This article mainly studies the problem of ground cloud classification and significantly improves the accuracy of ground cloud classification by applying an improved deep-learning method. The research results show that the method proposed in this article has a significant impact on the classification results of ground cloud images. These conclusions have important implications for providing new insights and future research directions in the field of ground cloud classification.