Image Database for Pose Hypotheses Generation

• Main Author: Nyqvist, Hanna
• Format: Others
• Language: English
• Published: Linköpings universitet, Reglerteknik 2012
• Subjects: database; loop closure; real-time; SLAM; visual images
• Online Access: http://urn.kb.se/resolve?urn=urn:nbn:se:liu:diva-81618

The presence of autonomous systems is becoming more and more common in today’s society.The contexts in which these kind of systems appear are numerous and the variations arelarge, from large and complex systems like autonomous mining platforms to smaller, moreeveryday useful systems like the self-guided vacuum cleaner. It is essential for a completelyself-supported mobile robot placed in unknown, dynamic or unstructured environments tobe able to localise itself and find its way through maps. This localisation problem is stillnot completely solved although the idea of completely autonomous systems arose in thehuman society centuries ago. Its complexity makes it a wide-spread field of reasearch evenin present days. In this work, the localisation problem is approached with an appearance based method forplace recognition. The objective is to develop an algorithm for fast pose hypotheses generationfrom a map. A database containing very low resolution images from urban environmentsis built and very short image retrieval times are made possible by application of imagedimension reduction. The evaluation of the database shows that it has real time potential becausea set of pose hypotheses can be generated in 3-25 hundreds of a second depending onthe tuning of the database. The probability of finding a correct pose suggestion among thegenerated hypotheses is as high as 87%, even when only a few hypotheses are retrieved fromthe database.