Intelligent quality control of reflective surfaces

Automated inspection of damage and defects

Previous solutions for visual quality control in industrial production are very complex to set up, require high investments and place high demands on the location (e.g. shielded rooms for robot arms).

Our Fraunhofer system for quality control of glossy or diffusely reflecting surfaces works completely automated and needs less than one minute for surface inspection. It categorizes the quality defects it finds using Deep Learning. The combination of deflectometry, i.e. the contact-free detection of reflective surfaces, conventional image recognition processes and Artificial Intelligence (AI) methods make our system so unique.

Intelligent surface inspection in three steps

Flexible design and easy integration

The AI-based visual quality control features a simple, mobile and cost-efficient hardware setup that does not require any special safety precautions. The system can operate independently of ambient light. This provides you with an easy integration option into existing productions or even mobile deployability - wherever you happen to need the system.

One hundred percent test coverage and our real-time analysis make it possible to quickly detect quality defects, such as paint inclusions, in industrial production and take appropriate action early in the production process. Our system is also ideally suited for the inspection and analysis of hail damage, e.g. for automotive appraisers.

Individually adaptable for your application

We would be happy to implement our surface inspection system, consisting of scanner, video processing and AI module, for your individual application. We also offer you the retrofitting of an AI module for your already existing image processing solution. In addition, we will be happy to advise you on optimizations to your existing system. Contact us and arrange your non-binding demo appointment.

Our surface inspection system - your advantages

Low hardware and maintenance costs

Retrofittable structure without safety precautions; mobile usable

100 percent test coverage

Artificial intelligence: Various error features can be trained in a self-learning manner


No expert knowledge required for operation

Individual adaptation of the components for your application

AI-based visual quality control in three steps

© Fraunhofer IAIS

1. Capture the object by video

The reflective object to be examined is passed under a scanner while being recorded with a video camera.

Accurate recording of the reflected scanner functions independently of changing light conditions. Special shielding of the system from external influences is therefore not necessary.

This even allows our technology to be used in halls or workshops for hail damage detection, for example.


2. 2D surface reconstruction

The video is immediately evaluated and the surface of the object is reconstructed in two dimensions. For this we use methods from classical image processing, which are characterized by their short runtime.

The component shape of the object does not have to be known for this.



3. Detection of the damage with Deep Learning

The damage is detected and classified on the basis of the 2D reconstruction using Artificial Intelligence methods (Deep Learning).

Various defect features, such as scratches, paint defects or dents down to the smallest dust inclusions, can be trained in a self-learning manner.

The defect features can be flexibly adapted and extended to your individual quality criteria.