Research On Visual Inspection Method Of IC Electronic Card Appearance Quality Defects

With the development of computer technology and the continuous progress of production technology,robots are more and more widely used in the industrial field to meet the needs of automated production.Nowadays,hi gher requirements are put forward for IC chip appearance quality,but most of the enterprises still stay in the traditional manual inspection stage for IC quality detection.This method has the shortcomings of slow speed,high cost and easy to be mischecke d,and it is difficult to meet the actual production requirements.Therefore,it is of great social and economic significance to develop a high-speed,high-precision and intelligent automatic detection system to replace manual detection.According to IC characteristics and inspection requirements,a visual inspection system for IC defects is designed on the basis of in-depth study of IC appearance quality defect detection algorithms.It can be used for on-line inspection,grabbing good products and packagin g,and eliminating defective products.The main work of this article includes the following:1.In terms of imaging,the key devices are selected through many experiments,and the camera and the strip white light source are installed on the oblique side of the manipulator to make a good imaging effect.At the same time,the problem of reflecting light on the IC surface is solved.Visual inspection realizes the functions of IC tray automatic feeding and robot grabbing and blanking automatic hot melt packaging.2.In image preprocessing,an adaptive median filter is proposed by comparing several image denoising techniques.The effect is better than that of ordinary median filter.Due to the over-enhancement of histogram equalization,a method combining histogram equalization and de-equalization is proposed to enhance IC images.In feature extraction,several edge detection algorithms are analyzed for gradient calculation and edge extraction,and an improved Canny operator is proposed.Otsu algorithm is used to determine the threshold.In the aspect of template matching,because of the false detection of gray value matching,the shape-based template matching algorithm is selected.In order to accelerate the matching speed,the similarity measure method is used to accelerate the calculation and image pyramid method to achieve fast registration.3.Defect detection and recognition.The defect is obtained by difference between the detected image and the good template image.Morphological processing is used to reduce the edge error so as to avoid misjudgment.Then Blob analysis is carried out to extract the speckle features.SVM classifier is designed,sample library is created,IC defect feature is extracted by the above method,and matching score is added as feature vector to train SVM to get image classifier.In the actual detection of IC,the extracted features are passed into the classifier for defect detection.Experiments show that the defect detection accuracy based on SVM is very high and it is suitable for actual production.4.System implementation.A raster IC scanning control strategy is used to reduce the scanning time.Zhang Zhengyou's calibration is used to provide position basis for robot grasping.Software and hardware logic tasks are designed.Softwar e development is based on Visual Studio 2010 and OpenCV library functions on Windows platform.Each part performs its own functions and complements each other.Finally,the whole system prototype is implemented and put into actual production.