<p>Lithological characteristics are the core influencing factors determining rock mass stability and drilling efficiency. Currently, commonly used testing and identification techniques (e.g., uniaxial compression tests, drilling sampling, and geophysical exploration) generally suffer from significant limitations including destructive sampling, high implementation costs, limited testing accuracy, and localized detection range. This paper integrates borehole imaging with digital image processing technology to achieve intelligent inversion of rock strength parameters, providing a more efficient technical approach for rock engineering stability assessment and drilling efficiency optimization. First, for borehole imaging data, a conditional generative adversarial network (cGAN) model is employed for image enhancement and feature extraction. RGB color space information is then utilized to establish standardized lithology classification criteria, providing fundamental data support for rock formation identification and strength prediction. Second, based on a multivariate Gaussian probability distribution model combined with Bayesian principles, a comprehensive approach was developed, which deeply couples rock strength with lithology classification, resulting in a multi-dimensional RGB-based strength prediction model. Finally, the accuracy and reliability of the proposed model were validated by comparing predictions with experimentally measured strength data. Results demonstrate that this novel rock strength prediction model achieves an accuracy of up to 95.57%, enabling precise identification of borehole rock formations and accurate assessment of rock strength parameters. This new prediction model, based on RGB color information, not only provides innovative methodologies for rock formation image analysis but also establishes a theoretical foundation for advanced lithological analysis, rapid assessment of rock mechanical properties, and the development of comprehensive geological models.</p>