The accurate classification of ground-based cloud is of great significance to many fields such as weather forecasting,aerospace and so on.In recent years,deep learning has achieved remarkable achievements in the classification of ground-based cloud.However,in addition to the visual features of ground-based cloud,the auxiliary features of ground-based cloud,namely the multi-modal information of ground-based cloud,also play an important role in the classification of ground-based cloud.In order to mine and integrate the multi-modal feature information of ground-based cloud,this study designs a ground-based cloud classification method based on dual-flow multi-modal multi-layer fusion network(DMMFN).Firstly,multi-modal information is separately transmitted into different sub-networks.Secondly,heterogeneous feature fusion is carried out in the feature layer.Finally,the model achieves a high accuracy rate of 85.70％on the multimodal ground-based cloud dataset.The experimental results show that the proposed DMMFN network model can effectively combine ground-based cloud multi-modal information with visual features,and improve the accuracy of ground-based cloud classification.
维普
万方数据