Cloud Feature Extraction and Classification of Meteorological Satellite Cloud Imagery

At present, meteorologists mainly adopt manual interpretation methods through satellite cloud imagery which are subjective and inefficient for weather analysis. Some researchers have proposed that computers can automatically classify cloud images by extracting the spectral features, texture features, spatial features, etc. However, the existing feature extraction and classification algorithms have problems of low precision, large computational complexity, and poor practicability. In this paper, the gray-scale co-occurrence matrix and the Gabor wavelet algorithms are proposed to extract the texture features of the three kinds of cloud images: cirrus cloud, cumuliform cloud and stratiform cloud, combined with the support vector machine classifier to classify cloud images. The results show that the approaches proposed in this paper can achieve more than 90% classification accuracy for the above three kinds of cloud images in a reliable manner.