Fakultät Maschinenbau
Refine
Document Type
- Article (2)
- conference proceeding (article) (2)
- Part of a Book (1)
Has Fulltext
- no (5)
Is part of the Bibliography
- no (5) (remove)
Keywords
- 3D depth camera (1)
- 3D scanning (1)
- Collaboration (1)
- Evaluation methods (1)
- Extrinsic error (1)
- Feature detection (1)
- Gesture recognition (1)
- Hough transform (1)
- Human-generated ground truth (1)
- Image processing (1)
Institute
Begutachtungsstatus
- peer-reviewed (3)
Lines are one of the basic features that are used to characterise the content of an image and to detect objects. Unlike edges or segmented blobs, lines are not only an accumulation of certain feature pixels but can also be described in an easy and exact mathematical way. Besides a lot of different detection methods, the Hough transform has gained much attention in recent years. With increasing processing power and continuous development, computer vision algorithms get more powerful with respect to speed, robustness and accuracy. But there still arise problems when searching for the best parameters for an algorithm or when characterising and evaluating the results of feature detection tasks. It is often difficult to estimate the accuracy of an algorithm and the influences of the parameter selection. Highly interdependent parameters and preprocessing steps continually lead to only hardly comprehensible results. Therefore, instead of pure trial and error and subjective ratings, a systematic assessment with a hard, numerical evaluation criterion is suggested. The paper at hand deals with the latter ones by using a human-generated ground truth to approach the problem. Thereby, the accuracy of the surveyed Kernel-based Hough transform algorithm was improved by a factor of three. These results are used for the tracking of cylindrical markers and to reconstruct their spatial arrangement for a biomedical research application.
As camera and projector hardware gets more and more affordable and software algorithms more sophisticated, the area of application for camera-projector configurations widens its scope. Unlike for sole camera calibration, only few comparative surveys for projector calibration methods exist. Therefore, in this paper, two readily available algorithms for the calibration of those arrays are studied and methods for the evaluation of the results are proposed. Additionally, statistical evaluations under consideration of different influencing factors like the hardware arrangement, the number of input images or the calibration target characteristics on the accuracy of the calibration results are performed. Ground truth comparison data is realized through a robotic system and structured light 3D scanning.
In smart manufacturing environments robots collaborate with human operators as peers. They even share the same working space and time. An intuitive interaction with different input modalities is decisive to reduce workload and training periods for collaboration. We introduce our interaction system that is able to recognize gestures, actions and objects in a typical smart working scenario. As key aspect, this article considers an empirical investigation of input modalities (touch, gesture), individual differences (performance, recognition rate, previous knowledge) and boundary conditions (level of automation) on user experience. Therefore, answers from 31 participants within two experiments are collected. We show that the arrangement of the human-robot collaboration (input modalities, boundary conditions) has a significant effect on user experience in real-world environments. This effect and the individual differences between participants can be measured utilizing recognition rates and standardized usability questionnaires.
Tightening quality requirements of industrial products involving manual assembly lead to the development of assisting workbenches with integrated functions to support workers performing these manual tasks. This contribution discusses a new approach to learning transitions of a finite state automaton representing the sequence of work tasks based on the video stream of a 3D depth camera. Preprocessed video data is fed into a three-stage classification scheme based on support vector machines. The results of the classification are then related to the state automation to trigger state transitions indicating the completion of a specific work task and the start of the next one. The proposed approach has been evaluated at an industrial assembly process of moderate complexity and shows very robust results with respect to disturbances caused by inaccurate object classification.
Steigende Anforderungen an die Qualität von zum Teil manuell gefertigten Produkten führen dazu, dass Handarbeitsplätze mit Assistenzsystemen für die Unterstützung der am Arbeitsplatz arbeitenden Mitarbeiterinnen und Mitarbeiter ausgestattet werden. Der Beitrag beschreibt einen neuen Ansatz, um mittels Verfahren des maschinellen Lernens die Objekterkennung sowie die Transitionen eines, den Arbeitsprozess repräsentierenden Zustandsautomaten eines solchen Systems einzulernen. Hierfür werden nach einer Vorverarbeitung Daten aus einer Tiefenkamera in drei Stufen durch Support Vector Machines (SVM) klassifiziert und das Ergebnis mit dem Zustandsautomaten verknüpft. Das Konzept wird an einem industriellen Montageprozess überschaubarer Komplexität evaluiert; es zeigt gute Ergebnisse hinsichtlich der Robustheit gegenüber Fehlern bei der Objektklassifikation.