Based On Pattern Recognition Of Regional Characteristics And Topology

The abroad application of CAD technology in pattern design has changed traditional pattern design mode and advanced the design efficiency. however, There is new request for pattern recognition because of an amount of pattern sample especially for tie sample which is various. So it is a problem that people can find the tie pattern designed or similar in the condition that the tie sample id provided. If it can do, time waste will be avoided and efficiency will be increased.This paper propounded a pattern recognition algorithm which is based on area features and topology on tie pattern's characteristics of few blocks and regular color pattern. The algorithm sorted the tie pattern using color block's area and space relation as its characteristics, and established corresponding recognition systems, which includes three main modules.(1) Pretreatment parts. It mainly included the selection of sample unit, pattern smoothness, the change from RGB image to gray image, merger of color block and gray transform. Through this part the scan image including lots of yawp changed to color block image and hasn't any relation with color match course.(2) Character abstracting part. An position message included rectangular feature and structure feature were used to describe the area of each block and the space relation of color blocks.(3) Classification part. According to the rule of the least of distance function, the rectangular feature and structure feature were compared with sample in the database separately to classify the pattern.On the base of Database with 200 recording, the recognition method propouned in this paper was prepared with other common methods and we got the conclusion that the recognition method based on area and structure feature is more effective to recognition of tie pattern.