Robust 3-D object recognition and pose estimation using 2-D image sequences
| Autor: | R. Otterbach |
|---|---|
| Rok vydání: | 1995 |
| Předmět: |
business.industry
Computer science 3D single-object recognition ComputingMethodologies_IMAGEPROCESSINGANDCOMPUTERVISION Cognitive neuroscience of visual object recognition 3D pose estimation Extended Kalman filter Feature (computer vision) Computer Science::Computer Vision and Pattern Recognition Segmentation Computer vision Artificial intelligence Cluster analysis business Pose |
| Zdroj: | Informatik aktuell ISBN: 9783540602934 DAGM-Symposium |
| Popis: | The paper presents a system for the recognition and pose estimation of 3-D objects, which relies on the analysis of 2-D image sequences. Based on feature correspondences in subsequent images an Extended Kalman filter recursively estimates 3-D contour images of the observed objects. In order to reduce the search complexity and the noise sensitivity, the recognition process is built on robust, contour-based 2-D algorithms. These techniques apply because of the previous segmentation of the 3-D contour image into plane curves. By pairwise matching of model and image contours hypotheses for the object’s pose are obtained. The verification computes globally consistent assignments of model and image features by combining similar pose hypotheses. Both the segmentation and the verification task are formulated as clustering problems and solved by means of a common algorithm in transformation space. With regard to industrial applications most importance has been attached to the modular design of the software and the experimental evaluation. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|