Study On The Recognition Technique Of Tire Surface Character Segmentation

Verifying the accuracy of the pre-production sample tire surface identification isvery important. It is not only affect the normal use, but also lead to severe accidents iftire surface identification did not consistantly with the actual information. It will takeas much as40minutes to human checking the tire surface identification currently. Ifapply machine-aided cognition to checking, the efficiency and accuracy of tire surfaceidentification will be improved Significant. As basis of machine-aided cognition tiresurface identification, the accuracy of tire surface Character Segmentation will be ofgreat importance.In this paper, the industrial camera captures the partial images as raw images. Thedeal with the raw images with Morphological dilation processing and Otsu thresholdsegmentation method to extract the inherent arc, then calculate the center of the arc vialeast square method. We use the arc and the center to rough positioning of the tiresurface identification in order to get d image region that need further segmentation.Then quick threshold self-adaptive threshold algorithm is selected to binary the tiremark regions images. After selected image noise elimination, the arc projection is usedto split the rows of multiple lines of characters. The radius projection is used to splitthe characters of single line of characters.In this paper, radial and arc directional scan methods are selected to handleidentifications which are distributed by arc on the tire surface. This can resolve pooradaptability problem via common scan methods. And then, use partial self-adaptivethreshold algorithm to resolve problem that foreground and background of theidentification are too similar to separate. Finally, use radius and arc projection mode toachieve single character segmentation.