Digital Twin Germany: Scalable LiDAR Classification for a National 3D Infrastructure

Airborne LiDAR data acquisition and classification form a foundational component of this national infrastructure.

Germany's Digital Twin Germany initiative is the most ambitious national geospatial programme in Europe. Led by the Federal Agency for Cartography and Geodesy (BKG), it has evolved from a feasibility study between 2020 and 2022 into full-scale implementation under the "DigiZ DE" project — with a clear objective: to build a consistent, highly precise, and interoperable 3D analysis platform covering the entire federal territory, aligned with broader European efforts such as Destination Earth.

The scale of that project is considerable. To make nationwide project tractable, production has been divided into multiple lots, each with defined geographic scope and technical requirements. BSF Swissphoto is responsible for Lot 2 — a substantial portion of the programme encompassing more than 63,000 km², including the islands of the Baltic Sea. The lot operates within a precisely defined perimeter: a hard boundary at the Polish border ensures no data is processed beyond the national limit, while a one-tile buffer is maintained throughout production to guarantee seamless coverage at every edge.

Lot 2 primarily includes the federal states of Mecklenburg-Western Pomerania (Mecklenburg-Vorpommern) and parts of Brandenburg.

Airborne LiDAR data acquisition and classification form a foundational component of this national infrastructure. High-density, standardised point cloud data enables consistent terrain modelling, object extraction, and analytical use cases across Germany.

Within this framework, BSF Swissphoto collaborates with Pointly for large-scale AI-driven classification.

The Use Case: AI-Supported Classification at Nationwide Scale

The Digital Twin Germany LiDAR project is a monumental task, covering a total federal territory of approximately 356,794 km². Within this national framework, a critical performance target was set: achieving a classification accuracy of over 95%. This high benchmark serves as the official quality requirement for data acceptance by the BKG.

For BSF Swissphoto, the integration of Pointly’s AI workflow has been essential in meeting these industrial-scale demands. As of early 2026, they have successfully classified over 34,000 km² within Lot 2. This represents more than half of their total assigned project area (approx. 63,000 km²), with all data reliably meeting the 95% accuracy threshold.

Data Density and Class Catalogue

The airborne LiDAR data is acquired at a density of approximately 40 points per square metre, enabling detailed identification of infrastructure and land cover elements. The class catalogue is described as complex and includes special object classes such as:

• High-voltage power lines

• Wind turbines

• Solar panels

These categories go beyond basic terrain and vegetation classes and reflect the analytical needs of a national digital twin platform.

Workflow Integration and Processing Performance

To meet the project’s demanding national-scale datasets and strict delivery timelines, the workflow developed with Pointly utilizes a sophisticated hybrid approach. This system seamlessly integrates traditional rule-based classification with AI-driven deep learning models, which are further refined through iterative training cycles. To ensure the highest data quality, the process incorporates advanced post-processing routines that significantly reduce the need for human intervention, requiring only minimal manual correction before final delivery.

Several AI training iterations were conducted in close coordination with the project team to ensure alignment with project-specific target structures. The objective was to achieve automatic classification requiring only limited manual intervention, which has been achieved according to BSF Swissphoto.

Processing performance is a defining feature of the workflow. The scalable cloud-based system enables:

• Processing speeds exceeding 3,000 km² within 24 hours

• Classification of a standard production block (7,687 tiles) within 3–4 days

This scalability ensures that classification throughput keeps pace with airborne acquisition and downstream modelling requirements.

An additional operational element is the integration of the Pointly API into BSF Swissphoto’s existing workflows, allowing seamless data exchange and automation within established production chains. This integration supports industrial-scale repeatability rather than isolated project-based processing

Quality Assurance and Model

Adaptability National digital twin projects require a high degree of consistency across diverse environments, ranging from varied terrain types and coastal zones to industrial areas and rural landscapes. To address these complexities, the classification system incorporates specialized deep learning training procedures and iterative model refinement. This approach is further enhanced by project-specific tuning to meet defined class structures and the use of transfer learning techniques, which allow models to adapt seamlessly to evolving requirements. Ultimately, the achieved accuracy level of over 95% reflects a successful balance between high-speed automation efficiency and rigorous quality control.

Strategic Relevance for Digital Twin Germany

In the Digital Twin Germany project, classification is not a peripheral processing step — it is the structural backbone that makes advanced geospatial applications possible. Accurate, automated categorization of point cloud data unlocks precise 3D object modeling, enables detailed analysis of infrastructure and energy assets, and supports the environmental and landuse simulations that planners and policymakers increasingly depend on. Without it, interoperability with European digital twin frameworks remains aspirational rather than achievable.

The collaboration between BSF Swissphoto and Pointly demonstrates what this looks like in practice: high-density airborne acquisition combined with scalable AI-driven workflows, capable of delivering consistent, high-accuracy results at the pace and scale a nationwide programme demands.