Research On Online High Speed Defect Detection System Of Empty Cans

The jar is a commonly used canned food packaging container,which can prolong the shelf life and shelf life.The market demand is huge,but some defects and foreign objects are unavoidable in its production process(the gap,deformation,long and short side of the pot mouth,the easy attachment of oil and iron dust at the bottom and the inner wall).These defects will affect the production process and greatly reduce the food in the tank.Seal and shelf life.It is not found in time and will be mixed with the liquid in the tank,which not only affects the quality of the product,but also poses a great threat to the safety of the consumer's life.In order to realize the on-line automatic testing of the empty tank and empty tank,this paper developed a high-speed defect detection system based on machine vision technology,and studied the key technical points.The main research results of this paper are summarized as follows:(1)According to the actual production environment of the jar empty tank,and combined with the key technical requirements proposed by the enterprise,the technical implementation scheme of the on-line high-speed detection system for the empty tank and empty tank is designed.The scheme involves machine vision technology,photoelectric sensing technology,precision mechanical transmission technology and so on,and is tested and verified by designing software system.In order to make the system have the best performance,we must consider several performance indicators such as system stability,rapidity,accuracy and real-time synchronicity.(2)According to the technology implementation plan of the detection system and the imaging characteristics of the empty cans,the hardware structure design and equipment selection are carried out.This paper mainly introduces the design and equipment selection of lighting system,image processing system and motion control system.The illuminant and the light source controller and the light avoiding mechanism are the important prerequisites for ensuring the stable imaging of the system.The hardware design of image processing system is mainly about the selection and calculation of camera parameters and lens parameters.The motion control system mainly includes the positioning and triggering devices composed of the conveyor belt,optical fiber sensor and I/O control card.It can remove the auxiliary hardware structure of the mechanism and alarm mechanism,and can handle the abnormal condition of the unqualified can and the feedback system in time.(3)According to the hardware design plan of the detection system,the algorithm of the cans,mouth and inner wall area is studied.In order to protect the image effect of the collection,this paper first uses the global image quality discrimination index PSNR and SSIM to analyze the image effect of the bottle mouth,the bottom of the bottle and the inner wall of the inner wall,and select the imaging scheme with better imaging effect.Secondly,after the preprocessing methods such as smoothing,filtering,denoising and histogram equalization are carried out,the defect detection algorithm is designed according to the image characteristics of the cans,the bottom and the inner wall image of the can.The algorithm of defect detection in the area of the canister is used to analyze the contour of the measurement area by using the maximum variance method(OTSU).Finally,the least square ellipse algorithm is used to fit the ellipse curve of the target.The ellipse is discretized and sampled,and the eccentricity and deviation analysis chart of the ellipse is calculated,which can effectively deal with all kinds of defects in the bottle mouth area.The defect detection algorithm in the bottom area of the tank mainly uses the Hof gradient method to divide the detection region into multiple concentric circles.Based on the connected domain pixel analysis method of two valued images,the point defect,line defect and surface defect are detected in the region.Internal wall defect detection algorithm,first through the Hof line transformation and the Hof gradient method to locate and segment the inner and outer ring of the inner wall image,white sealing edge and transparent glue seal edge and other areas.Due to vertical shooting,the image feature of the inner area is compressed by the middle and lower parts,so it is difficult to detect the algorithm.In this paper,an image extension method based on polar coordinate transformation is proposed to solve the problem of the lower part of the inner wall image.Finally,the two valued connected domain analysis method is used to locate and find the inner wall defects.(4)On the basis of the detection of the detection area algorithm,the host computer control software of this detection system mainly uses Open CV image processing class library,Balser camera class library,and on the Windows operating system,the Visual Studio 2013 is used to design the testing system software.The software architecture of the system is mainly designed by the Ribbon framework under MFC.At last,the performance of the test system is tested in the laboratory and the production site respectively.The results of the test data show that the detection system has a good detection effect when it runs at a speed of 10 cans of /s.The system runs stably and the accuracy rate reaches 99.89%.It basically meets the requirements of the production index.However,there are still some shortcomings in the system.When the detection speed is raised to 20 cans / seconds,the detection system will have a certain false trigger phenomenon.It needs to be further optimized.The solution can consider optimizing the system structure,replacing the computer with more computational performance and optimizing the algorithmic ability and so on.