| More and more company join in the competition of mobile phone market. The only way to stand out in the fierce competition is to improve the quality and the efficiency of manufacture. As we know, the quality of TFT-LCD can directly affect the user experience. Therefore, the defect detection of TFT-LCD is what the dissertation focus on. Majority of the work is finished by artificial inspection, which is time-consuming and not reliable. The method based on machine vision is on the rise, but most algorithms can’t satisfy the requirement of on-line inspection with low accuracy and efficiency.After analyzing the structure features and manufacturing process of TFT-LCD, the dissertation proposed a new cascade detection algorithm, which combines the advantages of rapidity and accuracy. It can solve most of the defect detection including dark point, light point, dark line and so on. At first, we designed the system of image acquisition to obtain the input image that has high contrast and low noise, then we implement the first step of cascade detection algorithm that using image feature of gray level difference of sub-image to segment the abnormal areal. The second step is based on phase only transform which correspond to the Discrete Fourier Transform(DFT), normalized by the magnitude. It can remove regularities like texture and noise. After that, we improve the method of setting regions of interest(ROI). To solve the problem of irregularity, we conduct the affine transformation and edge segmentation. And to solve the problem of non-uniformity, we divide the input image into small pieces, then conduct the cascade algorithms.At last, we compare the cascade algorithm with local threshold and directly Fourier Transform on the same platform with all 1800 input images. After evaluating the result based on the criteria of miss detection, false detection, accuracy and time-consuming, the cascade detection algorithm provides a better performance over all the above algorithms. |
PDF Full Text Request