Geomaterial segmentation method using multidimensional frequency analysis based on electrical resistivity tomography

Abstract

Abstract The determination of specific geotechnical materials in complex geotechnical environments is important for engineering construction and geological disaster risk management, e.g., for karst caves and fracture zones. As a fast and convenient field test method, electrical resistivity tomography is extensively applied in geotechnical engineering. The resistivity distribution of the underground rock mass and soil can be obtained after inversion of the apparent resistivity. To quantitatively determine the resistivity threshold of specific geotechnical materials, the multidimensional frequency analysis and threshold determination algorithm of image segmentation based on the structural similarity between resistivity grid data and image voxels is employed. The frequency of the resistivity data is evaluated comprehensively with regard to two aspects: the resistivity magnitude and neighboring difference. The neighboring difference is introduced to describe the variation degree of the resistivity in space. Moreover, the Otsu algorithm is adopted to determine the resistivity threshold for identifying categories of the geotechnical material. To validate the material segmentation method, two different cases with rock mass conditions and soil conditions are studied. For the underground cavities left by human mining activities in a rock mass, multidimensional frequency analysis can effectively reveal the frequency components of the cavity resistivity. For an ancient soil landslide with a narrow distribution range of the resistivity, multidimensional frequency analysis can not only quantitatively determine the resistivity threshold of the soil sliding body, but also identify the materials with small proportions in the whole area, e.g., the sliding zone, according to the correlation between the neighboring difference and the material properties. Multidimensional frequency analysis can provide a quantitative and effective segmentation criterion for electrical resistivity tomography results, exhibiting potential for applications in geomaterial investigations.