Algorithm Design And Application Research Of Label Defect Detection System

With the continuous development of the social economy,people pay more attention to the composition,origin,date of manufacture and expiration date of the product.Above all important information are recorded on labels.Labels are affected by various factors during the printing process,causing some quality problems.There are some spot defects,line defects,surface defects caused by multiple printing or less printing.Therefore,the quality detection of label is particularly important.This subject developed a set of automatic detection system for printed label quality,according to the actual use requirements.The system effectively avoided the problems of misdetection and missed detection,which was caused by artificial subjective factors.Therefore,the product detection rate was improved and the labels quality was ensured.This paper mainly completes the following work:1)The paper studied the defect detection theory about convolutional neural network and Caffe deep learning framework.On this basis,the overall design of the system combination with the detection requirements was presented.It mainly included requirements analysis,technical route implementation,hardware equipment research and selection,and software module design.2)Based on the research of principal component analysis(PCA)and support vector machine(SVM),a PCA-SVM label defect detection algorithm was proposed.The detection model was constructed by normalizing the data set and inputting it into the model to train and obtain the optimization parameters.The experimental results showed that the algorithm didn't need image preprocess and complex template creation.It was simple to implement and adaptable.3)In order to meet the requirements of higher recognition accuracy and recognition efficiency,an improved AlexNet network was proposed follow the Caffe framework.In the Linux system,the C++ language was used to build the network model,and then the recognition accuracy and recognition efficiency were compared and analyzed.It is concluded that the improved algorithm model has higherrecognition accuracy,and the model uses the original image data in the training process to enhance the scientificalness and objectivity of the data,improve the generalization ability of the model,which makes the application simple,and better meet the actual needs.4)Based on the Visual Studio 2015 development platform and OpenCV visual library,the software of the detection system was developed.Then the detection performance of the system was tested on the build test platform.The result showed that the accuracy rate of detection system met 96%,and the detection efficiency was297ms/p,which met actual production needs.In addition,a more intuitive human-computer interaction interface has been developed by Qt,which made the label defect detection system more perfect,and more convenient.