Accurate Prediction Of Reclaimed Soil Moisture In Mining Area Based On Ground Penetrating Radar

With the progress of human social civilization,the exploitation of mineral resources has gradually become the basic source of human production and life.Due to the long time,large-scale and high-intensity resource exploitation,it inevitably causes many environmental problems.Land reclamation has gradually become an important means to manage the mining environment,where the moisture content of the reclaimed soil is an important indicator to judge the effect of reclamation.Therefore,fast,accurate and nondestructive detection of reclamation soil moisture content distribution is of great significance for reclamation agricultural water resources management,reclamation effect monitoring and reclamation farmland hydrology research.Ground-penetrating radar is not only a geophysical technology,but also gradually evolves into a new type of near-earth remote sensing technology,which is characterized by the advantages of large-range continuous detection,rapid,nondestructive and low-cost for shallow subsurface media,making up for the shortcomings of traditional water content measurement methods.In order to explore the possibility of ground-penetrating radar in soil water content inversion for reclamation,this paper takes a subsidence reclamation area in Huaibei mining area as the research area,uses ground-penetrating radar as the data acquisition platform,analyzes the response relationship between ground-penetrating radar electromagnetic wave signal and soil water,couples the attribute information under different signal transformation and processing methods,uses various optimal parameter screening algorithms to explore the optimization of model variables and accuracy enhancement capability,and establishes The optimal soil volumetric moisture content inversion model based on ground-penetrating radar was developed,and the following main conclusions were drawn.(1)The time domain signal waveform trends of ground-penetrating radar acquired under different soil moisture conditions are basically the same,but there are certain differences in the numerical performance of parameters such as amplitude and energy;secondly,with the gradual increase of soil moisture content,the energy and amplitude intensity parameters gradually decrease,which is mainly due to the increase of soil moisture content leading to the rise of internal dielectric constant of soil,which makes the electromagnetic wave propagation in the process of soil In addition,by extracting the characteristic parameters of the original time domain signal,the Pearson correlation coefficient method was used to analyze the response characteristics of different characteristic parameters in the original time domain signal and the soil volumetric water content,and the results showed that the correlation between different characteristic parameters and the soil volumetric water content varied,and the energy parameter had the highest correlation.Finally,the prediction model based on the original time-domain signal parameters has low accuracy,which is difficult to meet the needs of actual production and life.(2)Different signal transformation and processing methods can effectively overcome the problems of inconspicuous correlation and missing key information in the original time-domain signal,and the correlation between radar signal and soil volumetric water content has been improved compared with the original time-domain signal.In addition,the inverse model of soil volumetric water content under different signal transformation methods was established by using regression model,and the accuracy of the model after Hilbert transform and linear FM Z-transform was improved compared with the model constructed by the original signal characteristics in the time domain.(3)Since the full attribute set of ground-penetrating radar has too many feature parameters and high data redundancy,Pearson correlation analysis is used,and 16effective feature parameters are screened by significance test and other methods;in addition,continuous projection algorithm,variable space shrinkage screening and optimal subset screening are used to optimize the feature set for parameter compression and optimization,and the results show that the optimal subset Secondly,the inversion model of soil volumetric moisture content under different screening algorithms was established by model construction methods,and the results showed that the accuracy of the limit learning machine moisture prediction model under the coupled optimal subset screening algorithm was the highest,with R~2 and RMSE reaching 0.89 and 0.54%;in addition,comparing different model construction methods,the optimal inversion model parameter sets were obtained by the optimal In addition,when comparing different model construction methods,the optimal set of inverse model parameters was obtained by the optimal subset screening algorithm,indicating that the optimal subset screening algorithm has strong adaptability in small sample data sets.(4)Analysis of the response relationship between the significantly correlated characteristic parameters and the volumetric water content of soil and their variation characteristics showed that the peak frequency and the average value of the early amplitude envelope could produce a good response to the unsaturated water content,while the instantaneous frequency,amplitude and energy could not reflect the water enrichment of soil well in the complex soil environment,and some methods need further research and calibration.In addition,the moisture distribution map obtained by ground-penetrating radar shows that the closer the study area is to the center of the subsidence,the higher the moisture enrichment,i.e.,the soil moisture distribution in the study area shows a trend of higher volumetric moisture content the closer it is to the center of the subsidence.Therefore,coal mining subsidence has a great impact on the moisture distribution and vegetation crop growth in the surrounding area,which should be taken seriously.Figure[27]Table[8]Reference[105]...