Title: Balanced feature matching in probabilistic framework and its application on object localisation

Authors: Ying Chen; Chunlu Ai

Addresses: Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, Wuxi city, Jiangsu Province, 214122, China; Key Laboratory of System Control and Information Processing (Ministry of Education), Shanghai Jiaotong University, Shanghai city, 200240, China ' Key Laboratory of Advanced Process Control for Light Industry (Ministry of Education), Jiangnan University, Wuxi city, Jiangsu Province, 214122, China

Abstract: A new algorithm of feature matching is proposed after balancing analysis of adjacency matrix of the matching model in a probabilistic framework. Considering all the interaction of the two feature point sets, a probabilistic model is established and solved using random walks with restart (RWR). To reduce the influence of deformation, and increase the accuracy of feature matching algorithm, a balancing analysis to the adjacency matrix of RWR is taken. Then an efficient method for bidirectional balance is presented, which makes the relevance weight between each two correspondence candidates balanced. The approach considers not only all the correspondence candidates of the two feature point sets, but also the geometrical relation between each pair of candidates. It improves the discriminative and accuracy performance of matching. Compared with other state-of-the-art algorithms, the method is more robust to outliers and geometric deformation, and is accurate in terms of matching rate in various matching applications, such as object localisation.

Keywords: probabilistic modelling; bidirectional balance; random walks with restart; RWR; object localisation; feature matching; outliers; geometric deformation; computer vision.

DOI: 10.1504/IJCAT.2013.052291

International Journal of Computer Applications in Technology, 2013 Vol.46 No.2, pp.91 - 100

Published online: 29 May 2013 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article