International SPIE Conference: Researchers of Fraunhofer IOF and the TU Ilmenau honored with “Best Paper Awards”

This year’s participation in the SPIE conference "Dimensional Optical Metrology and Inspection for Practical Applications XI" in Orlando was crowned with great success for two research teams from the Fraunhofer IOF and the TU Ilmenau. Both the “Orbbec Best Paper Award” as well as the “Orbbec Best Student Paper Award” were won by the researchers from Jena and Ilmenau for their innovative contributions in the field of optical metrology. 

From April 3 to 7 and after two years of virtual conferences, more than 200 representatives from the fields of optics, imaging systems, lasers, and advanced cameras met in Orlando for the largest technical exhibition in the field of imaging and sensing, “SPIE’s Defense + Commercial Sensing”. One part of the international event was the conference "Dimensional Optical Metrology and Inspection for Practical Applications XI", in which two teams from Fraunhofer IOF and TU Ilmenau participated. This symposium is focused on methods, analyses, and applications of optical metrology. In addition, the latest advances and developments in various industries were presented.

Particularly qualitative and innovative approaches are honored at the conference with the "Orbbec Best Paper Award" and the "Orbbec Best Student Paper Award". In this year the awards made their way home to Germany, more precisely to Jena and Ilmenau: Patrick Dietrich, Florian Siegmund, Christian Bräuer-Burchardt, Stefan Heist (Fraunhofer IOF), Gunther Notni (TU Ilmenau, Fraunhofer IOF) as well as Christina Junger (TU Ilmenau) were able to convince the eight-member program committee with their forward-looking research contributions.

Human-robot interaction booth makes cooperative work possible 

Patrick Dietrich, Florian Siegmund, Christian Bräuer-Burchardt, Stefan Heist und Gunther Notni received the „Orbbec Best Paper Award” for their paper "Human-robot interaction booth with shape-from-silhouette-based real-time proximity sensor”.

The paper discusses the innovative approach of an interaction booth for humans and robots so that they can cooperate spatially and temporally during a joint process section in the production flow. There is an increased safety risk for humans when interacting with a robot due to their immense force and speed. The human-robot interaction booth reduces this risk by constantly monitoring the positions of both parties within the booth and then adjusting the robot’s speed in relation to its proximity to the human. This monitoring is made possible by a multi-camera sensor that works according to the “shape-by-silhouette” principle; a shape reconstruction method that constructs 3D shapes of objects based on their silhouettes. 

Efficient learning methods for robot-assisted manufacturing processes

Furthermore, the “Orbbec Best Student Paper Award” was given to Christina Junger and Gunter Notni. Their paper "Optimisation of a stereo image analysis by densifying the disparity map based on a deep learning stereo matching framework" aimed at optimizing a stereo image analysis by densifying a depth image based on a deep learning stereo matching framework.

Stereovision is used in many applications, e.g., robotic manufacturing processes. Recently, several efficient stereo matching algorithms based on Deep Learning have been developed, including the Adaptive Aggregation Network (AANet/ AANet+) - an end-to-end learning algorithm. Challenges of stereo matching algorithms include objects with low texture or non-cooperative objects. The goal was to develop efficient learning methods for robotic manufacturing processes for cross-domain data streams to improve recognition tasks and process optimization. For this purpose, AANet+ was evaluated for its applicability and efficiency on a test data set with different measurement setups. The comparison with parallel measurement setups was able to show that AANet+ is capapble to robustly recognize texture-poor and optically non-cooperative objects.