An image recognition method for the deformation area of open-pit rock slopes under variable rainfall

Due to human mining action, relatively fragile open-pit mine rock slopes are prone to instability induced by heavy rain. Accurately identifying the information and area of deformation features on rock slopes is a key step for landslide disaster warning and prevention. This study simulates the deformation process of a large and steep rock slope in Dexing City, Jiangxi Province, China, under a variable rainfall event. The test results show that 1) the increased deformation region has positive relationships with increasing pore water pressure and water content values for the open-pit mine rock slope affected by heavy rainfall. 2) The infiltration of variable rainfall on the rock slope softens the weak interlayers and leads to failure of the slope toe first. Due to the action of gravity, the middle and upper parts slide and fail in sequence. In addition, an improved region growing segmentation method (IRGSM) is proposed based on the combination of multipixel seed points and point cloud coordinates for the image recognition of the deformation area of open-pit mine rock slopes. An error comparison with the original region growing segmentation method (ORGSM) shows that the average identification error in the X and Y directions by the method is reduced significantly (6.56% and 5.32% in IRGSM; 13.37% and 11.29% in ORGSM). This method may be applied to identify rock slope deformation in complex scenes with high precision.

Automated summary

This study simulated the deformation process of a large and steep rock slope in China under variable rainfall, finding that the increased deformation region is positively correlated with increasing pore water pressure and water content values, while the infiltration of rainfall softens weak interlayers and leads to failure of the slope toe first, followed by the middle and upper parts sliding and failing sequentially. The proposed improved region growing segmentation method showed a significantly reduced average identification error in X and Y directions compared to the original method, indicating its potential for high-precision identification of rock slope deformation in complex scenes.