<p>This thesis proposes SoPC (System on a Programmable<br />Chip) architectures for efficient embedding of vison-based<br />localization and obstacle detection tasks in a navigational<br />pipeline on autonomous mobile robots. The obtained<br />results are equivalent or better in comparison to state-ofthe-<br />art. For localization, an efficient hardware architecture<br />that supports EKF-SLAM's local map management with<br />seven-dimensional landmarks in real time is developed.<br />For obstacle detection a novel method of object<br />recognition is proposed - detection by identification<br />framework based on single detection window scale. This<br />framework allows adequate algorithmic precision and<br />execution speeds on embedded hardware platforms.</p> / <p>Ova teza bavi se dizajnom SoPC (engl. System on a<br />Programmable Chip) arhitektura i algoritama za efikasnu<br />implementaciju zadataka lokalizacije i detekcije prepreka<br />baziranih na viziji u kontekstu autonomne robotske<br />navigacije. Za lokalizaciju, razvijena je efikasna<br />računarska arhitektura za EKF-SLAM algoritam, koja<br />podržava skladištenje i obradu sedmodimenzionalnih<br />orijentira lokalne mape u realnom vremenu. Za detekciju<br />prepreka je predložena nova metoda prepoznavanja<br />objekata u slici putem prozora detekcije fiksne<br />dimenzije, koja omogućava veću brzinu izvršavanja<br />algoritma detekcije na namenskim računarskim<br />platformama.</p>
| Identifer | oai:union.ndltd.org:uns.ac.rs/oai:CRISUNS:(BISIS)101781 |
| Date | 02 December 2016 |
| Creators | Tertei Daniel |
| Contributors | Raković Mirko, Devy Michel, Borovac Branislav, Akil Mohamed, Obradović Đorđe, Piat Jonathan |
| Publisher | Univerzitet u Novom Sadu, Fakultet tehničkih nauka u Novom Sadu, University of Novi Sad, Faculty of Technical Sciences at Novi Sad |
| Source Sets | University of Novi Sad |
| Language | English |
| Detected Language | English |
| Type | PhD thesis |
Page generated in 0.0022 seconds