BREADER: AN INTERACTIVE MACHINE VISION SYSTEM FOR RECOGNITION OF GEOMETRICAL AND MATHEMATICAL STRUCTURES

摘要

We describe a system that uses image processing and computer vision techniques to discover and recognize geometric structures and mathematical symbols from images. The basic geometric constructions recognized include straight lines, polylines and circles. Interceptions between basic geometric construction are dealt with appropriated connectivity relation. Basic geometries may be aggregated to create composed objects. The system starts by segmenting the image in clusters of pixels. Mathematical (and other) symbols are recognized by extracting Fourier-like features of clusters, and using a simple classification algorithm. Mathematical structures are built by applying composition rules to a topological graph. The recognized object structures can be exported to a variety of vector graphic languages and type-setting systems, such as SVG. The recognition and image processing pipeline is highly configurable and can rely on an interactive GUI. This allows parameters to be set and object exporting options to be selected. The system is built in a modular way such that adding new types of object recognizers or expression composition rules is sim-ple.