Texture Features of Internal Cracks in Expansive Soil During Wet–Dry Cycles Using CT and Gray-Level Co-Occurrence Matrix

During wet–dry cycles, cracks in expansive soil evolve continuously, resulting in remarkable changes in the structural and textural characteristics of the soil mass. In this study, wet–dry cycle tests combined with CT scanning were carried out on undisturbed expansive soil collected from Hefei, and the gray-level co-occurrence matrix (GLCM) method was introduced to quantitatively extract the internal texture features of cracks from sequential CT images. Four GLCM parameters—angular second moment (ASM), contrast (CON), entropy (ENT), and correlation (COR)—were extracted at four directions (0°, 45°, 90°, and 135°) to investigate the texture evolution and directional anisotropy. The results indicate that cracks first initiate at weak positions near the top of the specimen and then extend downward, gradually developing into strip-shaped fractures and ultimately forming a three-dimensional crack network. With progressive crack development, the average gray value of CT images decreases from 198.06 to 137.38, and the distribution characteristics of gray values change significantly. To further quantify the spatial heterogeneity and directional anisotropy of the crack network, four GLCM texture parameters were extracted. Quantitatively, ENT and CON increase sharply with the number of wet–dry cycles: after the 5th cycle, ENT rises by 231–249% and CON by 317–335% across the four directions. Conversely, ASM and COR gradually decrease, with ASM declining by 25–29% and COR by 68–74%. All four parameters follow a logistic function with respect to the number of cycles (R2 > 0.99). Furthermore, CON and ENT exhibit statistically significant directional differences, enabling the identification of dominant crack development directions, whereas ASM shows a consistent but non-significant directional trend and COR is insensitive to directional variations. These findings demonstrate that GLCM parameters possess good applicability for the quantitative characterization of internal crack texture in expansive soil. This study provides an effective method for investigating crack evolution and offers a new perspective for CT image analysis of rock and soil mass.