| Soil salinization refers to the phenomenon that crops cannot grow and develop normally due to the high salt content of the soil;even part of the land has changed from cultivated land to wasteland,which has seriously affected land planting and food production.It is urgent to manage the problem of soil salinization rationally,and the first step to solve the problem is to monitor the overall situation of salinization,so as to achieve the purpose of overall planning.UAV technology can achieve fast,convenient,and fast effects when acquiring remote sensing information,and can obtain large-area images,so as to achieve efficient operation,achieve effective detection,real-time control,and finally propose corresponding solutions through real-time monitoring.And increase crop yields and reduce the harm of soil salinization.This paper uses soil salinization with different gradients in the Heat Irrigation District of Inner Mongolia as the research object.Through the remote sensing images and soil salt and moisture data collected from 2018 to 2019 in different growth periods of crops,the use of thermal infrared remote sensing is analyzed and discussed.How to monitor the soil salt and moisture more accurately with the image,analyze the correlation between the soil temperature during the bare soil period,the plant canopy temperature in July and August,and the soil water and salt,using single factor curve modeling and multiple linear regression models,Partial Least Squares Regression Model,Ridge Regression Model,Machine Learning BP Neural Network Model and ELM Extreme Learning Machine Model respectively construct different water and salt inversion models,and analyze and evaluate the accuracy.The main conclusions are as follows:(1)Through the analysis of the correlation between the surface temperature of the soil during the bare soil period and the soil salinity at different depths,it is found that the correlation between the surface temperature and the surface soil is higher than that of the deep soil;as the degree of salinization increases,the correlation between the surface temperature and the soil surface temperature is higher.The better the relevance;the normalization method uses the water-ground temperature difference as the standard of normalization treatment and has achieved better results.Between unary linear modeling and curve modeling,the use of cubic function operation has the best response characteristics for soil temperature inversion of soil salt content.(2)By combining the canopy temperature and the spectral reflectance of visible light in the 3 bands,and constructing an index through curve calculation,the soil water and salt content can be better inverted.Taking the index screened by gray correlation analysis method as the independent variable and the corresponding soil water and salt content as the dependent variable,based on three linear regression methods,the water and salt content inversion model was constructed,and it was found that ridge regression was used to invert soil moisture and soil salt.The best results are achieved every time,and the use of linear regression alone is the worst.(3)By combining the canopy temperature and the multi-spectral 6-band spectral reflectance to construct different indexes,taking the canopy temperature and the spectral correlation index as the independent variable,the water and salt content as the dependent variable,and using the machine learning method to construct the water and salt content Inversion model.Multivariate linear regression model,BP neural network model,ELM extreme learning machine three methods,the ELM extreme learning machine model has the best inversion effect,and the inversion effect of soil salt content at a depth of 20 cm is the best;but in the inversion of the surface layer When soil salt content,the BP neural network model has the best inversion effect;when inverting soil moisture content,both the surface and deep layers are the ELM extreme learning machine model with the best effect. |
PDF Full Text Request