Research On Rapid Defect Detection And Dynamic Counting System Of Beer Production Line

With the rapid development of the beer industry,in order to meet the high standards of modern industrial production lines,it is necessary to improve the original beer production line.The quality of beer packaging and the counting of beer bottles are important links in beer production and sales.Regarding the quality of beer packaging,such as misprints,missing prints,and blurs on the printing pattern of beer caps,as well as damage to the bottle mouth of beer bottles.For beer bottle counting,the traditional photoelectric method can only count beer bottles on a single lane.In response to market demand,a beer production line combining machine vision and image processing was designed based on the inspection of beer caps and bottle bottle mouths on the industrial production line and the counting of beer bottles on multiple production lines.Using digital images as carriers,Fast defect detection and dynamic counting system.According to the actual industrial needs,the research of the following algorithms is mainly completed:(1)Defect detection of beer cap pattern printing.Determine whether there are printed patterns on the beer cap and whether there are obvious defects in the patterns.Combining industrial real-time performance,an image feature statistical method is proposed to identify and detect the types of defects on the printing pattern of beer caps.At the same time,it is tested and compared with the SVM(Support Vector Machine)classifier and Alexnet model based on the feature parameters extracted in this paper,and the experimental results verify that the proposed method has better recognition rate.(2)Defect detection of beer bottle mouth.Determine whether there are obvious defects in the mouth of the beer bottle.In the real-time system,there is no need to mark the specific location of the beer bottle mouth defect.for this type of defect detection,it is proposed to perform image preprocessing and segmentation,then obtain the center of the beer bottle mouth by projection,and then perform the connected component of the target ring in the segmented image.The area area statistics and judgment to achieve bottle neck defect detection.At the same time,compared with the Alexnet classifier,the experimental results verify that the method has better recognition rate.(3)Beer bottle counting on multiple production lines.For the actual problem of inaccurate beer bottle counting in multiple production lines in the industry.An image segmentation method combining two-dimensional Otsu and EM(Expectation Maximization)algorithms and a beer bottle dynamic counting algorithm for multiple production lines are proposed.In the beer bottle image data set with complex background,the improved image segmentation method is used to segment the image data set toaccurately segment the objects in the image.Using the improved counting algorithm can improve counting accuracy and operating efficiency.For the segmented image,the hough transform and feature-based matching tracking algorithm are used to determine the center of each beer bottle end face and the pixel distance between two consecutive frames in the two frames before and after.Finally,the dynamic counting of beer bottles in multiple production lines is realized.Through counting experiments on beer bottle images collected from actual industrial multiple production lines,the experimental results show that the counting method is as accurate as 100%.Through many times in the industrial production site,on-site image acquisition,image processing and other operations,and the establishment of a friendly GUI(Graphical User Interface)interactive interface,easy to display the results after operation.Through experiments,we know that the machine vision-based beer production line rapid defect detection and dynamic counting system are designed,and the system is tested accordingly.The results show that the detection system can basically meet the needs of the production line.