CN110533657A - A kind of liquid crystal display appearance detecting method - Google Patents
A kind of liquid crystal display appearance detecting method Download PDFInfo
- Publication number
- CN110533657A CN110533657A CN201910821718.XA CN201910821718A CN110533657A CN 110533657 A CN110533657 A CN 110533657A CN 201910821718 A CN201910821718 A CN 201910821718A CN 110533657 A CN110533657 A CN 110533657A
- Authority
- CN
- China
- Prior art keywords
- defect
- boundary
- liquid crystal
- crystal display
- profile
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 239000004973 liquid crystal related substance Substances 0.000 title claims abstract description 32
- 230000007547 defect Effects 0.000 claims abstract description 101
- 238000001514 detection method Methods 0.000 claims abstract description 31
- 238000007689 inspection Methods 0.000 claims abstract description 27
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 24
- 230000011218 segmentation Effects 0.000 claims abstract description 14
- 238000004458 analytical method Methods 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 8
- 239000000284 extract Substances 0.000 claims description 6
- 238000010801 machine learning Methods 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 5
- 229910002056 binary alloy Inorganic materials 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 230000008859 change Effects 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 3
- 230000002452 interceptive effect Effects 0.000 claims description 3
- 238000002372 labelling Methods 0.000 claims description 3
- 238000013215 result calculation Methods 0.000 claims description 3
- 230000004069 differentiation Effects 0.000 claims 1
- 230000002950 deficient Effects 0.000 abstract description 10
- 238000004519 manufacturing process Methods 0.000 abstract description 5
- 238000012545 processing Methods 0.000 abstract description 5
- 238000010586 diagram Methods 0.000 description 14
- 238000012360 testing method Methods 0.000 description 4
- 230000006978 adaptation Effects 0.000 description 3
- 239000013598 vector Substances 0.000 description 3
- 238000005192 partition Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 206010000496 acne Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 238000005520 cutting process Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 230000009977 dual effect Effects 0.000 description 1
- 238000005530 etching Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration by the use of local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G06T5/70—
-
- 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
-
- 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
- G06T7/13—Edge detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/8874—Taking dimensions of defect into account
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/888—Marking defects
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8887—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges based on image processing techniques
-
- 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
- G06T2207/10004—Still image; Photographic 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
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
Abstract
A kind of liquid crystal display appearance detecting method calculates step, determining defects step including the inspection area segmentation step based on profile, detection algorithm in conjunction with multiscale space；Inspection area segmentation step based on profile mainly includes the case where following below scheme: (1) to each edge profile upper and lower, left and right with the presence or absence of being described, (2) method on boundary is determined, (3) the case where decision boundaries angle is right angle, (4) determine boundary；Detection algorithm calculates in conjunction with multiscale space the following steps are included: (1) generates multi-scale image sequence, and (2) detect the liquid crystal display defect under different scale；The following steps are included: (1) defect database marks, the 2D of (2) defective data is shown determining defects, and (3) learn boundary curve.The present invention checks that precision is high and speed is fast, has saved cost of labor, has improved detection and production efficiency.Overcome that prior art speed is slow, processing is complicated, defective missing inspection probability, determining defects inaccuracy disadvantage.
Description
Technical field
The present invention relates to liquid crystal display panel detection technique field, especially a kind of liquid crystal display appearance detecting method.
Background technique
Liquid crystal display panel presentation quality inspection is to guarantee the premise of liquid crystal display panel normal use.Mainly for not attaching polarizer
Liquid crystal display panel appearance checked, to detect scratched on panel, rub wound, sheet salient point, Pimple (dotted salient point),
The open defects such as Dimple (side concave point C, the side T), etching unevenness, Rimple (water ripples shape).Basic principle is to utilize linear array phase
Machine finally obtains the target detection object to liquid crystal display panel progress Image Acquisition, by algorithm analysis and lacks with the presence or absence of above-mentioned
It falls into.Traditional liquid crystal display panel appearance inspection method mainly includes setting inspection area, defects detection and defect three steps of classification.
Firstly, inspection area passes through setting at least three Mark (label), then inspection area is determined by way of template matching, (as schemed
Shown in 1), but as the resolution of panel physics is increasing, image is all typically now acquired using the line scan camera of 16K, is made
The method speed for obtaining template matching is slow, and processing is complicated.Secondly, the image of line scan camera actual photographed is in certain ruler
It is obtained under degree, there is no the characteristics of multiresolution, defect can not be detected according to the size adaptation of defect, so that there are defective
The probability of missing inspection.Finally, the threshold value of traditional determining defects is all hard -threshold, (as shown in Figure 7), the judgement inaccuracy of defect.
Summary of the invention
In order to overcome drawback present in available liquid crystal panel appearance quality examination, the present invention provides be mainly used in liquid
Crystal panel manufacturing enterprise uses, and replaces manual inspection with hardware check, using automatic segmentation algorithm, utilizes mark (label) position
It sets, by the big principle of pixel variation on boundary in grayscale image, two secondary black, the white disparity maps generated using field calculus of finite differences
Boundary profile is extracted, so that it is determined that detection region, and defect inspection is carried out using method of the traditional algorithm in conjunction with multiscale space
It looks into, accurate profile amount is provided by traditional Blob analysis (binary system), while being checked under multiple groups conditional parameter, enhanced
Checking ability, the method for determining defects based on machine learning obtain model parameter by the study of defect sample, determine most optimal sorting
Class curve, considerably increases determining defects accuracy, thus reaches and checks that precision is high and speed is fast, has saved cost of labor, mentioned
A kind of high liquid crystal display appearance detecting method of detection efficiency and production efficiency.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of liquid crystal display appearance detecting method, it is characterised in that using line scan camera, PC machine as detection instrument, line is swept
It retouches camera and is connected with PC machine through data line, and is interactive with system software information in PC, include the test zone based on profile in detection
Regional partition step, detection algorithm calculate step, determining defects step in conjunction with multiscale space；The test zone based on profile
Regional partition step mainly includes the case where following below scheme: (1) according to camera and image taking, to each edge profile upper and lower, left and right
With the presence or absence of the case where be described, (2) in the presence of boundary profile, the method for determining boundary, (3) decision boundaries angle is
The case where the case where right angle is not right angle includes fillet, chamfering etc., and (4) determine boundary；The detection algorithm and multiscale space
In conjunction with calculating the following steps are included: in (1) checking process, to the image being originally inputted, continuous boil down to it is original 1/4, generate
Multi-scale image sequence, (2) carry out inspection one by one to the image in multi-scale image sequence using binarization method, with detection
Liquid crystal display defect under different scale；The determining defects are the following steps are included: (1) defect database label, (2) defective data
2D show, (3) learn boundary curve.
Further, in the inspection area segmentation step based on profile, when being based in grayscale image, pixel becomes on boundary
When changing the big determining boundary of difference, since in gray level image, the gray value or difference value of boundaries on either side are changed greatly, rather than boundary
On gray scale or difference value variation slowly, boundary point can be extracted based on the above principles, also can be used field calculus of finite differences generation
Two secondary black, white disparity maps extract boundary profile；When decision boundaries angle being not right angle the case where, the boundary of two kinds of fillets, chamfering
Angle differentiates that process is as follows, draws a circle by radius of square side, using dot as starting point, to put on circle as terminal, every 20 degree, seeks
Boundary point is looked for, these points are finally fitted to circle, draws a circle by radius of square side, using dot as starting point, is to be put on circle
Terminal finds boundary point every 20 degree, and finally these point fittings are in line.
Further, during the detection algorithm calculates in conjunction with multiscale space, the liquid crystal display detected under different scale is lacked
Sunken step is as follows: (1) binarization method used to multi-scale image sequence, obtains the bianry image of multi-scale image sequence,
And binary map is generated using many condition parameter；(2) binary map is merged, finally obtains the strong condition two-value of a merging
The weak condition binary map of figure and a merging；(3) line defect is analyzed in binary map under combined weak condition, applies two-value first
Expansion, then two-value erosion removal noise is used, line is then reconnected, obtains doing the binary map that binary approach is analyzed；(4) two are carried out
The analysis of system method, and the characteristic quantity of the result calculating line defect according to binary approach analysis；(5) after having analyzed line but,
Remove the defect in the binary map obtained under the conditions of strong on line defect position；(5) analysis of block defect, point of block defect
Analysis be remove line defect it is strong under the conditions of obtained binary map on carry out；(6) remove line defect it is strong under the conditions of
To binary map on carry out Binary analysis, and the characteristic quantity of the result calculation block defect according to Binary analysis.
Further, in the defect database label of the determining defects step, according to the characteristic parameter area of defect, length
Flaw labeling is to be or be not by degree, width, contrast, shape.
Further, during the 2D of the defective data of the determining defects step is shown, optional two features are schemed to draw 2D,
Respectively as X-axis and Y-axis, it is shallow labeled as no that it is aterrimus labeled as being that each defect point, which is retouched on the 2 d image,
Black.
Further, in the study boundary curve of the determining defects step, it is based on machine learning algorithm support vector machines,
It is automatically learned boundary curve；After the expression of boundary curve and given defect sample data, sample is determined according to boundary curve
Be and be not label.
The medicine have the advantages that the present invention passes through grayscale image using mark (label) position using automatic segmentation algorithm
The big principle of pixel variation on middle boundary extracts boundary profile using two secondary black, the white disparity maps that field calculus of finite differences generates,
So that it is determined that detection region, and defect inspection has been carried out using method of the traditional algorithm in conjunction with multiscale space, pass through tradition
Blob analysis (binary system) provides accurate profile amount, while being checked under multiple groups conditional parameter, and checking ability is enhanced；Base
In the method for determining defects of machine learning, model parameter is obtained by the study of defect sample, determines optimal classification curve, significantly
Determining defects accuracy is increased, thus reaches and checks that precision is high and speed is fast, saved cost of labor, improved detection efficiency
With production efficiency mesh.It overcomes in the prior art, template matching method speed is slow, and processing is complicated, can not be according to defect
Size adaptation detect defect so that haveing the shortcomings that the probability of defective missing inspection, judgements of defect are inaccurate.Based on upper
It states, so the application prospect that the present invention has had.
Detailed description of the invention
Fig. 1 is traditional label setting schematic diagram.
Fig. 2 is the schematic diagram of fillet of the present invention, chamfering.
Fig. 3 is fillet profile detection schematic diagram of the present invention.
Fig. 4 is chamfering profile detection schematic diagram of the present invention.
Fig. 5 multi-scale image sequence diagram.
Fig. 6 is appearance testing result schematic diagram of the present invention.
Fig. 7 is traditional shortcoming decision logic schematic diagram.
Fig. 8 is present invention boundary curve model schematic diagram.
Fig. 9 is learning outcome curve synoptic diagram of the present invention.
Figure 10 is liquid crystal display panel image schematic diagram of the present invention.
Figure 11 is ROI (area-of-interest) schematic diagram of liquid crystal display panel of the present invention.
Figure 12 is the average gray of the average gray value and defect area in the external positive rectangle of the present invention in non-defective region
The absolute value of the difference schematic diagram of value.
Figure 13 is the wide schematic diagram of boundary rectangle of the present invention.
Figure 14 is the long schematic diagram of angle of the present invention, oblique rectangle.
Specific embodiment
Shown in Fig. 2,3,4,5,6,8,9,10,11, a kind of liquid crystal display appearance detecting method, using line scan camera, PC machine
As detection instrument, line scan camera is connected with PC machine through data line, and interactive with system software information in PC, detection method packet
Include the inspection area segmentation step based on profile, detection algorithm calculates step, determining defects step in conjunction with multiscale space；Institute
It states the inspection area segmentation step based on profile and mainly includes the case where following below scheme: (1) according to camera and image taking, to each
Edge profile upper and lower, left and right with the presence or absence of the case where be described, (2) in the presence of boundary profile, the method for determining boundary,
(3) the case where the case where decision boundaries angle is right angle is not right angle includes fillet, chamfering etc., and (4) determine boundary；It is described
Detection algorithm is calculated in conjunction with multiscale space the following steps are included: in (1) checking process, to the image being originally inputted, constantly
Boil down to it is original 1/4, generate multi-scale image sequence, (2) are using binarization method to the image in multi-scale image sequence
Inspection one by one is carried out, to detect the liquid crystal display defect under different scale；The determining defects are the following steps are included: (1) defect
The 2D of database flags, (2) defective data shows that (3) learn boundary curve.
Shown in Fig. 2,3,4, in the inspection area segmentation step based on profile, when being based in grayscale image, pixel changes on boundary
When the big determining boundary of difference, since in gray level image, the gray value or difference value of boundaries on either side are changed greatly, rather than on boundary
Gray scale or difference value variation slowly, boundary point can be extracted based on the above principles, also can be used field calculus of finite differences generation two
Secondary black, white disparity map extracts boundary profile (Fig. 2)；When decision boundaries angle being right angle the case where, two kinds of fillets, chamfering
Rim angle differentiates that process is as follows, draws a circle by radius of square side, using dot as starting point, to put on circle for terminal, every 20
Degree finds boundary point, these points are finally fitted to circle (Fig. 3)；A circle is drawn by radius of square side, using dot as starting point,
Put on circle as terminal, every 20 degree, boundary point is found, these point fittings are finally in line (Fig. 4).
Shown in Fig. 5, detection algorithm calculates in step in conjunction with multiscale space, when checking process, to the figure being originally inputted
Picture, continuous boil down to it is original 1/4, generate multi-scale image sequence (Fig. 5).Detecting algorithm can be in multi-scale image sequence
Image carries out inspection one by one and (traditional sauvola (binaryzation) method is used, to detect the defect under different scale；It uses
The reason of this algorithm is that human eye has the characteristics that multiresolution；Different scale images processing is parallel computation.Sauvola bis-
Value Method And Principle: belonging to one kind of local threshold binaryzation, and the width and height of local window can be used as variable and be configured, if
Assume that current pixel position is (r, c) for w*h, then the binarization threshold of current location are as follows:
Wherein u (r, c) indicates that the average gray value in this window ranges, σ (r, c) are corresponding standard deviations, and R represents possibility
Maximum standard.
In conjunction with traditional sauvola binaryzation and multi-scale image time series technique, the following (figure of liquid crystal display open defect is detected
5).There are following steps in this checking process: (1) multi-scale image sequence being obtained using Sauvola binarization method
Binary map is generated to the bianry image of multi-scale image sequence, and using many condition parameter.(2) binary map is merged, most
The strong condition binary map an of merging and the weak condition binary map of a merging are obtained eventually；There are three ways to merging, votes
Between method, binary map make with, between binary map make or；The logic that binary map merges is the same scale for generating many condition parameter
Layer binary map merges, different scale layer nonjoinder.(3) line defect is analyzed in the binary map under combined weak condition, herein mistake
It is expanded first using two-value in journey, then with two-value erosion removal noise, then reconnects line, thus obtained doing Blob analysis
Binary map.(4) Blob (binary system) analysis is carried out, and calculates the characteristic quantity of line defect according to the result of Blob analysis.(5) In
After having analyzed line defect, the defect in the binary map obtained under the conditions of strong on line defect position is removed, this is to divide
Block defect is analysed, such line defect will not influence the analysis of block defect.(6) analysis of block defect is clear in the analysis of block defect
In addition to line defect it is strong under the conditions of carry out in obtained binary map.(7) remove line defect it is strong under the conditions of obtained two-value
Blob analysis, and the characteristic quantity of the result calculation block defect according to Blob analysis are carried out on figure.The characteristic quantity of calculating such as following table
Lattice:
Fig. 7, shown in 8, determining defects include the following steps, defect database label: according to the characteristic parameter area of defect,
Flaw labeling is OK (YES) or NG (not being) by length, width, contrast, shape etc..The 2D of defective data is shown: optionally
Two features (such as the area of reaction flaw size with the contrast of reacting defect severity) are schemed to draw 2D, respectively as
X-axis and Y-axis；Each defect point is retouched on 2D image；It is aterrimus labeled as OK, is light/dark balance (figure labeled as NG
8)；Learn boundary curve: research machine learning algorithm support vector machines (SVM) is automatically learned boundary curve；Boundary curve
Expression；After given defect sample data, determine that the OK/NG of sample is marked according to boundary curve.It is as follows that SVM describes process:
A known data set(Fig. 7), sampling feature vectorsI.e.It is
D ties up the vector in real number space；Class label y ∈ { -1 ,+1 }, i.e. only two class samples；Due to needing to carry out to these two types of samples
Classification, therefore, target are to find optimum segmentation hyperplane, i.e., determine that the segmentation of maximum class interval is super flat according to training sample
Face.
If optimal hyperlane equation isThe distance of plane is arrived according to point, then sampleWith best hyperplaneThe distance between beStandardize to hyperplane, selects the sample so that nearest apart from hyperplaneIt is full
Foot'sWith b to get to standardization hyperplane.
The distance from nearest sample to edge is at this time
And class interval becomes
So far, problem is converted into searching so that the class interval i.e. maximized normal vector of (2) formulaIt willSubstitute into relational expressionB can be obtained.
(2) formula of maximization is equivalent to minimize:
In addition to this, there are also constraint conditions below:
This is a typical constrained extremal problem, and lagrange's method of multipliers solution can be used.
By being multiplied by a Lagrange's multiplier α to each of formula (4) constraint conditioni, this conditional extremum can be asked
Topic is converted into following free optimization problem, aboutB and αi(i=1,2 ..., N) minimizes L
Ask L pairsWith the partial derivative of b, and enable its be equal to zero, then have
(5) formula is unfolded
(6), (7) formula substitutes into (8) formula and obtains
Above formula withB is unrelated, only αiFunction, be denoted as
Constraint condition at this time are as follows: αi>=0 and
So far, lagrange duality problem is converted to, the convex quadratic programming problem about α
After solving α,It can be determined by the solution α of dual problem with b, be
Optimal classification function is
According to the above method, the classification of progress determines that result is as shown in Figure 9.
Shown in Fig. 5,6,10,11, when concrete application of the present invention, 1. using the liquid crystal display panel of line scan camera shooting clears outside
It sees image (Figure 10).2. firstly, the profile of search liquid crystal display panel appearance images obtains test zone according to contours extract algorithm
Domain and region of interest ROI (Figure 11)；Secondly, again cutting the inspection area of image, and require the width of the image cut
Degree is 256 multiple, to generate multi-scale image sequence well.3. generating multi-scale image sequence, it to be used for local binary
Change handles (Fig. 5).4. pair multi-scale image sequence uses the Sauvola binarization method of multi-parameter condition, obtain in different ginsengs
The bianry image of multi-scale image sequence under said conditions, then these bianry images are merged by bianry image folding
Bianry image sequence.5. pair bianry image sequence carries out Blob analysis, and calculates the feature of defect according to the result of Blob analysis
Amount.6. the characteristic quantity obtained according to step 5 substitutes into different defect sorting parameter models and determines final result.The present invention uses
Automatic segmentation algorithm uses neck by the big principle of pixel variation on boundary in grayscale image using mark (label) position
Two secondary black, the white disparity maps that domain calculus of finite differences generates extract boundary profiles, so that it is determined that detection region, and using traditional algorithm and more
The method that scale space combines has carried out defect inspection, provides accurate profile amount by traditional Blob analysis (binary system), simultaneously
It is checked under multiple groups conditional parameter, enhances checking ability, the method for determining defects based on machine learning passes through defect sample
This study obtains model parameter, determines optimal classification curve, and considerably increasing determining defects accuracy, (Fig. 6 is appearance detection
As a result), thus reach and check that precision is high and speed is fast, saved cost of labor, improved the mesh of detection efficiency and production efficiency
Ground.It overcomes in the prior art, template matching method speed is slow, and processing is complicated, can not be according to the size adaptation of defect
Defect is detected, so that haveing the shortcomings that the judgement inaccuracy of the probability of defective missing inspection, defect.
Basic principles and main features and advantages of the present invention of the invention have been shown and described above, for this field skill
For art personnel, it is clear that the present invention is limited to the details of above-mentioned exemplary embodiment, and without departing substantially from this spirit of the invention or
In the case where essential characteristic, the present invention can be realized in other specific forms.Therefore, in all respects, should all incite somebody to action
Embodiment regards exemplary as, and is non-limiting, the scope of the present invention by appended claims rather than on state
Bright restriction, it is intended that including all changes that fall within the meaning and scope of the equivalent elements of the claims in the present invention
It is interior.
In addition, it should be understood that although this specification is described in terms of embodiments, but not each embodiment is only wrapped
Containing an independent technical solution, this description of the specification is merely for the sake of clarity, and those skilled in the art should
It considers the specification as a whole, the technical solution in embodiment may also be suitably combined to form those skilled in the art can
With the other embodiments of understanding.
Claims (6)
1. a kind of liquid crystal display appearance detecting method, it is characterised in that using line scan camera, PC machine as detection instrument, line scanning
Camera is connected with PC machine through data line, and interactive with system software information in PC, includes the inspection area based on profile in detection
Segmentation step, detection algorithm calculate step, determining defects step in conjunction with multiscale space；The inspection area based on profile
Segmentation step mainly includes the case where following below scheme: (1) according to camera and image taking, to each edge profile upper and lower, left and right four
A direction with the presence or absence of the case where be described, (2) in the presence of boundary profile, the method for determining boundary, (3) decision boundaries angle
Whether right angle the case where, the case where not being right angle includes fillet, chamfering etc., and (4) determine boundary；The detection algorithm and more rulers
Spend space and combine and calculate the following steps are included: in (1) checking process, to the image being originally inputted, continuous boil down to it is original 1/
4, multi-scale image sequence is generated, (2) carry out inspection one by one to the image in multi-scale image sequence using binarization method,
To detect the liquid crystal display defect under different scale；The following steps are included: (1) defect database marks, (2) lack the determining defects
The 2D for falling into data shows that (3) learn boundary curve.
2. a kind of liquid crystal display appearance detecting method according to claim 1, which is characterized in that the inspection area based on profile
In segmentation step, in based on grayscale image on boundary when the big determining boundary of pixel variation, due in gray level image, boundary
The gray value or difference value of two sides change greatly, rather than borderline gray scale or difference value variation are slowly, based on the above principles can
Boundary point is extracted, two secondary black, the white disparity maps that the generation of field calculus of finite differences also can be used extract boundary profile；When decision boundaries angle
When the case where not being right angle, two kinds of fillets, the rim angle differentiation process of chamfering are as follows, draw a circle by radius of square side, with
Dot is starting point, put on circle as terminal, every 20 degree, finds boundary point, these points is finally fitted to circle, with square side
A circle is drawn for radius, using dot as starting point, put on circle as terminal, every 20 degree, boundary point is found, finally these points is fitted
It is in line.
3. a kind of liquid crystal display appearance detecting method according to claim 1, which is characterized in that detection algorithm and multiple dimensioned sky
Between combine calculate in, detect different scale under liquid crystal display defect the step of it is as follows: (1) to multi-scale image sequence use two-value
Change method obtains the bianry image of multi-scale image sequence, and generates binary map using many condition parameter；(2) by binary map into
Row merges, and finally obtains the strong condition binary map an of merging and the weak condition binary map of a merging；(3) in the weak item of merging
Line defect is analyzed under part in binary map, is expanded first using two-value, then with two-value erosion removal noise, then reconnects line, obtains
To the binary map for doing binary approach analysis；(4) binary approach analysis, and the result meter analyzed according to binary approach are carried out
Calculate the characteristic quantity of line defect；(5) after having analyzed line defect, position where line defect in the binary map obtained under the conditions of strong is removed
The defect set；(5) analysis of block defect, the analysis of block defect be remove line defect it is strong under the conditions of obtained binary map
Upper progress；(6) remove line defect it is strong under the conditions of obtained binary map on carry out Binary analysis, and according to binary system point
The characteristic quantity of the result calculation block defect of analysis.
4. a kind of liquid crystal display appearance detecting method according to claim 1, which is characterized in that the defect of determining defects step
In database flags, according to the characteristic parameter area of defect, length, width, contrast, shape, by flaw labeling be or
It is not.
5. a kind of liquid crystal display appearance detecting method according to claim 1, which is characterized in that the defect of determining defects step
During the 2D of data is shown, optional two features are schemed to draw 2D, and respectively as X-axis and Y-axis, each defect point is retouched in 2D image
Upper label is aterrimus, is light/dark balance labeled as no.
6. a kind of liquid crystal display appearance detecting method according to claim 1, which is characterized in that the study of determining defects step
In boundary curve, it is based on machine learning algorithm support vector machines, is automatically learned boundary curve；The expression of boundary curve is simultaneously given
After determining defect sample data, according to boundary curve determine sample whether label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910821718.XA CN110533657A (en) | 2019-09-04 | 2019-09-04 | A kind of liquid crystal display appearance detecting method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910821718.XA CN110533657A (en) | 2019-09-04 | 2019-09-04 | A kind of liquid crystal display appearance detecting method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110533657A true CN110533657A (en) | 2019-12-03 |
Family
ID=68666032
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910821718.XA Pending CN110533657A (en) | 2019-09-04 | 2019-09-04 | A kind of liquid crystal display appearance detecting method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110533657A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111627009A (en) * | 2020-05-27 | 2020-09-04 | 歌尔股份有限公司 | Screen detection method and device and head-mounted display equipment |
| CN113608378A (en) * | 2021-10-08 | 2021-11-05 | 深圳市绘晶科技有限公司 | Full-automatic defect detection method and system based on LCD (liquid crystal display) process |
-
2019
- 2019-09-04 CN CN201910821718.XA patent/CN110533657A/en active Pending
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111627009A (en) * | 2020-05-27 | 2020-09-04 | 歌尔股份有限公司 | Screen detection method and device and head-mounted display equipment |
| CN111627009B (en) * | 2020-05-27 | 2023-10-20 | 歌尔光学科技有限公司 | Screen detection method and device and head-mounted display equipment |
| CN113608378A (en) * | 2021-10-08 | 2021-11-05 | 深圳市绘晶科技有限公司 | Full-automatic defect detection method and system based on LCD (liquid crystal display) process |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108764257B (en) | Multi-view pointer instrument identification method | |
| CN107543828B (en) | Workpiece surface defect detection method and system | |
| US11551341B2 (en) | Method and device for automatically drawing structural cracks and precisely measuring widths thereof | |
| CN109523520B (en) | Chromosome automatic counting method based on deep learning | |
| CN109145830B (en) | Intelligent water gauge identification method | |
| US8331650B2 (en) | Methods, systems and apparatus for defect detection | |
| CN105957059B (en) | Electronic component missing part detection method and system | |
| CN106600600A (en) | Wafer defect detection method based on characteristic matching | |
| CN106529559A (en) | Pointer-type circular multi-dashboard real-time reading identification method | |
| CN108520514B (en) | Consistency detection method for electronic elements of printed circuit board based on computer vision | |
| CN105260749B (en) | Real-time target detection method based on direction gradient binary pattern and soft cascade SVM | |
| CN104392432A (en) | Histogram of oriented gradient-based display panel defect detection method | |
| CN111242899B (en) | Image-based flaw detection method and computer-readable storage medium | |
| CN111932511B (en) | Electronic component quality detection method and system based on deep learning | |
| CN108764312B (en) | Optimize multi objective dam defect image detecting method based on DS | |
| CN111652883B (en) | Glass surface defect detection method based on deep learning | |
| CN115496746A (en) | Method and system for detecting surface defects of plate based on fusion of image and point cloud data | |
| CN110533657A (en) | A kind of liquid crystal display appearance detecting method | |
| KR101782363B1 (en) | Vision inspection method based on learning data | |
| CN110660049A (en) | Tire defect detection method based on deep learning | |
| CN103852034A (en) | Elevator guide rail perpendicularity detection method | |
| CN115147363A (en) | Image defect detection and classification method and system based on deep learning algorithm | |
| CN110426395B (en) | Method and device for detecting surface of solar EL battery silicon wafer | |
| CN109447968A (en) | The metamorphic testing system of image-region growth algorithm | |
| CN113962929A (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20191203 |