Mobile museum guide based on fast SIFT recognition

Abstract

This article explores the feasibility of a market-ready, mobile pattern recognition system based on the latest findings in the field of object recognition and currently available hardware and network technology. More precisely, an innovative, mobile museum guide system is presented, which enables camera phones to recognize paintings in art galleries. After careful examination, the algorithms Scale-Invariant Feature Transform (SIFT) and Speeded Up Robust Features (SURF) were found most promising for this goal. Consequently, both have been integrated in a fully implemented prototype system and their performance has been thoroughly evaluated under realistic conditions. In order to speed up the matching process for finding the corresponding sample in the feature database, an approximation to Nearest Neighbor Search was investigated. The k-means based clustering approach was found to significantly improve the computational time. © 2010 Springer-Verlag Berlin Heidelberg.