Human eye is only sensitive to a small spectrum of light radiation (400-800nm). However, in many cases, perceiving a wider spectrum can be a real opportunity to better understand a scene. Indeed, whether it is for the measurement of physical parameters, or the tracking of targets in poor visibility conditions, additional information on different spectra is a real added value. Nevertheless, some companies offer multi-modal cameras that combine either adaptive filters with a broadband sensor or the combination of several sensors (visible, IR,...). But these systems are passive and only deliver one or more images in different frequency bands.
In this thesis, we aim to develop a smart camera integrating several sensors (Visible, IR, UV) as well as a processing unit to analyze in real time the different spectral images. The objective is to design an intelligent camera with a perception strategy according to the scene and conditions of capture. The first of the scientific challenges of this thesis will be the fusion of images obtained from different sensors with different resolutions. To achieve this, we could consider deep-learning methods that have shown their effectiveness. Unfortunately, these techniques are computationally intensive and are often not compliant with the constraints of an embedded system. This will be the second scientific challenge of this thesis. The final envisaged prototype looks like this type of camera (figure on right).
Keywords: Multimodal vision, Deep learning, Embedded system, Data fusion
This thesis will take place at the Institut Pascal in Clermont-Ferrand and more precisely within the DREAM team (dream.ispr-ip.fr). This team is specialized in embedded vision system integration and has a good skill in deep neural network integration on embedded targets.
Salary and Social Security
Interested candidates should send their application in English and the application should include the following information:
Superviser : François BERRY – firstname.lastname@example.org - https://dream.ispr-ip.fr/members/francois-berry/
Co- Superviser : Omar AIT AIDER – email@example.com - https://comsee.ispr-ip.fr/members/omar-ait-aider/Lue lisää
|Otsikko||PhD position - GhostCam: Making the invisible visible….|
|Job location||IMobS3 Laboratoire d'Excellence - Université Clermont Auvergne 4 Avenue Blaise Pascal TSA 60026 63178 AUBIERE cedex, 63178 AUBIERE|
|Julkaistu||toukokuuta 15, 2019|
|Viimeinen hakupäivä||Tehtävä täytetään mahdollisimman pian|
|Tieteenalat||Tekoäly,   Keinotekoinen neuroverkko,   Tietokonetekniikka,   Sähkötekniikka,   Laskennalliset tieteet,   Koneoppiminen,   Konenäkö,   Kuvan prosessointi  |