Preprocessing method of microstructure image of geotechnical materials

Abstract

Abstract The engineering properties of geotechnical materials are primarily controlled by the system state and overall behavior of their internal microstructures, and the complex physical and mechanical properties are the concentrated reflections of their microstructural characteristics. Therefore, it is necessary to study the microstructures of geotechnical materials. The development level of microstructural research is closely related to observation technology. In order to reveal the intrinsic relationship between the engineering properties of geotechnical materials and microstructures, the observation technology of microstructures must be continuously updated. At present, most of the test methods cannot continuously observe the changes of microstructures of geotechnical materials under loading, and cannot obtain the actual deformation and displacement information of particles and pore. Therefore, it is difficult to establish the soil stress model and constitutive relationship on the basis of the actual evolution of microstructures, which restricts the further development of relevant theories. In order to overcome the above shortcomings, in order to improve the accuracy and completeness of extracting the characteristic microstructural parameters of geotechnical materials, the clear parts of the same image sequence are extracted and reconstructed, and then high-definition images are obtained. Subsequently, the applicability of two commonly used threshold segmentation methods in the pretreatment of the micro-structure image of geotechnical materials is compared and analyzed. The results show that the multi-resolution technology of wavelet transform can effectively increase the signal-to-noise ratio, reduce gray variance and optimize the quality of the micro-structure digital image. The average gray level method and the maximum variance automatic threshold method are used to process the microstructure image of geotechnical materials, which can effectively segment the image and ensure the integrity of feature parameters extraction.