Severe Weather Intelligent Identification Technology Based on Artificial Intelligence Multisource Data

In view of the low accuracy of disaster weather recognition and high false alarm/missed alarm rates, this study proposes a multi-source data fusion technology based on artificial intelligence, constructs a multi-source data spatiotemporal alignment module based on gated recurrent units (GRU), and synchronously processes heterogeneous time series data. The improved ConvLSTM-Transformer hybrid model is used to extract multi-scale spatiotemporal features, and the dynamic feature fusion gated unit (DFGU) is designed to realize cross-modal collaborative modeling of radar and satellite data. Finally, the three-dimensional segmentation path of disaster weather targets is optimized by cascading 3D-CNN and conditional random field (CRF). The average recognition accuracy of this method reaches $\mathbf{9 2. 6 %}$, and the false alarm rate and false alarm rate are reduced to 7.77 % and 10.03 %, respectively. The results show that multi-source collaborative modeling improves the accuracy of disaster weather monitoring.