Recognition Of Hand-drawn Flowchart

Hand-sketched graphics on-line recognition is one of the research focuses for both national and foreign researchers in th(?) years. The tools of hand-sketched graphics recognition provide designer with the environment in which designers use head-storms to display their designable works. The research of hand-sketched graphics recognition is becoming more and more important. Hand-sketched graphics recognition is a multiple subjects' technique, and relates to computer graphics, Computer Aided Design,pattern recognition,man-machine interaction ect. This paper foucuses on the theory and technich of hand-sketched graphics recognition in aspect of flow chart. Based on the geometrical features of flow chart, the research applies line unit and graphics unit to recognize hand-sketched flow chart.The first step .of stroke's recognition process is the sampling points' pretreatment,ingluding reducing noises and polygonal approximating. Based on the three features: encirclement, inflexion,the ratio of length to width about stroke's external rectangle, strokes are classified roughly. And then different strokes of line-unit have different recognition algorithms.The precondition of graphics unit's recognition is the description of graphics unit's geometrical features. The main research about its is geometrical relation's deducible method and condition between line units of which graphics unit consists. the deformable template matching method is applied to recognize graphics unit. As for arrowhead recognition, some simple means is brought up.The research of hand-sketched flow chart on-line recognition entirely is mainly theoretical. According as the data in the data-base of hand-sketched graphics context, the research applies the rules of knowledge data-base to analyze grammar, to infer the geometrical relation in graphics and to fit the original hand-sketched flow chart to standard geometrical flow chart.The freehand sketch recognition prototype system is developed in the above algorithms and confirmed that recognizing the hand-sketched graphics is simple, rapid and precise. Some of the algorithms have also reference value for other domain's hand-sketched graphics.