Leveraging robust signatures for mobile robot semantic localization

JAVIER REDOLFI; JORGE SÁNCHEZ

Roma

Conferencia; CLEF 2012 Conference and Labs of the Evaluation Forum; 2012

Univ. La Sapienza, Roma

This paper describes the participation of the CIII UTN FRCteam in the ImageCLEF 2012 Robot Vision Challenge. The challengewas focused on the problem of visual place classification in indoor en-vironments. During the competition, participants were asked to classifyimages according to the room in which they were acquired, using theinformation provided by RGB and depth images only. We based our ap-proach on the Fisher Vector representation ?a robust signature recentlyproposed in the literature? and the use of efficient linear classifiers. Inorder to exploit the information provided by different information chan-nels, we adopted a simple fusion strategy and generated classificationscores for each image in the sequence. Two tasks were proposed duringthe competition: in the first, images had to be classified independentlyof one another while, in the second, it was possible to exploit the tem-poral continuity of the stream. For the first task, we adopted a simplethreshold based classification scheme. For the second, we considered theclassification of groups of images instead of single frames. These groups,i.e. temporal segments, were automatically generated based on the visualsimilarity of the images in the sequence. Our team ranked first on bothtasks, showing the effectiveness of the proposed schemes.