Introduction

3DSensorDB is a spatial database for storing, managing, and analyzing georeferenced 3D sensor data. The current implementation supports LiDAR-derived point clouds, with planned extensions for camera-based image data.

The system supports the ingestion of pose trajectories of the acquisition platform and its sensors during mobile or static recording sessions. Explicit representation of sensor pose (position and orientation) enables advanced analytical capabilities, particularly in combination with semantic environment models. Semantic models in OGC CityGML  can be imported into an automatically co-deployed 3DCityDB  instance, enabling spatio-temporal analytical queries across 3D sensor observations and semantic objects, as well as association and bidirectional enrichment procedures.

Alternatively, 3D point clouds can be indexed using only an octree, importing only the octants that contain key metrics such as point density and acquisition period. This enables reports and analyses at the octant level while minimizing database storage requirements.

As individual points are not always required and consume significant storage capacity, it is possible to import only aggregated cell information (octree octants) along with their attributes, such as point density and recording time periods. This approach is particularly useful for providing an overview of large-scale acquisition campaigns, including their spatial and temporal coverage.

For source code and issue tracking, refer to the repository .

Next Steps

• Getting Started — install and run the database using Docker
• Import — import point clouds, ROS2 bags, and CityGML datasets
• Process — generate sensor views and associate observations
• Export — export associated point cloud data
• Database Schema — table definitions and module overview