Study On Tire Mark Recognition Method Based On Prior Knowledge

Tire sidewall mark record parameters of a tire, it will mislead consumers if the markis mistaken, and then it may cause accidents. So in order to ensure that the surface of thetire mark comply with the design requirements before leaving factory, the manufactureoften produce a test tire to check up before mass production. Most of enterprises, test tire isinspect by human. Since the artificial test is time consuming and inaccuracy, it is difficultto adapt to the needs of large-scale production. As a result, the automatic test technology isneed for all producers. In the automatic test system, mark recognition is the core problemof testing, and the recognition rate affect the effect of test system directly.This paper mainly studies the method of tire mark recognition based on priorknowledge model. Firstly, analyses the design scheme of prior knowledge and presents thewhole frame of prior knowledge library; then the segmented region features are extractedby using the SIFT algorithm, completing establishment the standard of regional featuredatabase in prior knowledge library, and gives the region matching method; and then givescharacter feature extraction method based on the principal component analysis method;finally, use the character reconstruction method to recognize character, and improved thismethod, giving the recognition process using character recognition information database.This paper constructs a prior knowledge library, reducing identification difficult,and achieved good results; using the SIFT feature extraction method to match the markregion, and improving the accuracy of the mark region location; using the method ofprincipal component analysis in the character feature extraction, avoiding the effect ofbinaryzation; in character recognition, improving the reconstruction recognition method,improving the recognition accuracy. In the future, the scheme can adopt different recognition remodels in different categories of mark, so that the generality of the testmodel will greatly increase.