A cloud classification method based on information fusion of visible and infrared images

Correct observation for cloud has an indicative significance for rainfall forecasting and cloud classification technology is one of the problems in the field of automatic meteorological observation.In this paper,the cloud images collected with the whole sky imager instrument and the infrared imaging system were integrated for classification and measurement of cloud.Results show that on the basis of a great deal of cloud images collected from Beijing,Hangzhou and Lijiang weather stations,considering the features of cloud images and their indications for precipitation,the cloud shapes are divided into five categories including Clear,C_H,C_L,C_B and C_M.In addition,based on the 14 eigenvalues on colors and textures selected as the parameters for calculating cloud form and 552 cloud images used as the training samples,integrating a minimum distance algorithm with eigenvalue weighted method,the observed 500 samples are distinguished and divided into the above-mentioned 5 categories according to cloud form.Taking the standard cloud form classification as reference,the average accurate rate reaches 82%.This method combining the characteristics of visible light image with rich color and the advantage of infrared image in reducing the fog and haze interference shows a higher accuracy relative to the cloud measurement method with the single visible light sensor.This research makes a meaningful attempt in the field of information fusion of visible and infrared cloud image sensors.