Processing large data volumes

Preparing point clouds for visualization is time-consuming and computationally intensive, and is traditionally done as a postcollection processing step. To speed up this process, we have developed lidarserv, the first realtime-capable point cloud indexer. With lidarserv, point cloud data is converted to a visualization-optimized index during the acquisition process. This eliminates the need for any postprocessing, meaning that data can be visualized directly, while it is still being captured. In addition, lidarserv allows the efficient execution of arbitrary queries, including filtering by attribute values, bounding boxes, or levels of resolution. This is useful for analysis purposes or for reducing the amount of data that needs to be transmitted. lidarserv is available as open-source software: https://github.com/igd-geo/lidarserv

Publications:

• Bormann, P., Dorra, T., Stahl, B., & Fellner, D. W. (2022). Real-time Indexing of Point Cloud Data During LiDAR Capture. In Computer Graphics and Visual Computing (CGVC). https://doi.org/10.2312/cgvc.20221173
• Hermann, P., Krämer, M., Dorra, T., & Kuijper, A. (2024). Min-Max Modifiable Nested Octrees M3NO: Indexing Point Clouds with Arbitrary Attributes in Real Time. In Computer Graphics and Visual Computing (CGVC). https://doi.org/10.2312/cgvc.20241235

• Hermann, P., Krämer, M., Dorra, T., & Kuijper, A. (2025). Real-time indexing and visualization of LiDAR point clouds with arbitrary attributes using the M3NO data structure. In Computers & Graphics. https://doi.org/10.1016/j.cag.2025.104254.

The BauDNS project supports the energy-efficient refurbishment of building facades. Our software analyzes a point cloud scan of the existing building and automatically creates a high-precision CAD model. This supports architects in their work, enabling them to start planning without first having to model the building manually on the computer.

Our application uses semantic classification of the point cloud to correctly identify windows, doors and walls. This data is used to establish a logical relationship between the calculated geometries. Afterward, the final model can be exported via an IFC interface for other applications.

The Schwarzwald software can prepare point clouds of any size for web visualization. Schwarzwald is fast and uses existing hardware efficiently through intelligent parallelism. It supports both common frameworks, Potree and Cesium. Schwarzwald is available as open source software: https://github.com/igd-geo/schwarzwald

Publications:

Bormann, P. & Krämer, M. (2020). A System for Fast and Scalable Point Cloud Indexing Using Task Parallelism. In Smart Tools and Apps for Graphics - Eurographics Italian Chapter Conference

To develop our applications for highly efficient data processing, we rely heavily on modern development solutions such as the Rust programming language. In the area of point cloud processing, we have developed pasture, a software library for the development of applications for processing very large point cloud data sets.

pasture is available as open source software: https://github.com/igd-geo/pasture

Learn how to use pasture here: https://igd-geo.github.io/pasture_tutorial/

With our GeoToolbox, city models of any complexity can be converted from the CityGML format to various formats optimized for visualization. For example, city models can be converted to 3D Tiles Format and then displayed in a web application such as Cesium. In addition to simple colored city models, textured models are also supported. An efficient texture atlas is created for these models. Automatic creation of different levels of detail ensures that city models can be displayed on the web with optimum efficiency.

Another use case is conversion to volumetric formats, such as voxels, to represent city models in the video game Minecraft.