Vadstena: Automatic LOD2 Building Extraction

By Martina Bekrová
August 4, 2020

LOD2 building models is now one of the outputs for every dataset processed by Vadstena 3D Reality Capture. The LOD2 includes footprint, the roof shape, and height of reconstructed buildings. In addition our model now includes trees modelled using spherical harmonics; those will be described in a next blog post. Detailed LOD2 is useful for urban analysis such as estimation of solar exposure, classifying building types, or urban planning.

The Level of Detail is a vital concept for 3D city modelling. It indicates to what extent the 3D model corresponds with the real counterpart. The LOD0 is the simplest; it describes the footprint of a building and is represented by a 2D polygon. Consequent LODs are improving in terms of the complexity of objects in the geometric and semantic sense.

Melowntech’s LOD2 output is geospatially accurate and obtained fully automatically using Vadstena. It is extracted from the 3D mesh and AI-based land-cover. To extract the buildings, we start with the land cover to discover building positions and utilizing the 3D information we find the shape of particular building parts. Then we post-process the result to make the buildings fit more together and to reality.

The LOD2 gives accurate information about building position, height, footprint area, roof and shape. LOD2 buildings can be used for further automatic processing or visualization and navigation. Additional, advantage of LOD2 compared to 3D mesh, is data size because LOD2 data is a fraction of the 3D mesh.

Have a look at our recent results of the automatic LOD2 building extraction.

Bordeaux, France


Lucern, Switzerland


Melowntech’s vision is to create a digital counterpart to the real world. To achieve this vision, we are using Vadstena a fully automatic 3D natural and urban landscape reconstruction software system. Melowntech succeeds in remodeling the world accurately in the form of photorealistic 3D meshes, orthophotos, a digital terrain model and more, however, our long term goal is to describe the world fully semantically. AI-based Vadstena land-cover is now a stable part of Vadstena’ output and the first semantic description of the scene. LOD2 extraction is another step in how to achieve the full semantic world description.

Martina Bekrová, Machine Learning Specialist, Vadstena

Behind the pretty pictures, there is a real world. Every human can easily tell a house from a tree. Martina is delving deep into deep learning in order to pass these skills on to Vadstena 3D reality-capture system, to discover the meaning behind pixel values.