Crystal Wafer Automatic Inspection Technology Research

In the 21st century, mobile phones, computers and other electronic products play the more and more important role in people's lives. One thing that most electronic products have in common is that they have a CPU whose frequency means the processing speed of the devices, and the accurate frequency is obtained with the use of basic frequency source created by a special crystal oscillator. Crystal oscillator is widely used for its high accuracy, high reliability and low price. However, many defects in the crystal of crystal oscillator, such as broken corner, cracks and stain, have serious impact on the use of crystal oscillator. So how to detect the defects with high speed and accurate has tremendously significant in production of crystal oscillator.In this paper, the methods of detecting the defects are studied deeply which is based on digital processing, and a system using the methods was designed with the character of the chips. Firstly, the building of the capturing image system which includes the selections of light source and camera, design of illumination was introduced. Secondly, the control software of image capturing system and the way of its working with image processing software harmoniously are discussed in detail. Then the introduction and analysis of image preprocessing of chips'image were made. Finally, according to the kinds of defects detecting methods and implementations were provided in this paper.On the basis of the character of the chips, the detection scheme was designed to detect the edge of chips first, then to detect the surface. In order to check if the chip is rectangle, a way based on Hough Transform was given. And flaw in Hough Transform was found when using it to get the angles of chip's four sides, then the solution to this flaw was put forward. In the light of the chip's edge image, an improvement was made in image rotation when detecting the edge of chips, which can reduce the rotating time from 170ms to 47ms. For the sake of detecting very small broken corner, the longer sides of chip in image should be horizontally rotated. Because of this, the method that can make the longer sides be closed to horizontal line within 3 pixels was presented in the paper.