August 5, 2025
It’s The Outbox That Counts – Delivering Land Disturbance Reporting To Regulators
Outline’s ADM, AI-assisted technology including object detection, segmentation and feature extraction helps mining companies refine their Mining Closure Plans (MCPs).
What is Outline Global’s Automated Disturbance Mapping Technology?
Outline Global’s Automated Disturbance Mapping technology (ADM) provides mining operators with data and quick insights to refine their mine closure plans. Spatial data, including aerial and satellite imagery, can offer precise details regarding the locations and methods for implementing mitigation efforts or assessing the progress of ongoing initiatives.
Quickly deriving insights from substantial amounts of spatial data is possible today thanks to AI-assisted capabilities. That refers to AI systems designed to support humans in performing tasks that are often resource and time-intensive and/or repetitive. Specifically, ADM utilises the following AI-assisted technology:
• Object detection: identifying and locating objects within images or from high-definition elevation data.
• Feature extraction: automatically identifying and isolating meaningful characteristics (features) from digital images.
• Change detection: analysing time-lapsed images to identify area changes.
Three developments explain the popularity of machine learning today: the greater availability of vast amounts of data, more powerful and affordable computing resources and user-friendly tools that simplify the implementation of machine learning algorithms. Applied to mining, ADM enables mining operators to achieve compliance and reporting quicker thanks to workflows that encompass rapid capture of large data volumes, image processing, and the generation of derived insights.
The following examples demonstrate how ADM provides mining operators with site-specific insights and data to refine their mine closure plans:
• Object detection: finding disturbance features, such as exploration tracks, drill pads, and sump pits.
• Feature extraction: using pixel-based raster images as input files, the algorithm auto-generates vector layers containing AI-classified disturbance features such as pads and tracks.
• Change detection: identifying new disturbance instances through temporal comparison of disturbance layers.
ADM extracted pads & tracks
Manual (green) versus ADM (magenta)
How does Automatic Disturbance Mapping technology work?
ADM uses computer vision to help computers interpret images like humans do. Developing a model that identifies and extracts spatial features requires spatial datasets where such features are present and labelled to provide context for machine learning models.
Model training in machine learning is the process of teaching a machine learning algorithm to learn from data. For example, before ADM can “recognize” pads and tracks by itself, it needs to “see” substantial amounts of labelled images without making any mistakes. AI models must undergo training before being applied to new datasets.
Model training is an iterative process where a model is refined over time with new datasets to make it perform better, for example in managing edge cases, which are extreme or rare scenarios that can cause unexpected or incorrect behaviour in a model. In image analysis, this means features are either misclassified or missing from the results layer (and therefore not labelled at all).
Manual (green) versus ADM (magenta)
Manual verticies (green) versus ADM verticies (magenta)
Because such missing or mislabelled features can have a significant impact on environmental compliance reporting for mining areas, the AI model needs to be dependable and accurate. High-resolution imagery can prevent unwanted or unexpected model behaviour. High-resolution imagery with a small Ground Sample Distance (GSD) indicates that each image pixel represents a smaller area on the ground, resulting in a higher resolution and detail. Or in other words, 2cm GSD imagery is much higher resolution (pixels represent a smaller part of the earth’s surface) than 20cm GSD imagery.
Outline Global captures its own high-resolution aerial (RGB+Infrared) imagery and lidar datasets for mining customers. RGB and RGB+IR imagery offers high resolution and scale for the application, whereas lidar supplies a texturally detailed DEM (Digital Elevation Model) for AI detection. Outline Global has perfected systematic imagery/LiDAR capture for accurate and consistent automation, though it can use drone and satellite imagery when appropriate. This improves training data and makes the final optimized model more accurate and reliable, with a feedback loop to monitor its predictive performance.