Research On Imaging Of Dry Shrinkage Cracks On Farmland Soil Surface Based On Electrical Resistivity Technology

Under drought conditions,the surface of clayey soil is prone to water loss and shrinkage to produce cracks,and the impact of cracks is extremely wide.Exploring the development of fissures on the soil surface can help prevent or inhibit the generation of soil shrinkage fissures,thereby reducing the damage to the integrity of the soil by the fissures,and preventing the fissures from causing practical hazards such as preferential soil flow and instability of engineering foundations.For the research on the morphological characteristics of soil shrinkage cracks,most of them are based on experimental research,and the corresponding numerical model is relatively lacking,so it is impossible to predict the possible location of soil cracks.In this paper,a more efficient resistivity inversion algorithm is proposed to characterize the morphological characteristics of soil fissures by using the phenomenon that the resistivity characteristics of soil fissures can better reflect the distribution position and morphological characteristics of dry shrinkage fissures.The morphological development of invasive detection fractures provides ideal algorithms and means.In order to reveal the morphological development characteristics of dry-shrinkage cracks on the surface of farmland soil,this paper carried out a electrical resistivity technology(ERT)experimental study on the cracked farmland soil cracks under natural conditions,and applied the simultaneous successive linear estimation(SimSLE)coupled with hierarchical clustering analysis.SimSLE algorithm was used to invert the development of dry-shrinkage fissures on the surface of soil fissures.Digital image processing technology was used to perform binarization and skeleton extraction on the experimental images of farmland soil fissures and the images of simulated fissures.The Minkowski density(Minkowski)of the experimental fracture images and the simulated fracture images was used to quantify the morphological characteristics of the fracture network.At the same time,three evaluation indexes,R2,Index of Agreement IA,and Root Mean Square Error RMSE,were established to quantify the difference between the Minkowski density of the experimental fracture image and the simulated fracture image.The results show that:(1)The ERT test can provide the potential distribution of the soil measurement area in real time.It is easy to operate and can adapt to complex conditions.It provides a basis for inversion and simulation of fracture morphology images,and has good application value;(2)The inversion results of the SimSLE algorithm restore the morphological characteristics of the dry-shrinkage cracks on the surface of the farmland soil.The test crack image and the simulated crack image have a high degree of fit.The simulation results can reflect the morphological characteristics of the dry-shrinkage cracks in the farmland soil under natural conditions.The determination coefficient R2 of the area density,length density and Euler number density of the simulated test images are all greater than 0.6,and the Index of Agreement IA is greater than 0.9,which further illustrates the applicability and reliability of the model;(3)SimSLE coupled with hierarchical clustering analysis is used to simulate the morphological characteristics of dry-shrinkage cracks on the soil surface,and iteratively predicts the initial mean value,boundary information and covariance matrix of sitescale heterogeneous strata,and then uses them as the model for the next time the initial input value for the iteration.The inversion result of the new algorithm not only reproduces the shape and position of the fracture,but also makes the boundary information of the fracture and non-fracture more accurate,and the simulated fracture image and the experimental fracture image have high consistency.After clustering and partitioning,the determination coefficient R2 of the Minkowski density of simulated fracture morphological images increased by 9.76%～18.5%,the consistency index IA increased by 0.93%～1.47%,and the root mean square error RMSE decreased by 12.1%～21.1%.The model predicted The accuracy has been significantly improved.