Open Ph. D. position (2018-2021)
[bg_collapse view= »link » color= »#004480″ icon= »eye » expand_text= »Show More » collapse_text= »Show Less » ] ContexteUnmanned aerial vehicles (UAV) have made great progress and begun to play an increasingly important role in civilian and military applications recent years, such as aerial reconnaissance, surveillance, and rescue, due to its smaller body, lower power consumption, simpler operation, etc. However, the analysis of video imagery obtained from small UAV platform is always a difficult problem and challenging task for many reasons, including narrow field of view, rapid platform motion, image instability and the relatively small size of interest regions within the received image. A suitable general technique for automatic detection of visually salient regions from received video imagery, such as road people and vehicle, is thus needed in order to assist current human-performed manual analysis, manipulation and decision, as well as provide a wealth of valuable information for object recognition and tracking. During three years, the first objective of the PhD student is to develop algorithms based on deep learning for the saliency detection in the drone videos. In a second time, the PhD student can explore other interesting ways, e.g. the super-resolution image enhancement on the salient regions, region-of-interest image coding based on HEVC, etc. This PhD thesis topic has a strong relationship with our ANR ASTRID project (https://sites.google.com/insa-rennes.fr/projetanrastrid-dissocie) beginning in January 2018. The PhD student will have opportunities to collaborate with the postdocs involved in this project. This thesis is funded by the DGA (Directorate General of Armaments), but no confidentiality is involved. The publications are encouraged. Thesis descriptionUnmanned aerial vehicles (UAV) have made great progress and begun to play an increasingly important role in civilian and military applications recent years, such as aerial reconnaissance, surveillance, and rescue, due to its smaller body, lower power consumption, simpler operation, etc. However, the analysis of video imagery obtained from small UAV platform is always a difficult problem and challenging task for many reasons, including narrow field of view, rapid platform motion, image instability and the relatively small size of interest regions within the received image. A suitable general technique for automatic detection of visually salient regions from received video imagery, such as road people and vehicle, is thus needed in order to assist current human-performed manual analysis, manipulation and decision, as well as provide a wealth of valuable information for object recognition and tracking. This allows 1) bringing comfort to the action of observation and reducing human error (due to the observation fatigue after a long time of observation); 2) amplifying the efficiency of the corresponding operations by better detection (higher speed and precision); 3) facilitating the management of highly requested human resources. During three years, the first objective of the PhD student is to develop algorithms based on deep learning for the saliency detection in the drone videos. In a second time, the PhD student can explore other interesting ways, e.g. the super-resolution image enhancement on the salient regions, region-of-interest image coding based on HEVC, etc. This PhD thesis topic has a strong relationship with our ANR ASTRID project (https://sites.google.com/insa-rennes.fr/projetanrastrid-dissocie) beginning in January 2018. The PhD student will have opportunities to collaborate with the postdocs involved in this project. This thesis is funded by the DGA (Directorate General of Armaments), but no confidentiality is involved. The publications are encouraged. ContactThe PhD will take place jointly at INSA Rennes and IRISA, under the joint supervision of O. Déforges, O. Le Meur and L. ZHANG. Potential candidates are invited to contact them, as soon as possible: Requirements for the candidatesThe candidate should have an MSc degree or equivalent, with good notes. The candidate with a background of deep learning and image processing is preferred. He/She should have the nationality of an EU Member State or Switzerland. Other informationPhD start date: 1/10/2018. [/bg_collapse] |
