Recognition of On-line Hand Sketched Complex Diagrams

Abstract: The spread and availability of pen-based devices increased the research in free handwriting and hand sketched recognition. Most of systems in hand sketched diagram recognition are limited to a specific application or limited to specific shapes. Complex diagrams recognition requires not only understanding the primitive shapes but also determining relations between these shapes. This work presents a system for recognition of hand-sketched complex diagrams which create active and intelligent relations between different diagram components. A multi-layers framework is developed which consists of stroke layer, shape layer, and diagram layer. A distance function is used to measure the similarity between template shapes and user sketched shapes. In the diagram layer, many adjustments are applied to enhance the positions and sizes of shapes, increasing the harmony and regularity of shapes arrangement. This system has been tested using a set of primitive shapes and a set of complex diagram and high accuracy results have been obtained.

Keywords: Diagram recognition, free hand sketch, feature extraction, pattern recognition.