Radar echo rainfall intensity classification algorithm based on architecture search
Weather radar is an important means to monitor rainfall,which is very important to achieve accu-rate classification of rainfall intensity based on rainfall echo characteristics of weather radar.This paper pro-poses an Evolutionary Search Algorithm for CNN(EA-CNN)based on 3D attention module.The algorithm uses Genetic Algorithm(GA)to automatically search the optimal structure,quickly optimize the width and depth of CNN architecture,and accurately find the optimal CNN structure,which has strong generalization ability.The 3D attention module is introduced in the algorithm to effectively suppress the problem of inter-fering pixels in classification,which can accurately extract the rainfall characteristics of each intensity,and help relevant people to quickly obtain the rainfall classification results from the radar echoes.Experiments show that compared with four classical algorithms(KNN,SVM,VGG,ResNetl2),the proposed algorithm is more competitive in labor and material costs,and the classification accuracy is improved by 7.95％.
万方数据
维普
