Machine learning for machine vision systems is of growing importance to deal with the variability of industrial production. Learning systems are particularly required in the field of surface inspection and quality control. The final goal is to proceed from the detection of defective parts to avoiding defects by closing the feedback loop to the production process.

Adaptive components in machine vision system have reached a different level of technological maturity:

1. Camera systems for the acquisition of one or multiple images have already reached a high level of maturity. This also includes the pre-processing of the raw data. The TPScan and FScan technologies developed by PROFACTOR are examples of such integrated camera-based systems.

2. For detection and evaluation of the defects a set of machine learning tools has been developed, some of which are already in use by industry. Nevertheless there is a need for further development of machine learning methods used in conjunction with machine visions. Current challenges includes:

  • enhancing classification methods to increase their interpretability.
  • improving the usability of machine learning for end –user. They need to be able to understand the behaviour of machine learning systems, e.g. when introducing new defect classes.
  • reducing the effort needed for training. This can be achieve through active learning approaches that request user input only for selected samples.

3. Closing the feedback loop from quality control to the process is still a major challenge. This is particularly true for complex quality control systems, such as machine vision.

Closing the feedback loop requires data-driven models that are able to propose changes to process parameters that aim at avoiding defective parts. The models need to be continuously updated through incremental learning, that takes place in parallel to the normal production process. The selection of which samples are used for the trainings process and which ones are better left out, requires methods such as drift detection to assess the validity of the current models.

Projects

The factors of success in the context of the digital factory are the inclusion of human within the manufacturing process as well as the consideration of their individuality and experience. Assistive systems act an important role in this field. The challenges of a nearby collaboration of human and ro...+
Initial situation: Efficient production of carbon composite parts is an important topic for aerospace, automotive and other industries. In a draping process, carbon fibre textiles are shaped in order to produce parts with complex shape. This draping process results in complex deformation of carbon f...+
Die automatische Prüfung von Composite-Bauteilen gewinnt sowohl  in der Automobilindustrie als auch in der Luftfahrt zunehmend an Bedeutung. Während es bei den Produktionsverfahren substantielle Fortschritte gegeben hat, wird die Prüfung immer noch manuell durchgeführt, nimmt aber 30-50 Prozent...+
The INLINE project aims at the solution of key challenges to enable the implementation of a scalable manufacturing process for fuel cell systems. Current manufacturing processes rely on manual work that has substantial limits in terms of cycle times, costs and scalability. Developments will start wi...+
Holzfurnieroberflächen spielen als Leichtbau-Dekorteile für die Luftfahrtindustrie eine wichtige Rolle. Diese Oberflächen sind hochglänzend lackiert, deren Qualität wird im Rahmen der Abnahme von den Erstausrüstern (OEM) kritisch bewertet. Dabei werden Messgeräte eingesetzt, die Kennwerte üb...+
Non-destructive testing of components is an important auxiliary process step, not only in quality control but also in regular maintenance. The detection of cracks is currently done by using magnetic particle inspection, which is a decades-old, inefficient and ecologically undesirable process. There ...+
Machine vision in industrial quality control, e.g. in surface inspection, generates an enormous amount of data. These data are input for machine learning structures that reproduce human decision making. The setup and optimization of such machine learning structures in industrial environments require...+
The EU research project DARWIN aims to develop a robotic system with cognitive abilities, that include acting, learning and thinking. The robotic system also aims to develop an understanding of its environment together with the physical properties of the components to be handled. In order to underst...+

Publications

Heidl, S. Thumfart, E. Lughofer, C. Eitzinger, E. P. Klement; Machine Learning Based Analysis of Gender Differences in Visual Inspection Decision Making, Information Sciences, Vol. 224, pages 62-76, DOI: 10.1016/j.ins.2012.09.054, Mar 2013

Dittrich, T. Riklin-Raviv, G. Kasprian, R. Donner, P.C.Brugger, D. Prayer, G. Langs; A Spatio-Temporal Latent Atlas for Semi-Supervised Learning of Fetal Brain Segmentations and Morphological Age Estimation, Accepted for publication in Medical Image Analysis, 2013

Elkharraz, S. Thumfart, D. Akay, C. Eitzinger, B. Henson; Tactile texture features corresponding to human affective responses. Submitted to IEEE Transactions on Affective Computing

Heidl, S. Thumfart, E. Lughofer, C. Eitzinger, E. P. Klement; Machine Learning Based Analysis of Gender Differences in Visual Inspection Decision Making,Information Sciences, accepted, pre-press DOI: 10.1016/j.ins.2012.09.054

Grünauer, S. Zambal, K. Bühler; „Detektion von Koronararterien: Das Beste aus zwei Welten“, Bildverarbeitung für die Medizin (BVM):pp. 269-273, 2011

van Beilen, H. B ult, R. Renken, M. Stieger, S. T humfart, F. Cornelissen, V. Kooijman; Effects of Visual Priming on Taste-Odor Interaction, PLoS ONE 6(9): e23857, 2011, doi:10.1371/journal.pone.0023857

Heidl, C. Eitzinger, M. Gyimesi, F. Breitenecker; Learning over Sets with Recurrent Neural Networks: An Empirical Categorization of Aggregation
Functions, Mathematics and Computers in Simulation 82(3), pp. 442-449, doi:10.1016/j.matcom.2010.10.018, Nov 2011

Thumfart, R. H.A.H. Jacobs, E. Lughofer, C. Eitzinger, F. W. Cornelissen, W. Groissboeck, R. Richter, “Modeling human aesthetic perception of visual textures “, ACM Transactions on Applied Perception, Volume 8, Issue 5, Nov. 2011, doi:10.1145/2043603.2043609

Heidl, C. Eitzinger, M. Gyimesi, F. Breitenecker; Learning over Sets with Recurrent Neural Networks: An Empirical Categorization of Aggregation Functions, Mathematics and Computers in Simulation, ISSN 0378-4754, 2010

Groissboeck, E. Lughofer, S. Thumfart; Associating Visual Textures with Human Perceptions using Genetic Algorithms, Information Sciences, vol. 180, issue 11, pp. 2065-2084, doi:10.1016/j.ins.2010.01.035, 2010

H.A.H. Jacobs, R. Renken, S. Thumfart, F. W. Cornelissen; Different Judgments about Visual Textures Invoke Different Eye Movement Patterns, Journal of Eye Movement Research, 3(4):2, pp. 1-13, 2010

Tran, C. Eitzinger; ThermoBot – autonomous robotic system for thermographic detection of cracks. Workshop Proceedings of IAS-13, 13th Intl.Conf.on Intelligent Autonomous Systems, Padova (Italy) July 15-19,2014, ISBN 978-88-95872-06-3, pp.391-391

Eitzinger, S. Akkaladevi; Dexterous Assembler Robot Working with Embodied Intelligence, Workshop Proceedings of IAS-13, 13th Intl.Conf.on Intelligent Autonomous Systems, Padova (Italy) July 15-19,2014, ISBN 978-88-95872-06-3, pp.393-393

Eitzinger, K. Zhou; VALERI – Validation of Advanced, Collaborative Robotics for Industrial Applications. Workshop Proceedings of IAS-13, 13th Intl.Conf.on Intelligent Autonomous Systems, Padova (Italy) July 15-19,2014, ISBN 978-88-95872-06-3, pp.392-392

Eitzinger, A. Baghbanpourasl, S. Zambal; Image Processing Issues in Scanning Inspection Robots. Workshop Proceedings of IAS-13, 13th Intl.Conf.on Intelligent Autonomous Systems, Padova (Italy) July 15-19,2014, ISBN 978-88-95872-06-3, pp.394-402

Traxler, P. Thanner, G. Mahler; Temporal analysis for implicit compensation of local variations of emission coefficient applied for laser induced crack checking, 12th International Conference on Quantitative Infrared Thermography, Bordeaux, France, 7th-11th July 2014

Dittrich, T. Riklin-Raviv, G. Kasprian, R. Donner, P.C.Brugger, D. Prayer, G. Langs. A Spatio; Temporal Latent Atlas for Semi-Supervised Learning of Fetal Brain Segmentations and Morphological Age Estimation, Medical Image Analysis, Vol. 18(1), pp. 9-21, January 2014.

Alexander Walch, Christian Eitzinger; A combined calibration of 2D and 3D sensors, Proceedings of the VISAPP. 9th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, Lisbon, Portugal, 5th-8th Jan. 2014

Traxler, P. Thanner, P. Meyer Heyer; Design of and practical experience with a thermographic crack checking system using laser heating, 11th European conference on NDT, 2014 10 09 Prag, ISBN: 978-80-214-5018-9 by Brno University of Technology, http://www.ndt.net/events/ECNDT2014/app/content/Paper/166_Traxler.pdf

Traxler; Unterdrückung des Emissionsgradeinflusses in der Laser angeregten Rissprüfung, Tagungsband der ÖGfTh (Österreichische Gesellschaft für Thermografie), 26.9.2014 Eugendorf/Austria

Walch, C. Eitzinger. A combined calibration of 2D and 3D sensors, Proceedings of the VISAPP. 9th International Joint Conference on Computer Vision, Imaging and Computer Graphics Theory and Applications, Lisbon, Portugal, 5th-8th Jan. 2014

Heidl, S. Thumfart, and C. Eitzinger, Humans Differ; So Should Models. Systematic Differences Call for Per-Subject Modeling, ICAART 2012: Proceedings of the 4th Int. Conf. on Agents and Artificial Intelligence, pages 413-418, Vilamoura, Portugal, February 6th-8th, 2012

Heidl, S. Thumfart, E. Lughofer, C. Eitzinger, E. P. Klement; Classifier-based analysis of visual inspection: Gender differences in decision-making, Proc. of SMC 2010, IEEE Conference on Systems, Man and Cybernetics, pp. 113-120, Istanbul, Turkey, October 2010

Thumfart, J. Scharinger, C. Eitzinger; Pixel based Texture Mixing, Proc. of the 34th Workshop of the Austrian Association for Pattern Recognition, pp. 147-154, Zwettl, Austria, May 27-28th 2010

Henson, G. Elkharraz, S. Thumfart, D. Akay, C. Eitzinger; Machine vision approach to predicting affective properties of tactile textures, In Proceedings of the International Conference on Kansei Engineering and Emotion Research, KEER 2010, Paris, France, March 2- 4, ISBN 978-4-9905104-0-4, pp. 2261 – 2270, 2010.

Thumfart, W. Palfinger, M Stöger, C. Eitzinger; Accurate Fibre Orientation Measurement for Carbon Fibre Surfaces, accepted for presentation at CAIP 2013, York, UK, Aug 27-29th, 2013

Eitzinger, S.Ghidoni, E. Menegatti; ThermoBot: towards semi-autonomous, thermographic detection of crack, Proc. of the International Conference on Heating by Electromagnetic Source HES-13, pp. 461-468, Padua, May 21-24, 2013

Eitzinger, PROFACTOR, Steyr-Gleink, Österreich, G. Mahler, InfraTec, Dresden; Konzeption und Aufbau einer robotergestützten Plattform für optisch angeregte Wärmefluss-Thermografie. Presented at DGZFP, Thermographie-Kolloquium 2013, 26. – 27. September 2013, Leinfelden-Echterdingen

Traxler, PROFACTOR, Steyr-Gleink, Österreich, S. Koch, Institut Dr. Foerster, Reutlingen; Inline-Prüfung von warmgewalzten Stahlknüppeln mittels Wärmeflussthermographie, Presented at DGZFP, Thermographie-Kolloquium 2013, 26. – 27. September 2013, Leinfelden-Echterdingen

Thanner, G. Traxler, Design for Thermographic Crack Checking System using Laser Induced Heat Flux Technology, Presented at Factory Automation Conference 2012, Veszprem, Hungary, 21-22 May 2012 Proceedings of Factory Automation 2012, pages 122-125, Veszprem, Hungary

Thumfart, W. Palfinger, C. Eitzinger; Vision based sensors enabling automated production of composite material. In the Proc. of SAMPE / SEMAT 2012, Munich, May 24th – 25th, pp. 301 – 306, ISBN: 978-3-952 3565-6-2, 2012

Eitzinger, S. Thumfart: Optimizing Feature Calculation in Adaptive Machine Vision Systems, M. Sayed-Mouchaweh and E. Lughofer (eds.), Learning in Non-Stationary Environments: Methods and Applications, DOI 10.1007/978-1-4419-8020-5_13, Springer Science+Business Media New York 2012

S.Thumfart, PhD Thesis: Genetic Texture Synthesis. Johannes Kepler University Linz, Department of Computational Perception, Feb 2012

Dittrich; Ein Atlas der frühen Gehirnentwicklung. Published online at ORF Science, July 2013

Thanner: “Defect Avoidance, Machine-vision system catches defects in seamless steel tube production using linescan cameras and nearinfrared imaging“, Vision Systems Design (VSD) Magazin, 1.6. 2010

Wögerer, P. Thanner, G. Traxler: “Measurement of Material properties with Thermography“, FACTORY AUTOMATION 2011 Conference, Györ, Hungary, 24-26 May 2011

Wögerer, P. Thanner, G. Traxler: „Thermografic methods for online control for steel pipes“, FACTORY AUTOMATION 2011 Conference, Györ, Hungary, 24-26 May 2011

Petra Thanner „Mülltrennung mit Infrarottechnologie“, Newsletter E!AT aktuell, März 2010

Thumfart; “Pixel based Texture Mixing“, ÖAGM 2010 – 34th annual workshop of the Austrian Association for Pattern Recognition (AAPR) – Computer Vision in a Global Society, Zwettl, 28. Mai 2010

Thanner, W. Palfinger, “Qualitätssicherung von Carbonfaserteilen mit Bildverarbeitung
Handhabungstechnik – Der Schlüssel für eine automatisierte Herstellung von Composite-Bauteilen, Augsburg, 8. Juli 2010

Thanner, W. Palfinger, G. Traxler, “Wärmeflussauswertung für die induktiv angeregte Rissprüfung“, Thermografieforum Eugendorf, Eugendorf, 10. September 2010

Eitzinger; “Adaptive Produktion“, 25 Jahre Eureka, Linz, 7. Oktober 2010

Thanner; “EM80 – OIDIPUS, Optimized InGaAS Detectors for Imaging Applications and Industrial Spectroscopy“, 25 Jahre Eureka, Linz, 7. Oktober 2010

Heidl; “Classifier – based analysis of visual inspection: Gender differences in decision-making“, SMC2010, IEEE International Conference on Systems, Man and Cybernetics, Istanbul, 11. Oktober 2010

Thanner, G. Traxler; “Advanced Evaluation for Thermographic Crack Detection with Inductive Excitation for Steele Billets“, 20th Manufuturing Confernece, Budapest, 20. Oktober 2010

Traxler; „Automatisierte Inline-Prüfmöglichkeit mit aktiver Thermographie“, Seminar Wärmefluss-Thermographie, Erlangen, 4. November 2010

Petra Thanner; Defect Avoidance, Machine-vision system catches defects in seamless steel tube production using linescan cameras and near-infrared imaging,Vision Systems Design (VSD) Magazi, 1.6. 2010

Thanner, G. Traxler; Qualitätssicherung von Carbonfaserteilen mittels Bildverarbeitung, 8. Juli 2010, Handhabungstechnik – Der Schlüssel für eine automatisierte Herstellung von Composite-Bauteilen, Augsburg, Germany

Your Contact

Dr. Christian Eitzinger
Head of Machine Vision

+43 7252 885 250
christian.eitzinger@nullprofactor.at

We answer…

… your Questions