Integrating vocabulary clustering with spatial relations for symbol recognition

Abstract

This paper develops a structural symbol recognition method with integrated statistical features. It applies spatial organization descriptors to the identified shape features within a fixed visual vocabulary that compose a symbol. It builds an attributed relational graph expressing the spatial relations between those visual vocabulary elements. In order to adapt the chosen vocabulary features to multiple and possible specialized contexts, we study the pertinence of unsupervised clustering to capture significant shape variations within a vocabulary class and thus refine the discriminative power of the method. This unsupervised clustering relies on cross-validation between several different cluster indices. The resulting approach is capable of determining part of the pertinent vocabulary and significantly increases recognition results with respect to the state-of-the-art. It is experimentally validated on complex electrical wiring diagram symbols.

RSILC: Rotation- and Scale-Invariant, Line-based Color-aware descriptor

“…For example, the authors in [43,44] integrate spatial distribution of key-points by using shape context with texture features for food classification. In a similar fashion, spatial relations between the visual primitives (such as circles and corners) are integrated with statistical shape features for graphics recognition [45].…”

Local Enlacement Histograms for Historical Drop Caps Style Recognition

“…In the field of symbol recognition, the works of [9], [10] introduced bags-of-relations (BoR), an original way to produce vocabularies of spatial relations. The approach was applied on a well-controlled set of visual primitives specific to the application domain (e.g., circles, corners or extremities of symbols).…”

Bags of spatial relations and shapes features for structural object description

“…In the field of symbol recognition, recent works [21], [22] introduced bags-of-relations (BoR), an original way to produce composite vocabularies of spatial configurations. The approach was applied on a well-controlled set of visual primitives specific to the application domain (e.g., circles, corners or extremities).…”