Algorithm Research Of The Steel Billet Characters Recognition System On The Production Line

Optical character recognition technology(OCR)refers to turn the image from electronic equipment(such as a scanner or digital camera)into black and white dot matrix image file,and then get the information we want using some special features in the image,so that the image is transformed into a computer language.Optical character recognition technology is an important component of artificial intelligence and pattern recognition.In nowadays,this technology has been paid more and more attention and become popular in many fields,such as license plate number recognition,fingerprint recognition,face recognition,and has a huge influence in the military.The technology in this paper is a kind of character recognition based on the developing artificial intelligence in the steel industry.Characters recognition generally include pretreatment,characters positioning,characters segmentation,and characters recognition.This paper mainly studies the problems of positioning and segmentation of character recognition system in the production line of steel billet.We present a new method of positioning and segmentation because of the influence of temperature and light.Lay a good foundation of character recognition.The study in this paper is based on the project of heavy rail steel production in Wuhan Iron and Steel(Group)Company.Every billet number must be record in the production line.The billet number has three lines of white character;the target of the character recognition is to replace human resources into computer operations.There are a lot of influences in the recognition in the production line,it is hard to solve it.Aiming at these difficulties,we improved the previous character segmentation and location method,and proposed a new positioning method based on variance feature in this paper.The new judgement basis was the feature variance of the eight-character linear arrangement rule and the character height width ratio,which was unique in each steel billet.Screening the connected domain with this variance,the steel billet number could be located.With the location information of the connected domain,the character tilt could be rectified.Besides,a new circulation mechanism was added in addition to thecharacter projection segmentation.According to the number of the billet character,a new segmentation algorithm based on double termination condition is proposed.This method solved the problem of missing character,merged character and provided a good foundation for character recognition.Furthermore,a recognition method combined with feature recognition and template recognition was put forward.And we also proposed a recognition method based on the character corrosion and spot detection.The whole system is proposed to be analyzed and refined at the end of this article.The experiment shows that the correct rate of improved method is more than 95%,with stability.