classification,this paper also attempts to make this task and common classification of machine learning algorithms combined,and the three supervised learning algorithm,naive Bayesian algorithm,support vector machine(SVM),and.Decision Tree,applied to classify particles in the unit interval.In this paper,precipitation particles is classified mainly-as snowflake and graupel,and its result is tested with the help of Manual detection.Besides,the independent data has been searched to do further examination,proving the accuracy of classification algorithms.