GNSS-R Soil Moisture Inversion Based On Improved BP Neural Network Algorithmon

As an important parameter of ecological environment information,soil moisture plays an important role in the global climate and ecological environment,and plays a vital role in the study of ecological water cycle,vegetation water supply,and land carrying capacity.Soil moisture data can be used for vegetation growth monitoring,meteorological disaster monitoring,landslide prediction,mountain fire warning,etc.Therefore,real-time,accurate and long-term soil moisture information will be used for weather forecast,hydrological research,agricultural production,disaster monitoring,etc.The research provides important data support.At present,there are two main traditional ways to obtain soil moisture: one is to use professional remote sensing satellite means to obtain the soil moisture content,and the other is to obtain soil moisture through on-site sampling of soil moisture equipment.Both of these methods have certain drawbacks and defects.Although the former has a relatively large range of telemetry,it has a small spatial and temporal resolution and cannot meet the requirements of most applications.The latter mainly uses probes to detect soil moisture,and its detection range is relatively small,and often needs to be sampled in the field,which is time-consuming and laborious and cannot meet the long-term high time resolution requirement.Therefore,it is of great significance to study the technical methods of acquiring high spatial,high temporal resolution or real-time soil moisture data.In recent years,with the continuous development of global navigation and positioning technology,more and more satellites have appeared in the air for us to use.Under this background,the new GNSS-R soil moisture detection technology based on satellite signals has received extensive attention and the study.As a new type of microwave remote sensing technology,GNSS-R technology has its unique advantages,such as wide coverage,all-weather,real-time high resolution and other advantages.In addition,with the advent of the artificial intelligence boom,more and more machine learning has appeared in our lives.It is now widely used in all walks of life.Under this background,this topic will combine machine learning algorithms.The research on soil moisture inversion makes the soil inversion more reliable,accurate and intelligent.Based on the principles of GNSS-R inversion technology and satellite carrier signal characteristics,this paper has conducted systematic and exploratory research on soil moisture inversion methods and processes.The main research contents are as follows:(1)Summarize the advantages and disadvantages of GNSS-R and traditional soil measurement technology,and introduce in detail the carrier characteristics of satellite signals,the basic theory of multipath effect and the causes,the basic theory of signal-to-noise ratio,etc.The relationship between signal-to-noise ratio,multipath effect and satellite signal carrier signal is derived and analyzed.The basic theory of GNSS-R inversion technology is systematically introduced.(2)Discuss and study the processing methods of signal-to-noise ratio data and characteristic parameters,use polynomial fitting to separate reflected signals,spectrum analysis and least squares fitting to obtain parameters such as amplitude and phase.At the same time,we discuss the correlation between the characteristic parameters of signal-to-noise ratio and the environment,and study the influence of rainfall and elevation changes on the characteristic parameters.(3)In view of the complex modeling of traditional inversion models and the lack of extensiveness in the model,this paper combines the current machine learning boom to explore the use of neural network algorithms and genetic algorithms to study soil moisture inversion,Give full play to their advantages such as strong learning ability,outstanding memory ability,excellent anti-interference performance,and the ability to solve a large number of complex and difficult nonlinear problems.At the same time,in the research,the traditional BP neural network is easy to fall into the local minimum and cannot obtain the optimal solution in the inversion.The BP neural network optimized by genetic algorithm and the BP neural network optimized by wavelet are proposed to improve the inversion accuracy.Taking advantage of the powerful global search capability of genetic algorithms and the data processing advantages of wavelet functions,XB-BP inversion model and GA-BP inversion model are established.(4)Through the use of experimental data for training,learning and prediction of the three network models,the results show that in terms of prediction ability,the learning and prediction ability of wavelet BP neural network and genetic algorithm BP neural network is better than that of traditional offline model inversion prediction ability High,the genetically optimized BP network has the most stable and highest accuracy in soil moisture inversion prediction.