CN104732900B - Picture element flaw detection method and device - Google Patents
Picture element flaw detection method and device Download PDFInfo
- Publication number
- CN104732900B CN104732900B CN201310713365.4A CN201310713365A CN104732900B CN 104732900 B CN104732900 B CN 104732900B CN 201310713365 A CN201310713365 A CN 201310713365A CN 104732900 B CN104732900 B CN 104732900B
- Authority
- CN
- China
- Prior art keywords
- picture
- frame
- video
- registering
- flaw detection
- 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
Landscapes
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
A kind of picture element flaw detection method, comprises the following steps：Videoscanning is carried out to substrate to be measured；Extraction meets pre-conditioned frame of video picture；The frame of video picture and reference picture are carried out into the registering registering picture of acquisition；The registering picture and the reference picture are analyzed determination defect pixel.Above-mentioned picture element flaw detection method is by video capture and extraction meets pre-conditioned frame picture, being taken pictures relative to common camera can improve the speed of detection, by carrying out the registering quick detection that can be realized in the case of without high accuracy physical alignment and positioning motion scan to base board defect to be measured to frame of video picture and reference picture.Also provided is a kind of picture element flaw detection means.
Description
Technical field
The present invention relates to defect detecting technique, more particularly to a kind of picture element flaw detection method and device.
Background technology
Organic Light Emitting Diode (Organic Light-Emitting Diode, OLED) is active light-emitting display device,
Compared to Thin Film Transistor-LCD (the Thin Film Transistor Liquid Crystal of current main flow
Display, TFT-LCD), it has a high-contrast, wide viewing angle, low-power consumption, fast response time, the advantages of volume is thinner, is expected to
As the flat panel display of future generation after LCD.
In the manufacturing process of OLED, various bad, the uniformity of its picture display brightness can be caused because of the problem of technique
Guarantee is hardly resulted in, how to lift yield just turns into each manufacture commercial city very concern.First of all, it is necessary to be carried out to bad
Detection, finds out the root of bad generation because then just can targetedly carry out technologic improvement and lift yield.Traditional stream
Many on waterline to be detected using automatic optical detection device is bad to some in technique, it needs high-precision physics pair
Standard, because deviation of the alignment can directly affect the Detection results of pixel scale.And limit defects detection threshold value setting, if threshold
Value sets too small, and pixel comparison precision high is likely to result in the flase drop of large area, so select suitable threshold value relatively difficult,
And easily there is missing inspection.The image capturing system of its high-precision positioning motion scan and complexity limits its applied field simultaneously
Close.
The content of the invention
Based on this, it is necessary to propose it is a kind of without high accuracy physical alignment and positioning motion scan method for quick and
Device.
A kind of picture element flaw detection method, comprises the following steps:
Videoscanning is carried out to substrate to be measured；
Extraction meets pre-conditioned frame of video picture；
The frame of video picture and reference picture are carried out into the registering registering picture of acquisition；
The registering picture and the reference picture are analyzed determination defect pixel.
It is described that the registering picture and the reference picture are analyzed determination in one of which embodiment
The step of defect pixel, includes：
Sampled point in the reference picture determines the sample area in the registering picture；
The sample area is contrasted with the corresponding region in the reference picture, judge be in the sample area
It is no to have the defect pixel；
Determine the defect pixel of the sample area.
In one of which embodiment, it is described videoscanning is carried out to substrate to be measured the step of before, also including insertion
The step of corresponding filter.
It is described to carry out videoscanning to substrate to be measured and be specially that different wave length is used for multiple times in one of which embodiment
Filter repeat videoscanning is carried out to the substrate to be measured.
In one of which embodiment, also including the curve of spectrum of the acquisition sample area pixel, then according to institute
The uniformity for stating the curve of spectrum judges the position of abnormity point.
In one of which embodiment, by the pixel grey scale of the sample area and the corresponding area in the reference picture
The difference of the pixel grey scale in domain is more than the pixel of the predetermined threshold value as defect pixel.
In one of which embodiment, the extraction meets pre-conditioned frame of video picture specially extracted at equal intervals
The frame of video picture of the degree of correlation between picture in preset range.
In one of which embodiment, the pattern that videoscanning is carried out to substrate to be measured be substrate Move Mode or
Camera Move Mode.
In one of which embodiment, the substrate to be measured is at the uniform velocity moved by S types path and is realized to the base to be measured
Plate carries out videoscanning.
It is described that the frame of video picture and reference picture are carried out into registering acquisition registration figure in one of which embodiment
The step of piece, includes：
Calculate the degree of correlation between the frame of video picture and the reference picture；
The corresponding frame of video picture of the degree of correlation maximum and the reference picture appropriate section are mapped
As registering picture.
It is described that the frame of video picture and reference picture are carried out into registering acquisition registration figure in one of which embodiment
Piece includes：
Obtain a pair of characteristic points between the frame of video picture and the reference picture；
Obtained according to the feature point coordinatesIn d1And d2, whereinRepresent the coordinate (x, y) and the reference picture coordinate (x of the frame of video picture0,y0) it
Between affine transformation relationship, wherein a1、b1、a2And b2It is equal to 1；
According to formulaBy the frame of video picture translated accordingly and with it is described
Reference picture alignment is used as registering picture.
A kind of picture element flaw detection means, including:
Videoscanning unit, for carrying out videoscanning to substrate to be measured；
Picture extracting unit, pre-conditioned frame of video picture is met for extracting；
Registration unit, for the frame of video picture and reference picture to be carried out into the registering registering picture of acquisition；
Defect pixel determining unit, for the registering picture and the reference picture to be analyzed into determination defect
Pixel.
In one of which embodiment, the defect pixel determining unit includes：
Sample area determining unit, adopting in the registering picture is determined for the sampled point in the reference picture
Sample region；
Comparison unit, for the sample area to be contrasted with the corresponding region in the reference picture, judges institute
Whether state in sample area has the defect pixel；
Defect pixel acquiring unit, the defect pixel for obtaining the sample area.
In one of which embodiment, before videoscanning unit carries out videoscanning to substrate to be measured, insertion is corresponding
Filter.
In one of which embodiment, videoscanning unit carries out videoscanning specially video to the substrate to be measured
What the filter that scanning element is used for multiple times different wave length was repeated carries out videoscanning to the substrate to be measured.
In one of which embodiment, also including uniformity judging unit, the uniformity judging unit is used to obtain
The curve of spectrum of the sample area pixel, then the uniformity according to the curve of spectrum judge the position of abnormity point.
In one of which embodiment, by the pixel grey scale of the sample area and the corresponding area in the reference picture
The difference of the pixel grey scale in domain is more than the pixel of the predetermined threshold value as defect pixel.
In one of which embodiment, the extraction meets pre-conditioned frame of video picture specially extracted at equal intervals
The frame of video picture of the degree of correlation between picture in preset range.
In one of which embodiment, the pattern that videoscanning is carried out to substrate to be measured be substrate Move Mode or
Camera Move Mode.
In one of which embodiment, the substrate to be measured is at the uniform velocity moved by S types path and is realized to the base to be measured
Plate carries out videoscanning.
In one of which embodiment, the registration unit includes：
Correlation calculating unit, for calculating the degree of correlation between the frame of video picture and the reference picture；
First picture registration unit, for by the corresponding frame of video picture of the degree of correlation maximum and the reference
Picture appropriate section is mapped as registering picture.
In one of which embodiment, the registration unit includes：
Characteristic point acquiring unit, for obtaining a pair of characteristic points between the frame of video picture and the reference picture；
Parametric solution unit, for being obtained according to the feature point coordinatesIn d1With
d2, whereinRepresent the coordinate (x, y) and the reference picture coordinate of the frame of video picture
(x0,y0) between affine transformation relationship, wherein a1、b1、a2And b2It is equal to 1；
Second picture registration unit, for according to formulaThe frame of video picture is entered
The corresponding translation of row is simultaneously aligned as registering picture with the reference picture.
Above-mentioned picture element flaw detection method and device pass through video capture and extraction meets pre-conditioned frame picture, relatively
Being taken pictures in common camera can improve the speed of detection, by carrying out to frame of video picture and reference picture registering to realize
To the quick detection of base board defect to be measured in the case of without high accuracy physical alignment and positioning motion scan.
Brief description of the drawings
Fig. 1 is the flow chart of the picture element flaw detection method in one embodiment；
Fig. 2 is the structural representation of the Systems for optical inspection in one embodiment；
Fig. 3 is the schematic diagram that the camera in one embodiment is scanned to substrate to be measured；
Fig. 4 is the flow that frame of video picture and reference picture are carried out the registering registering picture of acquisition in one embodiment
Figure；
Fig. 5 is the stream that frame of video picture and reference picture are carried out the registering registering picture of acquisition in another embodiment
Cheng Tu；
Fig. 6 a to Fig. 6 d are to obtain frame of video picture in an embodiment and carry out the schematic diagram of registration；
Fig. 7 is the schematic diagram registering with reference picture of the frame of video picture in one embodiment；
Fig. 8 is to determine defect pixel the registering picture and the reference picture are being analyzed in Fig. 1
Flow chart；
Fig. 9 is the pixel grey scale comparative analysis schematic diagram to selection in one embodiment；
Figure 10 is the spectrum analysis schematic diagram of the pixel to choosing in one embodiment；
Figure 11 is the structural representation of the picture element flaw detection means in an embodiment；
Figure 12 is the structural representation of the registration unit in one embodiment；
Figure 13 is the structural representation of the registration unit in another embodiment；
Figure 14 is the structural representation of the defect pixel determining unit in Figure 11.
Specific embodiment
In one embodiment, as shown in figure 1, a kind of picture element flaw detection method, comprises the following steps:
S110, inserts corresponding filter.
As shown in Fig. 2 filter 50 can be filtered to detect the uniformity of material.The uniformity of material is influence
Product quality key factor.The individual components of material change even material type when changing, from appearance often without
Method is recognized.After filter 50 is filtered, reference light is set up to the spectrum analysis that passing material carries out multi-feature wavelength and is set a song to music
Line, then the curve of spectrum of detected materials and reference spectra curve comparison can just be detected the uniformity of the aspects such as material, technique.
If the process materials consistency problem that be directed to substrate to be measured 20 carries out spectral detection, then add before camera 60
Enter filter 50.Then substrate to be measured 20 is fixed on a mobile platform 10.Regulation micro optical system 40 focuses on to be measured
The layer to be detected of substrate 20.Light source 30 is broad spectrum light source, and light source 30 is used to provide illumination to micro optical system 40.Filtering dress
50 filter plates or tunable optic filter that can be specifically the characteristic wavelength for adding in imaging systems are put, filter 50 is used
In carrying out multispectral detection to the surface of substrate to be measured 20.The process of multispectral detection and analysis will be described in more detail below, if not
Need to carry out the consistency analysis of material, it is also possible to omit the step of inserting corresponding filter.
S120, videoscanning is carried out to substrate to be measured.
After camera 60 is focused on, the scope for detecting as needed determines original position and the path of camera video scanning, can be with
Using camera Move Mode or substrate Move Mode.For the ease of camera 60 in macro -graph and micro examination (plus microscope inspection
Look into) between switching, substrate Move Mode can be used.In a specific embodiment, as shown in figure 3, passing through S types path
At the uniform velocity move the videoscanning that substrate to be measured 20 is realized to substrate to be measured.
If the process materials consistency problem that be directed to substrate to be measured 20 carries out spectral detection, substrate to be measured 20 is carried out
Videoscanning is specially：The light intensity of light source 30 is adjusted to unanimously, using the filter 50 of correspondence different wave length, every time using not
The movement substrate 20 to be measured repeated after the filter 50 of co-wavelength carries out videoscanning.
S130, extraction meets pre-conditioned frame of video picture.
Video capture finish after by shoot video data process.By using video mode to substrate to be measured
20 are shot, and can greatly improve the speed of detection.
The frame number of many video capture devices collection per second is inversely proportional with the resolution ratio of every frame picture.For at a high speed
Mobile object, in order to see the details of object movement clearly, the temporal resolution of video is just very important.Present high speed phase
Machine, it is many it is per second can shoot million frames, to tackle high-speed capture the need for.In the present embodiment, it may be that more
Focus on the resolution ratio of frame of video picture, 24 frame per second can be even less than as frame number per second, because relative to substrate to be measured
20 rate travel should be enough.In addition, obtaining high-resolution picture can increase the precision of detection.Specific shooting
Video frame number per second should determine according to the speed of scanning.After the completion of video acquisition, video is converted into frame of video picture,
Then only need to process frame of video picture.
The degree of correlation met between pre-conditioned frame of video picture specially extracted at equal intervals picture is extracted in default model
The frame of video picture for enclosing.In a specific embodiment, the degree of correlation the regarding in preset range between extracted at equal intervals picture
The method of frequency frame picture includes：The degree of correlation between picture is calculated first, then by the degree of correlation and preset range between picture
Contrasted, selected the interval of satisfactory frame of video picture.In the present embodiment, preset range is that the selection degree of correlation is small
Picture, the preset range can reduce the repeating part between front and rear photograph, reduce it is unnecessary repeat, indirectly carry
The speed of scanning high.But if the degree of correlation between two width pictures sets too small, then adjacent frame of video picture it
Between will omit the pixels of many scannings, many pixel missing inspections can be caused, can so increase the error rate of detection.Therefore, phase
The preset range of Guan Du needs to be set as needed.
The computing formula of the degree of correlation between specific picture is C (x, y)=corr (X (x1, y1), Y (x2, y2)), its
Middle X (x1, y1) is the gray scale of the pixel that the first width picture middle position is set to (x1, y1) place, and Y (x2, y2) is the second width picture middle position
Be set to the gray scale of the pixel of (x2, y2), C (x, y) be two width pictures when shift position is (x, y) between the degree of correlation.Two width
The similarity of the content of picture is big, then the degree of correlation of this two width picture is just big.If the similarity of the content of two width pictures is small,
So the degree of correlation of this two width picture is just small.
S140, the registering registering picture of acquisition is carried out by frame of video picture and reference picture.
If necessary to process the frame of video picture changed, it is necessary to which these frame of video pictures are carried out into registration.Find out
The position of the pixel in frame of video picture, then carries out error analysis and material, process consistency analysis to related pixel.
In a specific embodiment, as shown in figure 4, frame of video picture and reference picture are carried out into registering acquisition registration
Picture includes：
S141, calculates the degree of correlation between frame of video picture and reference picture.
The degree of correlation between being calculated in frame of video picture and reference picture using above-mentioned formula.To the correlation of two width pictures
The calculating process of degree is also the coordinate of mobile two width pictures so as to the process matched to two width pictures.When two width pictures correspondence
When partial content similar portion is less, then the degree of correlation between the width picture of position two can be than relatively low.Constantly
Mobile two width pictures, when the content similarity of two width picture corresponding parts is higher, the degree of correlation between two width pictures
Can increase accordingly, until the appropriate section of two width pictures is completely superposed, its degree of correlation can also reach maximum.
S143, the corresponding frame of video picture of degree of correlation maximum and reference picture appropriate section are mapped as registration
Picture.
The maximum position of the degree of correlation of two width pictures, namely the part that corresponding two width picture appropriate section is overlapped.By phase
The frame of video picture and reference picture part picture answered are mapped, so that as registering picture.
In another specific embodiment, as shown in figure 5, frame of video picture and reference picture are carried out into registering acquisition matching somebody with somebody
Quasi- picture includes：
S142, obtains a pair of characteristic points between frame of video picture and reference picture.
The coordinate (x, y) of frame of video picture and reference picture coordinate (x0,y0) between affine transformation relationship beExpression formula according to affine transformation matrix needs to find out reference picture and frame of video figure respectively
Three pairs of characteristic points between piece, this three pairs of characteristic points have same characteristic features, to obtain affine transformation matrix.Then to frame of video figure
Piece carries out affine transformation and reaches registering purpose.But actually due to the setting of parallel sweep, reference picture and frame of video picture
Only exist translation transformation, i.e. a1、b1、a2And b21 is equal to, as long as obtaining d1And d2If, so finding out a pair of characteristic points.
S144, obtains related in the affine transformation matrix between frame of video picture and reference picture according to feature point coordinates
Parameter.
Can be obtained according to feature point coordinatesIn d1And d2。
S146：Frame of video picture is translated accordingly and is aligned according to affine transformation matrix and with reference picture as matching somebody with somebody
Quasi- picture.
After affine transformation matrix determines, frame of video picture and reference picture can just be mapped.According to affine transformation square
Frame of video picture is done corresponding translation by battle array just can be by frame of video picture registration.
In a specific embodiment, as shown in figures 6 a-6d, video V becomes frame of video picture V1, V2 ... by conversion
V10.Calculated by the degree of correlation between frame of video picture V1, V2 ... V10, then selected according to the default degree of correlation
Satisfactory frame of video picture F1, F2 and F3.Will by the method described by step S141-S143 or step S142-S146
Satisfactory frame of video picture F1, F2 and F3 and reference picture R carry out registration, obtain registering picture P1, P2 and P3.
By the way that frame of video picture and reference picture R carried out registering to obtain registering picture, it is to avoid high accuracy physical alignment
With the process of positioning motion scan, the speed of scanning is improve, reduce the complexity of scanning.In a specific embodiment
In, as shown in fig. 7, reference picture R and frame of video picture P1, P2, P3, P4 obtained by the filter 50 of different wave length and
P5 after translating registration, 1,2,3,4,5 this 5 reference points on reference picture R can different frame of video picture P1,
P2, P3, P4 and P5 find corresponding pixel.
S150, determination defect pixel is analyzed by registering picture and reference picture.Can be by individual element pair
The form of ratio determines defect pixel；Can also first subregion contrasted, find out defect area and determine the defect in the region again
Pixel.Specifically, as shown in figure 8, step S150 can include step S152 to step S156.
S152, the sampled point in reference picture determines the sample area in registering picture.With reference in registering picture
The corresponding pixel of sampled point in picture is referred to as corresponding points, the border circular areas with the corresponding points as the center of circle, or with the correspondence
Centered on point or angle point a certain rectangular area, or be characterized other graphics fields a little as sampling using the corresponding points
Region.The selection of figure and size can be preset according to the particular content to be detected, the selection, rectangle such as circle
Selection etc..
Whether S154, sample area is contrasted with the corresponding region in reference picture, judges there is scarce in sample area
Fall into pixel.The average gray of the corresponding region in the average gray of sample area and reference picture can for example be contrasted.
If gray scale difference exceedes setting value, then it is assumed that defective pixel in sample area.Above by zonule averaging method selected pixels
The method of gray scale can reduce the influence that registration error is brought, and reduce the possibility of flase drop, and accuracy of detection is improve indirectly.
S156, determines the defect pixel of sample area.By the pixel grey scale of sample area and the corresponding picture in reference picture R
The difference of the gray scale of element is more than the pixel of predetermined threshold value as defect pixel.In a specific embodiment, as shown in figure 9,
Abscissa is corresponding each pixel, and ordinate is gray value.Pixel 1,2,3,4,5 is in this 5 width frame figure of P1, P2, P3, P4 and P5
Gray scale in piece is most of all within threshold line compared with 1,2,3,4,5 gray scale in reference picture R of pixel.Except pixel
Beyond the scope of threshold line, and the pixel for exceeding threshold line scope is considered as 3 gray scale in a certain width frame of video picture
Defect pixel, therefore pixel 3 is defect pixel.
In one embodiment, it is specially using different filters if carrying out videoscanning to substrate to be measured 20
50, videoscanning is carried out to substrate to be measured 20 using what is repeated after the filter 50 of different wave length every time, obtain corresponding filtering
The spectrum of the characteristic wavelength of device 50.By obtaining the curve of spectrum of sample area pixel, then according to the consistent of the curve of spectrum
Property judges the position of abnormity point.As shown in Figure 10, abscissa is wavelength, and ordinate is gray scale, depicted in figure correspondence 1,2,3 this
The curve of spectrum of 3 pixels in different registering pictures, same curves are the different filters under same registering picture of same pixel
Gray scale under the corresponding wavelength that wave apparatus 50 are obtained, the different curves of same pixel for same pixel under different registering pictures not
With the gray scale under the corresponding wavelength that filter 50 is obtained.Uniformity according to the curve of spectrum can determine the pixel of exception,
Such that it is able to judge the position of abnormity point.Specifically, the curve of spectrum of pixel 3 wavelength be 550nm, 600nm, 650nm this three
Individual wavelength occurs in that exception.In wavelength, for 650nm, 700nm, the two wavelength occur in that exception to the curve of spectrum of pixel 1, therefore
Pixel 1 and 3 is considered as abnormity point, and its material and technique do not meet uniformity.This detection method can be with Alternative material
Uniformity makes classification and Detection, realizes the detection to process materials consistency problem.
Above-mentioned picture element flaw detection method is by video capture and extraction meets pre-conditioned frame picture, relative to common
Camera take pictures and can improve the speed of detection, carry out registering can realization without height by frame of video picture and reference picture
To the quick detection of base board defect to be measured in the case of precision physical alignment and positioning motion scan.By in sample area
Grey scale pixel value is analyzed, and can have what is be directed to be detected to defect, improves accuracy of detection.
In one embodiment, as shown in figure 11, a kind of picture element flaw detection means 100 include videoscanning unit 110,
Picture extracting unit 120, registration unit 130, defect pixel determining unit 140 and uniformity judging unit 150.
Videoscanning unit 110 is used to carry out videoscanning to substrate to be measured 20.Videoscanning is carried out to substrate to be measured 20
Pattern be substrate Move Mode or camera Move Mode.It is (plus micro- in macro -graph and micro examination for the ease of camera 60
Spectroscopy) between switching, substrate Move Mode can be used.In the present embodiment, base to be measured is at the uniform velocity moved by S types path
Plate 20 is realized carrying out videoscanning to substrate to be measured 20.Substrate to be measured 20 is shot by videoscanning unit 110, can be big
The big speed for improving detection.The frame number of many video capture devices collection per second is to be inversely proportional with the resolution ratio of every frame picture
's.For the object of high-speed mobile, in order to see the details of object movement clearly, the temporal resolution of video is just very important.It is existing
High speed camera, it is many it is per second can shoot million frames, to tackle high-speed capture the need for.In the present embodiment,
Should be the resolution ratio for more focusing on frame of video picture, 24 frame per second can be even less than as frame number per second, because relatively
In the rate travel of substrate to be measured 20 should be enough.In addition, obtaining high-resolution picture can increase the precision of detection.
The frame number that the video of specific videoscanning unit 110 shooting is per second should determine according to the speed of scanning.Video acquisition is completed
Afterwards, video is converted into frame of video picture, then only needs to process frame of video picture.
Picture extracting unit 120 is used for extraction and meets pre-conditioned frame of video picture.Specific picture extracting unit 120
Extract frame of video picture of the degree of correlation between picture in preset range in interval.In the present embodiment, preset range is related to choose
The small picture of degree, the preset range can reduce the repeating part between front and rear photograph, reduce it is unnecessary repeat,
What is connect improves the speed of scanning.But if the degree of correlation between two width pictures sets too small, then adjacent frame of video
The pixel of many scannings will be omitted between picture, many pixel missing inspections can be caused, can so increase the error rate of detection.Cause
This, the preset range of the degree of correlation needs to be set as needed.The computing formula of the degree of correlation between specific picture is C
(x, y)=corr (X (x1, y1), Y (x2, y2)), wherein X (x1, y1) are the picture that the first width picture middle position is set to (x1, y1) place
The gray scale of element, Y (x2, y2) is the gray scale of the pixel that the second width picture middle position is set to (x2, y2), and C (x, y) is in shift position
It is the degree of correlation between two width pictures when (x, y).The similarity of the content of two width pictures is big, then the degree of correlation of this two width picture
It is just big.If the similarity of the content of two width pictures is small, then the degree of correlation of this two width picture is just small.
If necessary to process the frame of video picture changed, it is necessary to which these frame of video pictures are carried out into registration.Find out
The position of the pixel in frame of video picture, then carries out error analysis and material, process consistency analysis to related pixel.
Registration unit 130 is used to for frame of video picture and reference picture to carry out the registering registering picture of acquisition.Defect pixel is true
Order unit 140 is used to for registering picture and reference picture to be analyzed determination defect pixel.
In one embodiment, as shown in figure 12, registration unit 130 includes the picture of correlation calculating unit 132 and first
Registration unit 134.
Correlation calculating unit 132 is used to calculate the degree of correlation between frame of video picture and reference picture.Relatedness computation
The degree of correlation between in the above-mentioned formula calculating frame of video picture of the use of unit 132 and reference picture.To the correlation of two width pictures
The calculating process of degree is also the coordinate of mobile two width pictures so as to the process matched to two width pictures.When two width pictures correspondence
When partial content similar portion is less, then the degree of correlation between the width picture of position two can be than relatively low.Constantly
Mobile two width pictures, when the content similarity of two width picture corresponding parts is higher, the degree of correlation between two width pictures
Can increase accordingly, until the appropriate section of two width pictures is completely superposed, its degree of correlation can also reach maximum.
First picture registration unit 134 is used for the corresponding frame of video picture of degree of correlation maximum and the corresponding portion of reference picture
Divide and be mapped as registering picture.The maximum position of the degree of correlation of two width pictures, namely corresponding two width picture appropriate section
The part of overlap.Be mapped for corresponding frame of video picture and reference picture part picture by the first picture registration unit 134, from
And as registering picture.
In another embodiment, as shown in figure 13, registration unit 130 includes characteristic point acquiring unit 131, parametric solution
Unit 133 and second picture registration unit 135.
Characteristic point acquiring unit 131 is used to obtain a pair of characteristic points between frame of video picture and reference picture.Parameter is asked
Solution unit 133 is used to be obtained according to feature point coordinatesIn d1And d2, whereinRepresent the coordinate (x, y) and reference picture coordinate (x of frame of video picture0,y0) between it is affine
Transformation relation.Expression formula according to affine transformation matrix needs to find out respectively three couples of spies between reference picture and frame of video picture
Levy a little, this three pairs of characteristic points have same characteristic features, to obtain affine transformation matrix.Then affine transformation is carried out to frame of video picture
Reach registering purpose.But actually due to the setting of parallel sweep, reference picture only exists translation transformation with frame of video picture,
That is a1、b1、a2And b21 is equal to, as long as obtaining d1And d2If, so finding out a pair of characteristic points.Second picture registration is single
Unit 135 is used for according to formulaFrame of video picture is carried out into corresponding translation and and reference picture
Alignment is used as registering picture.After affine transformation matrix determines, frame of video picture and reference picture can just be mapped.Second figure
Frame of video picture is done corresponding translation by piece registration unit 135 according to affine transformation matrix just can be by frame of video picture registration.
By the way that frame of video picture and reference picture R carried out registering to obtain registering picture, it is to avoid high accuracy physical alignment
With the process of positioning motion scan, the speed of scanning is improve, reduce the complexity of scanning.In a specific embodiment
In, as shown in fig. 7, reference picture R and frame of video picture P1, P2, P3, P4 obtained by the filter 50 of different wave length and
P5 after translating registration, 1,2,3,4,5 this 5 reference points on reference picture R can different frame of video picture P1,
P2, P3, P4 and P5 find corresponding pixel.
In one embodiment, as shown in figure 14, defect pixel determining unit 140 include sample area determining unit 142,
Comparison unit 144 and defect pixel acquiring unit 146.
The sampled point that sample area determining unit 142 is used in reference picture determines the sample region in registering picture
Domain.Pixel corresponding with the sampled point in reference picture is referred to as corresponding points in registering picture, the circle with the corresponding points as the center of circle
Shape region, or centered on the corresponding points or angle point a certain rectangular area, or be characterized a little with the corresponding points
Other graphics fields are used as sample area.The selection of figure and size can in advance be set according to the particular content to be detected
Fixed, such as circular selection, the selection of rectangle.Comparison unit 144 is used for the corresponding region in sample area and reference picture
Contrasted, judge in sample area whether defective pixel.Defect pixel acquiring unit 146 is used to obtain lacking for sample area
Fall into pixel.Specific defect pixel acquiring unit 146 is by the corresponding region in the pixel grey scale of sample area and reference picture
The difference of pixel grey scale is more than the pixel of predetermined threshold value as defect pixel.
In a specific embodiment, as shown in figure 9, abscissa is corresponding each pixel, ordinate is gray value.
Gray scale and pixel 1,2,3,4,5 of the pixel 1,2,3,4,5 in P1, P2, P3, P4 and P5 this 5 width frame picture are in reference picture R
Gray scale compared to major part all within threshold line.Except gray scale of the pixel 3 in a certain width frame of video picture is beyond threshold value
The scope of line, and the pixel for exceeding threshold line scope is considered as defect pixel, therefore pixel 3 is defect pixel.
If the process materials consistency problem that be directed to substrate to be measured 20 carries out spectral detection, in videoscanning unit
Before 110 pairs of substrates to be measured 20 carry out videoscanning, corresponding filter 50 is inserted.Videoscanning tool is carried out to substrate to be measured 20
Body is：The light intensity of light source 30 is adjusted to unanimously, using the filter 50 of correspondence different wave length, the filter of different wave length is used every time
The movement substrate 20 to be measured repeated after wave apparatus 50 carries out videoscanning.Uniformity judging unit 150 is used to obtain sample area
The curve of spectrum of pixel, then the uniformity according to the curve of spectrum judge the position of abnormity point.As shown in Figure 10, abscissa is
Wavelength, ordinate is gray scale, 1,2,3 this curves of spectrum of 3 pixels in different registering pictures of correspondence is depicted in figure, together
Gray scale under the corresponding wavelength that one curve is same pixel different filters 50 are obtained under the same registering picture, same pixel
The different curves corresponding wavelength that is same pixel different filters 50 are obtained under the different registering pictures under gray scale.According to
The uniformity of the curve of spectrum can determine the pixel of exception, such that it is able to judge the position of abnormity point.Specifically, pixel 3
In wavelength, for 550nm, 600nm, 650nm, these three wavelength occur in that exception to the curve of spectrum.The curve of spectrum of pixel 1 is in wavelength
650nm, 700nm the two wavelength occur in that exception, therefore pixel 1 and 3 is considered as abnormity point, and its material and technique do not meet
Uniformity.This detection method can make classification and Detection with the uniformity of Alternative material, realize to process materials uniformity
The detection of problem.
Above-mentioned picture element flaw detection means is by video capture and extraction meets pre-conditioned frame picture, relative to common
Camera take pictures and can improve the speed of detection, carry out registering can realization without height by frame of video picture and reference picture
Precision physical be aligned and positioning motion scan in the case of to the quick detection of the defect of substrate to be measured.
Embodiment described above only expresses several embodiments of the invention, and its description is more specific and detailed, but simultaneously
Therefore the limitation to the scope of the claims of the present invention can not be interpreted as.It should be pointed out that for one of ordinary skill in the art
For, without departing from the inventive concept of the premise, various modifications and improvements can be made, these belong to guarantor of the invention
Shield scope.Therefore, the protection domain of patent of the present invention should be determined by the appended claims.
Claims (22)
1. a kind of picture element flaw detection method, it is characterised in that comprise the following steps:
Videoscanning is carried out to substrate to be measured；
Extraction meets pre-conditioned frame of video picture；
The frame of video picture and reference picture are carried out into the registering registering picture of acquisition；
The registering picture and the reference picture are analyzed determination defect pixel.
2. picture element flaw detection method as claimed in claim 1, it is characterised in that described by the registering picture and the ginseng
Examining the step of picture is analyzed determination defect pixel includes：
Sampled point in the reference picture determines the sample area in the registering picture；
The sample area is contrasted with the corresponding region in the reference picture, is judged whether have in the sample area
The defect pixel；
Determine the defect pixel of the sample area.
3. picture element flaw detection method as claimed in claim 2, it is characterised in that video is carried out to substrate to be measured sweep described
Before the step of retouching, also including insert corresponding filter the step of.
4. picture element flaw detection method as claimed in claim 3, it is characterised in that described that videoscanning is carried out to substrate to be measured
Specially be used for multiple times different wave length filter repeat videoscanning is carried out to the substrate to be measured.
5. picture element flaw detection method as claimed in claim 4, it is characterised in that also including obtaining the sample area pixel
The curve of spectrum, then the uniformity according to the curve of spectrum judge the position of abnormity point.
6. picture element flaw detection method as claimed in claim 2, it is characterised in that by the pixel grey scale of the sample area with
The difference of the pixel grey scale of the corresponding region in the reference picture is more than the pixel of the predetermined threshold value as defect pixel.
7. picture element flaw detection method as claimed in claim 1, it is characterised in that the extraction meets pre-conditioned video
Frame picture is specially frame of video picture of the degree of correlation between extracted at equal intervals picture in preset range.
8. picture element flaw detection method as claimed in claim 1, it is characterised in that described that videoscanning is carried out to substrate to be measured
Pattern be substrate Move Mode or camera Move Mode.
9. picture element flaw detection method as claimed in claim 1, it is characterised in that by S types path it is at the uniform velocity mobile described in treat
Substrate is surveyed to realize carrying out videoscanning to the substrate to be measured.
10. picture element flaw detection method as claimed in any one of claims 1-9 wherein, it is characterised in that described by the video
The step of frame picture carries out registering acquisition registration picture with reference picture includes：
Calculate the degree of correlation between the frame of video picture and the reference picture；
The corresponding frame of video picture of the degree of correlation maximum and the reference picture appropriate section are mapped as
Registering picture.
11. picture element flaw detection methods as claimed in any one of claims 1-9 wherein, it is characterised in that described by the video
Frame picture carries out the registering registration picture that obtains with reference picture to be included：
Obtain a pair of characteristic points between the frame of video picture and the reference picture；
Obtained according to the feature point coordinates In d1And d2, wherein Table
Show the coordinate (x, y) and the reference picture coordinate (x of the frame of video picture0,y0) between affine transformation relationship, wherein a1、
b1、a2And b2It is equal to 1；
According to formula The frame of video picture is translated and with described with reference to figure accordingly
Piece alignment is used as registering picture.
A kind of 12. picture element flaw detection means, it is characterised in that including:
Videoscanning unit, for carrying out videoscanning to substrate to be measured；
Picture extracting unit, pre-conditioned frame of video picture is met for extracting；
Registration unit, for the frame of video picture and reference picture to be carried out into the registering registering picture of acquisition；
Defect pixel determining unit, for the registering picture and the reference picture to be analyzed into determination defect picture
Element.
13. picture element flaw detection means as claimed in claim 12, it is characterised in that the defect pixel determining unit bag
Include：
Sample area determining unit, the sample region in the registering picture is determined for the sampled point in the reference picture
Domain；
Comparison unit, for the sample area to be contrasted with the corresponding region in the reference picture, adopts described in judgement
Whether the defect pixel is had in sample region；
Defect pixel acquiring unit, the defect pixel for obtaining the sample area.
14. picture element flaw detection means as claimed in claim 13, it is characterised in that in videoscanning unit to substrate to be measured
Before carrying out videoscanning, corresponding filter is inserted.
15. picture element flaw detection means as claimed in claim 14, it is characterised in that videoscanning unit is to the base to be measured
Plate carry out videoscanning specially videoscanning unit be used for multiple times different wave length filter repeat to the base to be measured
Plate carries out videoscanning.
16. picture element flaw detection means as claimed in claim 15, it is characterised in that also including uniformity judging unit, institute
Uniformity judging unit is stated for obtaining the curve of spectrum of the sample area pixel, then according to the consistent of the curve of spectrum
Property judges the position of abnormity point.
17. picture element flaw detection means as claimed in claim 13, it is characterised in that by the pixel grey scale of the sample area
It is more than the pixel of the predetermined threshold value as defect pixel with the difference of the pixel grey scale of the corresponding region in the reference picture.
18. picture element flaw detection means as claimed in claim 12, it is characterised in that the extraction meets pre-conditioned regarding
Frequency frame picture is specially frame of video picture of the degree of correlation between extracted at equal intervals picture in preset range.
19. picture element flaw detection means as claimed in claim 12, it is characterised in that described video is carried out to substrate to be measured to sweep
The pattern retouched is substrate Move Mode or camera Move Mode.
20. picture element flaw detection means as claimed in claim 12, it is characterised in that at the uniform velocity mobile described by S types path
Substrate to be measured is realized carrying out videoscanning to the substrate to be measured.
The 21. picture element flaw detection means as any one of claim 12 to 20, it is characterised in that the registration unit
Including：
Correlation calculating unit, for calculating the degree of correlation between the frame of video picture and the reference picture；
First picture registration unit, for by the corresponding frame of video picture of the degree of correlation maximum and the reference picture
Appropriate section is mapped as registering picture.
The 22. picture element flaw detection means as any one of claim 12 to 20, it is characterised in that the registration unit
Including：
Characteristic point acquiring unit, for obtaining a pair of characteristic points between the frame of video picture and the reference picture；
Parametric solution unit, for being obtained according to the feature point coordinates In d1And d2, its
In Represent the coordinate (x, y) and the reference picture coordinate (x of the frame of video picture0,y0)
Between affine transformation relationship, wherein a1、b1、a2And b2It is equal to 1；
Second picture registration unit, for according to formula The frame of video picture is carried out into phase
The translation answered simultaneously is aligned as registering picture with the reference picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310713365.4A CN104732900B (en) | 2013-12-20 | 2013-12-20 | Picture element flaw detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201310713365.4A CN104732900B (en) | 2013-12-20 | 2013-12-20 | Picture element flaw detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN104732900A CN104732900A (en) | 2015-06-24 |
| CN104732900B true CN104732900B (en) | 2017-06-16 |
Family
ID=53456755
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201310713365.4A Active CN104732900B (en) | 2013-12-20 | 2013-12-20 | Picture element flaw detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN104732900B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105044127A (en) * | 2015-07-31 | 2015-11-11 | 深圳市星火辉煌系统工程有限公司 | OLED micro-display defect detection device and detection method |
| CN104992449B (en) * | 2015-08-06 | 2018-06-19 | 艾科芯（深圳）智能科技有限公司 | Information identification and surface defect online test method based on machine vision |
| CN106686372A (en) * | 2015-11-11 | 2017-05-17 | 天津三星电子有限公司 | Display terminal image display detection method and display terminal |
| CN108956550A (en) * | 2018-06-12 | 2018-12-07 | 华灿光电（浙江）有限公司 | A kind of method and apparatus of photoluminescence spectra processing |
| CN108844966A (en) * | 2018-07-09 | 2018-11-20 | 广东速美达自动化股份有限公司 | A kind of screen detection method and detection system |
| CN109256341B (en) * | 2018-08-10 | 2023-07-04 | 北京百度网讯科技有限公司 | Method and device for inspecting monocrystalline silicon solar cells |
| CN109360203B (en) | 2018-10-30 | 2021-12-03 | 京东方科技集团股份有限公司 | Image registration method, image registration device and storage medium |
| CN109508722A (en) * | 2018-11-08 | 2019-03-22 | 中交第二航务工程局有限公司 | Picture comparison method and picture Compare System based on gray value |
| CN110211102B (en) * | 2019-05-23 | 2021-01-19 | 杭州良淋电子科技股份有限公司 | Badminton racket broken line detection method for unmanned network renting and selling machine |
| CN110517231B (en) * | 2019-08-13 | 2023-12-22 | 云谷（固安）科技有限公司 | Method and device for detecting display edge of screen body |
| CN114266773B (en) * | 2022-03-02 | 2022-05-20 | 成都数联云算科技有限公司 | Display panel defect positioning method, device, equipment and storage medium |
| CN115908280B (en) * | 2022-11-03 | 2023-07-18 | 广东科力新材料有限公司 | Method and system for determining performance of PVC (polyvinyl chloride) calcium zinc stabilizer based on data processing |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101076720A (en) * | 2004-12-11 | 2007-11-21 | 株式会社Macron | Apparatus for inspecting backlight unit |
| CN101814264A (en) * | 2010-04-02 | 2010-08-25 | 苏州富鑫林光电科技有限公司 | CCD-type optical quality measurement device |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2001083474A (en) * | 2000-08-11 | 2001-03-30 | Sony Corp | Method for inspecting liquid crystal display panel |
| KR100955486B1 (en) * | 2004-01-30 | 2010-04-30 | 삼성전자주식회사 | A inspecting apparatus and inspecting method of display panel |
-
2013
- 2013-12-20 CN CN201310713365.4A patent/CN104732900B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101076720A (en) * | 2004-12-11 | 2007-11-21 | 株式会社Macron | Apparatus for inspecting backlight unit |
| CN101814264A (en) * | 2010-04-02 | 2010-08-25 | 苏州富鑫林光电科技有限公司 | CCD-type optical quality measurement device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN104732900A (en) | 2015-06-24 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104732900B (en) | Picture element flaw detection method and device | |
| Abdelhamed et al. | A high-quality denoising dataset for smartphone cameras | |
| CN104749184B (en) | Automatic optical detection method and system | |
| JP2022000652A (en) | High throughput and high resolution optical metrology for reflective and transmissive nano-photonic devices | |
| JP4061289B2 (en) | Image inspection method and apparatus | |
| US9813608B2 (en) | Fast auto-focus in imaging | |
| JP6594294B2 (en) | Image quality evaluation of microscopic images | |
| CN111145134B (en) | Block effect-based microlens light field camera full-focus image generation algorithm | |
| CN109406527B (en) | System and method for detecting fine appearance defects of micro camera module lens | |
| CN109872309A (en) | Detection system, method, apparatus and computer readable storage medium | |
| CN104749801B (en) | High Precision Automatic optical detecting method and system | |
| WO2013183471A1 (en) | Appearance inspection device and appearance inspection method | |
| CN109064439B (en) | Partition-based single-side light-entering type light guide plate shadow defect extraction method | |
| TW201516397A (en) | Bubble inspection system for glass | |
| CN106204466A (en) | A kind of self-adaptive solution method for Fourier lamination micro-imaging technique | |
| JP2005345290A (en) | Streak-like flaw detecting method and streak-like flaw detector | |
| US20150030230A1 (en) | Substrate inspection method, substrate manufacturing method and substrate inspection device | |
| WO2020164201A1 (en) | Defect detection method for tft array substrate | |
| CN109003228B (en) | Dark field microscopic large-view-field automatic stitching imaging method | |
| CN106773158A (en) | A kind of movable type seeks liquid crystal display pixel qualities analytical equipment and a method certainly | |
| KR102178319B1 (en) | Apparatus and method for brightness uniformity inspecting of display panel | |
| JP2004279296A (en) | Film thickness acquiring method | |
| Taylor | Implementing high resolution structured light by exploiting projector blur | |
| CN108280825A (en) | A kind of liquid crystal display emargintion detection method | |
| TWI477768B (en) | Method and apparatus for automatic optical inspection of flat panel substrate |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant |