Research Center Machine Learning

Image-based Quality Control

Machine Learning applications for automatic damage detection can improve production processes and reduce cost and time for quality control. Integrating expert knowledge or data from simulations or physical processes into the learning mechanism, we develop an image-based detection system for damage from hailstorms on vehicles. Insurance companies and their appraisers face the challenge of having to assess a large number of cases within a short time span. To facilitate the process, a mobile unit scans the car body’s damaged parts. Afterwards, our algorithms detect, classify, and measure the damages automatically. The technology is applicable to other areas of industrial quality control and damage assessment where smooth and reflective surfaces are produced, processed, or tested.

The image-based quality control system detects damages to the surface of a vehicle caused by hailstroms.