CN107274393B - Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection - Google Patents
Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection Download PDFInfo
- Publication number
- CN107274393B CN107274393B CN201710438495.XA CN201710438495A CN107274393B CN 107274393 B CN107274393 B CN 107274393B CN 201710438495 A CN201710438495 A CN 201710438495A CN 107274393 B CN107274393 B CN 107274393B
- Authority
- CN
- China
- Prior art keywords
- mrow
- image
- row
- msub
- mtd
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 53
- XUIMIQQOPSSXEZ-UHFFFAOYSA-N Silicon Chemical compound [Si] XUIMIQQOPSSXEZ-UHFFFAOYSA-N 0.000 title claims abstract description 36
- 229910052710 silicon Inorganic materials 0.000 title claims abstract description 36
- 239000010703 silicon Substances 0.000 title claims abstract description 36
- 230000007547 defect Effects 0.000 claims abstract description 40
- 238000000034 method Methods 0.000 claims abstract description 24
- 230000011218 segmentation Effects 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims abstract description 10
- 230000003044 adaptive effect Effects 0.000 claims abstract description 7
- 238000013100 final test Methods 0.000 claims abstract description 4
- 238000013101 initial test Methods 0.000 claims abstract description 3
- 230000009467 reduction Effects 0.000 claims description 10
- 238000012360 testing method Methods 0.000 claims description 10
- 230000002950 deficient Effects 0.000 claims description 8
- 230000007704 transition Effects 0.000 claims description 6
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000012805 post-processing Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 4
- 229910021421 monocrystalline silicon Inorganic materials 0.000 description 4
- 238000004458 analytical method Methods 0.000 description 3
- 238000000354 decomposition reaction Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 238000012217 deletion Methods 0.000 description 2
- 230000037430 deletion Effects 0.000 description 2
- 239000000686 essence Substances 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000015654 memory Effects 0.000 description 2
- 208000037656 Respiratory Sounds Diseases 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 239000013078 crystal Substances 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0004—Industrial image inspection
- G06T7/0008—Industrial image inspection checking presence/absence
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T3/00—Geometric image transformations in the plane of the image
- G06T3/40—Scaling of whole images or parts thereof, e.g. expanding or contracting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Quality & Reliability (AREA)
- Image Processing (AREA)
Abstract
The present invention proposes a kind of monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection, and its step is as follows:First, monocrystaline silicon solar cell piece surface image is pre-processed using image scaling and medium filtering;Secondly, a kind of grid line detection method is proposed, for deleting main gate line and secondary grid line in monocrystaline silicon solar cell piece surface image;Then, a kind of super-pixel segmentation and the method that adaptive thresholding is combined are proposed, for detecting defect area in without grating figure picture, obtains initial testing result figure;Finally, initial detecting result figure is post-processed by image scaling, obtains final testing result figure.The present invention not only requires relatively low to the acquisition quality of monocrystaline silicon solar cell piece surface image, and there is faster detection speed while keeping compared with high detection accuracy rate, it is significant to the quality inspection efficiency and ex factory pass rate of raising monocrystaline silicon solar cell piece.
Description
Technical field
The present invention relates to the technical field of the surface defects detection based on machine vision, is mainly based upon the list that grid line detects
Crystal silicon solar batteries piece detection method of surface flaw.
Background technology
The essence of monocrystaline silicon solar cell piece surface defects detection is whether to judge monocrystaline silicon solar cell piece surface
Existing defects and the position of positioning defect areas.In recent years, solar energy power generating turned into the master for solving problem of energy crisis
One of scheme is wanted, while the quality of solar cell tablet quality directly affects the efficiency of photovoltaic generation again, therefore, to solar-electricity
It is significant that pond piece surface carries out defects detection.
Solar battery sheet surface defects detection based on machine vision is current Main Trends of The Development, and prior art is equal
It is that various types of mathematical modeling is established to carry out defects detection according to the feature of defect, for example, based on Gradient Features, gathering
The mathematical modelings such as class algorithm, frequency-domain analysis, matrix decomposition and machine learning.However, all there is generalization ability deficiency in these methods
The problem of, i.e.,:The detection of a few class defects is only applicable to, for example, the detection method based on Gradient Features and clustering algorithm is applicable
Detected in the defects of crackle and disconnected grid type, the detection method based on frequency-domain analysis and matrix decomposition is more suitable for strip and distribution
The defects of more scattered, is detected.
The content of the invention
For the technology of the existing solar battery sheet detection method of surface flaw generalization ability deficiency based on machine vision
Problem, the present invention proposes the monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection, according to monocrystalline silicon
The characteristics of solar battery sheet surface texture, the master in monocrystaline silicon solar cell piece surface image is deleted by grid line detection
Grid line and secondary grid line, meanwhile, examined using the method that super-pixel segmentation and adaptive thresholding are combined in without grating figure picture
Survey defect area.
In order to solve the above-mentioned technical problem, the technical scheme is that:Mono-crystalline silicon solar electricity based on grid line detection
Pond piece detection method of surface flaw, it is characterised in that its step is as follows:One) using image scaling and medium filtering to monocrystalline silicon
Solar battery sheet surface image is pre-processed;Two) detected by grid line and delete monocrystaline silicon solar cell piece surface image
In main gate line and secondary grid line;Three) detected and lacked in without grating figure picture using super-pixel segmentation and adaptive thresholding method
Region is fallen into, obtains initial detecting result figure;Four) initial detecting result figure is post-processed by image scaling, obtained final
Testing result figure.
The pretreatment includes greyscale image transitions, image scaling and median filter method:
1) greyscale image transitions:The monocrystaline silicon solar cell piece surface image collected is converted into gray level image, obtained
The gray level image arrived is I1;
2) image scaling:By the gray level image I that size is m × n1Equal proportion scaling is carried out, it is k × g's to obtain resolution ratio
Zoomed image I2;Wherein, scaling is r ∈ (0,1), k=round (m × r), g=round (n × r), round () generation
Table rounds up computing, and m, n, k, g are nonnegative integers, and m, n represent gray level image I1Line number and columns, k, g represent scaling
Image I2Line number and columns;
3) image noise reduction:Using median filter method to zoomed image I2Noise reduction is carried out, removes what detection was had a great influence
Impulsive noise, the image tagged after noise reduction is I, and resolution ratio is k × g.
The median filter method is to zoomed image I2Carrying out the method for noise reduction is:Define the pixel that coordinate is (x, y)
Neighborhood be:
Nx,y(R)=(i, j) | | i-x |≤R, | j-y |≤R } (1)
Wherein, R is positive integer, represents the radius of neighborhood, and (i, j) represents the coordinate of pixel in neighborhood, Nx,y(R) represent
Using centered on pixel (x, y), radius as all pixels point in R neighborhood coordinate set;
Medium filtering is carried out to pixel (x, y), its result is:
I (x, y)=Median { I2(i,j),(i,j)∈Nx,y(R)} (2)
Wherein, I (x, y) be pixel (x, y) medium filtering after gray value, I2(i, j) is zoomed image I2Middle coordinate is
The gray value of the pixel of (i, j), Median { } are to take median operation;
The R=1 of formula (2);Utilize formula (2) traversal zoomed image I2All pixels point can obtain image I.
The side that the main gate line and secondary grid line deleted in monocrystaline silicon solar cell piece surface image are detected by grid line
Method is:
1) the gray scale sum of each row and columns of image I is calculated:
Wherein, xth row in I (x, y) representative image I, at y row pixel gray value, SHxRepresent all pictures of image I xth rows
The gray scale of element and SLyRepresent image I y row all pixels gray scale and;
2) position of main gate line and secondary grid line is detected:
Wherein, ZX represents the row coordinate set where main grid line position in image I, and ZY represents secondary grid line position in image I
The row coordinate set at place;If the gray scale and SH of xth rowxMore than or equal to the average value of all row gray scale sumsThen
Xth row is the row that main gate line is sitting in, and now, the coordinate " x " of this line is brought into set ZX;If y row gray scale and
SLyMore than or equal to the average value of all row gray scale sumsThen y row are the row that secondary grid line is sitting in, and now, this is arranged
Coordinate " y " bring into set ZY;
3) main gate line and secondary grid line are deleted:
According to the instruction of row coordinate in row coordinate set ZX, corresponding row is deleted from image I, realizes and deletes main grid
Line;According to the instruction of row coordinate in row coordinate set ZY, corresponding row are deleted from image I, realizes and deletes secondary grid line;Main grid
Image tagged after line and secondary grid line are deleted is image I', and its size is p × q, and p, q are positive integers, and p < k, q < g.
Defect area is detected in the no grating figure picture, the method for obtaining initial detecting result figure is:
1) super-pixel segmentation
Image segmentation, the area after segmentation are carried out to the image I' for deleting main gate line and secondary grid line using SLIC super-pixel algorithm
Field mark is:F={ F1,F2,...,FT, wherein, T represents the quantity of cut zone, each cut zone from the upper left corner of image to
The lower right corner is sorted successively, Fc, c ∈ [1, T] represent all pixels point that c-th of cut zone includes;
2) adaptive thresholding
A. each cut zone F is calculatedcAverage gray value be:
Wherein, I'(i, j) the i-th row in representative image I', at j row pixel gray value, M (Fc), c ∈ [1, T] are represented
Cut zone FcAverage gray value, | Fc| represent cut zone FcThe number of middle pixel;
B. defect area is determined by threshold decision:
D(Fc), c ∈ [1, T] represent cut zone FcTesting result, represent cut zone F if 0cBelong to defect area
Domain, otherwise cut zone FcBelong to non-defective region;
After obtaining the testing result of all cut zone, initial detecting result figure D is obtained, size is p × q.
The method of the post processing is:
Image scaling:Equal proportion image scale operation is performed in pretreatment stage, scaling is r ∈ (0,1), now will
Initial detecting result figure D carries out the inverse operation of equal proportion image scaling according to former ratio r, obtains final detection result image M, M
Size be h × l, h=round (p/r), l=round (q/r).M is bianry image, and the gray value of defect area is 0, non-to lack
The gray value for falling into region is 1.
The present invention is detected using grid line detection and super-pixel segmentation to monocrystaline silicon solar cell piece surface defect, energy
Enough to find rapidly and be accurately positioned the position of defect, advantage is:1) present invention is to monocrystaline silicon solar cell piece surface image
Acquisition quality it is less demanding, coloured image and gray level image, image resolution ratio be not less than 600 × 600 or so;2)
Detection speed of the present invention is very fast, and monolithic detection time is about 0.25s or so.The method can significantly improve monocrystaline silicon solar cell
The detection efficiency of piece surface defect, there is important meaning to the quality inspection efficiency and ex factory pass rate for improving monocrystaline silicon solar cell piece
Justice.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
There is the required accompanying drawing used in technology description to be briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
Fig. 1 is the structural representation of the present invention.
Embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present invention, the technical scheme in the embodiment of the present invention is carried out clear, complete
Site preparation describes, it is clear that described embodiment is only part of the embodiment of the present invention, rather than whole embodiments.Based on this
Embodiment in invention, the every other reality that those of ordinary skill in the art are obtained under the premise of creative work is not paid
Example is applied, belongs to the scope of protection of the invention.
As shown in figure 1, a kind of monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection, specifically
Implementation steps are as follows:
One) monocrystaline silicon solar cell piece surface image is pre-processed using image scaling and medium filtering.
Pretreatment:
1) greyscale image transitions:The monocrystaline silicon solar cell piece surface image collected is converted into gray level image, if
Image is gray level image in itself, then keeps constant, and the gray level image for marking to obtain is I1。
The effect of greyscale image transitions:The present processes design specifically for gray level image, and purpose is exactly to reduce
Amount of calculation, because coloured image is three passages, 3 times of amount of calculation, it is contemplated that the image of collection is likely to be coloured image,
Gray level image is likely to be, so first uniformly changing into gray level image.
2) image scaling:By the gray level image I that size is m × n1Equal proportion scaling is carried out, it is k × g's to obtain resolution ratio
Zoomed image I2;Wherein, scaling is r ∈ (0,1), k=round (m × r), g=round (n × r), round () generation
Table rounds up computing, and m, n, k, g are nonnegative integers, and m, n represent gray level image I1Line number and columns, k, g represent scaling
Image I2Line number and columns.By the gray level image I that the size of input is 624 × 6241Equal proportion scaling is carried out, scaling is
0.2, obtain the zoomed image I that size is 125 × 1252。
The purpose of image scaling is to reduce amount of calculation.Because being found by many experiments, resolution ratio (size) exceedes
600 × 600 image, Detection results are influenceed after narrowing down to 125 × 125 less, amount of calculation can also to be substantially reduced, so
And if the original resolution inherently 125*125 of collection image, it is unworkable, because the quality of this image compares high score
Resolution narrows down to that an equal amount of image is far short of what is expected, because the definition of full resolution pricture is higher, even narrows down to same
Size, definition image still more less than original resolution are good.Therefore, the resolution ratio of image not preferably less than 600*600, although
Also to be reduced during actual calculating below, but image scaling reduces amount of calculation.
3) image noise reduction:Using median filter method to zoomed image I2Carry out noise reduction and obtain image I, remove to detecting shadow
Ring larger impulsive noise.
Define coordinate is for the pixel neighborhood of a point of (x, y):
Nx,y(R)=(i, j) | | i-x |≤R, | j-y |≤R } (1)
Wherein, R is positive integer, represents the radius of neighborhood, and (i, j) represents the coordinate of pixel in neighborhood, Nx,y(R) represent
Centered on pixel (x, y), radius is the coordinate set of all pixels point in R neighborhood, such as:Nx,y(1) represent with (x,
Y) 3 × 3 neighborhoods centered on, Nx,y(2) 5 × 5 neighborhoods centered on (x, y) are represented, by that analogy.
Medium filtering is carried out to pixel (x, y), its result is:
I (x, y)=Median { I2(i,j),(i,j)∈Nx,y(1)} (2)
Wherein, I (x, y) be pixel (x, y) medium filtering after gray value, I2(i, j) is zoomed image I2Middle coordinate is
The gray value of the pixel of (i, j), Median { } is takes median operation, i.e.,:All elements in set are sorted by size,
The element value of centre is taken as output result.
Utilize formula (2) traversal zoomed image I2All pixels point can obtain image I, I size and I2It is identical, and
125×125。
Two) main gate line and the secondary grid line in monocrystaline silicon solar cell piece surface image are deleted using grid line detection method.
Grid line is deleted:
1) the gray scale sum of each row and columns of image I is calculated:
Wherein, xth row in I (x, y) representative image I, at y row pixel gray value, SHxRepresent all pictures of image I xth rows
The gray scale of element and SLyRepresent image I y row all pixels gray scale and.
2) position of main gate line and secondary grid line is detected:
Wherein, ZX represents the row coordinate set where main grid line position in image I, and ZY represents secondary grid line position in image I
The row coordinate set at place.If the gray scale and SH of xth rowxMore than or equal to the average value of all row gray scale sumsThen
Xth row is the row that main gate line is sitting in, and now, the coordinate " x " of this line is brought into set ZX;If y row gray scale and
SLyMore than or equal to the average value of all row gray scale sumsThen y row are the row that secondary grid line is sitting in, and now, this is arranged
Coordinate " y " bring into set ZY.
By input picture it can be found that the whether thicker main gate line of line direction, or secondary grid thinner on column direction
Line, its gray value is all apparently higher than non-grid region, thus it is possible to by calculating the gray scale on row and column direction and to detect it
, if the gray scale of grid line and non-grid region be averaged, it is clear that the larger grid region of gray scale is naturally larger than average.
3) main gate line and secondary grid line are deleted:
First, according to the instruction of row coordinate (row where main gate line) in row coordinate set ZX, by corresponding row from image I
Middle removal, it is deletion main gate line to perform this operation;Then, according to row coordinate (row where secondary grid line) in row coordinate set ZY
Instruction, corresponding row are removed from image I, it is to delete secondary grid line to perform this operation;Image mark after grid line deletion
Image I' is designated as, resolution ratio is 117 × 90.
Three) defect area is detected in without grating figure picture using super-pixel segmentation and adaptive thresholding method, obtained
Initial detecting result figure.
Defective area detection
1) super-pixel segmentation
For more accurate positioning defect areas, the present invention carries out image segmentation using super-pixel algorithm to image I',
Result queue after segmentation is:F={ F1,F2,...,F800, wherein 800 represent the quantity of cut zone, each cut zone is from figure
The upper left corner of picture is sorted successively to the lower right corner, Fc, c ∈ [1,800] represent all pixels point that c-th of region includes.The present invention
Super-pixel segmentation can be realized by SLIC (Simple Linear Iterative Clustering) super-pixel algorithm
(ACHANTA R,SHAJI A,SMITH K,et al.SLIC superpixels compared to state-of-the-
art superpixel methods.IEEE transactions on pattern analysis and machine
intelligence,2012,34(11):2274-2282.)。
2) adaptive thresholding
A. each cut zone F is calculatedcAverage gray value:
Wherein, I'(i, j) the i-th row in representative image I', at j row pixel gray value, M (Fc), c ∈ [1,800] generations
Table cut zone FcAverage gray value, | Fc| represent cut zone FcThe number of middle pixel.
B. defect area is determined by threshold decision:
D(Fc), c ∈ (1,800) represent cut zone FcTesting result, represent cut zone F if 0cBelong to defect
Region, otherwise cut zone FcBelong to non-defective region;
After obtaining the testing result of all cut zone, you can initial detecting result figure is obtained, labeled as D, size 117
× 90, it is clear that D is bianry image, and defective area grayscale value is 0, and area free from defect gray value is 1.
Four) initial detecting result figure is post-processed by image scaling, obtains final testing result figure.
Post processing:
Image scaling:Equal proportion image scale operation is performed in pretreatment stage, scaling 0.2 now will be initial
Testing result figure D carries out the inverse operation of equal proportion image scaling according to former ratio 0.2, obtains final detection result image M, M's
Size is 585 × 450, wherein, 585=round (117/0.2), 450=round (90/0.2).M is bianry image, defect area
The gray value in domain is 0, and the gray value in non-defective region is 1.Above by input picture reduce purpose be in order to reduce amount of calculation,
Here amplify again according to former ratio, it is therefore an objective to allow observer is clearer to observe defect area.
The present invention be used for implement hardware environment be:Intel (R) Core (TM) i5CPU 3.2G computers, 8GB internal memories,
1GB video memory video cards, the software environment of operation are:Matlab R2014b and Windows 7.The selected monocrystalline silicon sun of experiment
Energy cell piece surface image is the gray level image of a width resolution ratio 624 × 624.The present invention is 624 × 624 in 100 width resolution ratio
Monocrystaline silicon solar cell piece surface image on be tested checking, it is as a result as follows:
1) using the technology of the present invention can be detected with fast speed in monocrystaline silicon solar cell piece surface image
Defect, the average detected time of each image is 0.25s or so.
2) the defects of can obtaining 99% using the present invention, detects accuracy, wherein, defects detection accuracy is defined as judging
Monocrystaline silicon solar cell piece surface image of the quantity of correct monocrystaline silicon solar cell piece surface image with participating in detection
The ratio between total quantity.Table 1 show the testing result statistics of 100 width test images:All detection is correct for 40 defective images,
Only have 1 width image in 60 width zero defect images by flase drop, be that a small amount of spot on cell piece surface is erroneously interpreted as the reason for flase drop
(surface blot is also considered as defect to defect by some producers for requiring strict, and by this standard, detection accuracy of the invention is
100%).Such flase drop can receive for producer, because for producer, find out all sun for including defect
Energy cell piece is mostly important, and a small amount of zero defect cell piece can be received by misjudgement.
The testing result statistics of the width test image of table 1 100
The foregoing is merely illustrative of the preferred embodiments of the present invention, is not intended to limit the invention, all essences in the present invention
God any modification, equivalent substitution and improvements made etc., should be included in the scope of the protection with principle.
Claims (5)
1. the monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection, it is characterised in that its step is as follows:
One) monocrystaline silicon solar cell piece surface image is pre-processed using image scaling and medium filtering;Two) grid are passed through
Main gate line and secondary grid line in monocrystaline silicon solar cell piece surface image are deleted in line detection;Three) using super-pixel segmentation and certainly
Adapt to thresholding method and detect defect area in without grating figure picture, obtain initial detecting result figure;Four) image scaling is passed through
Initial detecting result figure is post-processed, obtains final testing result figure;
Described detected by grid line deletes main gate line in monocrystaline silicon solar cell piece surface image and the method for secondary grid line is:
1) the gray scale sum of pretreated each row and columns of image I is calculated:
<mrow>
<msub>
<mi>SH</mi>
<mi>x</mi>
</msub>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>g</mi>
</munderover>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>x</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>3</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
<mrow>
<msub>
<mi>SL</mi>
<mi>y</mi>
</msub>
<mo>=</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<mi>I</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mi>y</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>g</mi>
<mo>&rsqb;</mo>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>4</mn>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
Wherein, xth row in I (x, y) representative image I, at y row pixel gray value, SHxRepresent image I xth row all pixels
Gray scale and SLyRepresent the gray scale and k, g representative image I line number and columns of image I y row all pixels;
2) position of main gate line and secondary grid line is detected:
<mrow>
<mi>Z</mi>
<mi>X</mi>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>Z</mi>
<mi>X</mi>
<mo>&cup;</mo>
<mo>{</mo>
<mi>x</mi>
<mo>}</mo>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>SH</mi>
<mi>x</mi>
</msub>
<mo>&GreaterEqual;</mo>
<mfrac>
<mn>1</mn>
<mi>k</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>x</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>k</mi>
</munderover>
<msub>
<mi>SH</mi>
<mi>x</mi>
</msub>
<mo>,</mo>
<mi>x</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>Z</mi>
<mi>X</mi>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>o</mi>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>r</mi>
<mi>w</mi>
<mi>i</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>5</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
<mrow>
<mi>Z</mi>
<mi>Y</mi>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>Z</mi>
<mi>Y</mi>
<mo>&cup;</mo>
<mo>{</mo>
<mi>y</mi>
<mo>}</mo>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<msub>
<mi>SL</mi>
<mi>y</mi>
</msub>
<mo>&GreaterEqual;</mo>
<mfrac>
<mn>1</mn>
<mi>g</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>y</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>g</mi>
</munderover>
<msub>
<mi>SL</mi>
<mi>y</mi>
</msub>
<mo>,</mo>
<mi>y</mi>
<mo>&Element;</mo>
<mo>&lsqb;</mo>
<mn>1</mn>
<mo>,</mo>
<mi>g</mi>
<mo>&rsqb;</mo>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mi>Z</mi>
<mi>Y</mi>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>o</mi>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>r</mi>
<mi>w</mi>
<mi>i</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>6</mn>
<mo>)</mo>
</mrow>
<mo>;</mo>
</mrow>
Wherein, ZX represents the row coordinate set where main grid line position in image I, and ZY is represented in image I where secondary grid line position
Row coordinate set;If the gray scale and SH of xth rowxMore than or equal to the average value of all row gray scale sumsThen xth row
It is the row that main gate line is sitting in, now, the coordinate " x " of this line is brought into set ZX;If the gray scale and SL of y rowyGreatly
In the average value equal to all row gray scale sumsThen y row are the row that secondary grid line is sitting in, now, by the seat of this row
Mark " y " is brought into set ZY;
3) main gate line and secondary grid line are deleted:
According to the instruction of row coordinate in row coordinate set ZX, corresponding row is deleted from image I, realizes and deletes main gate line;Root
According to the instruction of row coordinate in row coordinate set ZY, corresponding row are deleted from image I, realizes and deletes secondary grid line;Main gate line and
Image tagged after secondary grid line is deleted is image I', and its size is p × q, and p, q are positive integers, and p < k, q < g.
2. the monocrystaline silicon solar cell piece detection method of surface flaw according to claim 1 based on grid line detection, its
It is characterised by, the pretreatment includes greyscale image transitions, image scaling and median filter method:
1) greyscale image transitions:The monocrystaline silicon solar cell piece surface image collected is converted into gray level image, obtained
Gray level image is I1;
2) image scaling:By the gray level image I that size is m × n1Equal proportion scaling is carried out, obtains the scaling figure that resolution ratio is k × g
As I2;Wherein, scaling is r ∈ (0,1), k=round (m × r), and g=round (n × r), round () represent four houses
Five enter computing, and m, n, k, g are nonnegative integers, and m, n represent gray level image I1Line number and columns, k, g represent zoomed image I2's
Line number and columns;
3) image noise reduction:Using median filter method to zoomed image I2Noise reduction is carried out, the pulse being had a great influence to detection is removed and makes an uproar
Sound, the image tagged after noise reduction is I, and resolution ratio is k × g.
3. the monocrystaline silicon solar cell piece detection method of surface flaw according to claim 2 based on grid line detection, its
It is characterised by, the median filter method is to zoomed image I2Carrying out the method for noise reduction is:Define the pixel that coordinate is (x, y)
Neighborhood be:
Nx,y(R)=(i, j) | | i-x |≤R, | j-y |≤R } (1)
Wherein, R is positive integer, represents the radius of neighborhood, and (i, j) represents the coordinate of pixel in neighborhood, Nx,y(R) represent with picture
Centered on vegetarian refreshments (x, y), radius be R neighborhood in all pixels point coordinate set;
Medium filtering is carried out to pixel (x, y), its result is:
I (x, y)=Median { I2(i,j),(i,j)∈Nx,y(R)} (2)
Wherein, I (x, y) be pixel (x, y) medium filtering after gray value, I2(i, j) is zoomed image I2Middle coordinate for (i,
J) gray value of pixel, Median { } are to take median operation;
The R=1 of formula (2);Utilize formula (2) traversal zoomed image I2All pixels point can obtain image I.
4. the monocrystaline silicon solar cell piece detection method of surface flaw according to claim 1 based on grid line detection, its
It is characterised by, detects defect area in the no grating figure picture, the method for obtaining initial detecting result figure is:
1) super-pixel segmentation
Image segmentation is carried out to the image I' for deleting main gate line and secondary grid line using SLIC super-pixel algorithm, the region mark after segmentation
It is designated as:F={ F1,F2,...,FT, wherein, T represents the quantity of cut zone, and each cut zone is from the upper left corner of image to bottom right
Angle is sorted successively, Fc, c ∈ [1, T] represent all pixels point that c-th of cut zone includes;
2) adaptive thresholding
A. each cut zone F is calculatedcAverage gray value be:
<mrow>
<mi>M</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mrow>
<mo>|</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
<mo>|</mo>
</mrow>
</mfrac>
<munder>
<mo>&Sigma;</mo>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
<mo>&Element;</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
</mrow>
</munder>
<msup>
<mi>I</mi>
<mo>&prime;</mo>
</msup>
<mrow>
<mo>(</mo>
<mi>i</mi>
<mo>,</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>7</mn>
<mo>)</mo>
</mrow>
</mrow>
Wherein, I'(i, j) the i-th row in representative image I', at j row pixel gray value, M (Fc), c ∈ [1, T] represent cut section
Domain FcAverage gray value, | Fc| represent cut zone FcThe number of middle pixel;
B. defect area is determined by threshold decision:
<mrow>
<mi>D</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>M</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>></mo>
<mfrac>
<mn>1</mn>
<mi>T</mi>
</mfrac>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>c</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>T</mi>
</munderover>
<mi>M</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>F</mi>
<mi>c</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>o</mi>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>r</mi>
<mi>w</mi>
<mi>i</mi>
<mi>s</mi>
<mi>e</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>-</mo>
<mo>-</mo>
<mo>-</mo>
<mrow>
<mo>(</mo>
<mn>8</mn>
<mo>)</mo>
</mrow>
</mrow>
D(Fc), c ∈ [1, T] represent cut zone FcTesting result, represent cut zone F if 0cBelong to defect area,
Otherwise cut zone FcBelong to non-defective region;
After obtaining the testing result of all cut zone, initial detecting result figure D is obtained, size is p × q.
5. the monocrystaline silicon solar cell piece detection method of surface flaw according to claim 1 based on grid line detection, its
It is characterised by, the method for the post processing is:
Image scaling:Equal proportion image scale operation is performed in pretreatment stage, scaling is r ∈ (0,1), now will be initial
Testing result figure D carries out the inverse operation of equal proportion image scaling according to former ratio r, obtains the big of final detection result image M, M
Small is h × l, h=round (p/r), l=round (q/r);M is bianry image, and the gray value of defect area is 0, non-defective area
The gray value in domain is 1.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710438495.XA CN107274393B (en) | 2017-06-12 | 2017-06-12 | Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710438495.XA CN107274393B (en) | 2017-06-12 | 2017-06-12 | Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107274393A CN107274393A (en) | 2017-10-20 |
CN107274393B true CN107274393B (en) | 2018-03-23 |
Family
ID=60066527
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710438495.XA Active CN107274393B (en) | 2017-06-12 | 2017-06-12 | Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107274393B (en) |
Families Citing this family (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109003332B (en) * | 2018-06-25 | 2022-12-06 | 重庆交通大学 | Asphalt pavement surface texture simulation system and simulation method thereof |
TWI689869B (en) | 2018-12-28 | 2020-04-01 | 致茂電子股份有限公司 | Image inspection method |
CN109872303B (en) * | 2019-01-16 | 2021-04-23 | 北京交通大学 | Surface defect visual detection method and device and electronic equipment |
CN112038254A (en) * | 2019-05-15 | 2020-12-04 | 天津科技大学 | Automatic wafer detection and marking device based on machine vision technology and design method |
CN110222784B (en) * | 2019-06-13 | 2020-05-26 | 郑州轻工业学院 | Solar cell defect detection method integrating short-term and long-term depth features |
CN110443278B (en) * | 2019-07-02 | 2022-02-15 | 广州大学 | Method, device and equipment for detecting thickness abnormality of grid line of solar cell |
CN115049645B (en) * | 2022-08-12 | 2022-11-04 | 瀚能太阳能(山东)集团有限公司 | Solar cell panel surface defect detection method |
CN115115643B (en) * | 2022-08-30 | 2022-11-15 | 一道新能源科技(衢州)有限公司 | Method for detecting production defects of photovoltaic cell panel |
CN115272306B (en) * | 2022-09-26 | 2022-12-16 | 太阳谷再生资源(江苏)有限公司 | Solar cell panel grid line enhancement method utilizing gradient operation |
CN115330768B (en) * | 2022-10-12 | 2023-02-28 | 江苏跃格智能装备有限公司 | Quality grading method for solar cell panel |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2013026255A (en) * | 2011-07-15 | 2013-02-04 | Npc Inc | Apparatus and method for inspecting defect of solar battery |
CN103258206A (en) * | 2012-11-28 | 2013-08-21 | 河海大学常州校区 | Silicon solar cell surface defect detection and identification method |
CN103872983A (en) * | 2014-04-04 | 2014-06-18 | 天津市鑫鼎源科技发展有限公司 | Device and method for detecting defects on surface of solar cell |
CN105281665A (en) * | 2015-09-29 | 2016-01-27 | 中国电子科技集团公司第四十一研究所 | Portable solar cell defect detection device and method |
CN106296695A (en) * | 2016-08-12 | 2017-01-04 | 西安理工大学 | Adaptive threshold natural target image based on significance segmentation extraction algorithm |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP5243785B2 (en) * | 2007-12-28 | 2013-07-24 | 日清紡ホールディングス株式会社 | Solar cell inspection apparatus and solar cell defect determination method |
-
2017
- 2017-06-12 CN CN201710438495.XA patent/CN107274393B/en active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2013026255A (en) * | 2011-07-15 | 2013-02-04 | Npc Inc | Apparatus and method for inspecting defect of solar battery |
CN103258206A (en) * | 2012-11-28 | 2013-08-21 | 河海大学常州校区 | Silicon solar cell surface defect detection and identification method |
CN103872983A (en) * | 2014-04-04 | 2014-06-18 | 天津市鑫鼎源科技发展有限公司 | Device and method for detecting defects on surface of solar cell |
CN105281665A (en) * | 2015-09-29 | 2016-01-27 | 中国电子科技集团公司第四十一研究所 | Portable solar cell defect detection device and method |
CN106296695A (en) * | 2016-08-12 | 2017-01-04 | 西安理工大学 | Adaptive threshold natural target image based on significance segmentation extraction algorithm |
Non-Patent Citations (1)
Title |
---|
基于机器视觉的太阳能电池片表面缺陷检测研究现状及展望;钱晓亮等;《北京工业大学学报》;20170131;第43卷(第1期);引言,第3.3部分 * |
Also Published As
Publication number | Publication date |
---|---|
CN107274393A (en) | 2017-10-20 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107274393B (en) | Monocrystaline silicon solar cell piece detection method of surface flaw based on grid line detection | |
CN109460753B (en) | Method for detecting floating object on water | |
CN109145830B (en) | Intelligent water gauge identification method | |
Zhang et al. | Solar cell surface defect detection based on improved YOLO v5 | |
CN109919934B (en) | Liquid crystal panel defect detection method based on multi-source domain deep transfer learning | |
CN109636772A (en) | The defect inspection method on the irregular shape intermetallic composite coating surface based on deep learning | |
CN110689011A (en) | Solar cell panel defect detection method of multi-scale combined convolution neural network | |
CN111310756B (en) | Damaged corn particle detection and classification method based on deep learning | |
CN111223093A (en) | AOI defect detection method | |
CN110222784A (en) | Fusion in short-term with it is long when depth characteristic solar battery sheet defect inspection method | |
WO2023207104A1 (en) | Ground penetrating radar tunnel lining quality inspection method based on self-supervised learning | |
CN109191421A (en) | Cylindricality lithium battery periphery pit visible detection method | |
Liu et al. | Surface defect detection of solar cells based on feature pyramid network and GA-faster-RCNN | |
CN110455827A (en) | A kind of smart new energy solar panel qualification detection system and method | |
CN112991264B (en) | Method for detecting crack defect of monocrystalline silicon photovoltaic cell | |
Ying et al. | Automatic detection of photovoltaic module cells using multi-channel convolutional neural network | |
CN111553898A (en) | Fabric defect detection method based on convolutional neural network | |
CN114612406A (en) | Photovoltaic panel defect detection method based on visible light and infrared vision | |
CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
CN116205876A (en) | Unsupervised notebook appearance defect detection method based on multi-scale standardized flow | |
CN114627062A (en) | Solar cell surface defect detection method for improving Faster R-CNN | |
Li et al. | Wafer crack detection based on yolov4 target detection method | |
CN102722718B (en) | Method for classifying cells | |
Abdullah-Vetter et al. | Localization of defects in solar cells using luminescence images and deep learning | |
CN110533657A (en) | A kind of liquid crystal display appearance detecting method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |