Building facades consist of a combination of different geometric structures. Large wall surfaces are often approximately planar, windows and doors form openings, balconies and cornices project outward from the plane, while loggias and recesses lie behind it.

Building edges occur where the spatial orientation of a surface changes or a façade plane ends. The AI reliably detects these discontinuities and converts them into geometrically stable CAD lines.

Balconies and loggias are three-dimensional structures that differ spatially from the main facade plane. The depth information of the point cloud enables automatic delineation of these elements. The detected geometries are spatially assigned to the respective facade surface.

The transformation from point cloud to usable CAD geometry follows a structured process with several sequential processing steps. Each step builds on the previous one, enabling controlled, traceable processing.

Point clouds often contain more than just the building itself. Other objects are also captured during the scanning process. These structures are automatically classified as much as possible and separated from the actual facade geometry.

The main surfaces of the building are analyzed geometrically. Through plane fitting, normal vectors, and spatial continuity, a geometric base model of the building envelope is created — the foundation for all subsequent recognition steps.

The detected point groups are converted into CAD-ready geometries. Depending on the element, different basic geometric types are created that can be used directly in CAD and BIM systems.

Pure object detection is not enough for professional CAD output. The generated geometries must be topologically consistent — lines must be continuous, surfaces closed, and duplicate objects removed. Only after this cleanup is the result directly usable in CAD software.

In addition to classic 2D facade views, the detected geometry can also be built in three dimensions. Planar facade surfaces are positioned spatially, and windows, doors, and other elements are assigned to their respective surfaces. Depending on the use case, different levels of modeling detail are possible.

Depending on the project, the detected geometries can then be assigned to structured BIM objects. This creates a reliable geometric foundation for further BIM processes.

The required modeling depth is defined before the project begins. This ensures that not every detail is modeled unnecessarily — only the information needed for the actual use case is generated. This targeted approach reduces effort and costs.

An existing building facade is rarely perfectly flat, vertical, or right-angled. For renovation and prefabrication, these deviations can be of critical importance – an idealized CAD model alone is not sufficient here.

Our solution can therefore not only generate idealized CAD geometries. It can also compare the actual surface with a geometric reference and precisely quantify deviations.

For new facade systems, it is critical to know how far the existing surface projects outward or recedes inward relative to a planned reference plane. These distances can be calculated directly from the point cloud — across the entire surface and with high spatial resolution.

Scan-to-CAD is particularly relevant for projects where facade elements are industrially prefabricated. Production requires reliable information about the actual as-built condition — not the planned target condition.

Point clouds only contain areas that were actually captured by the sensor. In practice, various objects can obstruct parts of a facade. Reflective or transparent surfaces also produce problematic measurement values.

Depending on the project, different capture methods can be combined to reduce occlusions and gather additional information. The choice of capture method influences coverage, point density, and available supplementary data.

Transparent quality assurance is essential for building facade detection as well. For each detected element, various quality metrics can be calculated. This ensures no uncontrolled AI output — uncertain areas can be specifically flagged for review.

The AI handles repetitive analysis and digitization work. Professional oversight remains an integral part of the process. This approach combines the scalability of automated processing with the reliability of human expertise.

Three-dimensional output of the detected facade geometry as a direct foundation for further construction, clash detection, or visualization. The spatial positioning of all elements is fully preserved.

The point cloud is retained as a geometric reference. Upon request, a classified point cloud can also be provided, in which each point is assigned a semantic class – such as wall, window, balcony, vegetation, or ground.

The classified point cloud enables visual quality control and can serve as the basis for further automated processing steps.

Optionally, quality information can be delivered alongside the geometric output. The QA data enables targeted post-processing and documents the reliability level of each detected element.

Point clouds of buildings can contain detailed information about architecture, construction, access points, and technical installations. Protecting this information is a core component of our solution.

Customer data is used exclusively for the agreed-upon processing. Use for training or further development of AI models does not occur without the customer's explicit consent.

Automated processing can be comprehensively documented. This ensures that at any time it remains traceable how the final CAD or BIM result was derived from the point cloud — a critical aspect for auditability, quality assurance, and project documentation.

Depending on the project and IT infrastructure, the solution can be operated with varying security requirements. The technical measures can be adapted to the organizational and regulatory requirements of the customer.

Modern laser scanners capture a building in a short amount of time with enormous detail. The decisive next step is to automatically convert this data into usable planning information. Our AI does not turn the point cloud into a mere visualization — it generates structured CAD- and BIM-ready geometries from it.