A Crack Segmentation Model for Coal-Rock CT Images Based on Explicit Visual Prompting and Its Applications

To address the challenge of accurately identifying fine fractures in CT scan images of coal and rock, this study proposes a fracture segmentation model based on explicit visual prompting for coal and rock CT images (EViP-CTCrack). The model is validated on a self-constructed coal and rock CT image dataset, CTrock. EViP-CTCrack primarily consists of components such as a residual mixed connection convolution module, a cross-attention upsampling module, a multi-representative vector classifier, and an explicit visual prompting generator. Experimental results demonstrate that EViP-CTCrack achieves an average intersection over union (IoU) of 88.1% and an average precision of 94.4% on the CTrock dataset, yielding promising fracture segmentation performance. Finally, the model is applied to fracture recognition in mining boreholes, where a porosity-compressive strength equation is established, facilitating the rapid estimation of uniaxial compressive strength.