A Hypergraph-Embedded Convolutional Neural Network for Ice Crystal Particle Habit Classification

In the field of weather modification, it is important to accurately identify ice crystal particles in clouds. When habits are correctly identified, cloud structure can be further understood and seeding other methods modification used change microstructure cloud. Consequently, phenomena changed at an appropriate time support human production quality life. However, morphology varied. Traditional particle classification based on expert experience, which subjective unreliable for identification categories by threshold setting. addition, existing deep learning faced with problem improving performance datasets unbalanced sample distributions. Therefore, we designed a Convolutional Neural Network (CNN) embedded hypergraph convolution module, named Hy-INet. The module effectively capture information from hypergraphs constructed local global feature spaces learn features small samples that have numbers. Experimental results demonstrate proposed method achieve superior task habits.