一种改进AlexNet模型气溶胶颗粒分类方法

Improved AlexNet Model for Aerosol Particle Classification

大气环境污染问题频繁的发生,对社会的经济、城市的发展以及人们的生活都产生了极大的负面影响.如何有效的改善大气环境污染现状,提升大气环境质量,是需要亟待解决的问题.本文根据沈阳地区大气污染特征,将大气气溶胶颗粒划分为7种类别,分析每种类别气溶胶颗粒质谱图的特征.对深度学习分类模型AlexNet进行相应改进优化,利用现有的被命名的质谱图信息,完成气溶胶颗粒成分分类模型的训练与建立.在气溶胶颗粒监测分析的过程中,实现气溶胶颗粒自动分类代替人工分类的过程,将识别质谱图的工作自动化,极大的节省了人力资源.测试结果显示改进的AlexNet分类模型准确率达到95%,在气溶胶监测工作中,可以用于辅助完成气溶胶颗粒的污染特征分析以及来源解析.

The frequent occurrence of air pollution has a great negative impact on social economy,urban development and people′s life.How to effectively improve the status of atmospheric pollution and promote the quality of atmospheric environment is an urgent problem to be solved.According to the characteristics of air pollution in Shenyang,atmospheric aerosol particles are divided into seven categories,and the characteristics of mass spectra of each category are analyzed.AlexNet,the deep learning classification model,was improved and optimized.The existing mass spectrum information was used to complete the training and establishment of aerosol particle composition classification model.In the process of monitoring and analyzing aerosol particles,the automatic classification of aerosol particles instead of manual classification is realized,which can automatically identify mass spectra and greatly save human resources.The test results show that the accuracy of the improved AlexNet classification model reaches 95%,which can be used to assist in the analysis of aerosol pollution characteristics and source apportionment.