Analysis And Research On Weilding Defect Detection Based On Machine Vision

With the development of industry in recent years,manual detection in complex and dangerous working environments can no longer meet the necessary detection requirements.Therefore,in order to improve the efficiency of industrial assembly line detection,reduce the false detection rate caused by manual detection,and be more objective and reliable.To evaluate the welding quality,this paper applies the related technology of machine vision to the field of welding defect detection,designs the image acquisition system of coaxial and low-angle light sources,and develops specific parameters for the characteristic parameters of the pad solder joint and the bridge wire.Detection algorithm realizes welding defect detection and defect classification.This subject uses the bridge wire of the automobile airbag detonator as the welding defect detection object,and the research is mainly carried out from the following aspects:(1)Based on the image acquisition system of machine vision,this article designs two different image acquisition methods on the image acquisition system: the first is the acquisition method under the illumination of the coaxial light source,which is used to detect the number of solder pads;the second It is the image acquisition under the illumination of a low-angle light source,which is used to inspect the welding quality of the bridge wire.The experimental results show that the two light sources are driven by 8.6V and 9.4V voltages,which can obtain a higher contrast between the feature area and the background color,and provide a clear feature image for the later image feature parameter extraction.(2)Based on machine vision-based solder joint detection,this paper first performs image enhancement,image filtering and morphological processing of the characteristic area on the collected images,and obtains the dynamic area fitting circle of the left solder joint and the right solder joint of the pad,and then performs the image Rotate,detect the coordinate distribution of the solder joints in the characteristic area,and determine the discrete number.The research results show that the normal detonator welding product has only two solder joints,and the lateral coordinate of the first solder joint is in the dynamic range of 705?753,and the lateral coordinate of the second solder joint is in the dynamic range of 995-1023.All of them are normal bridge wire welding products.(3)Based on the feature detection of bridge wire by machine vision,this paper first performs second-order differential edge detection on the feature area of bridge wire,calculates the total length of the edge contour of the bridge wire,and then uses the straight line detection algorithm under Hough transform to detect the number of straight lines of the bridge wire.And according to the calculated area,length,width and maximum circumscribed rectangle of the bridge wire in the characteristic area,the defect type of thedetonator welding is determined.The experimental results show that the number of straight line detection of normal bridge wires is 1,and the size of the largest circumscribed rectangle is 245×15?255×16.The test results of the 10 randomly selected samples in this paper are all qualified products,the similarity measurement is above 0.9800,the detection accuracy of the computer test reaches 98.10%,and the single detection time is about 1.9seconds.The detection accuracy and detection time have reached the technical indicators and can be used in actual production applications on the assembly line.Based on the experimental results entitled "Machine Vision-based Welding Defect Detection Analysis and Research",this article shows that the detonator for normal welding has only two welding points,and the position coordinates of the left and right welding points are within a certain range.If the abscissa is not The scope of this standard means that the detonator has welding defects;the number of normal welding bridge wire linear inspections is 1,and the similarity measurement is more than 98% of products without welding defects;the inspections in this article can be obtained from the analysis of the inspection data The algorithm has more advantages in detection accuracy and detection efficiency than the defect detection of Sherlock software commonly used in the industry.It can be seen that the bridge wire defect detection algorithm in this paper has practical application value in the field of machine vision,and it is also useful for the next step of defect detection error analysis.Laid a good foundation.