Clouds, as one of the major meteorological phenomena, play a profound role in climate predictions and services. Most existing methods only utilize the visual sensors for ground-based cloud classification, which neglects other important characteristics of clouds. Research published in EURASIP Journal on Wireless Communications and Networking utilizes the multimodal information collected from weather station networks for ground-based cloud classification and proposes a novel method named deep multimodal fusion, or DMF. In order to learn the visual features, researchers trained a convolutional neural network model to obtain the sum convolutional map by using a pooling operation across all the feature maps in deep layers. Afterward, they employed a weighted strategy to integrate the visual features with multimodal features. Researchers validated the effectiveness of the proposed DMF on the multimodal ground-based cloud dataset, and the experimental results demonstrated the proposed DMF achieves better results than the state-of-the-art methods.
Source: https://jwcn-eurasipjournals.springeropen.com/articles/10.1186/s13638-018-1062-0 (By SpringerOpen)