Detection And Classification Of Mura Defect For TFT-LCD Based On Computer Vision

With the development of LCD drive technology,TFT-LCD(Thin Film Transistor Liquid Crystal Display)is welcomed by more and more consumers,and its applications are also constantly expanding.In order to meet the needs of the market,TFT-LCD is developing with bigger size,thinner panel and higher resolution,so the risk of Mura defects is greater.At present,the main detection method is using the naked eye detection,so the detection efficiency is low,and it is difficult to classify and quantify the defects.In order to improve the productivity and the quality of the products,it is urgent to study the fast,accurate and stable method for automatic detection of Mura defects.Through the principle analysis,algorithm design and simulation around the background texture suppression,correction of uneven brightness,segmentation and quantification and classification of defects,the method of automatic Mura defect detection and defect classification based on machine vision is studied.The main work is as follows:Firstly,dealing with the background texture,the real Gabor filter is used to suppress background texture.This paper designs the main parameters and the image fusion method,finally verifies this algorithm through the simulation experiments which shows that the proposed algorithm can effectively suppress the background texture.Secondly,aimed at uneven brightness of the LCD image,the image brightness correction method based on PCA background fitting is proposed.This paper designs the uneven brightness correction method,and the algorithm is validated by simulation experiment which shows that this algorithm can effectively correct the uneven brightness.Thirdly,the Mura defect segmentation method is studied.In order to overcome the traditional active contour level set segmentation algorithm's drawback,the bilateral filter and initial contour optimization method is used to improve the algorithm.Through simulation experiment,compared with the traditional active contour level set algorithm,the modified algorithm has significant advantages on the segmentation result and time performance.Fourthly,according to the significance of the LCD screen defect classification for process improving,the classification method of LCD screen defects is studied.The support vector machine is used to realize the classification of point defects,line defects and Mura defects.The simulation results show that the method has good classification accuracy.Finally,the Mura defect detection and defect classification process is established,and the images of the LCD panel is collected as the experimental samples.Then the background texture suppression,the brightness correction,the Mura defect segmentation and quantization,and the LCD screen defect classification are verified by experiments.The 66 samples in the experiment were successfully detected by the modified active contour level set segmentation method.The defect classification accuracy can be up to 98.3% by the support vector machine.