CN113034464B - Visual real-time detection method for defects of liquid crystal display under multiple backgrounds - Google Patents
Visual real-time detection method for defects of liquid crystal display under multiple backgrounds Download PDFInfo
- Publication number
- CN113034464B CN113034464B CN202110306033.9A CN202110306033A CN113034464B CN 113034464 B CN113034464 B CN 113034464B CN 202110306033 A CN202110306033 A CN 202110306033A CN 113034464 B CN113034464 B CN 113034464B
- Authority
- CN
- China
- Prior art keywords
- defect
- defects
- lcd
- gray
- area
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000007547 defect Effects 0.000 title claims abstract description 194
- 238000000034 method Methods 0.000 title claims abstract description 37
- 239000004973 liquid crystal related substance Substances 0.000 title claims abstract description 17
- 230000000007 visual effect Effects 0.000 title claims abstract description 17
- 238000011897 real-time detection Methods 0.000 title abstract description 9
- 238000001514 detection method Methods 0.000 claims abstract description 57
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000000694 effects Effects 0.000 claims description 15
- 239000002245 particle Substances 0.000 claims description 15
- 238000010586 diagram Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 8
- 238000012935 Averaging Methods 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 5
- 230000009466 transformation Effects 0.000 claims description 4
- 238000005520 cutting process Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000005215 recombination Methods 0.000 claims 1
- 230000006798 recombination Effects 0.000 claims 1
- 238000003709 image segmentation Methods 0.000 abstract description 3
- 230000000877 morphologic effect Effects 0.000 abstract description 3
- 230000011218 segmentation Effects 0.000 description 18
- 239000011159 matrix material Substances 0.000 description 5
- 238000011179 visual inspection Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 3
- 238000003705 background correction Methods 0.000 description 3
- 230000002950 deficient Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 238000007689 inspection Methods 0.000 description 3
- 238000004519 manufacturing process Methods 0.000 description 2
- 241001270131 Agaricus moelleri Species 0.000 description 1
- 230000001174 ascending effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 125000001475 halogen functional group Chemical group 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000009776 industrial production Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G02—OPTICS
- G02F—OPTICAL DEVICES OR ARRANGEMENTS FOR THE CONTROL OF LIGHT BY MODIFICATION OF THE OPTICAL PROPERTIES OF THE MEDIA OF THE ELEMENTS INVOLVED THEREIN; NON-LINEAR OPTICS; FREQUENCY-CHANGING OF LIGHT; OPTICAL LOGIC ELEMENTS; OPTICAL ANALOGUE/DIGITAL CONVERTERS
- G02F1/00—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics
- G02F1/01—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour
- G02F1/13—Devices or arrangements for the control of the intensity, colour, phase, polarisation or direction of light arriving from an independent light source, e.g. switching, gating or modulating; Non-linear optics for the control of the intensity, phase, polarisation or colour based on liquid crystals, e.g. single liquid crystal display cells
- G02F1/1306—Details
- G02F1/1309—Repairing; Testing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/40—Image enhancement or restoration using histogram techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/70—Denoising; Smoothing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/194—Segmentation; Edge detection involving foreground-background segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10024—Color image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30108—Industrial image inspection
- G06T2207/30121—CRT, LCD or plasma display
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Nonlinear Science (AREA)
- Chemical & Material Sciences (AREA)
- Crystallography & Structural Chemistry (AREA)
- Optics & Photonics (AREA)
- Quality & Reliability (AREA)
- Testing Of Optical Devices Or Fibers (AREA)
- Liquid Crystal (AREA)
Abstract
The invention discloses a visual real-time detection method for defects of a liquid crystal display under multiple backgrounds, which comprises the steps of firstly preprocessing an image to be detected by using algorithms such as image equalization, histogram equalization and the like; three different algorithms are provided for six different defects, and the method can better adapt to defect detection under different backgrounds, and specifically comprises the following steps: aiming at the problems of fuzzy edge, low contrast, complex background texture and the like in LCDMura and scratch defects, a maximum inter-class variance method and morphological operation are adopted to carry out defect image segmentation under a pure color display background; dividing and marking the ROI edge defect under the grid display background by adopting a maximum inter-class variance method and a 4-connectivity judgment criterion, and judging whether the 32-level gradient color display background has color difference or not by dividing the gray value variance and the mean value of sub-regions for the color deviation defect. Experimental results show that the algorithm can well realize real-time detection of the defects of the LCD, and has better applicability and engineering practical value.
Description
Technical Field
The invention relates to a visual real-time detection method for defects of a liquid crystal display under multiple backgrounds, belonging to the field of visual detection of defects of semiconductors.
Background
In recent years, as the demand for LCD panels has increased year by year, competition among manufacturers has become particularly intense, and improvement in product quality and production efficiency has become an important factor for manufacturers to maintain superiority. In the production process of the liquid crystal display, the liquid crystal material has various defects of different types on the LCD panel due to spatial distribution difference and concentration difference, short circuit and open circuit of scanning lines, impurities in the liquid crystal and the like, so that the visual effect and the user experience of the product are influenced, and the product quality cannot be guaranteed. At present, the defect detection of the liquid crystal display is mainly finished by adopting visual detection and machine vision. Visual inspection requires a professional to make an empirical judgment on a defect sample library. Although visual inspection can be well adapted to different types and sizes of display panel inspection, the standard is difficult to unify measurement, and inspection is time-consuming and has poor stability. With the rapid development of the field of visual inspection, methods for defect detection and identification by using the technology are also emerging continuously, and the visual inspection technology mainly compares images acquired by a camera with qualified product data in a sample library and makes judgment based on the comparison information. Although the current visual inspection technology has achieved great success in the defect inspection industry, the problem of detecting and identifying LCD defects is still a problem in the industry, and needs to be continuously improved and perfected.
Disclosure of Invention
The invention provides a visual real-time detection method for defects of a liquid crystal display under multiple backgrounds, which can be effectively used for judging whether the defects exist or not by preprocessing a data set and fusing three different detection algorithms; and further performing visual display.
The technical scheme of the invention is as follows: a visual real-time detection method for defects of a liquid crystal display under multiple backgrounds comprises the following specific steps:
step 3, detecting whether the image in the experimental data set has defects by using three different detection algorithms; wherein, three different detection algorithms include: LCD Mura and scratch defect detection algorithm, LCD edge defect detection algorithm and LCD color level error defect algorithm.
The preprocessing is specifically to perform averaging, gaussian filtering, gamma transformation and histogram equalization in sequence.
The equalization specifically comprises the following steps: continuously acquiring three-frame images f through an industrial camera on one or more display backgrounds to be detected 1 、f 2 、f 3 And averaging the three frames of images to obtain f average 。
The display defects are classified into six different defects, and specifically include: spotMura, line Mura, region Mura, tone scale, scratch, and edge defect.
The step 3 is specifically as follows:
LCD Mura and scratch defect detection algorithm: marking a defect area by aiming at one or more images in an experimental data set under a display background through the relation between pixel variance and a threshold value; aiming at the defect area, obtaining an image with a corrected background through subtraction operation of pixels and pixel mean values; dividing the corrected image into defect regions by a maximum inter-class variance method; carrying out feature extraction by using particle morphology, and calculating the number of extracted defects to obtain a visual defect effect diagram; judging whether one or more defects of spot Mura, region Mura, line Mura and scratch defects exist according to the visual defect effect graph; wherein the plurality is at most six;
LCD edge defect detection algorithm: aiming at images in an experimental data set under six display backgrounds, defect areas can not be extracted by adopting LCD Mura and scratch defect detection algorithms, a grid is used as a detection background of LCD edge defects, grid edges are intercepted and recombined, a maximum inter-class variance method is adopted to segment a recombined grid, and defect characteristics are extracted after particle morphology processing to obtain a visual defect effect graph;
LCD color level error defect algorithm: if the defect area is not extracted by the LCD edge defect detection algorithm, the 32-level gray gradient is used as a display background, and whether the LCD has the color level deviation defect or not is judged by dividing the mean value and the standard deviation of the gray values of the subareas.
The LCD edge defect detection algorithm specifically comprises the following steps: adopting the grid as a display background, extracting the grid in the picture from the picture of the preprocessed LCD grid background through pixel value information, and intercepting the left edge and the upper edge of the grid as an ROI 1 Region, cutting right and lower edges as ROI 2 Region, intercepting ROI 1 And ROI 2 The regions are respectively placed in a new template, grids in the two regions are segmented by a maximum inter-class variance method, and defect features are extracted after particle morphology processing to obtain a visual defect effect diagram: if ROI 1 And ROI 2 If the number of the particles of the area marks is equal to 1, the edge of the LCD has no defects; otherwise, judging that the LCD edge has defects.
The LCD color level error defect algorithm specifically comprises the following steps: dividing the gray scale background into 2 equal parts along the vertical direction, and respectively representing the 2 equal parts as a U area and an L area:
in the U area, if the gray level mean value in each row in the traversal frame is sequentially increased along with the increasing of the gray level, the standard deviation is 0; the difference between the gray value mean values in two adjacent traversal frames is 8, and the standard deviation is 4; if the gray values in the same gray scale are the same and the standard deviation is 0, the color scale of the region has no deviation defect; otherwise, the color level deviation exists in the area;
in the L area, if the gray average value in the traversal frame in each row is sequentially reduced along with the increasing of the gray level, the standard deviation is 0; the difference between the mean values of the gray values in two adjacent traversal frames is 8, and the standard deviation is 4; if the gray values in the same gray scale are the same and the standard deviation is 0, indicating that the color scale of the region has no deviation defect; otherwise, the region has a color level deviation.
Dividing the U area and the L area into N equal sub-areas along the horizontal direction, uniformly dividing the sub-areas into M lines along the vertical direction, and traversing the N multiplied by M sub-areas in the U area and the L area by adopting an H multiplied by W traversal frame; the pixel values of the sub-regions of the m-th row and n-th column of the U and L regions are respectively expressed as
And
where M is {1, …, M } and N is {1, …, N }, and the corresponding M × N sub-regions have the corresponding gray pixel mean value of
And standard deviation of
Wherein I is belonged to { U, L },
the beneficial effects of the invention are: firstly, preprocessing an image to be detected by using algorithms such as image equalization, histogram equalization and the like; three different algorithms are provided for six different defects, and the method can better adapt to defect detection under different backgrounds, and specifically comprises the following steps: aiming at the problems of fuzzy edge, low contrast, complex background texture and the like existing in LCD Mura and scratch defects, a maximum Inter-Class Variance method (Inter-Class Variance) and morphological operation are adopted to carry out one or more defect image segmentation under a pure color display background; dividing and marking the ROI edge defect under the grid display background by adopting a maximum inter-class variance method and a 4-connectivity judgment criterion, and judging whether the 32-level gradient color display background has color difference or not by dividing the gray value variance and the mean value of sub-regions for the color deviation defect. Experimental results show that the algorithm can well realize real-time detection of the defects of the LCD, has better applicability and engineering practical value, effectively solves the defect detection of the display under the complex background and reduces the problem of missed detection of the previous algorithm.
Drawings
FIG. 1 is a flow chart of the present invention;
FIG. 2 is a flow chart of the pretreatment;
FIG. 3 is a graph of pretreatment results;
FIG. 4 is a ROI area map;
FIG. 5 is a diagram of a gradual background;
FIG. 6 is a graph of the results of various algorithm defect segmentation;
FIG. 7 is a diagram illustrating the segmentation effect of edge defects;
FIG. 8 illustrates the effect of grid background defect segmentation;
FIG. 9 is a diagram of the color level deviation existing in the U region;
FIG. 10 is a graph showing the color level shift in the L region;
FIG. 11 shows the level deviation defect detection result of U region;
fig. 12 shows the detection result of the level shift defect in the L region.
Detailed Description
Example 1: as shown in fig. 1-12, a method for visually detecting defects of a liquid crystal display under multiple backgrounds in real time comprises the following steps:
step 3, detecting whether the image in the experimental data set has defects by using three different detection algorithms (if the defects can not be detected by using the three detection algorithms, the display is a qualified product, otherwise, the display has defects); wherein, three different detection algorithms include: LCD Mura and scratch defect detection algorithm, LCD edge defect detection algorithm and LCD color level error defect algorithm.
Still further, the present invention provides the following implementation:
the industrial production process of the invention is as follows: firstly, a defect data set of a display under one or more display backgrounds is acquired, and the specific acquisition process is as follows: the industrial personal computer controls the motion trail of the two-degree-of-freedom rectangular coordinate mounting frame through PLC control, and then the industrial camera is sent to a specified position. And adjusting the position of the CCD industrial camera through the two-degree-of-freedom mounting rack according to the model of the display which is produced in the industrial field. Under the irradiation of a sufficient shadowless light source, the CCD industrial camera collects the LCD display panel image on the double speed chain, converts the image into an electrical signal, and converts it into a digital image via an image acquisition card inserted in the PC. Inputting a digital image into a computer and detecting according to the method of the invention: if the display panel is found to have defects, a signal is sent to the PLC through the digital I/O card, the display panel is conveyed to the lower layer of the double-speed chain by controlling the vertical lifting platform, and the display panel is conveyed to an unqualified display panel storage area. At the moment, the PLC can synthesize the auxiliary sensor, track the position of the display panel and control the vertical lifting device to accurately convey the defect display panel to the underground layer of the double-speed chain, so that the shunting of the defect display panel is achieved.
Further, the method of the invention is implemented as follows:
the display defect data set is shot by a camera to obtain display defect images under different backgrounds, and the data set collected by the embodiment can be divided into six defects of spotMura, line Mura, region Mura, color order, scratch and edge defect.
The data preprocessing in step 1 is to improve the accuracy and detection efficiency of real-time detection, and each acquired LCD picture to be detected is preprocessed by using picture equalization, gaussian filtering, gamma transformation and histogram equalization, and the preprocessing flow is as shown in fig. 2. Continuously acquiring three-frame pictures f to be detected of one/multiple display backgrounds through an industrial camera 1 、f 2 、f 3 And averaging the three frames of images to obtain f average The method can avoid serious random noise caused by overhigh camera ISO value, complex field environment and the like; after the picture of the LCD image to be detected is equalized, Gaussian operators, gamma transformation and histogram equalization are sequentially adopted for processing, so that the filtering of high-frequency noise and the enhancement of image brightness and contrast are realized, the important information in the image is highlighted, and the quality of the image is changed. The images under one or more backgrounds are initially acquired for subsequent processing, so that the accuracy of defect detection can be improved, and the qualification rate can be improved. The results of the data preprocessing are shown in fig. 3 ((a) a graph showing equalization, (b) a graph showing gaussian filtering, (c) a graph showing after gamma conversion, and (d) a graph showing after histogram equalization processing).
The three algorithms proposed for the six defects in the step 3 are an LCD Mura and scratch defect detection algorithm, an LCD edge defect detection algorithm and an LCD color level error defect detection algorithm.
The step 3 provides the main steps of extracting the defect regions of the LCD Mura and scratch defect detection algorithms in three different algorithms:
1) the LCD image is equally divided into m × n sub-regions of c × c pixels, and the c value is reasonably selected according to the defect size. The variance σ of each subregion is calculated by equation (1) 2 ：
In the formula:
p i ＝n i n; i is a gray value; u is the mean value of the grey values of the subareas; p is a radical of formula i Is the probability of gray value i in the sub-region; n is a radical of an alkyl radical i The number of pixels with the gray value i; n is the total number of sub-region pixels.
2) Establishing a matrix F with m rows and n columns, setting the initial value of each element in the matrix F to be 0,will be the variance σ 2 Greater than a threshold value T f Has a value of 1 for the corresponding position element of the sub-region in the matrix F, and a threshold value T f According to the gray pixel value of the defect.
3) Traversing the matrix F, marking the connected region of the F median value 1 as a defect, corresponding the element coordinates of four corner points of the connected region in the matrix F to the LCD image, and obtaining a square region formed by the four points as a defect region to be extracted.
When in segmentation, only the extracted defect area needs to be processed, so as to solve the problem that the small-area defect is difficult to segment accurately and quickly. In order to enhance the real-time performance of a vision detection system and eliminate the influence of uneven illumination on an image, a Background correction algorithm (Background correction algorithm) is introduced in the image segmentation process. And obtaining the image after background correction through subtraction operation of the pixels and the pixel mean value. And (4) dividing the corrected image into the defect regions by a maximum Inter-Class Variance method (Inter-Class Variance). The size of the defective region of the LCD is C x C pixels, the sub-regions have different gray levels of L, denoted as {0,1,2,3 … L-1}, and are classified into C, D two parts by assuming that a threshold value t is selected, and C is defined by the gray level values of [0, t ]]D is composed of gray values at t +1, L-1]All pixels of (1). Let P C (t) and P D (t) represents the probability of occurrence of class C and class D, respectively, and can be expressed as
μ C (t) and μ D (t) represent the mean gray values of class C and class D, respectively, expressed as
And
the optimal threshold T of the maximum between-class variance method can be calculated by the maximum between-class variance,
wherein, the between-class variance can be expressed as:
the formula for the optimal threshold T can be simplified as:
and dividing the defects in the sub-regions from the image by the optimal threshold T, performing feature extraction by using morphological processing, and calculating the number of the extracted defects.
Aiming at images in an experimental data set under six display backgrounds, defect regions can not be extracted by adopting an LCD Mura and scratch defect detection algorithm, and then the defect regions are extracted by adopting an LCD edge defect detection algorithm of three different algorithms (namely, if the LCD Mura and scratch defect detection algorithm can realize detection under one background, only one background is used, if one background can not be used, more than 2 backgrounds are adopted, 6 backgrounds are at most adopted, if 6 backgrounds can not be realized, the LCD edge defect detection algorithm is adopted), the method mainly comprises the following steps:
edge defects are as follows: broken lines, halos, gravity Mura, light leakage, etc.
1) Extracting the grid in the picture from the picture of the preprocessed LCD grid background through pixel value information, and intercepting the left edge and the upper edge of the grid as ROI 1 Region, cutting right and lower edges as ROI 2 Region and ROI to be intercepted 1 And ROI 2 The regions are placed in a new template, respectively, as shown in fig. 4; ROI alignment by maximum inter-class variance 1 And ROI 2 And dividing the grids in the area, and then calculating the pixel value of the particles to realize particle labeling so as to judge whether the edge of the LCD has defects. The particle marking method mainly judges the ROI according to a 4-connectivity judgment criterion 1 And ROI 2 Whether the regions are labeled as the same type of particle. The searching algorithm of the image grain mark mainly comprises the following steps:
1) ROI (region of interest) 1 And ROI 2 Region(s)The scanning is divided into 12 lines, starting from the first line of the image, a sequence of non-0 pixels is formed into a block, and the flag value is set for each block in its ascending order.
2) Scanning the blocks in all the lines line by line starting from the second line, and if the blocks are not communicated with all the blocks in the previous line, setting a new mark value to the blocks; if it is connected to only one block in the previous row, assigning the block to the tag value of that block in the previous row; if it has connectivity to more than two blocks in the previous row, the smallest token value in these connected blocks is assigned to these connected blocks.
3) Step 2 is repeated until all pixels are marked.
Respectively aligning ROI in the new template according to connectivity judgment criterion 1 And ROI 2 Judging whether the number of the particles marked by the area is equal to 1 or not, and judging whether the edge of the LCD has defects or not: if the number of marked particles is equal to 1, the grid lines at the edges are completely divided, and the defects of light leakage, image covering and the like do not exist; otherwise, the LCD edge defect can be judged.
The main steps for providing the defects of the LCD color gradation error defect detection algorithm in the three different algorithms in the step 3 are as follows:
1) dividing the 32-level gray level into a plurality of sub-areas, and judging whether defects exist in the display background or not by calculating the mean value and the variance in the sub-areas.
2) As shown in fig. 5, the gray scale background is divided into 32 equal parts (equal to the number of gray scale levels) along the horizontal direction, and is divided into 2 equal parts along the vertical direction, which are respectively represented by U and L, and the two regions are uniformly divided into 17 rows along the vertical direction, so that 544 sub-regions can be obtained respectively. Then, the U and L regions are divided into N equal sub-regions in the horizontal direction and are simultaneously divided uniformly into M rows in the vertical direction, where N-32 and M-17. And traversing N multiplied by M sub-regions in the U region and the L region by adopting a traversal frame with the size of H multiplied by W. In addition, pixel values of sub-regions of m-th row and n-th column of the U and L regions are respectively expressed as
And
where M ∈ {1, …, M } and N ∈ {1, …, N }, and the corresponding M × N sub-regions have corresponding grayscale pixel mean of
And standard deviation of
Wherein I is belonged to { U, L },
after the average value and the standard deviation of 544 sub-regions of the U and L regions are calculated respectively, whether the LCD display panel has the color level deviation defect can be judged according to the pixel average value and the standard deviation between pixels of adjacent sub-regions. In particular, in the U region, if the mean value of the gray pixels in each sub-region from left to right is
Are sequentially increased, i.e.
And the difference between the gray average values of any adjacent subregions is 8, and the difference between the standard deviations of any adjacent subregions is 4, the U region in the LCD display panel can be judged to have no color gradation error defect; similarly, in the L region, if the mean of gray pixels in each sub-region from left to right
Successively decrease, i.e.
If the difference between the mean values of the gray-scale pixels of the adjacent subregions is 8 and the difference between the standard deviations of the adjacent subregions is 4, the L region in the LCD display panel has no color gradation error defect; otherwise, it can be considered that the LCD display panel has the color gradation error defect.
The experimental results obtained by the LCD Mura and scratch defect detection algorithm in the three different detection algorithms in the step 3 are shown in FIG. 6 and Table 1 (in FIG. 6, (a) -preprocessing image, (b) -Clustering, (c) -inversion, (d) -Metric, (e) -Moments, (f) -Otsu, (g) -artificial segmentation). The analytical procedure was as follows:
TABLE 1 segmentation results of the segmentation algorithms
1) It can be seen from the splitting effect of the spot Mura that the Otsu splitting algorithm in the LCD Mura and scratch defect detection algorithms corresponding to table 1 can more accurately split the defects, and the other four algorithms cannot be effectively split, mainly because the Otsu splitting algorithm has a larger variance between the two types when the difference between the pixel numbers is larger, the splitting effect is better.
2) The 5 defects are respectively segmented through a 5-segmentation algorithm, and as can be seen from the segmentation effect, the Otsu segmentation algorithm can completely segment the 5 defects and has a small difference with a manual segmentation result;
3) by comparing the segmentation results, Otsu can be suitable for segmenting the Mura and the scratch defects, other algorithms only have good segmentation effects on the scratches and the line Mura defects, and the segmentation effects on spot Mura and region Mura are poor; therefore, the Otsu algorithm is more practical to select for Mura and scratch defects, and the segmentation range is wider.
The experimental results obtained by the LCD edge defect detection algorithm in the three different detection algorithms in the step 3 are shown in FIG. 7 (in FIG. 7, (a) -preprocessing image, (b) -Clustering, (c) -inversion, (d) -Metric, (e) -Moments, (f) -Otsu). The analytical procedure was as follows:
1) as can be seen from the comparison graph of the segmentation result of the edge defect, for the segmentation of the edge defect, no matter using any one of the segmentation algorithms in fig. 7, the edge defect cannot be well segmented, and the segmentation effect is poor, so that the background and the foreground of the LCD image cannot be clearly distinguished. Therefore, when the edge defect has a large overlap with the gray level of the background, the above 5-division algorithm cannot accurately separate the object from the background.
2) As shown in fig. 8, a grid is used as a display background, regions ROI1 and ROI2 are cut out and respectively placed in a new template, grids in the two regions are segmented by an Otsu algorithm in an LCD edge defect detection algorithm, and after 4 connectivity judgment criteria particles are labeled, the ROI is found 2 The grids in the area cannot be marked into a whole, and more short lines exist, so that the existence of edge defects of the LCD can be judged.
Fig. 11 and 12 show experimental results obtained by the LCD color level error defect detection algorithm in the three different detection algorithms of step 3. The analysis process is as follows:
1) the pseudo tone scale error defect image is shown in fig. 9 and 10. In fig. 9, 32-level gray scales are used as a display background, a scale error defect exists in the U region, and a scale error defect exists in the L region in fig. 10. Fig. 11 and 12 are calculation results of the gradation pixels of fig. 9 and 10, respectively.
2) Fig. 11(a) and 11(b) depict the gray pixel mean and standard deviation of the second row in fig. 9 in the case of no-defect and defect. Fig. 11(c) and 11(d) illustrate the pixel mean values and standard deviations of the 16 th gray scale in fig. 9 under non-defective and defective conditions. Fig. 11 shows the tendency of the average value of the pixels of the second row increasing linearly from left to right in the case of no defect, and the standard deviation in each sub-area is 0. In the case of no defect, the average value of the pixel of the 16 th gray scale is 128, and the standard deviation in the sub-area is 0. Further, it is easily seen from the graph that, in the case of a defect, there is a significant fluctuation in the average value and standard deviation of the pixels in the 14 th to 19 th gray levels, and there is a significant fluctuation in the average value and standard deviation of the pixels in the 1 st to 5 th rows of the 16 th gray levels, indicating that the LCD display panel has a defect in the color level error, and that the defect is located in the 1 st to 5 th rows and the 14 th to 16 th levels of the color level image U region.
3) Fig. 12(a) and 12(b) depict the pixel mean and standard deviation, respectively, for the 10 th row in fig. 10, with and without defects. Fig. 12(c) and 12(d) depict the pixel mean and standard deviation of the 18 th gray level in fig. 10 in the case of no defect and defect, respectively. This result shows that in the defect-free case, the pixel mean values of the 10 th row decrease linearly from left to right, with the standard deviation of the 10 th row being equal to 0. In the defect-free case, the average value of the pixels of the 18 th gray level is equal to 120, and the standard deviation of the 18 th gray level is 0. Further, in the case of a defect, the pixel average value and the standard deviation of 16 th to 20 th gray levels fluctuate, which indicates that the LCD display panel has a color level defect, and the color level defect is located in the 5 th to 11 th rows and 16 th to 20 th levels of the color level image L region.
While the present invention has been described in detail with reference to the embodiments shown in the drawings, the present invention is not limited to the embodiments, and various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art.
Claims (6)
1. A method for visually detecting defects of a liquid crystal display under multiple backgrounds in real time is characterized by comprising the following steps: the method comprises the following specific steps:
step 1, carrying out image preprocessing on a defect data set of a display under one or more acquired display backgrounds to obtain an experimental data set;
step 2, classifying the defects of the display in the experimental data set into six different defects;
step 3, detecting whether the image in the experimental data set has defects by using three different detection algorithms; wherein, three different detection algorithms include: LCD Mura and scratch defect detection algorithm, LCD edge defect detection algorithm and LCD color level error defect algorithm;
the display defects are classified into six different defects, which specifically include: spotMura, line Mura, region Mura, tone scale, scratch and edge defect;
the step 3 is specifically as follows:
LCD Mura and scratch defect detection algorithm: marking a defect area by aiming at one or more images in an experimental data set under a display background through the relation between pixel variance and a threshold value; aiming at the defect area, obtaining an image with a corrected background through subtraction operation of pixels and pixel mean values; dividing the corrected image into defect regions by a maximum inter-class variance method; carrying out feature extraction by using particle morphology, and calculating the number of extracted defects to obtain a visual defect effect diagram; wherein the plurality is at most six;
LCD edge defect detection algorithm: aiming at images in an experimental data set under six display backgrounds, defect regions cannot be extracted by adopting LCD Mura and scratch defect detection algorithms, taking a grid as a detection background of LCD edge defects, intercepting grid edges for recombination, segmenting a recombined grid by adopting a maximum inter-class variance method, and extracting defect characteristics after particle morphology processing to obtain a visual defect effect diagram;
LCD color level error defect algorithm: if the defect area is not extracted by the LCD edge defect detection algorithm, the 32-level gray gradient is used as a display background, and whether the LCD has the color level deviation defect or not is judged by dividing the mean value and the standard deviation of the gray values of the subareas.
2. The method for visually inspecting defects of a liquid crystal display under multiple backgrounds in real time according to claim 1, wherein: the preprocessing is specifically to perform averaging, gaussian filtering, gamma transformation and histogram equalization in sequence.
3. The method for visually inspecting defects of a liquid crystal display under multiple backgrounds according to claim 2, wherein: the equalization specifically comprises the following steps: continuously acquiring three frames of images f for one/more display backgrounds to be detected through an industrial camera 1 、f 2 、f 3 And averaging the three frames of images to obtain f average 。
4. The method for visually inspecting defects of a liquid crystal display under multiple backgrounds in real time according to claim 1, wherein: the LCD edge defect detection algorithm specifically comprises the following steps: taking the grid as a display background, extracting the grid in the picture from the picture of the LCD grid background after pretreatment through pixel value information, and intercepting the left edge and the upper edge of the grid as an ROI 1 Region, cutting right and lower edges as ROI 2 Region, intercepting ROI 1 And ROI 2 The regions are respectively placed in a new template, grids in the two regions are segmented by a maximum inter-class variance method, and defect features are extracted after particle morphology processing to obtain a visual defect effect diagram: if ROI 1 And ROI 2 If the number of the particles of the area mark is equal to 1, the edge of the LCD has no defect; otherwise, judging that the LCD edge has defects.
5. The method for visually inspecting defects of liquid crystal display in multiple backgrounds according to claim 1, wherein: the LCD color level error defect algorithm specifically comprises the following steps: dividing the gray scale background into 2 equal parts along the vertical direction, and respectively representing the 2 equal parts as a U area and an L area:
in the U area, if the gray level mean value in each row in the traversal frame is sequentially increased along with the increasing of the gray level, the standard deviation is 0; the difference between the mean values of the gray values in two adjacent traversal frames is 8, and the standard deviation is 4; if the gray values in the same gray scale are the same and the standard deviation is 0, indicating that the color scale of the region has no deviation defect; otherwise, the color level deviation exists in the region;
in the L area, if the gray average value in the traversal frame in each row is sequentially reduced along with the increasing of the gray level, the standard deviation is 0; the difference between the mean values of the gray values in two adjacent traversal frames is 8, and the standard deviation is 4; if the gray values in the same gray scale are the same and the standard deviation is 0, indicating that the color scale of the region has no deviation defect; otherwise, the region has a color level deviation.
6. The method for visually inspecting defects of a liquid crystal display under multiple backgrounds according to claim 5, wherein: dividing the U area and the L area into N equal sub-areas along the horizontal direction, uniformly dividing the sub-areas into M lines along the vertical direction, and traversing the N multiplied by M sub-areas in the U area and the L area by adopting an H multiplied by W traversal frame; the pixel values of the sub-regions of the m-th row and n-th column of the U and L regions are respectively expressed as
And
where M ∈ {1, …, M } and N ∈ {1, …, N }, and the corresponding M × N sub-regions have corresponding grayscale pixel mean of
And standard deviation of
Wherein, I is epsilon { U, L },
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110306033.9A CN113034464B (en) | 2021-03-23 | 2021-03-23 | Visual real-time detection method for defects of liquid crystal display under multiple backgrounds |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110306033.9A CN113034464B (en) | 2021-03-23 | 2021-03-23 | Visual real-time detection method for defects of liquid crystal display under multiple backgrounds |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113034464A CN113034464A (en) | 2021-06-25 |
| CN113034464B true CN113034464B (en) | 2022-07-26 |
Family
ID=76472842
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110306033.9A Active CN113034464B (en) | 2021-03-23 | 2021-03-23 | Visual real-time detection method for defects of liquid crystal display under multiple backgrounds |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113034464B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113628189B (en) * | 2021-08-11 | 2023-10-24 | 西安工程大学 | Rapid strip steel scratch defect detection method based on image recognition |
| CN114862801A (en) * | 2022-05-11 | 2022-08-05 | 南京巨鲨显示科技有限公司 | Method, system, device and storage medium for detecting display performance of medical display |
| CN117269197B (en) * | 2023-11-21 | 2024-02-02 | 昆明理工大学 | Method and device for detecting surface quality of polycrystalline silicon raw material |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1928535A (en) * | 2006-09-07 | 2007-03-14 | 哈尔滨工业大学 | Machine vision based LCD spot flaw detection method and system |
| JP2008170325A (en) * | 2007-01-12 | 2008-07-24 | Seiko Epson Corp | Stain flaw detection method and stain flaw detection device |
| JP2014164221A (en) * | 2013-02-27 | 2014-09-08 | Sharp Corp | Defect detection method for display panel and defect detection device for display panel |
| CN106228532A (en) * | 2016-07-05 | 2016-12-14 | 中科院成都信息技术股份有限公司 | A kind of lcd screen sub-pixel defect inspection method |
| CN106650770A (en) * | 2016-09-29 | 2017-05-10 | 南京大学 | Mura defect detection method based on sample learning and human visual characteristics |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP5865707B2 (en) * | 2012-01-06 | 2016-02-17 | 株式会社キーエンス | Appearance inspection apparatus, appearance inspection method, and computer program |
-
2021
- 2021-03-23 CN CN202110306033.9A patent/CN113034464B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1928535A (en) * | 2006-09-07 | 2007-03-14 | 哈尔滨工业大学 | Machine vision based LCD spot flaw detection method and system |
| JP2008170325A (en) * | 2007-01-12 | 2008-07-24 | Seiko Epson Corp | Stain flaw detection method and stain flaw detection device |
| JP2014164221A (en) * | 2013-02-27 | 2014-09-08 | Sharp Corp | Defect detection method for display panel and defect detection device for display panel |
| CN106228532A (en) * | 2016-07-05 | 2016-12-14 | 中科院成都信息技术股份有限公司 | A kind of lcd screen sub-pixel defect inspection method |
| CN106650770A (en) * | 2016-09-29 | 2017-05-10 | 南京大学 | Mura defect detection method based on sample learning and human visual characteristics |
Non-Patent Citations (3)
| Title |
|---|
| 《In-TFT-array-process micro defect inspection using nonlinear principal component analysis》;Yi-Hung Liu等;《Int J Mol Sci》;20091120;全文 * |
| 《TFT-LCD Mura缺陷机器视觉检测方法》;毕昕等;《机械工程学报》;20100630;全文 * |
| 《基于机器视觉的液晶屏Mura缺陷检测方法》;钱基德等;《计算机科学》;20180630;全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113034464A (en) | 2021-06-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113034464B (en) | Visual real-time detection method for defects of liquid crystal display under multiple backgrounds | |
| CN110286126B (en) | Wafer surface defect regional detection method based on visual image | |
| CN111161243B (en) | Industrial product surface defect detection method based on sample enhancement | |
| CN111179225B (en) | Test paper surface texture defect detection method based on gray gradient clustering | |
| CN110389127B (en) | System and method for identifying metal ceramic parts and detecting surface defects | |
| CN109978839B (en) | Method for detecting wafer low-texture defects | |
| CN107845087A (en) | The detection method and system of the uneven defect of liquid crystal panel lightness | |
| CN111915704A (en) | Apple hierarchical identification method based on deep learning | |
| CN108961230B (en) | Identification and extraction method for structural surface crack characteristics | |
| CN110443791B (en) | Workpiece detection method and device based on deep learning network | |
| CN110648330B (en) | Defect detection method for camera glass | |
| CN113706490B (en) | Wafer defect detection method | |
| CN114119591A (en) | Display screen picture quality detection method | |
| CN114910480A (en) | Wafer surface defect detection method based on machine vision | |
| CN112308854A (en) | Automatic detection method and system for chip surface flaws and electronic equipment | |
| CN113034488A (en) | Visual detection method of ink-jet printed matter | |
| CN110781913A (en) | Zipper cloth belt defect detection method | |
| CN115078365A (en) | Soft package printing quality defect detection method | |
| CN117115171B (en) | Slight bright point defect detection method applied to subway LCD display screen | |
| CN110530894B (en) | Light guide plate bright spot defect detection method | |
| CN108898584B (en) | Image analysis-based full-automatic veneered capacitor welding polarity discrimination method | |
| CN113192061B (en) | Extraction method and device of LED package appearance detection image, electronic equipment and storage medium | |
| CN111161264A (en) | Method for segmenting TFT circuit image with defects | |
| CN114705696A (en) | Defect detection method based on digital image processing and automatic optical detection | |
| CN115546141A (en) | Small sample Mini LED defect detection method and system based on multi-dimensional measurement |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |