Printed Words Defect Detection Of Medicinal Glass Bottle Based On The Images Registration

With the continuous development of economy,people's quality of life is constantly improving,and the safety of drugs is more and more concerned by the state.In order to prevent the production date and the production lot number of the drug being modified.State Food and Drug Administration requires the appearance printing detection process of the medicine bottle to be completed by the pharmaceutical packaging manufacturer.At present,pharmaceutical bottle manufacturers have not equipped with the equipment to detect the defects of the bottle,and the detection is completed artificially.The limitation of artificial detection is very large,and it is not able to carry out the tasks of quality guarantee.In this case,a complete set of defects detection system for the printed word of glass bottle is designed.(1)According to the requirements of the market for the printed words detection of medicinal glass bottle,the overall architecture of the system is designed,including mechanical structure unit,image acquisition and processing unit,and the control unit.It is completed that the selection of the PLC,the acquisition equipment and the motor,the design of the light source and the lighting device,and the design of other hardware.(2)According to the principle and common methods of image registration,aiming at the characteristics of the medicine glass bottle printing words defect detection,the feature-based image registration method is selected.In image registration,feature extraction and feature matching are the two most important steps.The Harris corner detection algorithm and the SIFT(Scale Invariant Feature Transform)feature vector with high extraction precision are selected in feature points extraction.In feature points matching,Harris algorithm is matched with the normalized cross correlation method(NCC),the SIFT algorithm is matched with the Euclidean distance,and both of them use RANSAC(RANdom Sample Consensus)out-of-the-box points algorithm to match accurately.By comparing and analyzing the results,the SIFT algorithm with high matching accuracy is selected.In printed words defect detection of medicinal glass bottle,the difference image is obtained by image difference,and then the difference image is binarized to the binary image,finally the isolated points are removed by the eight neighborhoods.By calculating the ratio of the pixels number with the 0 gray value in the isolated points removed image and in the original image,it is determined whether the bottle is qualified,the experiment shows that the method can meet the requirements of actual production.(3)According to the experimental design and hardware selection of the system,the experimental platform of printed words defect detection of medicinal glass bottle is set up,the parameters of the system can be set through the working interface.The selected image processing method is applied to the platform,and the defect of medicinal glass bottle printing word can be detected in real time.The effectiveness of the algorithm is verified by the experimental results,the design goal of the detection system is realized,and detection project.of the system is completed.