Combinaison de caméras thermique et couleur pour la segmentation cibles/arrière-plan en environnement non contrôlé
|Advisor:||Maldague, Xavier; Prévost, Donald|
|Abstract:||Most existing video monitoring systems only use the visible range of the electromagnetic spectrum. Moreover, approaches proposed until now for automated target detection and tracking are often too computationally expensive for real-time applications and suffer significant performance degradation in outdoor environment. The purpose of this thesis is to make the most of thermal and visible information in order to improve efficiency of detection and tracking of moving target in uncontrolled environment. Within the framework of this project, only one motionless acquisition node is considered. A node includes a LWIR thermal camera (8 - 12 m) and a color electro-optical camera. Three main technical challenges are tackled: development of an hybrid thermal / visible acquisition platform optimized for video monitoring applications in outdoor environment, segmentation of moving targets in uncontrolled environment, and finally, combination of thermal and color information. For every aspect, special attention is addressed to computational efficiency to reach a minimum processing rate of 10 frames per second. It is worth to mention that this thesis has mainly been oriented along needs of INO's Vision program, conferring a very concrete and practical aspect to the research. Hence, developed algorithms have been continually assessed and improved in the context of real applications. Notions related to thermal / visible images registration have lead to development and manufacturing of the INO's VIRXCam camera, while detection algorithms are the grounding elements of INO's AWARE video monitoring library, used since 2007 by an INO's partner interested in positioning of moving targets on airport aprons, taxiways and runways. This thesis regroups most notions to be considered in the development of an outdoor video monitoring system. Both hardware and software aspects are addressed. We hope that this document will be used as a work of reference for those interested in the development of algorithms devoted to real-time detection of moving targets in uncontrolled environment, or interested in the design of hybrid thermal/electro-optical acquisition platform.|
|Document Type:||Thèse de doctorat|
|Open Access Date:||18 April 2018|
|Collection:||Thèses et mémoires|
All documents in CorpusUL are protected by Copyright Act of Canada.