Research On Automatic Inspection Of LCD Defects

The manufacturing of LCD has been developing rapidly and entering a new stage of standardization and large scale. LCD defects inspection is regarded as a key in the production. At present, manual inspection is major method used by most manufactur-ers, which has a number of drawbacks including human subjectivity and high cost. An automatic LCD defects detection system based on machine vision and image process-ing is proposed in this paper. The system can overcome many of disadvantages and offer opportunities to improve quality and reduce cost.Studies on LCD defects inspection are presented in this paper. A threshold seg-mentation method based on Weber’s Law is used to detect most common defects in LCD including point and line. One class of defects includes a variety of blemishes, called mura in the field, which performance as low contrast, non-uniform brightness regions and typically larger than single pixels. It is hard to distinguish mura defect-s from background, so several works have been developed recently. In this paper, SVD method and DCT method are used to estimate the background of LCD images respectively. Different detecting methods are proposed for different mura types. The subtracted images are processed by thresholding and statistical methods to detect point mura defects. Two methods are proposed to detect region mura defects, including s-tatistic method based on moving window and edge detection. Edge detection method has good performance in line mura defects inspection. Several edge methods are ex-perimented on the LCD images. The experiments results shows that Canny operator is optimum for detecting line mura defects.The system is designed based on machine vision and image processing imple-ments detection on three types of LCD products. The research works contribute to the developing of automatic defect detection in other products, and offer opportuni-ties to bring automatic LCD defects detection system into application in the domestic industries.