Neural Network Based Steel Coil Printing Number Research On Identification Method

As the marking of steel coil,the number of steel coil spary printing records the production batch number,specification,model number and other product information of steel coil,which is the main basis of product logistics tracking.Therefore,automatic identification of steel coil number is one of the core technology that enterprises need to solve urgently in intelligent production.In this thesis,the optical character automatic recognition(OCR)technology is used to automatically identify the number printed on steel coils.It is designed a comprehensive image preprocessing algorithm combining image gray value processing and image morphology processing in this problem.A rectification algorithm,cylindrical expansion method,is proposed to correct the distortion of the cylinder image.The scanning method is used to cut the single character image in the steel roll number printing image.The feature vector information of single-character image is obtained by counting the direction gradient histogram of single-character image,which is used as the input of number recognition process.This thesis designs a BP neural network for automatic character recognition.According to the character of a single character,neurons in input layer,hidden layer and output layer of BP neural network and the parameter setting of neural network are discussed in detail.The training and recognition process of BP neural network is analyzed,and the character recognition rate is improved by adjusting the network parameters.Finally,the simulation test results of the experiment in this thesis under the MATLAB platform show that the design in this thesis can effectively extract the character information in the steel coil number printing image with high speed,high recognition rate and high practical value.