作者: Alexey Kostin , Josef Kittler , William Christmas
DOI: 10.1016/J.PATREC.2004.10.020
关键词:
摘要: Object recognition using graph-matching techniques can be viewed as a two-stage process: extracting suitable object primitives from an image and corresponding models, matching graphs constructed these two sets of primitives. In this paper we concentrate mainly on the latter issue graph matching, for which derive technique based probabilistic relaxation labelling. The new method was evaluated standard data sets, SOIL-47 COIL-100, in both objects must recognised variety different views. results indicated that our is comparable with best other current techniques. potential also demonstrated challenging examples cluttered scenes.