CD Laboratory for People and Object Surface Authentication

This CD Laboratory investigates various approaches for the most accurate and efficient recognition of persons and objects in contexts as diverse as the authentication of event visitors and the detection of counterfeit branded products.

In many previous scenarios, biometric person recognition and the recognition of physical objects based on their (micro) surface structure differ fundamentally, even though both are based on image processing and computer vision methods and have their roots in pattern recognition and machine learning: Typically, the former is used to recognise different instances of humans (i.e. different individuals), for example for access control at sports stadiums or amusement park rides, while the latter tends to differentiate between completely different types of objects (such as cars and aeroplanes, or cars and the road they are driving on).

However, when use cases such as distinguishing between original and counterfeit branded products are added, object recognition becomes more complex, as it now has to compare two instances of the same object type. And while here the result is at least a binary classification as ‘original’ or ‘counterfeit,’ it is even more challenging to compare different instances of the same product within an automated production chain.

The CD Laboratory is therefore conducting research in various directions to contribute to the further development of both person and object recognition. In the former area, for example, more elegant options for admission to sports stadiums are to be made possible: Instead of long waiting times, during which visitors have to queue at automated gates and then be granted access one by one via facial recognition, visitors will be visually tracked while queuing so that facial recognition is already complete by the time they reach the gate – significantly speeding up the process.

For this approach, facial recognition must therefore be integrated into a multi-sensory, multi-face tracking system: In order to find the right balance between minimal computing power and decision-making delays on the one hand and maximum recognition accuracy on the other, various innovative approaches are being implemented and compared in the CD Laboratory. For research in the field of object recognition, on the one hand, a deliberately highly realistic setup is used with smartphone cameras: This should enable end users to verify whether a product is original or counterfeit (which can be potentially life-threatening, especially in the field of medicine) using their own mobile phones. However, extensive research is also needed to ensure that mobile phone cameras with their integrated significant post-processing of acquired imagery can perform this complex task. On the other hand, visual surface structure of human teeth will be investigated and procedures will be developed to discriminate different teeth of a single person and teeth of different persons by considering their visual surface structures.

The work of the CD Laboratory will therefore benefit a wide range of people and economic and industrial sectors – from production companies to consumers, from event visitors to organisers – in terms of convenience, efficiency and security.