CN109615612A - A kind of defect inspection method of solar panel - Google Patents
A kind of defect inspection method of solar panel Download PDFInfo
- Publication number
- CN109615612A CN109615612A CN201811380796.2A CN201811380796A CN109615612A CN 109615612 A CN109615612 A CN 109615612A CN 201811380796 A CN201811380796 A CN 201811380796A CN 109615612 A CN109615612 A CN 109615612A
- Authority
- CN
- China
- Prior art keywords
- matrix
- image
- solar panel
- rank
- indicates
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 37
- 230000007547 defect Effects 0.000 title claims abstract description 35
- 238000007689 inspection Methods 0.000 title claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 102
- 238000012545 processing Methods 0.000 claims abstract description 20
- 238000011084 recovery Methods 0.000 claims abstract description 12
- 238000012937 correction Methods 0.000 claims abstract description 10
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 8
- 230000008569 process Effects 0.000 claims abstract description 8
- 230000002950 deficient Effects 0.000 claims abstract description 7
- 230000011218 segmentation Effects 0.000 claims abstract description 7
- 238000012360 testing method Methods 0.000 claims abstract description 6
- 238000003708 edge detection Methods 0.000 claims abstract description 4
- 238000001914 filtration Methods 0.000 claims abstract description 4
- 238000005192 partition Methods 0.000 claims abstract description 4
- 238000001514 detection method Methods 0.000 claims description 13
- 230000008901 benefit Effects 0.000 claims description 6
- 238000013459 approach Methods 0.000 claims description 4
- 238000012887 quadratic function Methods 0.000 claims description 3
- 238000012797 qualification Methods 0.000 claims description 3
- 238000013139 quantization Methods 0.000 claims description 3
- 239000004577 thatch Substances 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 2
- 230000008439 repair process Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- NJPPVKZQTLUDBO-UHFFFAOYSA-N novaluron Chemical compound C1=C(Cl)C(OC(F)(F)C(OC(F)(F)F)F)=CC=C1NC(=O)NC(=O)C1=C(F)C=CC=C1F NJPPVKZQTLUDBO-UHFFFAOYSA-N 0.000 description 2
- 238000010248 power generation Methods 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 239000000969 carrier Substances 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000008602 contraction Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 239000012535 impurity Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000007781 pre-processing Methods 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/13—Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
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 invention discloses a kind of defect inspection methods of solar panel, it is realized by building the measuring system constituted by test stand, grey area array cameras, telecentric lens, annular light source and objective table, steps are as follows: shooting the picture of solar panel, the expansion process in horizontal and vertical direction is carried out respectively, edge detection is carried out using canny operator, using Hough transform to image rotation, the empty value of rotated image is repaired;By filtering and noise reduction, luminance proportion processing is carried out using gamma correction, then even partition image;Using robust principal component algorithm, the image after segmentation is converted to matrix D, and resolve into low-rank matrix A and sparse matrix E, according to APG gradient approximate algorithm, recovery processing is carried out to matrix D;The reverse single wafer image that returns is carried out to low-rank matrix A and sparse matrix E to operate, and handles to obtain the solar panel image of defect by low-rank matrix decomposition technique, and then detect defective locations.
Description
Technical field
The present invention relates to technical field of vision detection, and in particular to a kind of defect inspection method of solar panel.
Background technique
In the converting system of solar energy and electric energy, the main carriers of solar power generation are solar battery sheets, its matter
One of the main reason for amount is influence solar energy generating efficiency.Since cell piece is chronically exposed in air, by sunshine, rainwater,
The influence of the environmental factors such as temperature, air quality, cell piece surface have a degree of breakage, and surface generates defect, so that
The decrease of power generation of solar energy, therefore need regularly to detect the solar panel of solar power plant, Check sees electricity
The quality and case of surface defects of pond piece, and to solar battery that is underproof in actual operation or being damaged
Piece takes timely measure and is replaced.Currently, solar battery sheet surface defect mostly uses artificial detection, traditional people's range estimation
Examination needs to expend a large amount of man power and material, while the result detected is easy the subjective factor influence of examined people again, and detects
Efficiency is lower.The present invention proposes a kind of defect inspection method of solar panel based on machine vision.
Summary of the invention
The purpose of the present invention is to solve drawbacks described above in the prior art, provide a kind of defect of solar panel
Detection method.
The purpose of the present invention can be reached by adopting the following technical scheme that:
A kind of defect inspection method of solar panel, the defect inspection method include the following steps:
S1, the image for shooting a width solar panel, carry out at horizontal expansion processing and vertical expansion original image
Reason, carries out edge detection using canny operator, rotates image according to Hough transform, then uses closest interpolation algorithm reparation
The cavity value of rotated image;
S2, background is removed by Threshold segmentation, reuses median filter to the solar panel after angle correct
Image filtering then carries out luminance proportion processing to image using gamma correction, then carries out even partition to image,
Obtain single target solar battery sheet detection image;
S3, using robust principal component algorithm, the image after segmentation is converted to matrix D, and resolve into low-rank matrix A and dilute
It dredges matrix E and recovery processing is carried out to matrix D according to APG gradient approximate algorithm;
S4, reverse return single wafer image operation is carried out to low-rank matrix A and sparse matrix E, obtained by low-rank square
The zero defect solar cell wafer and defective solar cell wafer of battle array decomposition technique processing, to detect solar energy
The defective locations of solar panel.
Further, the process of the cavity value of rotated image is repaired in the step S1 using closest interpolation algorithm
It is as follows:
It calculates separately and point P (x0, y0) Euclidean distance of four points closed on, it will be with point P (x0, y0) nearest integer sits
The grey scale pixel value of punctuate (x, y) is taken as P (x0, y0) point grey scale pixel value.
Further, the image of the solar panel shot in the step S1 is gray level image, with background color
Difference is obvious, edge clear.
Further, Hough transform process is as follows in the step S1:
First by (ρ, θ) space quantization, wherein ρ indicates that in polar coordinate system with a distance from origin, θ indicates the angle in polar coordinate system
Degree, obtains two-dimensional matrix M (ρ, θ), M (ρ, θ) is an accumulator, initial value 0, M (ρ, θ)=0;
To borderline each point (Xi, Yi), bring all quantized values of θ into ρ=xcos θ+ysin θ, wherein x and y table
The abscissa and ordinate for showing two-dimensional Cartesian coordinate system, calculate corresponding ρ, and by accumulator add 1, M (ρ, θ)=M (ρ, θ)+
1;
To borderline each point (Xi, Yi) after processing, analyze M (ρ, θ) if M (ρ, θ) > T and mean that there are one
Line segment, M (ρ, θ) indicate that the fitting parameter of this line segment, T are a nonnegative integer, are determined by the priori knowledge of target in image
Fixed;
Line segment in image is by M (ρ, θ) and (Xi, Yi) determine jointly, then breaking portion is connected.
Further, gamma correction is edited by the gamma curve to image in the step S2, realization pair
The method that image carries out non-linear tone editor, detects the dark parts and light-colored part in signal, and increases the two ratio
Greatly, wherein gamma correction definition is formulated:A is a non-negative real number, VinIndicate input picture
Contrast, VoutIndicate the contrast value after processing output;As γ < 1, become larger in the dynamic range of low gray level areas, it is high
The dynamic range in gray value region becomes smaller, and then picture superposition, the dynamic range as γ > 1, in low ash angle value region
Become smaller, the dynamic range in high gray value region becomes larger, and then picture contrast reduces.
Further, the step S3 process is as follows:
By n unit target solar battery picture X of segmentation1,X2,…XnBy column vector I1,I2…InIt indicates, it is to be measured
Image array D is indicated are as follows: D=[vec (I1)|vec(I2)|…|vec(In)]∈Rm×n, wherein m indicates each wafer images
Pixel value, vec (In) indicate n-th of chip image array, Rm×nIndicate the real number matrix that row size is m and column size is n;
Using robust principal component algorithm, testing image matrix D is resolved into low-rank matrix A and sparse matrix E, the matrix D
∈Rm×n, it is by a low-rank matrix A ∈ Rm×nWith a sparse matrix E ∈ Rm×nIt constitutes, i.e. D=A+E, wherein low-rank square
Battle array A is the data matrix not polluted by noise or defect, and sparse matrix E is small noise and the defect represented in the picture;
Restore to be found the nearest r dimensional linear subspace projection mapped in matrix D, definition is such as by the matrix D destroyed
Under: min rank (A)+λ ‖ E ‖, wherein objective function is the zero norm ‖ E ‖ of the order rank (A) and sparse matrix E of low-rank matrix A,
λ indicates weight shared by noise;
Rank of matrix is replaced with the nuclear norm approximation of matrix again, 1 norm of matrix is carried out approximate zero norm instead of matrix, obtained
To recovery matrix A0, conversion results are as follows: min rank (A)+λ ∑ij|Eij|, wherein ∑ij|Eij| it indicates in sparse matrix E
The sum of absolute value of all elements;
Using APG gradient approximate algorithm, completes the recovery to matrix D and handle.
Further, APG gradient approximate algorithm is in two qualifications x ∈ H and L (X)=b in the step S3
Under, ming (x) is found out, wherein H indicates to meet | (x, y) | the value of all (x, y) under the conditions of≤‖ x ‖ ‖ y ‖, b is invariant,
G (x) is objective function, and L (X) indicates gradient qualified function, on the section that it is defined, for arbitrary x1And x2, and it is any real
Number λ ∈ (0,1), meets g (λ x1+(1-λ)x2)≤λg(x1)+(1-λ)x2, the recovery problem of low-rank matrix A is approximately as described below:
Relaxation expression formula is as follows:
The wherein ‖ of penalty function f (x)=0.5 L (X)-b ‖2, relaxation parameter μ > 0;
Then it is approached with a separable quadratic function F (X), obtains following formula:
WhereinG (x)=‖ A ‖+λ ‖ E1‖, Y indicate the target point of approach,<>
Indicate dot product,Indicate the total derivative of f, LfIt is the general happiness thatch constant of benefit, ‖ E1‖ indicates ∑ij|Eij|;
Then it acquires
Wherein Ak+1And Ek+1The A and E of+1 iteration of kth are respectively indicated, argmin μ ‖ A ‖ is indicated when μ ‖ A ‖ is minimized
A, μ > 0 is relaxation parameter, operand is adjusted by adjusting μ, to adjust velocity of approch, whereinAndIndicate the matrix A power of G kth time iteration,Indicate matrix E times of G kth time iteration
Power;
Again by pairSVD decomposition is carried out, is iterated until convergence, you can get it final Approaching Results A and E.
The present invention has the following advantages and effects with respect to the prior art:
1, the present invention rotates image using Hough transform, needs to repair picture in the preprocessing part of image
Cavity value caused by rotation, the empty value that this method design uses, which is filled up, can allow the distortion rate of image lower, reinforce the later period
To the precision of the post-processing of image segmentation.
2, the detection method of surface flaw that uses of the present invention is using rectangle recovery algorithms, and wherein approach can be bright for accelerating gradient
It is aobvious to reduce the number of iterations, improve iteration precision, while the number of singular value decomposition is reduced, the rapidity of increased identification;
Detailed description of the invention
Fig. 1 is the flow chart of the defect inspection method of solar panel disclosed in the present invention;
Fig. 2 is that the structure of detection device used in the defect inspection method of solar panel disclosed in the present invention is shown
It is intended to;
Fig. 3 is to repair showing for empty value after the defect inspection method of solar panel disclosed in the present invention rotates image
It is intended to.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is
A part of the embodiment of the present invention, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art
Every other embodiment obtained without making creative work, shall fall within the protection scope of the present invention.
Embodiment
The present invention provides a kind of defect inspection method of solar panel, the visible detection method by by test stand 5,
Colored area array cameras 1, telecentric lens 2, pedestal light source 4 and objective table 3 build the measuring system realization of composition, the measuring system group
Can referring further to Figure 2 at the specific location structural relation of component, the hardware installation of measuring system should meet colour plane battle array phase
Machine 1, telecentric lens 2, objective table 3 and pedestal light source 4 axis parallel, and objective table 3 is installed on the field depth of telecentric lens 2
It is interior.
Using the measuring system as the defect inspection method of detection instrument the following steps are included:
S1, the image for shooting a width solar panel carry out angular adjustment, then carry out to original image horizontal swollen
Swollen processing connects the edge in horizontal direction, then carries out vertical expansion processing, connection to original image
Edge in vertical direction carries out edge detection using canny operator, finds rotation angle using Hough transform,
Then image is rotated, after rotation, is worth using the cavity that closest interpolation algorithm repairs rotated image.
Wherein, process such as Fig. 3 that the cavity value of rotated image is repaired using closest interpolation algorithm, is calculated separately and point
P(x0, y0) Euclidean distance of four points closed on, it will be with point P (x0, y0) nearest rounded coordinate point (x, y) grey scale pixel value
It is taken as P (x0, y0) point grey scale pixel value, it obtains angle and adjusts correct solar panel image.
Wherein, the image of the solar panel of shooting is gray level image, obvious with background color difference, edge clear.
Wherein, Hough transform is first by (ρ, θ) space quantization, and wherein ρ is indicated in polar coordinate system with a distance from origin, θ table
Show the angle in polar coordinate system, obtain two-dimensional matrix M (ρ, θ), M (ρ, θ) is an accumulator, initial value 0, M (ρ, θ)=0.
Again to borderline each point (Xi, Yi), bring all quantized values of θ into ρ=xcos θ+ysin θ, wherein x and y indicates two dimension
The abscissa and ordinate of rectangular coordinate system calculate corresponding ρ, and accumulator are added 1, M (ρ, θ)=M (ρ, θ)+1.Then,
It will be to borderline each point (Xi, Yi) after processing, analysis M (ρ, θ) is meant that if M (ρ, θ) > T there are a line segment,
This line segment is meaningful, and M (ρ, θ) means that the fitting parameter of this line segment.T indicates a nonnegative integer, this is non-negative
Integer is determined by the priori knowledge of target in image.Finally, the line segment in image is by M (ρ, θ) and (Xi, Yi) common
Determining, then breaking portion is connected again.
S2, background is removed by Threshold segmentation, reuses median filter to the solar panel after angle correct
Image filtering goes impurity point to interfere, and then, carries out luminance proportion processing to image using gamma correction, solves solar energy
Then the influence of brightness irregularities caused by solar panel further groove carries out even partition to image and obtains to 6 equal part of ranks
Single target solar battery sheet detection image.
Wherein, gamma correction is edited by the gamma curve to image, carries out non-linear tone volume to image
The method collected, detects the dark parts and light-colored part in signal, and increases the two ratio, to improve picture contrast
Effect.Its definition can be indicated with lower formula:Wherein, A is a non-negative real number, usual situation
Lower A=1, VinIndicate the contrast of input picture, VoutIndicate the contrast value after processing output.As γ < 1, in low ash degree
The dynamic range in region becomes larger, and the dynamic range in high gray value region becomes smaller, and then picture superposition, such gamma
Correction course is referred to as gamma contraction, and as γ > 1, the dynamic range in low ash angle value region becomes smaller, high gray value region
Dynamic range becomes larger, and reduces the picture contrast in low ash angle value region, improves the picture contrast in high gray value region, reaches
To Histogram equalization.
S3, using robust principal component algorithm, the image after segmentation is converted to matrix D, and resolve into low-rank matrix A and dilute
It dredges matrix E and recovery processing is carried out to matrix D according to APG gradient approximate algorithm;
By n unit target solar battery picture X of segmentation1,X2,…XnBy column vector I1,I2…InIt indicates.It is to be measured
Image array D is indicated are as follows: D=[vec (I1)|vec(I2)|…|vec(In)]∈Rm×n, wherein m indicates each wafer images
Pixel value, vec (In) indicate n-th of chip image array, Rm×nIndicate the real number matrix that row size is m and column size is n;
Then, using robust principal component algorithm, matrix D is resolved into low-rank matrix A and sparse matrix E, matrix D ∈ Rm×n, it be by
One low-rank matrix A ∈ Rm×nWith a sparse matrix E ∈ Rm×nIt constitutes.That is D=A+E.Wherein, low-rank matrix A is not made an uproar
Sound or the data matrix of defect pollution, sparse matrix E is small noise and the defect represented in the picture.By the square destroyed
Battle array D restores, and to find the nearest r dimensional linear subspace projection mapped in matrix D.It is defined as follows: min rank (A)+λ ‖ E ‖,
Wherein objective function is zero norm of the sum of ranks sparse matrix E of low-rank matrix A, and λ indicates weight shared by noise.Again with matrix
Nuclear norm approximation replaces rank of matrix, and 1 norm of matrix carrys out approximate zero norm instead of matrix, and be restored matrix A0.Its turn
It is as follows to change result: min rank (A)+λ ∑ij|Eij|, wherein ∑ij|Eij| indicate sparse matrix E in all elements absolute value it
With;Finally, completing the recovery to matrix D using APG gradient approximate algorithm and handling;
Wherein, APG gradient approximate algorithm is to find out ming (x), wherein H at two qualifications x ∈ H and L (X)=b
Indicate to meet | (x, y) | the value of all (x, y) under the conditions of≤‖ x ‖ ‖ y ‖, b is invariant, and g (x) is objective function, L (X)
Gradient qualified function is indicated, on the section that it is defined, for arbitrary x1And x2, and any real number λ ∈ (0,1), meet g (λ
x1+(1-λ)x2)≤λg(x1)+(1-λ)x2, the recovery problem of low-rank matrix A is approximately as described below:
Under normal circumstances, relaxation expression formula is as follows:
The wherein ‖ of penalty function f (x)=0.5 L (X)-b ‖2, relaxation parameter μ > 0;
Then it is approached with a separable quadratic function F (X), obtains following formula:
WhereinG (x)=‖ A ‖+λ ‖ E1‖, Y indicate the target point of approach,<>
Indicate dot product,Indicate the total derivative of f, LfIt is the general happiness thatch constant of benefit, ‖ E1‖ indicates ∑ij|Eij|;
Then it acquires
Wherein Ak+1And Ek+1The A and E of+1 iteration of kth are respectively indicated, argmin μ ‖ A ‖ is indicated when μ ‖ A ‖ is minimized
A, μ > 0 is relaxation parameter, operand is adjusted by adjusting μ, to adjust velocity of approch, whereinAndIndicate the matrix A power of G kth time iteration,Indicate matrix E times of G kth time iteration
Power;
Again by pairSVD decomposition is carried out, is iterated until convergence, you can get it final Approaching Results A and E.
S4, reverse return single wafer image operation is carried out to low-rank matrix A and sparse matrix E, obtained by low-rank square
The zero defect solar cell wafer and defective solar cell wafer of battle array decomposition technique processing, to detect solar energy
The defective locations of solar panel.
Wherein, low-rank matrix A and sparse matrix E is carried out from singular value decomposition by testing image matrix D in step S3.
In conclusion the defect inspection method of solar panel provided by the invention saves big in contrast to artificial detection
The man power and material of amount, the result for avoiding detection are easy the subjective factor influence of examined people again, have to raw on assembly line
The next solar panel Detection accuracy height of output, fireballing advantage.
The above embodiment is a preferred embodiment of the present invention, but embodiments of the present invention are not by above-described embodiment
Limitation, other any changes, modifications, substitutions, combinations, simplifications made without departing from the spirit and principles of the present invention,
It should be equivalent substitute mode, be included within the scope of the present invention.
Claims (7)
1. a kind of defect inspection method of solar panel, which is characterized in that the defect inspection method includes following step
It is rapid:
S1, the image for shooting a width solar panel carry out horizontal expansion processing to original image and vertical expansion are handled, adopt
Edge detection is carried out with canny operator, image is rotated according to Hough transform, is then rotated using closest interpolation algorithm reparation
The cavity value of image afterwards;
S2, background is removed by Threshold segmentation, reuses median filter to the solar panel image after angle correct
Filtering then carries out luminance proportion processing to image using gamma correction, then carries out even partition to image, obtain
Single target solar battery sheet detection image;
S3, using robust principal component algorithm, the image after segmentation is converted to matrix D, and resolve into low-rank matrix A and sparse square
Battle array E carries out recovery processing to matrix D according to APG gradient approximate algorithm;
S4, reverse return single wafer image operation is carried out to low-rank matrix A and sparse matrix E, obtained by low-rank matrix point
The zero defect solar cell wafer and defective solar cell wafer for solving technical treatment, to detect solar battery
The defective locations of plate.
2. a kind of defect inspection method of solar panel according to claim 1, which is characterized in that the step
The process for repairing the cavity value of rotated image in S1 using closest interpolation algorithm is as follows:
It calculates separately and point P (x0, y0) Euclidean distance of four points closed on, it will be with point P (x0, y0) nearest rounded coordinate point
The grey scale pixel value of (x, y) is taken as P (x0, y0) point grey scale pixel value.
3. a kind of defect inspection method of solar panel according to claim 1, which is characterized in that the step
The image of the solar panel shot in S1 is gray level image, obvious with background color difference, edge clear.
4. a kind of defect inspection method of solar panel according to claim 1, which is characterized in that the step
Hough transform process is as follows in S1:
First by (ρ, θ) space quantization, wherein ρ indicates that in polar coordinate system with a distance from origin, θ indicates the angle in polar coordinate system,
It obtains two-dimensional matrix M (ρ, θ), M (ρ, θ) is an accumulator, initial value 0, M (ρ, θ)=0;
To borderline each point (Xi, Yi), bring all quantized values of θ into ρ=xcos θ+ysin θ, wherein x and y indicates two
The abscissa and ordinate for tieing up rectangular coordinate system, calculate corresponding ρ, and accumulator is added 1, M (ρ, θ)=M (ρ, θ)+1;
To borderline each point (Xi, Yi) after processing, analyze M (ρ, θ) if M (ρ, θ) > T and mean that there are a lines
Section, M (ρ, θ) indicate that the fitting parameter of this line segment, T are a nonnegative integer, are determined by the priori knowledge of target in image
's;
Line segment in image is by M (ρ, θ) and (Xi, Yi) determine jointly, then breaking portion is connected.
5. a kind of defect inspection method of solar panel according to claim 1, which is characterized in that the step
Gamma correction is edited by the gamma curve to image in S2, realizes the side that non-linear tone editor is carried out to image
Method detects dark parts and light-colored part in signal, and increases the two ratio, wherein formula is used in gamma correction definition
It indicates:A is a non-negative real number, VinIndicate the contrast of input picture, VoutIndicate processing output
Contrast value afterwards;As γ < 1, becoming larger in the dynamic range of low gray level areas, the dynamic range in high gray value region becomes smaller,
And then picture superposition, as γ > 1, the dynamic range in low ash angle value region becomes smaller, the dynamic model in high gray value region
It encloses and becomes larger, and then picture contrast reduces.
6. a kind of defect inspection method of solar panel according to claim 1, which is characterized in that the step
S3 process is as follows:
By n unit target solar battery picture X of segmentation1,X2,…XnBy column vector I1,I2…InIt indicates, testing image
Matrix D indicates are as follows: D=[vec (I1)|vec(I2)|…|vec(In)]∈Rm×n, wherein m indicates the pixel of each wafer images
Value, vec (In) indicate n-th of chip image array, Rm×nIndicate the real number matrix that row size is m and column size is n;
Using robust principal component algorithm, testing image matrix D is resolved into low-rank matrix A and sparse matrix E, matrix D ∈ Rm ×n, it is by a low-rank matrix A ∈ Rm×nWith a sparse matrix E ∈ Rm×nIt constitutes, i.e. D=A+E, wherein low-rank matrix A is
The data matrix not polluted by noise or defect, sparse matrix E is small noise and the defect represented in the picture;
Restore to be found the nearest r dimensional linear subspace projection mapped in matrix D by the matrix D destroyed, be defined as follows: min
Rank (A)+λ ‖ E ‖, wherein objective function is that zero norm ‖ E ‖, the λ expression of the order rank (A) and sparse matrix E of low-rank matrix A is made an uproar
Weight shared by sound;
Rank of matrix is replaced with the nuclear norm approximation of matrix again, 1 norm of matrix is carried out approximate zero norm instead of matrix, obtained extensive
Complex matrix A0, conversion results are as follows: min rank (A)+λ ∑ij|Eij|, wherein ∑ij|Eij| indicate own in sparse matrix E
The sum of absolute value of element;
Using APG gradient approximate algorithm, completes the recovery to matrix D and handle.
7. a kind of defect inspection method of solar panel according to claim 6, which is characterized in that the step
APG gradient approximate algorithm is to find out ming (x) at two qualifications x ∈ H and L (X)=b in S3, and wherein H indicates to meet |
(x, y) | the value of all (x, y) under the conditions of≤‖ x ‖ ‖ y ‖, b are invariant, and g (x) is objective function, and L (X) indicates gradient limit
Function is determined, on the section that it is defined, for arbitrary x1And x2, and any real number λ ∈ (0,1), meet g (λ x1+(1-λ)x2)
≤λg(x1)+(1-λ)x2, the recovery problem of low-rank matrix A is approximately as described below:
Relaxation expression formula is as follows:
The wherein ‖ of penalty function f (x)=0.5 L (X)-b ‖2, relaxation parameter μ > 0;
Then it is approached with a separable quadratic function F (X), obtains following formula:
WhereinG (x)=‖ A ‖+λ ‖ E1‖, Y indicate the target point of approach, and<>indicates point
Product,Indicate the total derivative of f, LfIt is the general happiness thatch constant of benefit, ‖ E1‖ indicates ∑ij|Eij|;
Then it acquires
Wherein Ak+1And Ek+1The A and E of+1 iteration of kth are respectively indicated, argmin μ ‖ A ‖ indicates the A, μ when μ ‖ A ‖ is minimized
> 0, it is relaxation parameter, operand is adjusted by adjusting μ, so that velocity of approch is adjusted, whereinAndIndicate the matrix A power of G kth time iteration,Indicate matrix E times of G kth time iteration
Power;
Again by pairSVD decomposition is carried out, is iterated until convergence, you can get it final Approaching Results A and E.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811380796.2A CN109615612A (en) | 2018-11-20 | 2018-11-20 | A kind of defect inspection method of solar panel |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811380796.2A CN109615612A (en) | 2018-11-20 | 2018-11-20 | A kind of defect inspection method of solar panel |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109615612A true CN109615612A (en) | 2019-04-12 |
Family
ID=66004007
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811380796.2A Pending CN109615612A (en) | 2018-11-20 | 2018-11-20 | A kind of defect inspection method of solar panel |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109615612A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110232705A (en) * | 2019-05-17 | 2019-09-13 | 沈阳大学 | A kind of reversed low-rank sparse learning objective tracking of fusion fractional order variation adjustment |
CN111915552A (en) * | 2020-06-02 | 2020-11-10 | 华南理工大学 | Method for detecting internal defects of solar cell |
CN112150423A (en) * | 2020-09-16 | 2020-12-29 | 江南大学 | Longitude and latitude sparse mesh defect identification method |
CN113470014A (en) * | 2021-08-31 | 2021-10-01 | 江苏裕荣光电科技有限公司 | Fault detection method of solar cell panel based on artificial intelligence |
CN113570708A (en) * | 2021-07-30 | 2021-10-29 | 重庆市特种设备检测研究院 | Defect three-dimensional modeling method and device and computer readable storage medium |
CN116577671A (en) * | 2023-07-12 | 2023-08-11 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
CN116773548A (en) * | 2023-08-21 | 2023-09-19 | 泓浒(苏州)半导体科技有限公司 | Wafer surface defect detection method and system |
CN118521591A (en) * | 2024-07-25 | 2024-08-20 | 常熟理工学院 | Visual detection method and system for surface defects of optoelectronic device |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101960612A (en) * | 2007-12-28 | 2011-01-26 | 日清纺绩株式会社 | Solar battery inspecting apparatus and method for judging solar battery defect |
CN102113138A (en) * | 2008-08-01 | 2011-06-29 | 欧瑞康太阳Ip股份公司(特吕巴赫) | Method for manufacturing a photovoltaic cell structure |
CN102609951A (en) * | 2012-03-29 | 2012-07-25 | 重庆大学 | Method for detecting shading part of photovoltaic cell panel |
CN102736009A (en) * | 2011-04-04 | 2012-10-17 | 三星电机株式会社 | Method and apparatus for inspecting solar cell |
CN107798326A (en) * | 2017-10-20 | 2018-03-13 | 华南理工大学 | A kind of profile visual detection algorithm |
-
2018
- 2018-11-20 CN CN201811380796.2A patent/CN109615612A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101960612A (en) * | 2007-12-28 | 2011-01-26 | 日清纺绩株式会社 | Solar battery inspecting apparatus and method for judging solar battery defect |
CN102113138A (en) * | 2008-08-01 | 2011-06-29 | 欧瑞康太阳Ip股份公司(特吕巴赫) | Method for manufacturing a photovoltaic cell structure |
CN102736009A (en) * | 2011-04-04 | 2012-10-17 | 三星电机株式会社 | Method and apparatus for inspecting solar cell |
CN102609951A (en) * | 2012-03-29 | 2012-07-25 | 重庆大学 | Method for detecting shading part of photovoltaic cell panel |
CN107798326A (en) * | 2017-10-20 | 2018-03-13 | 华南理工大学 | A kind of profile visual detection algorithm |
Non-Patent Citations (1)
Title |
---|
邱洁: ""基于图像处理的电池片表面缺陷检测方法研究与应用"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110232705A (en) * | 2019-05-17 | 2019-09-13 | 沈阳大学 | A kind of reversed low-rank sparse learning objective tracking of fusion fractional order variation adjustment |
CN110232705B (en) * | 2019-05-17 | 2023-05-12 | 沈阳大学 | Reverse low-rank sparse learning target tracking method integrating fractional order variation adjustment |
CN111915552A (en) * | 2020-06-02 | 2020-11-10 | 华南理工大学 | Method for detecting internal defects of solar cell |
CN112150423A (en) * | 2020-09-16 | 2020-12-29 | 江南大学 | Longitude and latitude sparse mesh defect identification method |
CN112150423B (en) * | 2020-09-16 | 2024-04-30 | 江南大学 | Method for identifying defects of warp and weft sparse meshes |
CN113570708A (en) * | 2021-07-30 | 2021-10-29 | 重庆市特种设备检测研究院 | Defect three-dimensional modeling method and device and computer readable storage medium |
CN113470014A (en) * | 2021-08-31 | 2021-10-01 | 江苏裕荣光电科技有限公司 | Fault detection method of solar cell panel based on artificial intelligence |
CN116577671A (en) * | 2023-07-12 | 2023-08-11 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
CN116577671B (en) * | 2023-07-12 | 2023-09-29 | 中国华能集团清洁能源技术研究院有限公司 | Battery system abnormality detection method and device |
CN116773548A (en) * | 2023-08-21 | 2023-09-19 | 泓浒(苏州)半导体科技有限公司 | Wafer surface defect detection method and system |
CN116773548B (en) * | 2023-08-21 | 2023-11-07 | 泓浒(苏州)半导体科技有限公司 | Wafer surface defect detection method and system |
CN118521591A (en) * | 2024-07-25 | 2024-08-20 | 常熟理工学院 | Visual detection method and system for surface defects of optoelectronic device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109615612A (en) | A kind of defect inspection method of solar panel | |
CN104156536B (en) | The visualization quantitatively calibrating and analysis method of a kind of shield machine cutter abrasion | |
CN109615653B (en) | Leakage water area detection and identification method based on deep learning and visual field projection model | |
CN108520537B (en) | Binocular depth acquisition method based on luminosity parallax | |
CN113744270A (en) | Unmanned aerial vehicle visual detection and identification method for crane complex steel structure surface defects | |
CN109360246B (en) | Stereoscopic vision three-dimensional displacement measurement method based on synchronous subarea search | |
CN103048331B (en) | Printing defect detection method based on flexible template registration | |
CN107677216B (en) | A kind of multiple abrasive grain three-dimensional appearance synchronous obtaining methods based on photometric stereo vision | |
CN109063663B (en) | Thick cloud detection and removal method for time sequence remote sensing image from coarse to fine | |
CN109410207A (en) | A kind of unmanned plane line walking image transmission line faultlocating method based on NCC feature | |
CN101806700B (en) | Method and system for measuring soil infiltration capability | |
CN106525864B (en) | Light source compensation device and compensation method for image gray scale | |
CN108665468B (en) | Device and method for extracting tangent tower insulator string | |
CN114973032B (en) | Deep convolutional neural network-based photovoltaic panel hot spot detection method and device | |
CN106023193B (en) | A kind of array camera observation procedure detected for body structure surface in turbid media | |
CN109712084A (en) | A kind of image repair method, image repair system and flat panel detector | |
CN113628261B (en) | Infrared and visible light image registration method in electric power inspection scene | |
CN110110131B (en) | Airplane cable support identification and parameter acquisition method based on deep learning and binocular stereo vision | |
CN112906689B (en) | Image detection method based on defect detection and segmentation depth convolutional neural network | |
CN110009601A (en) | Large-Scale Equipment irregular contour detection method of surface flaw based on HOG | |
CN112668754B (en) | Power equipment defect diagnosis method based on multisource characteristic information fusion | |
CN109949725A (en) | A kind of AOI system image grayscale standardized method and system | |
CN106027911B (en) | A kind of in-orbit focus adjustment method of the spaceborne transmission of visible light type camera of earth observation | |
CN113935971A (en) | Method and device for detecting surface defects of composite material | |
CN115546716A (en) | Binocular vision-based method for positioning fire source around power transmission line |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190412 |
|
WD01 | Invention patent application deemed withdrawn after publication |