Study On Tire Surface Mark Character Segmentation

Before mass production, tire manufacturers usually produce a specimen to checkwhether the tire mark is correct. If some marks are not correct or not consistent with thetire, it can cause bad influence of using the wrong tire. It takes about40minutes to identifyall the tire marks on one tire accurately. It is necessary to develop a computer-aidedidentification system to improve the accuracy and the efficiency. The segmentation of tiremark regions is the basis of the identification system. It is important to make a rightsegmentation before identification.In this paper, the partial images captured by industrial camera are used as raw images.After doing pretreatment such as graying and denoising, images are ready to be processed.Considering about the feature of images to be processed, the Otsu threshold segmentationand the expansion in morphology are used to extract an arc on the image. According to thecoordinates of points on the arc, center and radius are calculated through out the leastsquare method. Coarse positioning of the tire mark regions is done by the center and radius.Then existing binary algorithms are compared, and a local adaptive threshold, QuickThresholding Algorithm, is selected to binary the tire mark regions images. The arcprojection is used to split the rows of multiple lines of characters. The radius projection isused to split the characters of single line of characters.Firstly, in this paper, arc scan and radius scan are used to solve the question that tiremarks are arrayed arc-shaped. Common scan methods are not suitable for this question.Secondly, the other question, which foreground and background are too similar to separate,is solved by using the local adaptive threshold, Quick Thresholding Algorithm. Finally,segmentation based on arc projection and radius projection is achieved the desired effect.In the further researching, the binary algorithm is an important orientation.