RT Journal Article
T1 Online Context-based Object Recognition for Mobile Robots
A1 Ruiz-Sarmiento, José Raúl
A1 Guenther, Martin
A1 Galindo-Andrades, Cipriano
A1 González-Jiménez, Antonio Javier
A1 Hertzberg, Joachim
K1 Robots autónomos
AB This work proposes a robotic object recognitionsystem that takes advantage of the contextual information latentin human-like environments in an online fashion. To fully leveragecontext, it is needed perceptual information from (at least) aportion of the scene containing the objects of interest, which couldnot be entirely covered by just an one-shot sensor observation.Information from a larger portion of the scenario could stillbe considered by progressively registering observations, but thisapproach experiences difficulties under some circumstances, e.g.limited and heavily demanded computational resources, dynamicenvironments, etc. Instead of this, the proposed recognitionsystem relies on an anchoring process for the fast registrationand propagation of objects’ features and locations beyond thecurrent sensor frustum. In this way, the system builds a graphbasedworld model containing the objects in the scenario (bothin the current and previously perceived shots), which is exploitedby a Probabilistic Graphical Model (PGM) in order to leveragecontextual information during recognition. We also propose anovel way to include the outcome of local object recognitionmethods in the PGM, which results in a decrease in the usuallyhigh CRF learning complexity. A demonstration of our proposalhas been conducted employing a dataset captured by a mobilerobot from restaurant-like settings, showing promising results.
YR 2017
FD 2017-03-31
LK http://hdl.handle.net/10630/13408
UL http://hdl.handle.net/10630/13408
LA eng
NO Universidad de Málaga. Campus de Excelencia Internacional Andalucía Tech.
DS RIUMA. Repositorio Institucional de la Universidad de Málaga
RD 25 ene 2026