Research Article
Research on Typhoon Identification of FY-4A Satellite Based on CNN-LSTM Model
@INPROCEEDINGS{10.1007/978-3-031-31733-0_11, author={Wenqing Feng and Xinyu Pi and Lifu He and Jing Luo and Ouyang Yi and Qiming Cao and Zihang Li and Zhao Zhen}, title={Research on Typhoon Identification of FY-4A Satellite Based on CNN-LSTM Model}, proceedings={Smart Grid and Innovative Frontiers in Telecommunications. 7th EAI International Conference, SmartGIFT 2022, Changsha, China, December 10-12, 2022, Proceedings}, proceedings_a={SMARTGIFT}, year={2023}, month={5}, keywords={Typhoon Cloud System Identification CNN-LSTM Hybrid Model Feature Extraction FY-4A Satellite}, doi={10.1007/978-3-031-31733-0_11} }- Wenqing Feng
Xinyu Pi
Lifu He
Jing Luo
Ouyang Yi
Qiming Cao
Zihang Li
Zhao Zhen
Year: 2023
Research on Typhoon Identification of FY-4A Satellite Based on CNN-LSTM Model
SMARTGIFT
Springer
DOI: 10.1007/978-3-031-31733-0_11
Abstract
Typhoons are one of the most serious natural disasters, which are extremely destructive and pose a great threat to the safe operation of power grids. To improve the risk warning and pre-control capability of power grid operation under typhoon weather, this paper proposes a typhoon cloud system identification method based on a two-dimensional convolutional neural network (CNN) and long short-term memory (LSTM) network. First, the spectral features are selected according to the physical characteristics of clouds, combined with the square field of point clouds as the spatial information of point clouds to construct a sample library of typhoon cloud blocks; then, the spatial features are automatically extracted by the convolutional neural network; Finally, the LSTM network extracts the spatial local difference features and the time series features of continuous changes of a typhoon cloud system to provide multi-angle features for satellite cloud map to identify typhoon cloud system. Combined with the multi-channel scanning imaging radiometer AGRI (Advanced Geostationary Radiation Imager) data in the geostationary Fengyun-4 meteorological satellite (FY-4A), the monitoring and research of typhoon weather in Guangdong Province, China, is taken as an example. The experimental results show that compared with the Faster-RCNN method of abstracting typhoon features for identification, the CNN-LSTM model-based typhoon identification method can achieve a more detailed division between typhoon and non-typhoon regions based on the multidimensional features of the cloud system in typhoon regions, and achieve better identification results.
