Ground Penetrating Radar(gpr)based Shallow Surface Soil Water Content Detection Method

Ground-penetrating radar is a new technique for detecting soil moisture at the mesoscale in the field.This paper introduces the use of ground-penetrating radar to detect soil moisture content information at the shallow surface（0-50cm）of gangue-filled reclamation sites,and applies different detection methods,divided into three main parts:the first part is to detect soil moisture content by doing buried steel pipe experiments to create a reflective interface;the second part is the use of ground-penetrating radar and electrical methods combined with the method of gangue-filled reclamation sites The second part is to use the combination of ground-penetrating radar and electrical method to detect the soil moisture content of the coal gangue-filled reclamation site,the two together determine the reflective interface and compare it with the actual depth of the measured profile on site to make the results more convincing,and finally combine with the Topp formula to calculate the soil moisture content;the third part is to use the combination of ground-penetrating radar and machine learning method to model between the instantaneous parameters of the two detection results before and after rainfall and the actual measured soil moisture content,and establish BP neural network model,GA-BP neural network and RF model,and finally the merits of each model are evaluated.The results show that:（1）In the buried steel pipe experiments,the four different instantaneous parameter combinations were not effective in inverting the soil water content in the 0-15cm depth range,and there was a large error between the inverse soil water content and the measured value,making it difficult to accurately invert the soil water content at this depth.In the depth range of15-30cm,except for the phase amplitude ratio model,which still has a large error between the inversion value and the measured value at 15-20cm depth,the other three models can accurately and effectively invert the soil water content,among which the frequency-amplitude ratio combination method has the best inversion effect among the four methods.（2）In the combination of ground-penetrating radar and electrical methods of detection,the use of ground-penetrating radar combined with network parallel electrical techniques can effectively determine the soil-gangue interface.The average velocity is determined by the soil depth,and the water content is the average water content at different depths.TDR measures the water content at each current depth,so the relative error with the actual data measured by TDR is relatively large,and the average water content at different depths can be processed as the average water content at different depths by taking the average of the measured water content at different depths.In a certain depth range,the relative error between VMC_Toppand AVMC_TDRis smaller and the correlation is higher as the depth increases.（3）The GA-BP neural network model outperforms the BP neural network model and the random forest model in terms of accuracy of the period prediction results,both in the natural ground and in the prediction results of the ground after precipitation,and the prediction results after simulated precipitation are better than those of the natural ground.Therefore,the GPR method combined with the GA-BP neural network model is more suitable for shallow surface soil water content detection where the soil water content is high.Figure[38]Table[15]Reference[74]...