Find a copy online
Links to this item
theses.fr Accès au texte intégral
Find a copy in the library
Finding libraries that hold this item...
Details
Genre/Form: | Thèses et écrits académiques |
---|---|
Material Type: | Document, Thesis/dissertation, Internet resource |
Document Type: | Internet Resource, Computer File |
All Authors / Contributors: |
Emmanuel Schmitt; Patrick Charpentier; Vincent Bombardier; Université de Nancy I.; École doctorale IAEM Lorraine - Informatique, Automatique, Électronique - Électrotechnique, Mathématiques de Lorraine.; Centre de recherche en automatique (Nancy). |
OCLC Number: | 494474067 |
Notes: | Titre provenant de l'écran-titre. |
Description: | 1 online resource |
Responsibility: | Emmanuel Schmitt ; sous la direction de Patrick Charpentier et de Vincent Bombardier. |
Abstract:
Our works concern the study of a fuzzy sensor adapted to the color identification on wooden boards. The proposed method aims particularly to take into account the human subjectivity concerning the color perception and to provide results in the user vocabulary. The application field (timber industry) imposes certain constraints. Indeed, the color classes are not separated (non-strict boundaries) and are represented with few samples. Then, it results from it, imprecisions and uncertainties in the output class definition. After a state of the art on the techniques of image processing, of pattern recognition, and on the structure of intelligent sensors, our works are exposed along two axes: from the sensors to the measurements, and from the measurements to the decision. Firstly, we have evaluated and corrected the perturbations linked to the environment of the sensor (temperature, ageing, ...). Then, we have determined the most discriminating colorimetric space, and generated the characteristic vector composed of interpretable attributes allowing to identify the colors. From these data, we have developed the Fuzzy Reasoning Classifier based on a fuzzy linguistic rule mechanism aggregating conjunctive rules according to the Larsen model. Finally, for the multi-sensor systems, a fuzzy operator of data merging is used. The use of this fuzzy sensor has shown the good behaviour of the system compared to the real-time industrial constraints, and an improvement of 10% of the recognition rates.
Reviews

