Automated quality control of reflective surfaces

The quality control of reflective surfaces in production pushes conventional measurement technology to its limits. Alternatives such as manual visual inspections or technically complex testing systems have disadvantages: the former are time-consuming and subjective; the latter are characterized by high costs and low flexibility. These weaknesses are addressed by the Fraunhofer Institutes IAIS and IPT in the project AKIBO. The innovative testing system to be implemented in the project is characterized by its modular, scalable design with low-cost hardware. This is made possible by combining traditional image processing and Artificial Intelligence.

© Fraunhofer IAIS
The arc-shaped scanner detects surface damage, e.g., on car bodies.
© Fraunhofer IAIS
Dr. Theresa Bick is developing the technology in a miniature replica.
© Fraunhofer IAIS
The system is suitable, for example, for the detection of paint inclusions or hail damage.

The automation of manual visual inspection of defects on component surfaces in industrial production has so far only been achieved by very complex solutions. These require high investments both in acquisition and maintenance and place high demands on the location (e.g., shielded rooms for robot arms). In addition, defects on highly reflective surfaces can only be detected by these systems with difficulty or not at all. The system developed by the Fraunhofer institutes IAIS and IPT, on the other hand, offers low hardware and maintenance costs, can be used flexibly, and can be retrofitted into existing production facilities. Among other things, the solution is capable of inspecting reflective surfaces such as painted components, even working under the influence of scattered light (e.g., ceiling lighting in a hall).

Classical image processing in combination with Artificial Intelligence

The special characteristic of our technology is the combination of classical image processing and Artificial Intelligence. The system uses the advantages from both worlds – fast, approximate algorithms from classical image processing and the powerful methods of Deep Learning. The classical image processing methods are used for data preprocessing. For the detection and classification of surface anomalies, we rely on Convolutional Neural Networks (CNNs). This enables the system to learn different defect features (scratches, dents, paint gradients, ...) and thus to distinguish and localize them.

Incorporating expert knowledge with Informed Machine Learning

A particular challenge in the application of Deep Learning in industrial production is the acquisition collection of annotated training data. For this purpose, we rely on Informed Machine Learning: by incorporating expert knowledge, the system is able to make the data-hungry neural networks feasible. The amount of training data and the annotation effort can be kept comparatively low. Due to the latest developments of our system, not only highly reflective but also diffusely reflective surfaces can be tested. Thus, the system can be used for a wide variety of test objects. Currently, defects of a size of 0.1mm on 1m component size can be detected. Industrial use is currently being tested in the project based on various use cases in cooperation with partners from production. The goal is to increase the technological maturity of the system.

Individual solutions for individual challenges

Is your company facing the challenge of automating and reliably implementing your quality control of reflective surfaces? We offer a cost-effective, flexible inspection system that is suitable for both highly and diffusely reflective surfaces. We look forward to getting in touch with you and working on a solution together.

Video: The intelligent quality control of reflective surfaces

Privacy warning

With the click on the play button an external video from is loaded and started. Your data is possible transferred and stored to third party. Do not start the video if you disagree. Find more about the youtube privacy statement under the following link:

The video shows the intelligent quality control of reflective surfaces in action. Dr. Theresa Bick, one of the developers, also talks about the underlying components, the uncomplicated application possibilities and the advantages of the AI-based technology.

Project info

This project is funded

by the Fraunhofer CCIT Research Center Machine Learning.