Skip to Main content Skip to Navigation

Perception de scène par un système multi-capteurs, application à la navigation dans des environnements d'intérieur structuré

Abstract : This thesis falls within the framework of perceptual navigation. Our goal is to study and design an interpretation system for indoor environment scenes, observed by a multi-sensor system that combines ultrasonic sensors and RGB camera. The proposed system can be used to equip smart device for assisting blind people, or for robot operating in furnished spaces. In a scene interpretation system, the acquisitions made by the sensors present real world restrictions and are affected by imperfections, which should be taken into account rather than ignored. Their inclusion in our interpretation system was carried out by using the possibility theory when modeling the acquired data. The adopted models are possibility distributions. The analysis and interpretation of the acquired scene was then based on this possibilistic knowledge. The navigation aid system proposed in this work describes the surrounding scene according to a simplistic model, and divides the intercepted field by the sensors into three main areas, namely: front, left and right. It provides the user with information about the distance between him and the detected objects, the material rigidity of these objects, as well as their position in the scene (object on the left, object in front, object on the right). The performance of the proposed interpretation system is evaluated using the "NA_System" prototype, developed by the "Cybernics team" of the "CEM_Lab" laboratory of the National Engineering School of Sfax (ENIS). The obtained results are encouraging and show the effectiveness of the possibility theory as a framework for representing data acquired from different sensors. The proposed scene interpretation strategy has been proved to be effective in inte grating information from multiple knowledge sources. In the data processing chain for scene analysis and interpretation, new approaches are proposed, especially in attribute selection, salient object detection, classification, fusion and registration data from two sources.
Document type :
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Monday, January 10, 2022 - 2:13:39 PM
Last modification on : Friday, August 5, 2022 - 2:54:52 PM
Long-term archiving on: : Monday, April 11, 2022 - 11:09:19 PM


Version validated by the jury (STAR)


  • HAL Id : tel-03519333, version 1


Marwa Chakroun. Perception de scène par un système multi-capteurs, application à la navigation dans des environnements d'intérieur structuré. Modélisation et simulation. Ecole nationale supérieure Mines-Télécom Atlantique; École nationale d'ingénieurs de Sfax (Tunisie), 2021. Français. ⟨NNT : 2021IMTA0278⟩. ⟨tel-03519333⟩



Record views


Files downloads