| Solder electromagnetic jet technology is characterized by both high positioning and dimensional accuracy requirements.In this paper,the application of machine vision in solder electromagnetic jet technology is mainly studied.Most efforts are devoted to the construction of machine vision system model,the process of camera calibration,mark point recognition and location and BGA ball quality inspection.As a result,a visual positioning and inspecting system is integratedly develop,which is to be used on solder electromagnetic spray platform.The overall design of the machine vision system is carried out,aiming at achieving positioning and inspecting functions in the electromagnetic injection system.A three-axis orthogonal manipulator is selected to be the actuator.The basic parameters of the camera,lens and light source hardware are calculated.Then a machine vision hardware platform was built with parts selected.According to the functional requirements of controllor software,the structural and functional module design is fulfiled.In order to describe the image principle and process of the camera in the system accurately,the image pixel coordinate,the image physical coordinate,the camera coordinate and the world coordinate are defined,basing on the actuator scheme of the three-axis orthogonal manipulator.Then camera linear imaging model and nonlinear distortion model are established.The camera’s internal parameter matrix,external parameter matrix and nonlinear distortion coefficient are obtained.Nextly,the camera’s internal and external parameters are calibrated.Basing on the active vision camera calibration method,the calibration formulas are deduced,and the calibration methods are improved.In this progress,firstly the internal and external parameter matrixs of the ideal linear model of the camera are estimated.Then the nonlinear distortion of the lens is corrected.Finally all camera parameters are optimized in the iterative convergence progress.To obtain the ideal mark feature,smooth filters,binarization and morphology are used to solve the typical salt and pepper noise and gaussian noise in images of marks.In the experiment,mark images with niose intensity range from 1% to 50% are observed and analysed.Meanwhile,the peak signal to noise ratio(PSNR)is used to be the evaluating indicator of the filter in performance.The experiment indicates that the median filter has the most significant filtering denoising effect on low density noise.On the basis of this,the mark point feature is identified and located through two methods,such as template matching based on the gray value and shape matching based on the shape feature.As a result,the positioning deflection error of the PCB is corrected.In the study of BGA ball quality inspection,some defects may occur in the preparation process of the BGA solder ball.The morphological differences among these defects are observed and analysed.The roundness of the ball and the region area are extracted to be the evaluation scale of the shape and size characteristics respectively.The balls are divided into four types,such as qualified,multi solder,less solder and burr.Furthermore,a two-dimensional gaussian mixture model classifier is proposed with four single gaussian models.It is used to classify the quality of the solder ball.The GMM model parameters are estimated and iteratively solved by EM algorithm.The classifier is trained with a sample set containing 68 solder balls,which leads to a good performance.Then a test is carried out on this GMM classifier with a test set consisting of 408 solder balls.The result shows that the total false alarm rate is 2.94% while the accuracy rate is 97.06%,and the missing rate is zero.The experimental results declare that the GMM classifier is effective against the ball defect.At last,a visual positioning and inspecting system for solder electromagnetic spray is accomplished.Mark images processing and recognition program as well as the core processing program of BGA ball recognition and classification are developed in Halcon.Then it is integrated with user graphical interface in Visual Studio.As a result,a machine vision positioning and inspecting system is finally realized and successfully applied in the solder electromagnetic jet platform. |
