Thesis supervisor: Tamás Szirányi
Location of studies (in Hungarian): Department of Material Handling and Logistics Systems Abbreviation of location of studies: ALRT
Description of the research topic:
a) Preliminaries: Pedestrian detection in 2D, 3D point cloud analysis and recognition
b) Research goals: Looking at the available 3D models (as Google Street view) and the quickly developing on-board vehicle vision systems, we face the problems of handling big-data and learning systems in 3D.
• On board laser scanner on car-top (e.g. Velodyne LIDAR HDL-64/32/16), and/or multicamera systems for streaming 3D point clouds and panoramic vision (http:// http://velodynelidar.com/), makes it possible to generate high density 3D models of the environment.
Autonomous navigation, including the autonomous vehicles on roads and low-altitude flying vehicles in the air should work in dynamically changing environment. In this changing environment most of the appearing objects are identical or similar to some previous one in similar position/location, others, however, could be new items or some previous objects in very different positions. This area is the topic of intensive research of automotive industry and the technology of intelligent transportation systems. The continuous learning and the scalable object database needs new and sophisticated training methodology; presently, Deep Learning technology may solve similar computation intensive training tasks.
c) Tasks:
• 3D point clouds registered to the images of optical cameras and other sensors;
• Partial recognition of the environmental structures and objects;
• Fusion of different image sources from the same or different vehicles;
• Semantic interpretation of the scanned environment and events;
• 3D models of the environment and possible objects.
d) Equipments: cameras, Lidars and computer systems of the laboratory; flying UAV robot with visual grabbing and positioning instruments.
e) Planned scientific results: Publications in the leading robotics, image processing and autonomous driving journals and conferences. Key-words: robot vision, robot vision based communication, autonomous driving, sensor network, 3D recognition, simultaneous localization and mapping (SLAM), graph theory.
Required language skills: English Further requirements: Expected background: good math, programming in MATLAB and/or C/C++
Number of students who can be accepted: 2
Deadline for application: 2017-10-12
2024. IV. 17. ODT ülés Az ODT következő ülésére 2024. június 14-én, pénteken 10.00 órakor kerül sor a Semmelweis Egyetem Szenátusi termében (Bp. Üllői út 26. I. emelet).
Login
