A Unified Database Schema for Geometric and Semantic Data: Continuous Volumetric Stocktaking in Gravel Quarries

In mining and quarry operations, continuous and accurate monitoring of material volumes is imperative for operational efficiency and inventory management. Current methods often rely on standalone, offline tools, necessitating intermittent data collection and processing, introducing delays and potential inaccuracies. This paper presents a novel unified database schema that seamlessly integrates geometric and semantic data, facilitating real-time monitoring and volume estimation of material heaps in gravel quarries. Utilizing the PostGIS extension of PostgreSQL, our schema allows for the robust storage and management of spatial data alongside easy integration with open-source geospatial tools like QGIS. We have developed interfaces for visual SLAM algorithms (openVSLAM and maplab), 2D LiDAR maps, and proprietary semantic maps for intralogistics use cases, enabling data aggregation in a common reference frame or with stored transformation between different reference frames. An application of this schema for continuous stocktaking of gravel heaps is demonstrated. Our system executes a volume estimation pipeline by defining Regions of Interest (ROIs) within the quarry and employing point cloud data generated via Structure from Motion (SfM) or LiDAR. This pipeline includes ROI extraction from point clouds, ground plane detection, mesh generation, and volume estimation, with the results written back to the database as meta-information. We validate our approach by comparing the volume estimates obtained with a commercially available offline tool, Pix4D, showcasing the efficacy and precision of our system for continuous, real-time monitoring and volume estimation in mining and quarry environments.