Visual Image Detection Algorithm And System For Surface Defect Of Can Lids

With the introduction of the national "Smart Manufacturing 2025" strategy,intelligent industrial automation production lines have become a trend.China is a big beverage country.Accelerating the intelligent upgrade of beverage production lines has become an important application for promoting intelligent manufacturing equipment.Machine vision inspection and control technology is the key to improving the equipment performance on the beverage production line,which can improve the traditional shortcomings such as slow detection,low efficiency,and poor accuracy due to manual identification.This topic aims at the defect detection of the can lid of the packaging product in the can production line.Based on the machine vision technology,a complete defect detection system for can lid is designed.The main research contents and work of this article are summarized as follows:First of all,this article summarizes the significance of the current beverage industry production line;introduces the composition and application of the machine vision system;analyzes the current research status of the quality inspection of cans at home and abroad;and summarizes the main methods for surface defects detection of cans..Secondly,the hardware composition of the lid visual defect detection system was introduced in detail for the needs of the lid detection in the beverage production line.In the mechanical structure,the design of the transmission mechanism and defective reject device;in the electrical control system,put forward the control program based on PC and PLC;from the lighting program,camera and lens selection of several aspects of the analysis,set up a suitable Visual inspection platform.In order to better detect the defects on the lid of the can,the key technologies for the enhancement,segmentation and positioning of the lid image were explored and studied in depth.The positioning algorithm of the jar lid is the core part of the machine vision detection algorithm.Based on the positioning of the lid image,the advantages and disadvantages of the two methods of the center of gravity method and the least squares method are analyzed and compared.Based on the two methods,a method is proposed.An improved least squares method,which firstly uses the center of gravity method to pre-position the can lid image,then divides the can lid into four and a half rings according to the result of the pre-positioning,and uses the least squares fitting method to obtain four half rings.The radius and the center of the circle are analyzed based on the residual statistics,and the fitting parameters with the statistics greater than 3.0 are discarded.Finally,the final center and radius of the circle are obtained by weighting the remaining half-ring fitting parameters.Aiming at the variety of defects in the lid and the complex structure,a sub-region defect detection method was proposed.The lid lid inspection area was divided into circular and annular areas.For the circular area,the defect detection method based on Blob analysis was adopted.In the detection of defects in the annular area,an algorithm based on the least square fitting of the vertical gray projection curve is proposed.The fitting effect of the gray projection curve is judged based on the residual statistic of the fitting curve and the presence or absence of a defect is analyzed based on this.In addition,this paper further researches the method of block-based PCA and BP neural network to realize the detection of the whole can lid defect,and compares and analyzes the sub-region defect detection algorithm proposed in this paper and the algorithm based on block PCA and BP neural network.Finally,a software platform for the visual detection system of can lid surface defects was designed and developed.Experiments show that the detection accuracy of the can lid defect detection system can reach more than 96%,and the average detection time of a single can lid is 18.6ms,which can meet the needs of the beverage industry production line.