Symbol Recognition Using a Concept Lattice of Graphical Patterns

Abstract. In this paper we propose a new approach to recognize symbols by the use of a concept lattice. We propose to build a concept lattice in terms of graphical patterns. Each model symbol is decomposed in a set of composing graphical patterns taken as primitives. Each one of these primitives is described by boundary moment invariants. The obtained concept lattice relates which symbolic patterns compose a given graphical symbol. A Hasse diagram is derived from the context and is used to recognize symbols affected by noise. We present some preliminary results over a variation of the dataset of symbols from the GREC 2005 symbol recognition contest.

• Type: Conference
• Year: 2009
• Where: GREC
• Authors: Marçal Rusiñol, Karell Bertet, Jean-Marc Ogier, Josep Lladós
• Keywords: Boundary Moment Invariants, Computer Graphics, Concept Lattice, Graphical Patterns, GREC 2009