Photoelectric Control System For Billet Recongnition

With the mergers and acquisitions performance of WISCO and Baosteel,overcapacity crisis will be effectively alleviated.However,the production must be affected by cutting worker,so the efficiency of automation products will play an important role in the production line.In the heavy rail steel production line,steel billet characters represent different types,which need to enter the furnace.Once there are errors in the recognition,they will bring certain risks to product.Before this,the identification number of steel billet mainly depends on manual identification,and then reported to huge industrial control room by radio.It brings huge pressure to the workers,and the production efficiency is not high,therefore the billet recognition system came into being.In this paper,because the billet character is printed or handwritten,instead of neat marking machine production,the identification process is particularly difficult.For example,uneven illumination,inclinations and stickiness,make the detection process difficult.The existing character recognition algorithms can not effectively identify such characters,and the non-standard character recognition technology encounters bottlenecks.In view of the above problems,this paper has done some research on billet image preprocessing and character recognition.This paper mainly completed the following works:Firstly,preprocessing the collected billet image.As the light of the production line is dark and uneven gray image of the collected billet,this paper first balanced image processing to avoid light caused by subsequent processing errors..Then,according to the characteristics of the billet image,an improved filtering algorithm is adopted.The algorithm effectively removes the noise of the image and achieves the effect of preserving the target information.Secondly,billet character recognition algorithm.After the image preprocessing is finished,we extract the image edges.However,the single direction edge detectionalgorithm can not achieve the desired effect,so some improvements are made on the basis of the traditional algorithm.On the basis of preserving the original image information,the four connected domain method is used to find the seed points twice,and the billet image is quickly scanned to reduce the processing time of the billet image in order to meet the production requirement.For the case of sticky characters in the project,a multi-stick character segmentation algorithm based on statistical discrimination is proposed in this used to connect the broken target information.Then refine the processing to extract the skeleton.Finally we deburring,to avoid image distortion.After analyzing the advantages and disadvantages of the current recognition algorithm,this paper adopts a method that combine region template matching method and character feature extraction.It improve the recognition rate and recognition time.In the experiment,each step has been done to ensure the effectiveness of the algorithm.The experimental results show that the equalization adopted in this paper can effectively avoid the influence of illumination problems.The improved median filtering method has a significant effect on image noise processing.The proposed algorithm is based on statistical discrimination which can effective separated the adhesion character,and with good robustness.