Research On Fog Identification Based On Himawari 8 Satellite Remote Sensing Data

Heavy fog is a common natural weather phenomenon and a natural disaster that cannot be ignored.In recent years,with the rapid development of the economy,the fog has increasingly affected people's production and life.Therefore,researchers are paying more and more attention to the monitoring and identification of heavy fog.With the rapid development of science and technology,satellite remote sensing technology is becoming more and more mature.Remote sensing data has faster update,wider detection range and higher timeliness than traditional ground data.Therefore,more and more fields use remote sensing satellite technology.The data used in this article for the Himawari-8 satellite data,whether it is from the quality of the cloud map,the frequency of interception,the channel,and the sharpness are greatly improved compared with the previous generation satellite.Therefore,the use of Himawari-8 satellite data for fog monitoring and identification research can improve the recognition performance of fog.In this paper,we must first extract and label the fog data before conducting research.According to the latitude and longitude of the ground station,the location of the ground station corresponding to Anhui Province is found,and then the latitude and longitude matching is performed with the satellite dat,and the satellite data at the position is extracted,and the fog of the position is judged according to the visibility of the ground station.In this paper,the fog recognition is mainly studied by two kinds of algorithms.The first type of algorithm is based on the fog recognition of traditional machine learning methods,and the second type of algorithm is based on deep learning fog recognition.The main research contents are as follows:1)Fog identification based on machine learning classification method.Under the condition of balancing samples and unbalanced samples,a variety of traditional machine learning classification algorithms are used for fog identification.Among the algorithms used are support vector machines,naive Bayes,decision trees,and so on.Experimental verification Under the balanced sample,the machine learning classification algorithm has a good effect on the detection of fog.In the case of unbalanced samples,the data expansion using the Synthetic Minority Oversampling Technique(SMOTE)algorithm can also effectively improve the recognition accuracy of fog.2)Fog identification based on classification method based on deep learning.This paper builds a network model,optimizes network parameters,and extracts relevant features from the data.Experiments show that in the case of balanced samples,the convolutional neural network has better recognition ability for fog,and the effect is higher than that of traditional machine learning.Under the unbalanced sample,the SMOTE algorithm combined with the convolutional neural network can also improve the recognition accuracy of the fog,and its recognition performance also exceeds the traditional machine learning classification algorithm.