core samples. The application under discussion makes use of custom-built algorithms for hyperspectral segmentation and fracture detection. These were designed to support downstream assessments such as Rock Mass Rating (RMR) evaluation. The lithological segmentation module, trained using convolutional neural networks (CNNs), demonstrated high concordance with expert classification and maintained robust performance even when applied to hyperspectral data from sites not included in the training set. This underscores its generalizability and practical applicability across a range of geological settings. For RQD measurement, the application achieved an average absolute difference of 11.8 compared to manual measurements across multiple core images. When converted into RMR scoring, this corresponded to an average deviation of approximately 3 points. Notwithstanding this discrepancy, the system demonstrated a reduction in measurement time of more than 50%, thus substantiating its practical utility in scenarios necessitating expeditious field assessments. The integration of sophisticated algorithms within a user-friendly interface has been demonstrated to engender a substantial reduction in operational workload and cognitive demands, thereby expanding the application's usability to encompass non-specialist users. The system is a scientifically grounded and field-ready tool that has the capacity to enhance the accuracy and efficiency of geological evaluations in a range of applied contexts, including mining, tunnelling, and civil engineering. ACKNOWLEDGEMENTS The authors thank Y. Shinohara of data-harness for support with the implementation of the Python script as a web application. The authors also thank N. Otsuka, a technical staff member of the Faculty of Engineering, Hokkaido University, for support with the implementation of the programming work. The researchers would also like to acknowledge the Science and Technology Research Partnership for Sustainable Development (SATREPS), Japan Science and Technology Agency (JST)/Japan International Cooperation Agency (JICA) for the support to the JapanKazakhstan SATREPS Knight Project during this research and Japan International Cooperation Agency (JICA) Kizuna Program. REFERENCES Alzubaidi, L., Zhang, J., Humaidi, A. J., Al-Dujaili, A., Duan, Y., Al-Shamma, O., Santamaría, J., Fadhel, M. A., Al-Amidie, M., & Farhan, L. (2021). Review of deep learning: concepts, CNN architectures, challenges, applications, future directions. Journal of Big Data 2021 8:1, 8(1), 53-. https://doi.org/10.1186/s40537-021-00444-8 Bioucas-Dias, J. M., Plaza, A., Camps-Valls, G., Scheunders, P., Nasrabadi, N. M., & Chanussot, J. (2013). Hyperspectral remote sensing data analysis and future challenges. IEEE Geoscience and Remote Sensing Magazine, 1(2), 6–36. https://doi.org/10.1109/MGRS.2013.2244672 Cheng, X. (2024). A Comprehensive Study of Feature Selection Techniques in Machine Learning Models. SSRN Electronic Journal. https://doi.org/10.2139/ssrn.5154947 Figueroa Barraza, J., López Droguett, E., & Martins, M. R. (2021). Towards Interpretable Deep Learning: A Feature Selection Framework for Prognostics and Health Management Using Deep Neural Networks. Sensors 2021, Vol. 21, 21(17). https://doi.org/10.3390/s21175888
RkJQdWJsaXNoZXIy MTM0Mzk2