Research On Defect Detection Of Syringe Screen Printing Based On Machine Vision

Medical appliance manufacturing is the key to standardization and efficiency in the pharmaceutical industry,and the most common medical appliance is the syringe.The silkscreen,as part of the syringe pillar,represents its own information.Most of the silkscreen is key information such as numbers,scales and characters with special implications,so the printing quality of the silkscreen will directly determine whether the syringe is usable or not.The traditional defect detection algorithm for the surface information of flat objects has been comparatively mature,but the detection algorithm for the surface information of columnar objects is not mature,and the columnar surface defect detection algorithm is not fully applicable to different columnar objects.The traditional surface information detection algorithm for flat objects does not take into account the information characteristics on the three-dimensional space with the information acquisition method on the three-dimensional space,then it cannot be applied directly on the syringe.Therefore,the study of a machine vision-based detection algorithm for syringe defects has a strong application value.In this thesis,the main research object is syringe,and the method of syringe column surface screen printing defect detection is explored.Compared with other columnar items columnar surface labels,the syringe columnar surface silkscreen features studied in this system are fewer,containing only key information such as numbers,scales,characters,etc.In view of the existence of features with high similarity as well as the difficulty of image matching and stitching,it is necessary to improve the real-time performance of the system while ensuring the detection effect,and to complete the following work in conjunction with actual production:Firstly,for the phenomenon of blurred edges of screen printing content,a variety of image enhancement methods are compared,and the Sobel sharpening method is finally used to improve the edges of screen printing content.For the problem of background interference with approximate edges,a method based on blurred subordinate degree function is proposed to segment the image.Secondly,this thesis proposes an optimized matching algorithm based on the scale invariance and sliding window aiming at the low efficiency and high incorrect matching rate of the traditional matching algorithm.By transforming the scale and using sliding window matching,the thesis improves the efficiency of matching and increases the correct rate of matching on the basis of the original matching algorithm.Thirdly,in the process of defect detection,the thesis firstly deals with the difference between the screen-printed content of the stitched picture and the beginning location of the screen-printed content of the stencil picture,and corrects the picture by second cropping and stitching.Then it designs the error detection criteria and identification methods for different defect categories.Finally,the defect detection is carried out according to the binary map dissimilarity results of the standard map and the test map to achieve the purpose of effective identification of non-conforming products.Finally,the thesis performs multi-threaded acceleration of the system,allowing certain steps in the image processing to be executed in parallel,shortening the time consumed by the defect detection system.The thesis finalized the design of a machine vision-based syringe screen printing defect detection system.After testing,the system has a false detection rate of less than5%,a leakage rate of less than 3%,and a detection efficiency of 3 pcs/sec.