Design Of QFN Chips Surface Detection System Based On Machine Vision

Surface detection technology based on machine vision is an important part of semiconductor packaging quality detection.In this paper,a system of chips surface detection based on machine vision is designed by using the 2mm X 1.5mm QFN chips,it aims to judge the defect type of the printed characters and the chips direction.Firstly,based on requiring analysis,the overall design of the chip surface detection system can be determined by five aspects,which include structure design,lighting design,industrial camera selection,lens selection and software development platform.Secondly,image preprocessing is desigened to solve the problem of the noise of the digital image,including image filtering,image segmentation and image morphological processing,each step of the image processing will be compared and analysed,the paper proposes a improved OTSU segmentation algorithm based on gray histogram,and the processing effect is improved.Thirdly,image preprocessing is desigened to solve the problem of judgement of printed characters defect type,which inlcude grayscale feature extraction,characters segmentation,image matching and the HU moment features extractation.The methods of character segmentation are improved,and a weighted template matching algorithm based on reducing the image detection area is proposed.The image processing accuracy and speed are improved.Fourthly,A chip orientation recognition algorithm based on Hough transform is proposed.the target area of the chip orientation pin is placed by reducing the image detection area,so that,the influences of image noise and scratches are reduced,and the robustness of image processing are improved.Finally,the software system is completed,including UI design,database design and multithread design.It will take about 58ms of the image processing of chip character defect type judgment,and take about 28ms of the image processing of chip orientation recognition.