The Research Of Micro-imaging Inspection Algorithms And Key Technology For High Density Flexible IC Packaging Substrate

The high-density Flexible Integrated Circuit Packaging Substrates(FICS)are widely used in compact,light and movable electronic products.Since the fabrication of FICS is complicated,then the defect in any step would cause a great loss.For such high-density FICS,the line width and line distance are as thin as 10 um or less,then a microscope is needed to acquire the FICS.With the development of the chip manufacturing technology,the line width and the line distance are reduced.The defect detection becomes more difficult than the traditional defect detection.When the images of FICS are collected by the micro-imaging technology,the texture structure and defect of the substrate are simultaneously magnified.It leads to the similarity between the texture structure and the defect,e.g.,oxidation defect.However,the existing methods frequently misjudge the texture structure of the substrate and the defect.It is urgent to develop some new detection methods to meet the requirements in industrial production.To satisfy these demands,new methods focused on the key technical problems of automatic defect detection of FICS by using the microscope are proposed.The main research contents and innovations are concluded in the following aspects.(1)As unknown industrial noises exist in micro-imaging substrate images detecting,a filtering template of weighted average neighborhood closed curve based on topological mapping is proposed.The experimental results show that our filtering template is better than the existing filters in effectiveness.(2)In the oxidized defect substrate images from low magnification microscope,the oxidative defect detection algorithm based on topological mapping is proposed to solve the problems that traditional visual detection methods are easy to seriously misjudge the black background as a defect and heavily rely on the standard template.This algorithm is not only insensitive to the background texture and illumination of FICS,but also has strong antiinterference ability and robustness.The effectiveness of the proposed method is demonstrated by some experiments of noise images,enhanced images and defect images.(3)In the oxidized defect substrate images of high magnification microscope,an algorithm based on differential geometry is proposed to solve the problems that the traditional visual detection methods are easy to seriously misjudge the black background and substrate texture structure as defects and rely heavily on the standard template.First,a sample fitting model is established.Then a defect detection model is built.Under the same environment,the singletime modeling of the fitted sample can realize the rapid detection of the defect substrate.In addition,the existing sample fitting model and related parameters can be directly used in subsequent defect detection,which greatly reduces the detection time and highly meets the realtime requirements of industrial inspection.(4)In the ink foreign matter defects substrate images of micro-imaging,an algorithm based on differential geometry is proposed to solve the problems that the traditional visual detection methods are prone to misjudge the black background and film background as defects and the images have low contrast.First,a curvature-based sample fitting model is established.And then a segmentation linear function based on image gray value analysis is proposed to improve the contrast of the image.Finally,a probability-based defect detection model is established.In the same environment,the relevant parameters of the sample fitting model and segmentation linear function can be directly applied to the subsequent image,which greatly reduces the time of subsequent defect detection,thereby improving the overall efficiency of the defect detection.The practical applications prove that the proposed algorithm is not only with superior precision,but also meets the industrial real-time requirements for rapid production of flexible substrates.Based on the previous results,the application of the four algorithms are realized and the effects are verified.Furthermore,experimental data shows that the defect detection methods proposed in this thesis could meet the industrial requirements.