Study On Automatic Detecting Technique For The IC Wafer By Machine Vision

As the CD of the requirement of ULSI/VLSI IC is smaller and smaller, and defect density is lower and lower, the various defects in IC manufacture occur more frequently. The present production technology of integrated circuit may already reach 0.10 micron and following level, there are various defect hard to distinguish the graph of silicon chip from the defect of silicon chip, which is a very complex work. Because of the influence of various artificial factors and the restriction of IC precise property, current detection methods of most enterprises are hard to catch up with their products detection quality and efficiency.This paper discussed intelligent detecting technology for IC wafer based on machine vision .The major contents are some key problems of intelligent detection for IC wafer by machine vision. There are following four aspects: 1. Defect classification and recognition of IC wafer image; 2.Extracting efficiently the extra material defect features from an IC real image in the IC real manufacturing process based on mathematical morphology; 3.Recognition of IC wire holes by matching algorithm which is based on image features; 4.Control strategy of detection platform for IC wafer and calibration technology in detection. In this paper, go deep into analyze the difficulty technology of IC wafer detection from algorithm theory and application, and have gotten innovation research in the above aspects.In allusion to the inherent characteristic of IC wafer, this paper has analyzed the orientation of IC wafer, detecting principle and the difficulty of automatic analysis. What is more, it put forward the new control strategy of detection platform, combined raster scan with vector scan, which could make the orientation of detection platform faster and more accurate. The strategy has solved a lot of shortages in conventional detection methods.The method of extracting extra material defect from the IC wafer image by mathematical morphology is an innovative, which is comparatively efficient and accurate for profile and position extraction. At first, verify the algorithm principle by extracting the...