In order to reduce the loss of meteorological disasters,a dynamic early warning method of meteorological disasters based on dynamic regions of interest and big data is proposed.This paper builds a dynamic early warning framework for mete-orological disasters in the client/server mode.The system collects meteorological-related information and disaster subject infor-mation,and processes them by the data processing layer.The fuzzy C-means clustering algorithm is improved to extract the dy-namic area of interest of the satellite cloud image,determine the meteorological disaster scene and the possible affected area.Three improved deep shrinkage self-encoding networks are used to extract the meteorological factors,environmental factors,and disaster-affected subject factors in the area of interest respectively.feature,and then construct a Softmax classifier for scene recognition,determine the degree of correlation between the scene and various disaster-causing factors,and combine the support vector machine to determine the level and type of meteorological disasters to achieve dynamic early warning of meteoro-logical disasters.The experimental results show that the clustering effect of this method is outstanding,and it can identify the types and grades of meteorological disasters.
关键词
动态感兴趣区域/大数据/气象灾害/动态预警/致灾因素
Key words
dynamic region of interest/big data/meteorological disaster/dynamic early warning/disaster causing factor
维普
万方数据