AI-Enabled Slope Upgrading Works: From Static Archives to Operational Intelligence
DOI:
https://doi.org/10.21467/proceedings.8.2.9Abstract
The planning of slope upgrading works under Hong Kong’s Landslip Prevention and Mitigation Programme requires reliable early-stage cost and schedule forecasting, together with efficient review of historical district archive records. In practice, this process suffers from limited structured expenditure data and the fragmented nature of scanned archival documents. This paper presents an AI-enabled workflow using a novel AI system that integrates explainable estimator based on structured slope attributes with a vision-based information retrieval pipeline for district archive PDFs. The estimator combines Random Forest and XGBoost models to generate bounded cost and schedule predictions with explicit confidence indicators. The information retrieval pipeline employs cascaded CLIP and VLM-based page classification, grid-label extraction, and robust spatial reconstruction to identify map pages relevant to a target slope location. The effectiveness of the proposed approach is validated using 360 historical slope upgrading works records and 799 archival image samples. Overall, this AI system, combining both estimator and information retrieval pipeline, transforms static archival resources into operational intelligence, improving planning consistency, reducing manual document review, and supporting more transparent early-stage decision-making.
References
Common Spatial Data Infrastructure (CSDI), 2026. Catalogue of Slopes maintained by Geotechnical Engineering Office of Civil Engineering and Development Department [WWW Document]. CSDI. URL
https://portal.csdi.gov.hk/geoportal/#metadataInfoPanel
GEO, 2011. GEO Publication No. 1/2011.
GEO, 2013. The New Priority Ranking Systems for Man-made Slopes and Retaining Walls (GEO Report No. 284).
GEO, 2025a. Past Notable Landslides [WWW Document]. Hong Kong Slope Safety (HKSS) Website. URL
https://hkss.cedd.gov.hk/hkss/en/facts-and-figures/past-notable-landslides/index.html
GEO, 2025b. Slope Information System [WWW Document]. Hong Kong Slope Safety (HKSS) Website. URL
https://hkss.cedd.gov.hk/hkss/en/facts-and-figures/slope-information-system/sis/index.html
Hong Kong Special Administrative Region (HKSAR) Government, 2025. The Chief Executive’s 2025 Policy Address Supplement.
Stevenson, W.J., 2015. Operations management, Twelfth edition. ed, The McGraw-Hill Series in operations and decision sciences. McGraw-Hill Education, New York.
Wu, K., 2014. A Unified View on Planning, Scheduling and Dispatching in Production Systems.
Breiman, L., 2001. Random forests. Machine learning, 45(1), pp.5-32.
Chen, T. and Guestrin, C., 2016. XGBoost: A scalable tree boosting system. In Proceedings of the 22nd acm sigkdd international conference on knowledge discovery and data mining (pp. 785-794).
Fischler, M.A. and Bolles, R.C., 1981. Random sample consensus: a paradigm for model fitting with applications to image analysis and automated cartography. Communications of the ACM, 24(6), pp.381-395.
Radford, A., Kim, J.W., Hallacy, C., Ramesh, A., Goh, G., Agarwal, S., Sastry, G., Askell, A., Mishkin, P., Clark, J. and Krueger, G., 2021. Learning transferable visual models from natural language supervision. In International conference on machine learning (pp. 8748-8763). PMLR.
Downloads
Published
Issue
Section
License
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.
