摘要
基于降雨图像数据,依据降雨量划分不同的降雨强度;结合深度神经网络理论建立降雨强度识别模型,对降雨强度进行实时监测与预警.首先,通过福州市8个气象站点获取降雨图像及其对应的降雨量数据;其次,依据降雨强度对降雨图像进行分类,共分为6个等级,并将数据分为白天图像和晚上图像2个数据集;最后,采用DenseNet深度卷积神经网络构建降雨强度识别模型.结果表明:(1)各气象站点降雨强度的识别精度均高于80%,识别精度不存在明显差异;(2)白天降雨图像的识别精度高于晚上;(3)白天和晚上图像存在特征差异,使其识别精度在不同网络层数上的变化趋势不一致;(4)数据量不均衡将会影响模型总体的识别精度.表明基于降雨图像的DenseNet降雨强度识别模型具有良好的数据适应性及准确的识别结果.
To improve the accuracy of real-time rainfall intensity identification,rainfall intensity identification model based on deep neural network theory was proposed based on rainfall images and precipitation record.Firstly,rainfall images and corresponding rainfall data were retrieved from 8 meteorological stations in Fuzhou City.Then the rainfall were classified into 6 intensities and images were divided into daytime and nighttime images.Finally,DenseNet deep convolutional neural network was used to construct a rainfall intensity identification model.The experimental results showed that identification accuracies at each meteorological station were higher than 80%,and they were not significantly different from each other.Accuracies of daytime rainfall images exceeded nighttime ones,leading to different trends in identification accuracy at various network layers.However,unbalanced data volume affected the overall classification accuracy of the model.In summary,DenseNet rainfall intensity classification model based on rainfall images is of higher adaptability to data and higher real-time recognition accuracy.
作者
洪思弟
赖绍钧
林志玮
丁启禄
刘金福
HONG Sidi;LAI Shaojun;LIN Zhiwei;DING Qilu;LIU Jinfu(College of Computer and Information Science,Fujian Agriculture and Forestry University,Fuzhou,Fujian 350002,China;Fuzhou Meteorological Bureau,Fuzhou,Fujian 350014,China;College of Forestry,Fujian Agriculture and Forestry University,Fuzhou,Fujian 350002,China;Forestry Post-Doctoral Station of Fujian Agriculture and Forestry University,Fuzhou,Fujian 350002,China;Cross-Strait Nature Reserve Research Center,Fujian Agriculture and Forestry University,Fuzhou,Fujian 350002,China;Key Laboratory of Fujian Universities for Ecology and Resource Statistics,Fuzhou,Fujian 350002,China)
出处
《福建农林大学学报（自然科学版）》
CSCD
北大核心
2020年第4期567-576,共10页
Journal of Fujian Agriculture and Forestry University:Natural Science Edition
基金
福州市科技局社会发展项目(2018-S-109)
海峡博士后交流资助计划
中国博士后科学基金面上项目(2018M632565)。
