| Backlight is the key light-emitting device behind liquid crystal display(LCD),and its luminous effect will directly affect the visual effect of liquid crystal display module(LCM).Because of the complex structure and many processes in the manufacturing process of back light source,various visual defects can not be avoided.Among them,Mura defect has low contrast,fuzzy edge and different shape,which makes it difficult to detect Mura defects.At present,it mainly relies on the traditional artificial vision detection(HVI)method,but it has the disadvantages of strong subjectivity,low efficiency,lack of objective quantitative index,and so on.Therefore,it is urgent to develop a stable and efficient method for Mura defect detection.Based on this,this thesis proposes an improved defect detection algorithm based on CV model,which can overcome the uneven gray level of sample image caused by sample collection process,and effectively improve the detection accuracy and efficiency of Mura defect.The main research and contributions are as follows:(1)Because of the variety of back light sources and the defects of the edge,the shape of the interested area is irregular,which leads to the problem that the interested area can not be extracted accurately.In this thesis,ROI extraction method based on Kirsch operator and tangent space transformation is proposed.Firstly,the edge detection is carried out by using adaptive Kirsch operator,and the edge is precisely located by edge refinement algorithm.Then,the obtained edge points are transformed into tangent space,and then the obtained tangent space curve is fitted by angle projection,and then the coordinate transformation points are obtained.Finally,the corrected back light source area is accurately extracted by perspective transformation.(2)In view of the problem that the active contour model is sensitive to the initial contour,and the defect location is not determined due to the random location of the defect,this thesis proposes a defect region pre judgment algorithm based on background reconstruction.First,the background of the back light source is reconstructed by B-spline fitting method,then the difference operation is done with the original image to get the difference image,and the morphological processing of the difference image is carried out.Finally,the extreme point is obtained by edge extraction,and the initial contour coordinate is determined.The experimental results show that the algorithm can obtain the appropriate initial contour effectively.(3)In view of the background gray level inequality,CV model can not accurately segment Mura defects.This thesis presents a horizontal set segmentation model which combines global regional information(CV)and local fitting information(LIC)as fidelity terms.Firstly,the offset field correction term and CV region fitting term are introduced to fuse by adaptive weight coefficient.Then,energy penalty term and length term are added to avoid repeated initialization of horizontal set function,and improve the speed and smoothness of curve evolution.Meanwhile,according to the minimum perceptible contrast curve of Mura defect,the characteristic term of Mura defect is constructed and introduced into energy functional equation to improve the detection accuracy of Mura defect.Through the experimental test,it is proved that the model has a high detection accuracy for Mura defects of the back light source. |
