FMQF Technical Record 5 - Automated identification of mining features using Light Detection and Ranging (LiDAR) data – Stage 3
 |
| |||||
Product description:Executive Summary:A new dataset of mine shafts and surface workings (mining features) has been developed as part of the Former Mines and Quarries Framework (FMQF) program. The FMQF aims to deliver a state-wide management framework for abandoned and legacy mines and quarries on Crown land in Victoria. This report details the application of a bespoke automated process, developed by WSP Global Inc. to detect former mining features on Crown land in the Cobungra-Dargo region using LiDAR data. The method builds on previous successful applications across the Golden Plains, Greater Melbourne and Central Goldfields regions. Potential mining features were identified from LiDAR point cloud data and LiDAR-derived Digital Elevation Models (DEMs) using DEM sink and noise LiDAR detectors. A subset of these features was then manually labelled and used to train a machine learning image classification model, enabling automated classification of similar features within the LiDAR dataset. A random forest model was used to filter the detections, utilising a range of statistical, geological and geographical attributes to determine each feature’s probability of representing a mining feature. Low-probability features were removed, improving the accuracy and reliability of the final dataset. Model performance was assessed using recall and precision, which were measured at 77% and 95% respectively. Using this process, 20,695 mining features were identified across the Cobungra-Dargo region and incorporated into the FMQF consolidated database, a significant increase from the 2,852 features previously recorded in the historical database. In total, 224,185 features have been added to the database through LiDAR interpretation and machine learning over the course of the FMQF program. This study demonstrates the value of applying innovative techniques to improve the accuracy of historical data and uncover previously unrecorded mining features. Bibliographic Reference: Silver, E.R., Dang, L.H., Herley, S.S., Riley, C.P., Eid, R. & Tran, L.V., 2026. Automated identification of mining features using Light Detection and Ranging (LiDAR) - Stage 3 Cobungra-Dargo region, Victoria. Former Mines and Quarries Framework Technical Report 5. Geological Survey of Victoria, Department of Energy, Environment and Climate Action, 19 pp. Download: The downloadable version of this report is supplied as (PDF 8 MB), accessible version (DOCX 4 MB) and Attachment A1 data (ZIP 3.8 MB). Related products: | ||||||
