Study On Key Technology Of 3D Solder Paste Inspection And Application Based On Machine Vision

With appearance of SMT(Surface Mounted Technology),electronic components gradually tend to be miniaturized and dense.Solder paste printing,as the first step in SMT,if comprehensively detected,such as whether there is tin missed or a lack of tin,bias or bridge exist,so that pseudo soldering and short circuit can be avoided.And the detection can find the adverse trends of product early so as to improve production efficiency and reduce the cost.SPI(Solder Paste Inspection)based on computer vision,need to use grating,light source of laser and CCD(Charge Coupled Device)camera integrating different algorithms to reconstruct topography.Using the information of height,area and volume,we can find the defect in solder paste in time.We have investigated and analysis the principle of laser triangulation measurement and PMP(Phase Measuring Profilometry).In view of the 3D SPI equipment using laser triangulation principle has high speed and good 3D effect,but poor repeatability,meanwhile the result of detection can be easily changed with the shake of surroundings and gearing.PMP lighted by structured light has good repeatability and reproducibility,however the measured volume is smaller than the real value.We develop a new method to do 3D solder paste inspection.Firstly,we build a flexible and adjustable platform for solder paste inspection.We adopted a new marble platform base,to ensure the solid and stability of the fuselage.Meanwhile we adopted a high speed motion control system and designed an algorithm of speed interpolation to realize the precise trajectory.Lighting system using high brightness of LED(Light Emitting Diode)can form uniform diffusion light with special spherical structure,so as to prevent the diffuse reflection of PCB and guarantee the clear imaging.Camera system using a flat field correction based on two reference method,can eliminate the heterogeneity of the CCD image.With white balance,PCB image rendered true color.In order to decrease the times of scanning,reduce wear of mechanical components,extend the life of the system and make it run reliably and stably,field of view was set as 44 mm by 420 mm,the pixels of the whole image is 3648 by 13750,almost 5 million.Secondly,using the constructed 3D SPI platform,to obtain the two-dimensional information of the solder paste,we explored segmentation based on the global and local threshold and 2-d maximum entropy threshold,region growing and other traditional methods.Using machine learning methods,such as k-means clustering,Gaussian mixture model,Naive Bayesian and so on,according to the features of gray scale,brightness contrast and region information,we successfully segment out the region of solder paste in the complex background.We obtain the depth of the solder paste image pixels based on binocular vision,and we also reconstruct 3D topography of solder paste based on OpenGL.Finally,3D SPI system performance has been test,mainly including the accuracy and precision,as well as the stability,linear,bias and repeatability and reproducibility etc.Results show that the combined measuring tool system variation is less than 10%,and number of categories that can be distinguished is more than 5.All in all,the system is a good measuring system.