Research And Design Of TFT Defect Detection System Based On Machine Vision

As a new generation of flexible display technology,AMOLED display screen has many advantages,such as low power consumption,high contrast,bright color,flexible foldable and so on.However,due to cost reasons,AMOLED display screen can not completely replace LCD display.The main reason for the high cost of AMOLED display is that the yield of TFT section is too low,so it is important to study the TFT defect detection technology based on machine vision for improving the yield of AMOLED display Because of the high resolution requirement of display screen,the TFT layout line width of AMOLED screen is in micron level at present,and the defect of micron level may lead to the scrap of product,so the detection accuracy must reach micron level,and because of the need of industrial production,the algorithm must meet the requirements of real-time and robustness.At present,image preprocessing has to meet the requirements of real-time and robustness.Neither theory nor defect detection algorithm can meet the above requirements in micron-scale defect detection.Therefore,in order to achieve micron-scale TFT defect detection,it is necessary to study the TFT image preprocessing and defect detection algorithm.In order to solve this problem,this paper builds a TFT image acquisition platform,and studies the key technology of TFT defect detection based on machine vision,mainly from three aspects:image acquisition platform,image preprocessing and defect detection.The main research contents are as follows:(1)Aiming at the requirement of micron-level detection accuracy for TFT defect detection,and considering the characteristics of TFT substrate with many film layers and high reflectivity,a TFT defect image acquisition system with high accuracy and robustness is built through the research and design of cameras,lenses,light sources,driving devices and software.Experiments show that the system fully meets the requirements of TFT image defect detection(2)Aiming at the problem that the current image preprocessing algorithms can not meet the requirements of TFT image preprocessing for AMOLED display screen in terms of real-time,robustness and processing quality,an optimized multi-image average filtering algorithm and a matching algorithm based on local template matching are proposed as the preprocessing algorithms for TFT images.Like preprocessing algorithm,Experiments show that the image preprocessing algorithm introduced in this paper meets the real-time,robustness and processing quality requirements of TFT image preprocessing.(3)Aiming at the problem that the current defect detection and classification algorithms can not meet the accuracy and timeliness of TFT image micron defect detection and classification of AMOLED display screen,this paper proposes an optimized differential image defect detection method and a defect classification algorithm based on BP neural network.These two algorithms are used to solve the problem of TFT image defect detection and classification.Experiments show that the optimized algorithm of difference image defect detection and defect classification based on BP neural network meet the requirements of accuracy and timeliness of TFT image defect detection and classificationThe research results show that the key technology of defect detection based on machine vision in this paper has achieved good results in the construction of defect image acquisition system,image preprocessing,defect detection and classification.It can well complete the micron defect detection of AMOLED screen in TFT section and fully meet the needs of production.