Subproject: Automatic image analysis for damage detection


Visually recognizable damage to the surface of a structure is an indicator of possible impairment of load-bearing capacity and serviceability. The time-consuming and cost-intensive manual inspection of structures can be significantly supported by image analysis. Automated methods allow the processing of a large number of high-resolution images, add objectivity to the inspection process and enable complete and standardized documentation.

Project objectives

In this subproject, high-resolution images from building surveys are analyzed automatically. The main focus will be on crack detection on concrete surfaces. However, in the further development of the machine learning methods investigated, care must be taken to ensure that transfer to the detection of other anomalies on building surfaces, e.g. rust plumes, is relatively easy.
The anomalies detected in the image must be appropriately located and made visible on the object surface both for the purpose of interactive visualization in the cloud-based software environment and for final documentation and evaluation. For this purpose, the sensor orientation determined by Structure-from-Motion (SfM) can be used to directly relate individual image contents to the reconstructed 3D object model by means of raycasting.


An image database of annotated cracks is created as a training basis for supervised machine learning, allowing successive training and a continuous increase in recognition rates. A supervised machine learning method with a network architecture specifically adapted for crack analysis is provided.
In addition, a learned crack model directly applicable for detection is developed. The automatic anomaly detection algorithms will be integrated into the cloud-based software platform and made available for structural inspection along with 3D location of damage.

Project duration

     06/2020 – 09/2022


Projekt partners

Funding program

The project supported by the Free State of Thuringia was co-financed by funds of the European Union within the framework of the European Regional Development Fund (ERDF).