Block-based cloud classification with statistical features and distribution of local texture features

[摘要] This work performs cloud classificationon all-sky images. To deal with mixed cloud types in one image, we proposeperforming block division and block-based classification. In addition toclassical statistical texture features, the proposed method incorporateslocal binary pattern, which extracts local texture features in the featurevector. The combined feature can effectively preserve global information aswell as more discriminating local texture features of different cloud types.The experimental results have shown that applying the combined featureresults in higher classification accuracy compared to using classicalstatistical texture features. In our experiments, it is also validated thatusing block-based classification outperforms classification on the entireimages. Moreover, we report the classification accuracy using differentclassifiers including the k-nearest neighbor classifier, Bayesian classifier,and support vector machine.

[学科分类] 几何与拓扑