Henan Mongol Autonomous Grassland-like Haruka Sensation Number Stationary Super-Division Method Research

Remote sensing images can quickly obtain a wide range of ground information in different bands and play an important role in agricultural and animal husbandry monitoring.Compared with satellite remote sensing,UAV images have the characteristics of high resolution.The high-resolution images collected by UAVs are used to correct satellite remote sensing images to improve the data accuracy of satellite remote sensing.Compared with the traditional super-resolution algorithms of deep learning,deep learning super-resolution processing has become a key issue of research.This paper uses high-resolution UAV images for low-resolution satellite remote sensing images.The grassland vegetation index NDVI is corrected,and the super-resolution research of satellite remote sensing images is carried out.The main contents of the research work mainly include three aspects:1.Data set production.The data set construction in this paper is divided into two parts: UAV remote sensing images and satellite remote sensing.The construction process of the UAV remote sensing data set is as follows: In July 2020,16 alpine grassland areas in Henan County,Qinghai Province were used to carry multispectral images.A total of 900 multispectral remote sensing images were collected by the camera,and preprocessing was performed to obtain the UAV multispectral remote sensing data set.Satellite remote sensing dataset: The 45-spoke landsat8 remote sensing data and the 4 pieces of Gaofen-1 remote sensing data from January 2019 to December2020 were preprocessed to complete the construction of the satellite remote sensing dataset.2.Use UAV images to correct satellite remote sensing images.For the prediction of NDVI of alpine grassland vegetation,satellite remote sensing data with a wide data coverage area is used,and high-resolution UAV remote sensing data is used to make up for the low resolution of satellite remote sensing.Affected by the external environment such as topography and landforms,it is insufficient to collect data in a large area.Based on the landsat8 satellite remote sensing and unmanned aerial vehicle remote sensing data,this study carried out the correction of the NDVI of the alpine grassland vegetation index.3.Methods for super-resolution reconstruction of remote sensing images.Based on the image super-resolution algorithm SRGAN,this paper improves the SRGAN algorithm for better performance of the super-resolution reconstruction effect.The improved method is divided into two parts:(1)Since the remote sensing images of the alpine steppe landsat8 vary greatly,we propose the SRCSRGAN network which consists of three parts: classifier,generator and discriminator.The input remote sensing images are classified and preprocessed,and then each of the four types is separately sent to the image super-resolution algorithm for super-resolution processing,so as to improve the network’s learning ability for each type,adding the classification preprocessing module to compare Images generated from unjoined networks have better quality.(2)For the improved super-resolution network model,Smooth L1 Loss is used as a new loss function,and TVloss is added to constrain image noise,thereby improving the super-resolution effect.