Research On Cloud Cover Model Prediction Based On High Resolution Satellite Archive Image

With the development of remote sensing technology,more and more remote sensing applications require high-quality,high-resolution satellite remote sensing data to provide spatial support.At the same time,the cloudiness prediction of satellite imaging with high spatial resolution and high temporal resolution such as ZY-3 and GF-1 has gradually become a research hotspot.However,the difference in resolution between the ZY-3 and GF-1 data sets and the traditional ISCCP data set and FY-2E data set makes the cloud cover prediction model based on ISCCP,FY-2E and other data no longer suitable for ZY-3,GF-1,which restricts the wide application of high-resolution satellite archived image data in cloud amount prediction.Therefore,the research of cloud cover prediction model based on ZY-3 and GF-1 data sets is of great significance for high-resolution satellite data acquisition and application research.The time series prediction model based on historical cloudiness data largely depends on the density of the data set.This article mainly focuses on the multi-scale data sets of ZY-3 and GF-1,and deeply analyzes the characteristics of high-resolution data sets from the perspective of time series research.And for the cloud data of highresolution images,a cloud-based time series combined prediction model based on dualtree complex wavelet decomposition is established.The main research contents and work are as follows:1)A high-resolution cloud data sequence decomposition method is proposed.On the basis of convolutional neural network preprocessing of historical cloud data,the decomposition of high-resolution cloud data is completed,and two types of cloud data are obtained: low-frequency trend time series and high-frequency random time series.The dual-tree complex wavelet can retain the overall change information of cloud data but better retain the randomness of high-frequency information.It also avoids the increase in the proportion of random item data brought by the increase of the data set,resulting in overfitting.The follow-up cloud amount prediction work laid the foundation.2)A combination model of cloud cover prediction is proposed.Based on the time series decomposition work,two prediction models are designed for low-frequency trend information and high-frequency random information,which are respectively applicable to different time series information,calculate the cloudiness prediction value,and reconstruct to obtain the final cloudiness prediction result,Realized the mid-tolong-term prediction based on high-resolution historical cloud data,and solved the problem that the traditional model is not suitable for high-resolution data sets such as ZY-3.3)Apply cloud cover prediction results to satellite imaging planning.The influencing factors of cloud cover are included in satellite imaging planning,so that the cloud cover within a certain period of time in the region of interest is predicted,and guidance is provided for satellite imaging planning and aerial photography operation plans.