Metal Material Is Imprinted Bump Small Class Character Recognition Research

Character recognition is a traditional lesson in the pattern recognition. The recognition of the automobile brand, handwriting and printing word have made very great progress, but the image of the indistinctive color word have not get enough attention than the forenamed, so that its' research productions and applications are far more behind. The reason is that the form of indistinctive color word depends on the protuberant word template that presses on the materials. The representative examples are angle iron, tyre and label. Improving the recognition rate of the indistinctive color word have very great meaning in the application of industry, for example metallurgy, architecture and material, and also have much reference for the automobile brand recognition research and so on.This paper put forward a method that can realize the auto recognition of the protuberant character on angle iron. The paper research on the pretreatment of the protuberant angle iron image and the extraction, classifying of the word combined the optics.At first, this paper adopts many image processing ways to extract single character in the pretreatment such as smooth, filter and bring forward an auto-segmentation way base on the canny operator etc. Second, a very strong adaptive way of 13 point combined the periphery contour that extracts the word's features is being introduced into this paper. At last, the BP network is being used in the recognition of the words.The results prove the way in this paper has strong adaptability.