The Image Processing In TFT- LCD Defect Detection System

With the development of computer science, digital image processing has been widely used in industrial production, such as automatic detection and identification of the assembly parts, quality of printed circuit board testing, defect detection of fabric in textile industry and so on. Compared with the traditional manual detection, image processing based on optics is higher real-time, more accurate and stable. Nowadays, TFT-LCD has replaced CRT successfully and become a necessity in field of display. Although the production environment demand is really high, low quality of LCD still can’t be avoided. So the defect detection of LCD is an important part in the process of LCD production. Traditional detection that is based on human being’s sight has seriously reduced production efficiency. Thus, how to realize the intelligent LCD defect detection has become an important research field.The methods for LCD detection can be divided into two categories, defect detection based on electrical and optical properties. Image processing based on optics is simple, low cost and easy to work with production line. The main problem is how to realize the detection algorithm. As there are so many different kinds of defects, the design of algorithm becomes much more difficult. In this paper, we put forward the detection algorithm, and realize it on windows in the view of characteristics among point, line and Mura defect: Point defect requires high precision, and needs to locate the specific place of a single liquid crystal tube. The core process depends on how to make the grid accurate. Line defect is relatively simple, by using median filter to remove background and analyzing gray scale of the horizontal and vertical direction, we can get position of the defect. Mura defect is the most difficult, because characteristics of the defect are not obvious at all. What’s more, it can be easily affected by the environment. In this paper, we suppress the effect of background and enhance the prospect, by using Gabor filter to eliminate the image. Then increase the contrast of Mura areas by using gray level transformation, to get position of the defects.This paper first introduces the basic structure of the LCD intelligent detection system, and discusses the image processing methods used in the detection algorithm. Then it introduces the three kinds of defect detection processes and the corresponding experiments in details. Finally, it analyzes advantages and disadvantages of the algorithm including the follow-up work.