Abstract: Due to the complexity and abruptness of deep engineering disasters, the mechanical parameters of rocks, particularly uniaxial compressive strength (UCS), play a critical role in stability analysis, risk assessment, support parameter optimization, and intelligent construction. Achieving rapid quantitative characterization of rock strength is therefore essential. A novel approach for the rapid quantitative characterization of rock UCS from a mineralogical perspective is proposed, utilizing laser-induced breakdown spectroscopy (LIBS) technology, thereby introducing a mineral-to-strength conversion methodology. A spectral database for granite and metamorphic sandstone is constructed using LIBS, and the elemental and mineral datasets of the rocks are acquired by X-ray fluorescence (XRF) and X-ray diffraction (XRD) tests, respectively. The mass fraction of mineral components is analyzed via the support vector regression (SVR) algorithm. In the end, a mineral-strength conversion model is established to calculate the UCS from the predicted values of mineral component concentrations, and its rationality and scientific validity are validated by the standard mechanical tests. Results indicate that the coefficient of determination R² of each element in the spectral-elemental prediction model is between 0.96 and 0.99, and the model can effectively obtain the content of each element; the elemental-mineral prediction model has the best prediction effect of quartz and aluminum phosphate, followed by feldspathic and mica mineral groups, and good prediction effect of chlorite and turbidite zeolite as well. The multiple linear regression model (R² = 0.850 2) is more suitable than the least squares method (R² = 0.719 6) for the nonlinear relationship between the predicted value of mineral mass fraction and uniaxial compressive strength, and the introduction of the correction coefficients can be more effective in realizing the accurate conversion of minerals-strengths. Finally, the mineral-strength conversion model is used to conduct rapid batch testing of the UCS of rocks in different mileage of working faces at the project site, and combined with the Kriging interpolation technique and Matlab technology to realize the 3D continuous visualization of the UCS at the project site. The research results provide theoretical basis for rapid, accurate and intelligent disaster source identification and dynamic optimization of surrounding rock support parameters in deep engineering to improve the engineering problem of limited timeliness of traditional mechanical tests.