Research On Surface Defect Detection Method Of Metal Can End In Beverage Production Line

The detection of surface defects of metal can end in beverage production lines is an important part of container quality inspection and a key link to ensure product quality.As a traditional method for detecting surface defects of can ends,manual detection is easily affected by subjective factors,and the detection accuracy is low and the timeliness is poor.With the development of intelligent manufacturing technology with intelligent equipment as the core,machine vision-based surface defect detection technology can identify defects by analyzing and processing images,instead of manual detection,and real-time and accurate image processing technology realizes can end The key to surface visual inspection.In this paper,the metal canned container can end of the beverage production line is taken as the research object,and the high-speed and high-precision detection method of the surface defect of the can end is developed for the defects such as dirt,deformation and scratches on the surface.The main research contents and innovations are as follows:Aiming at the phenomenon that the can end image may have more interference points and cause the tank lid positioning error,this paper proposes a multicircumference weight fitting and positioning method based on edge point constraint.The method scans the three edge points through the center of gravity of the can end,and rounds the outer edge points to obtain the initial center,and then constrains the remaining edge points by the initial center,and performs weight fitting to obtain the optimal center position of the can end.The experimental results show that the method can overcome the influence of the interference point to achieve accurate positioning of the can end.For the 1296×966 can end image,the average positioning error is less than 2.5 pixels,and the execution time is less than 5 milliseconds.Aiming at the phenomenon that the irregular texture interference and uneven illumination caused by the central region cause high false detection rate,a central region defect detection based on global adaptive LBP feature is proposed.The method performs multi-scale circular LBP feature sampling on the central region to obtain local texture information,which can effectively overcome the influence of uneven illumination on the surface of the can end.By comparing the global adaptive binary pattern calculation,the texture comparison of the sampling points in the neighborhood can be The irregular complex texture interference in the center of the can end is suppressed.The experimental results show that the defect detection accuracy reaches 97.64%,which can better identify the defects in the central region.For the detection of the defect of the annular region of the metal can end,a defect detection method based on SVM classification is adopted.The extracted annular region is radially expanded and vertically projected,Gaussian second-order differential filtering and first-order difference are performed on the projection curve,and then the four-dimensional features are extracted for the projection curve to perform SVM training to obtain a classifier,and the defect of the annular region is identified by the classifier..The method can replace the manual selection threshold,and has high adaptability and robustness.The comprehensive correct rate for the defect detection of the ring region is 98.64%,and the execution time is less than 5 milliseconds.The metal can end surface defect detection method designed in this paper can effectively identify the surface defects that may exist in each area of the can end.The comprehensive detection accuracy is over 98%,which is more accurate than the previous can end surface defect detection method.Practical value.