Research Of IC Defect Detection Based On Machine Vision

With the rapid development in recent years in domestic and international microelectronics manufacturing, the demand for IC(Integrated Circuit) is greatly growing in industry and the surface quality of IC is the key factor that affects the quality of IC, good IC defect detection work is particularly important. But the current domestic IC defect detection technology is mostly relatively backward, stays in manual and semi-automatic detection phase,the detection efficiency is low and the quality of detection can not be guaranteed, it’s unable to adapt to the modern generation needs.On the basis of a careful study of the IC common defects, this thesis does a systematic and in-depth research on IC surface defects detection technology, mainly reflects in the following three aspects:(1) Studies the IC pins coplanarity detection methods and puts forward using grating projection method combined with digital image processing technology to measure IC pins coplanarity. Using industrial cameras to capture structured light image reflected by IC surface and pins, respectively using image enhancement technology, Otsu threshold segmentation,morphological skeleton method, least square method to accomplish the refinement of structured light, calculates the IC pin coplanarity based on the experimental model after the camera calibration.(2) Proposes a new method to determine subpixel edge locations using the third-order gray moment to overcome the traditional edge detection methods of poor positioning accuracy,weak ability to resist noise shortcomings. Using Otsu threshold segmentation algorithm in positioning, gray moment coarse positioning, gray moment respectively to determine the sub-pixel precision positioning edge position and analyses the positioning accuracy.(3) Studies the common defect types of IC chips, mainly including glue residue, fracture surface scratch, pin fracture and cuts. Defect detection algorithms are designed aimed at these defects respectively and compared with other detection methods. Using homomorphic filtering to eliminate the phenomenon of uneven illumination in the image, research is mainly focused on the ROI(Area Of Interest) localization algorithm, a new positioning method is put forward based on the morphology corrosion and Blob connected domain analysis, the purpose is to realize the accurate rapid positioning and segmentation of the defect images.The IC defect detection algorithms proposed in this thesis can accurately detect the main defects of IC and can meet the practical needs of industry. It’s also an application andvalidation of the visual inspection and digital image processing theory. Therefore, this thesis research has integrated value of theory and practice.