Optimized intelligent algorithm for classifying cloud particles recorded by a Cloud Particle Imager

• DOI: 10.1175/jtech-d-21-0004.1
• Publication Date: 2021-07-16T15:53:53Z
• Authors: Zepei Wu; Shuo Liu; Delong Zhao; Ling Yang; Zixin Xu; Zhipeng Yang; Dantong Liu; Tao Liu; Yan Ding; Wei Zhou; Hui He; Mengyu Huang; Ruijie Li; Deping Ding
• Keywords: Data Processing

Abstract Cloud particles have different shapes in the atmosphere. Research on cloud particle plays an important role analyzing growth of ice crystals and microphysics. To achieve accurate efficient classification algorithm crystal images, this study uses image-based morphological processing principal component analysis, to extract features images apply intelligent algorithms for Particle Imager (CPI). Currently, there are mainly two types ice-crystal methods: one is mode parameterization scheme, other artificial intelligence model. Combined with data feature extraction, dataset was tested ten classifiers, highest average accuracy 99.07%. The fastest speed real-time test 2,000 images/s. In actual application, should consider speed, because order millions. Therefore, a support vector machine (SVM) classifier used study. SVM-based optimization can classify into nine classes 95%, blurred frame 100%, This method has relatively high faster than classic neural network new could be also applied physical parameter analysis