CN112233101A - Photovoltaic cell panel quality evaluation method and system based on artificial intelligence - Google Patents

Photovoltaic cell panel quality evaluation method and system based on artificial intelligence Download PDF

Info

Publication number
CN112233101A
CN112233101A CN202011157344.5A CN202011157344A CN112233101A CN 112233101 A CN112233101 A CN 112233101A CN 202011157344 A CN202011157344 A CN 202011157344A CN 112233101 A CN112233101 A CN 112233101A
Authority
CN
China
Prior art keywords
defect
image
cell
evaluation
color
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.)
Withdrawn
Application number
CN202011157344.5A
Other languages
Chinese (zh)
Inventor
钟竞
曾忠英
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Individual
Original Assignee
Individual
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Individual filed Critical Individual
Priority to CN202011157344.5A priority Critical patent/CN112233101A/en
Publication of CN112233101A publication Critical patent/CN112233101A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/40—Analysis of texture
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10024—Color image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/10—Image acquisition modality
    • G06T2207/10048—Infrared image
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20092—Interactive image processing based on input by user
    • G06T2207/20104—Interactive definition of region of interest [ROI]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection

Landscapes

  • Engineering & Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The invention relates to the field of artificial intelligence, in particular to a photovoltaic cell panel quality evaluation method and system based on artificial intelligence, wherein the method comprises the following steps: acquiring a battery assembly image and a battery panel color image; acquiring a battery piece defect segmentation image of a battery assembly image; classifying the defect types of the battery pack images; constructing a shape defect evaluation model to obtain a cell shape defect evaluation index value; constructing a texture defect evaluation model to obtain a texture defect evaluation index value of the battery piece; obtaining a color defect evaluation index value of the battery piece; and obtaining a single cell quality evaluation value by weighting and summing the cell shape defect evaluation index value, the cell texture defect evaluation index value and the cell color defect evaluation index value, and adding all the cell quality evaluation values to obtain a cell panel quality evaluation value, so as to evaluate the quality of the cell panel. The embodiment of the invention can quickly and efficiently evaluate the quality of the battery plate and improve the accuracy.

Description

Photovoltaic cell panel quality evaluation method and system based on artificial intelligence
Technical Field
The invention relates to the field of artificial intelligence, in particular to a photovoltaic cell panel quality evaluation method and system based on artificial intelligence.
Background
With the rapid development of science and technology, the energy consumption is increasing day by day, and solar energy has an important position in a long-term energy strategy as renewable new energy, and the utilization of solar energy is more and more popularized in the future. Meanwhile, quality detection of the solar cell panel is also more important, and if the solar cell panel has quality problems, the power generation efficiency is reduced, and even the use safety of the solar cell panel is damaged.
At present, the quality detection of the solar cell panel mainly adopts a visual inspection method to detect grid lines, defects, flatness and the like of each cell. In practice, the inventors found that the above prior art has the following disadvantages:
some quality problems cannot be observed by naked eyes, the visual inspection method wastes manpower, the efficiency is low, misjudgment can be caused, and inconvenience is brought to quality evaluation of the solar cell panel.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a photovoltaic cell panel quality evaluation method and system based on artificial intelligence, and the adopted technical scheme is as follows:
in a first aspect, an embodiment of the present invention provides an artificial intelligence-based photovoltaic cell panel quality assessment method, including the following steps:
acquiring a battery assembly image and a battery panel color image;
acquiring a battery piece defect segmentation image of a battery assembly image;
classifying the defect types of the battery assembly image, wherein the defect types comprise n shape defect types and m texture defect types;
constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a battery piece shape defect evaluation index value Sd；
Constructing a texture defect evaluation model according to the texture defect type and the battery piece defect segmentation image to obtain a battery piece texture defect evaluation index value Td；
Performing cell hue analysis and cell boundary color difference analysis on the cell panel color image to obtain a cell color defect evaluation index value Cd；
Evaluating the index value S by evaluating the shape defect of the battery piecedAnd a battery piece texture defect evaluation index value TdAnd a cell color defect evaluation index value CdThe weighted sum of the two cell masses obtains the mass of each cellEvaluating the value M, and adding the quality evaluation values of all the battery plates to obtain a quality evaluation value M' of the battery plate, and evaluating the quality of each battery plate.
Preferably, the index value S for evaluating the shape defect of the battery piecedObtained from the following equation:
Figure BDA0002743176540000021
wherein S' represents the number of shape defective pixels, SOTotal number of pixels, W, representing a segmented image of a defect in a cellxAnd representing the weight of the x-th type shape defect type of the battery piece.
Preferably, the index value T for evaluating the texture defect of the battery piecedObtained from the following equation:
Figure BDA0002743176540000022
wherein S' represents the number of texture defect pixels, WyAnd representing the weight of the y-th texture defect type of the battery piece.
Preferably, the index value C for evaluating the color defect of the celldThe acquisition steps are as follows:
obtaining a single battery plate color image of the battery plate color image, and converting the battery plate color image into an HSV space;
performing hue analysis on the HSV image:
Figure BDA0002743176540000023
wherein, P represents a hue analysis evaluation value, O represents the pixel number of the image in the designated interval of the H channel, and N represents the total pixel number of the HSV image;
analyzing the color difference of the cell boundaries:
Figure BDA0002743176540000024
wherein, BdRepresenting a cell boundary color difference evaluation value, Hi、HjRespectively represent the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively representing an ith ROI area and a jth ROI area V channel image, wherein the ROI areas are interested areas divided based on four corner points of the cell, and mean is an average value of all pixels of the image;
color defect evaluation index value C of the entire battery piecedObtained from the following equation:
Cd＝P+Bd
preferably, the specific steps of evaluating the quality of each panel are:
and dividing the quality evaluation value M' of the battery panel into at least two segmentation intervals according to a preset threshold value, so that each segmentation interval corresponds to a corresponding quality level.
In a second aspect, another embodiment of the present invention provides an artificial intelligence-based photovoltaic cell panel quality assessment system, which includes the following modules:
the image acquisition module is used for acquiring a battery assembly image and a battery panel color image;
the image segmentation module is used for acquiring a battery piece defect segmentation image of the battery assembly image;
the defect classification module is used for classifying the defect types of the battery pack images, wherein the defect types comprise n shape defect types and m texture defect types;
the shape defect evaluation module is used for constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a battery piece shape defect evaluation index value Sd；
The texture defect evaluation module is used for constructing a texture defect evaluation model according to the texture defect type and the battery piece defect segmentation image to obtain a battery piece texture defect evaluation index value Td；
Color defect evaluation module for batteryPerforming cell hue analysis and cell boundary color difference analysis on the plate color image to obtain a cell color defect evaluation index value Cd；
A panel quality evaluation module for evaluating the index value S by evaluating the shape defect of the celldAnd a battery piece texture defect evaluation index value TdAnd a cell color defect evaluation index value CdThe quality evaluation value M of each cell is obtained through weighted summation, and the quality evaluation values of all the cells are added to be used as the quality evaluation value M' of the cell panel, and the quality of each cell panel is evaluated.
Preferably, the shape defect evaluation module further comprises a shape defect evaluation index value calculation module for obtaining a cell shape defect evaluation index value Sd：
Figure BDA0002743176540000031
Wherein S' represents the number of shape defective pixels, SOTotal number of pixels, W, representing a segmented image of a defect in a cellxAnd representing the weight of the x-th type shape defect type of the battery piece.
Preferably, the texture defect evaluation module further includes a texture defect evaluation index calculation module for obtaining a texture defect evaluation index value T of the battery celld：
Figure BDA0002743176540000032
Wherein S' represents the number of texture defect pixels, WyAnd representing the weight of the y-th texture defect type of the battery piece.
Preferably, the color defect evaluating module further includes:
the image conversion module is used for obtaining a single battery piece color image of the battery piece plate color image and converting the battery piece color image into an HSV space;
the hue analysis evaluation module is used for performing hue analysis on the HSV image:
Figure BDA0002743176540000041
wherein, P represents a hue analysis evaluation value, O represents the pixel number of the image in the designated interval of the H channel, and N represents the total pixel number of the HSV image;
the boundary color difference evaluation module is used for analyzing the boundary color difference of the battery pieces:
Figure BDA0002743176540000042
wherein, BdRepresenting a cell boundary color difference evaluation value, Hi、HjRespectively represent the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively representing an ith ROI area and a jth ROI area V channel image, wherein the ROI areas are interested areas divided based on four corner points of the cell, and mean is an average value of all pixels of the image;
a color defect evaluation index calculation module for calculating color defect evaluation index value C of the whole celld：
Cd＝P+Bd
Preferably, the panel quality evaluation module further includes a segment evaluation module, configured to divide the panel quality evaluation value M' into at least two segment intervals according to a preset threshold, so that each segment interval corresponds to a corresponding quality level.
The invention has the following beneficial effects:
1. according to the invention, the quality of the battery piece can be judged only by fully utilizing the camera and calculating the obtained image without manual judgment, so that the accuracy is improved.
2. According to the invention, the color difference of the image is effectively analyzed through the HSV color space, so that the color quality of the battery plate can be accurately evaluated.
3. The invention adopts a double-path deep neural network structure, so that the network is easier to converge and has stronger robustness.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions and advantages of the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and other drawings can be obtained by those skilled in the art without creative efforts.
FIG. 1 is a block diagram of an artificial intelligence based photovoltaic panel quality assessment system according to an embodiment of the present invention;
fig. 2 is a flowchart of a method for evaluating the quality of a photovoltaic cell panel based on artificial intelligence according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a panel of some type;
FIG. 4 is a schematic diagram of a battery slice after dividing boundary ROI regions at four corner points of the battery slice;
FIG. 5 is a flow diagram of a fully connected network;
fig. 6 is a structural diagram of a system for evaluating quality of a photovoltaic cell panel based on artificial intelligence according to an embodiment of the present invention.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a method and a system for evaluating the quality of a photovoltaic cell panel based on artificial intelligence according to the present invention, with reference to the accompanying drawings and preferred embodiments. In the following description, different "one embodiment" or "another embodiment" refers to not necessarily the same embodiment. Furthermore, the particular features, structures, or characteristics may be combined in any suitable manner in one or more embodiments.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
The following describes a specific scheme of the photovoltaic cell panel quality evaluation method and system based on artificial intelligence in detail with reference to the accompanying drawings.
Referring to fig. 1 and fig. 2, fig. 1 shows a framework diagram of a system for evaluating quality of a photovoltaic cell panel based on artificial intelligence according to an embodiment of the present invention, fig. 2 shows a flowchart of a method for evaluating quality of a photovoltaic cell panel based on artificial intelligence according to another embodiment of the present invention, and the method includes the following specific steps:
and step S001, acquiring a battery pack image and a battery piece color image.
An infrared camera in the electroluminescent tester shoots an image of the battery pack, and a common camera acquires a color image of the battery piece.
Electroluminescence (EL) is a light emitting phenomenon generated when a substance is excited by corresponding electric energy under the action of a certain electric field. When the battery pack emits light, the battery pack image is collected through the infrared camera, the infrared camera used in the embodiment of the invention has higher spatial resolution, and the accuracy of subsequent battery piece defect detection is improved.
And step S002, obtaining a battery piece defect segmentation image of the battery pack image.
The battery piece defect segmentation image is obtained through a semantic segmentation network model, in the embodiment of the invention, the semantic segmentation network model adopts an Encoder-Decoder structure, the types of encoders are many, and in order to take efficiency into consideration, the Encoder-Decoder in the embodiment of the invention adopts a skip-level connection structure. In other embodiments, the semantic segmentation network model may also adopt networks such as ICNet and deep to extract features, or other semantic segmentation network models that can achieve the same effect.
The specific training process is as follows:
1) the collected battery pack image is sliced, as shown in fig. 3, each battery plate is composed of a plurality of battery pieces 30, the subfissure type on each battery piece is conveniently analyzed, and the size of each battery piece is determined according to the type of the battery plate. The slice image is then normalized, i.e. the picture matrix is changed to a floating point number between 0,1, for better convergence of the model.
2) The network label data pixel categories are divided into three categories, namely normal, shape defect and texture defect, and the three categories are sequentially replaced by numbers 0-2.
3) Inputting the processed image data and the tag data which is subjected to one-hot coding into a network for training.
The battery piece defect detection encoder performs feature extraction on input data, inputs the normalized near-infrared image data and outputs a defect detection feature map;
and the battery piece defect detection decoder performs up-sampling on the defect detection characteristic graph, performs pixel-level classification and outputs a probability graph for battery piece defect segmentation.
4) The loss function is a cross-entropy loss function.
5) And obtaining a battery piece defect segmentation image by performing argmax operation on the network output battery piece defect segmentation probability map.
And step S003, classifying the defect types of the battery pack image, wherein the defect types comprise n shape defect types and m texture defect types.
In the embodiment of the invention, the defect classification network model adopts an Encoder-Decoder structure, and the battery piece defect type classification Encoder applies an EfficientNet image classification network for feature extraction, so that the network obtains excellent performance with lower calculation cost.
The specific training process is as follows:
1) the slice images processed in step S002 are simultaneously input as a defect classification network model.
2) The label data of the network is the defect type of the battery piece of the current image, wherein the shape defect comprises five categories of normal, hidden crack, broken grid, burn and fragment, and is replaced by a number of 0-4. The texture defects comprise four categories of normal, spot, wheel mark and fingerprint, and are sequentially replaced by numbers 0-3.
3) Inputting the processed image data and the tag data which is subjected to one-hot coding into a network for training.
And the battery plate defect type classification encoder performs feature extraction on input data, inputs the normalized near-infrared image data and outputs a defect classification feature map.
And (4) carrying out correlation (joint operation) on the defect classification characteristic diagram extracted by the battery sheet defect type classification encoder and the defect detection characteristic diagram extracted by the battery sheet defect detection encoder in the step S002, changing the defect classification characteristic diagram into a one-dimensional vector through a Flatten operation, sending the one-dimensional vector into a full-connection network, wherein the full-connection network plays a role of mapping the characteristics to a sample marking space, and outputting the probability of each shape defect type and the probability of each texture defect type of the battery sheet in the current image.
4) The final classification function of the fully connected network uses softmax. The specific form of the fully-connected network is shown in fig. 5, and belongs to a multi-label fully-connected neural network.
5) The loss function is a cross-entropy loss function.
6) The output result is a probability value, and the tag is obtained through argmax operation.
Step S002 and step S003 are two networks in parallel, and step S002 plays a role in assisting the training and accelerating the convergence in step S003. The characteristic diagram received by the full-connection network is the defect detection characteristic diagram generated by the joint battery piece defect detection encoder, so that the training of the battery piece defect detection encoder is supervised by two items, and the defect detection characteristic diagram generated by the battery piece defect detection encoder is beneficial to the training of the full-connection network.
Step S004, constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a shape defect evaluation index value Sd。
As an example, in the embodiment of the present invention, the shape defects are classified into cracks, burns, chips, and broken grids, which are mostly caused by errors in the production process.
The shape defect evaluation model is as follows:
Figure BDA0002743176540000071
Sdfor the shape defect evaluation index value, S' is the number of shape defect pixels, SOTotal number of pixels, W, representing a segmented image of a defect in a cellnAnd representing the weight of the nth type of shape defect of the battery piece.
There are four defect classes for shape defects, with burnout, chipping, and subfissure being the most severe, and second, the weights in the present embodiment are assigned as follows [0.35, 0.30, 0.25, 0.10 ].
Step S005, a texture defect evaluation model is constructed according to the texture defect type and the battery piece defect segmentation image, and a texture defect evaluation index value T is obtainedd。
As an example, texture defects are classified as blobs, fingerprints, or wheel marks in embodiments of the present invention, which may be caused by excessive machine pressure or manual mishandling.
The texture defect evaluation model is as follows:
Figure BDA0002743176540000072
Tdfor texture defect evaluation index value, S' is the number of texture defect pixels, WmAnd representing the weight of the m-th class texture defect type of the battery piece.
There are three defect classes for texture defects, blob, wheel print, fingerprint, and in the present embodiment, the weights are assigned as follows [0.6, 0.25, 0.15 ].
Since the weights of the defects are different, in step S002, for a plurality of defects of the same type existing in a single cell, as an example, the cell has defects of broken grids and burnt shapes, and the defect with the highest weight among the defects of the same type should be labeled, that is, the burnt with the highest weight in the above example should be labeled.
Step S006, carrying out cell slice hue analysis and cell slice on the collected color imageAnalyzing the color difference of the boundary to obtain a color defect evaluation index value Cd。
The color defects are divided into abnormal colors, abnormal or uneven colors of corner areas, and are probably caused by uneven chemical reaction during film coating.
The color defect assessment method specifically comprises the following steps:
1) the method comprises the steps of carrying out white balance processing on a color image, namely an RGB image, of a battery piece collected by a common camera to enable the color of the image to be richer and more real, and then carrying out color space transformation to convert the color image into an HSV space.
The HSV conversion steps are as follows:
a) the original image is normalized, i.e. the range becomes [0,1 ].
b)V＝max(R，G，B)
c)
Figure BDA0002743176540000081
d)
Figure BDA0002743176540000082
e) The calculation result may appear as H < 0, so the following calculation is performed:
f)
Figure BDA0002743176540000083
HSV is a relatively intuitive color model in which the color parameters are: hue (H, Hue), Saturation (S, Saturation), lightness (V, Value), and Value ranges are:
0≤H≤360，0≤S≤1，0≤V≤1
the hue channel of the HSV space can well represent the hue difference of the image, and has a great effect on analyzing the color defect of the image.
2) And carrying out hue analysis on the HSV image.
Counting the number of pixels of an H channel of the image in different intervals, and calculating to obtain the proportion of each interval:
Figure BDA0002743176540000091
p is a hue analysis evaluation value, O is the number of pixels of the image in the designated interval of the H channel, and N is the total number of pixels of the HSV image.
For photovoltaic cells, the dark blue color of a polysilicon cell is the most common color, and the black color of single crystal silicon. As an example, in the embodiment of the present invention, a polysilicon cell is selected for analysis, and the analysis method is as follows:
since the interval hue is close to blue [220,260], the pixel value range intervals selected in the embodiment of the present invention are [0, 220 ], (260,360 ].
Counting the number a of pixels of the H channel pixel value in the value domain interval to obtain a pixel proportion
Figure BDA0002743176540000092
PaNamely, the color deviation of the whole image can be reflected, and the larger the deviation is, the larger the color difference of the battery plate is.
3) And analyzing the color difference of the cell boundaries.
As shown in fig. 4, boundary ROI area division is performed based on four corner points of the cell, to obtain four ROI areas 40, and the boundary color unevenness is measured by the following formula:
Figure BDA0002743176540000093
Bdrepresenting a cell boundary color difference evaluation value, Hi、HjI.e. representing the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively represent the ith and jth ROI area V-channel images. mean is the average of all pixels of the image.
Figure BDA0002743176540000094
Is the normalization operation of the H-channel image because of S, V-passThe pixel value ranges are all [0,1]]。
Evaluation value B of color difference of battery plate boundarydThe larger the size, the more uneven the border color.
4) Color defect estimation value C of battery piece as a wholedObtained from the following equation:
Cd＝P+Bd
in the embodiments of the present invention, P ═ Pa。
For convenience of description, the parallel steps S002 and S003, and the parallel steps S004 and S005 are used as the shape texture defect model, and the parallel steps S006 are used as the color defect model. In the embodiment of the invention, the execution sequence of the shape texture defect model and the color defect model is not sequential or simultaneous, and the shape texture defect model can be executed first and then the color defect model can be executed; or executing the color defect model first and then executing the shape texture defect model; the shape texture defect model and the color defect model can be executed simultaneously, and the effect of the embodiment of the invention is not influenced.
Step S007 of evaluating an index value S by evaluating the shape defect of the celldAnd a battery piece texture defect evaluation index value TdAnd a cell color defect evaluation index value CdThe quality evaluation value M of each cell is obtained through weighted summation, and the quality evaluation values of all the cells are added to be used as the quality evaluation value M' of the cell panel, and the quality of each cell panel is evaluated.
The method comprises the following specific steps:
1) obtaining a single battery piece quality evaluation value M:
M＝Sd*Ws+Td*Wt+Cd*Wc
wherein, WsWeight of shape defect, WtWeight, W, representing texture defectscRepresenting the weight of the color defect. Generally, shape defects have the greatest effect on the cell, followed by texture and finally color. The empirical values are [0.5, 0.35, 0.15 respectively]。
2) And adding the quality evaluation values of all the battery plates to obtain the total quality evaluation value M' of the whole battery plate.
3) Dividing the battery plate quality evaluation value M' into at least two segmentation intervals according to a preset threshold value, and enabling each segmentation interval to correspond to a corresponding quality level:
as an example, the embodiment of the present invention is divided into three segment intervals, and in other embodiments, segmentation may be performed according to actual situations:
the first partition [0,1], M' is for [0,1], and the quality is considered to be excellent.
The second segment interval (1, 2), M' e (1, 2), is considered to be of good quality.
Third segmentation [2, + ∞), M' ∈ [2, + ∞), the quality was considered poor.
According to the embodiment of the invention, the battery plate image is acquired through the EL technology and a common camera, a double-path deep neural network structure is adopted, so that the network is easier to converge and stronger in robustness, the defects of the shape, the texture and the color are scored, finally, the quality evaluation of the battery plate is realized through the combination of the shape defect evaluation, the texture defect evaluation and the color defect evaluation, the evaluation accuracy is improved, and the method is fast and efficient. And the quality of the battery piece can be judged only by fully utilizing the camera and calculating the obtained image without manual judgment, so that the labor is saved.
Based on the same inventive concept as the method, the invention further provides a photovoltaic cell panel quality evaluation system based on artificial intelligence.
Referring to fig. 6, the system includes an image acquisition module 101, an image segmentation module 102, a defect classification module 103, a shape defect evaluation module 104, a texture defect evaluation module 105, a color defect evaluation module 106, and a panel quality evaluation module 107.
Specifically, the image obtaining module 101 is configured to obtain a battery assembly image and a battery panel color image; the image segmentation module 102 is configured to obtain a battery piece defect segmentation image of the battery assembly image; the defect classification module 103 is used for classifying the defect types of the battery pack image, wherein the defect types comprise n types of shape defect typesAnd m texture defect types; the shape defect evaluation module 104 is used for constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a battery piece shape defect evaluation index value Sd(ii) a The texture defect evaluation module 105 is configured to construct a texture defect evaluation model according to the texture defect type and the battery piece defect segmentation image, and obtain a battery piece texture defect evaluation index value Td(ii) a The color defect evaluation module 106 is configured to perform cell hue analysis and cell boundary color difference analysis on the cell panel color image to obtain a cell color defect evaluation index value Cd(ii) a The panel quality evaluation module 107 is used for evaluating an index value S by evaluating the shape defect of the celldAnd a battery piece texture defect evaluation index value TdAnd a cell color defect evaluation index value CdThe quality evaluation value M of each cell is obtained through weighted summation, and the quality evaluation values of all the cells are added to be used as the quality evaluation value M' of the cell panel, and the quality of each cell panel is evaluated.
Preferably, the shape defect evaluation module further comprises a shape defect evaluation index value calculation module for obtaining a cell shape defect evaluation index value Sd：
Figure BDA0002743176540000111
Wherein S' represents the number of shape defective pixels, SOTotal number of pixels, W, representing a segmented image of a defect in a cellxAnd representing the weight of the x-th type shape defect type of the battery piece.
Preferably, the texture defect evaluation module further includes a texture defect evaluation index calculation module for obtaining a texture defect evaluation index value T of the battery celld：
Figure BDA0002743176540000112
Wherein S' represents the number of texture defect pixels, WyWeight for expressing y-th texture defect type of battery slice。
Preferably, the color defect evaluating module further includes:
the image conversion module is used for obtaining a single battery piece color image of the battery piece plate color image and converting the battery piece color image into an HSV space;
the hue analysis evaluation module is used for performing hue analysis on the HSV image:
Figure BDA0002743176540000113
wherein, P represents a hue analysis evaluation value, O represents the pixel number of the image in the designated interval of the H channel, and N represents the total pixel number of the HSV image;
the boundary color difference evaluation module is used for analyzing the boundary color difference of the battery pieces:
Figure BDA0002743176540000121
wherein, BdRepresenting a cell boundary color difference evaluation value, Hi、HjRespectively represent the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively representing an ith ROI area and a jth ROI area V channel image, wherein the ROI areas are interested areas divided based on four corner points of the cell, and mean is an average value of all pixels of the image;
a color defect evaluation index calculation module for calculating color defect evaluation index value C of the whole celld：
Cd＝P+Bd
Preferably, the panel quality evaluation module further includes a segment evaluation module, configured to divide the panel quality evaluation value M' into at least two segment intervals according to a preset threshold, so that each segment interval corresponds to a corresponding quality level.
In summary, in the embodiment of the invention, the image acquisition module acquires the cell image by using the EL technology and the common camera, the image segmentation module, the defect classification module, the shape defect evaluation module, the texture defect evaluation module and the color defect evaluation module perform shape, texture and color defect scoring, and finally the quality evaluation of the cell panel is realized by the cell panel quality evaluation module, so that the evaluation accuracy is improved, and the method is fast and efficient. And the quality of the battery piece can be judged only by fully utilizing the camera and calculating the obtained image without manual judgment, so that the labor is saved.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (10)

1. A photovoltaic cell panel quality evaluation method based on artificial intelligence is characterized by comprising the following steps:
acquiring a battery assembly image and a battery panel color image;
acquiring a battery piece defect segmentation image of the battery assembly image;
classifying the defect types of the battery assembly image, wherein the defect types comprise n shape defect types and m texture defect types;
constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a battery piece shape defect evaluation index value Sd；
Constructing a texture defect evaluation model according to the texture defect type and the battery piece defect segmentation image to obtain a battery piece texture defect evaluation index value Td；
Performing cell hue analysis and cell boundary color difference analysis on the cell panel color image to obtain a cell color defect evaluation index value Cd；
Evaluating an index value S by evaluating the shape defect of the battery piecedAnd the index value T for evaluating the texture defects of the battery piecedAnd the cell color defect evaluation index value CdThe quality evaluation value M of each cell is obtained through weighted summation, and the quality evaluation values of all the cells are added to be used as the quality evaluation value M' of the cell panel, and the quality of each cell panel is evaluated.
2. The artificial intelligence-based photovoltaic cell panel quality assessment method according to claim 1, wherein said cell shape defect assessment index value SdObtained from the following equation:
Figure FDA0002743176530000011
wherein S' represents the number of shape defective pixels, SOThe total number of pixels, W, of the segmented image of the battery piece defectxAnd representing the weight of the x-th type shape defect type of the battery piece.
3. The artificial intelligence based photovoltaic cell panel quality assessment method according to claim 2, wherein said cell texture defect assessment index value TdObtained from the following equation:
Figure FDA0002743176530000012
wherein S' represents the number of texture defect pixels, WyAnd representing the weight of the y-th texture defect type of the battery piece.
4. The artificial intelligence based photovoltaic cell panel quality assessment method according to claim 1, wherein said cell color defect assessment index value CdThe acquisition steps are as follows:
obtaining a single cell color image of the cell plate color image, and converting the cell color image into an HSV space;
performing hue analysis on the HSV image:
Figure FDA0002743176530000021
wherein P represents a hue analysis evaluation value, O represents the pixel number of an image in a specified interval of an H channel, and N represents the total pixel number of the HSV image;
analyzing the color difference of the cell boundaries:
Figure FDA0002743176530000022
wherein, BdRepresenting a cell boundary color difference evaluation value, Hi、HjRespectively represent the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively representing an ith ROI area and a jth ROI area V channel image, wherein the ROI areas are interested areas divided based on four corner points of the cell, and mean is an average value of all pixels of the image;
the cell being integralColor defect evaluation index value CdObtained from the following equation:
Cd＝P+Bd
5. the artificial intelligence based photovoltaic cell panel quality assessment method according to claim 1, wherein the specific steps of assessing the quality of each cell panel are as follows:
and dividing the quality evaluation value M' of the battery panel into at least two segmentation intervals according to a preset threshold value, so that each segmentation interval corresponds to a corresponding quality level.
6. A photovoltaic cell panel quality evaluation system based on artificial intelligence is characterized by comprising the following modules:
the image acquisition module is used for acquiring a battery assembly image and a battery panel color image;
the image segmentation module is used for acquiring a battery piece defect segmentation image of the battery assembly image;
the defect classification module is used for classifying the defect types of the battery pack image, wherein the defect types comprise n types of shape defect types and m types of texture defect types;
a shape defect evaluation module for constructing a shape defect evaluation model according to the shape defect type and the battery piece defect segmentation image to obtain a battery piece shape defect evaluation index value Sd；
The texture defect evaluation module is used for constructing a texture defect evaluation model according to the texture defect type and the battery piece defect segmentation image to obtain a battery piece texture defect evaluation index value Td；
A color defect evaluation module for performing cell hue analysis and cell boundary color difference analysis on the cell panel color image to obtain a cell color defect evaluation index value Cd；
A panel quality evaluation module for evaluating the index value S by evaluating the shape defect of the celldAnd the index value T for evaluating the texture defects of the battery piecedAnd stationThe index value C for evaluating the color defect of the battery piecedThe quality evaluation value M of each cell is obtained through weighted summation, and the quality evaluation values of all the cells are added to be used as the quality evaluation value M' of the cell panel, and the quality of each cell panel is evaluated.
7. The system according to claim 6, wherein the shape defect estimation module further comprises a shape defect estimation index value calculation module for obtaining a cell shape defect estimation index value Sd：
Figure FDA0002743176530000031
Wherein S' represents the number of shape defective pixels, SOThe total number of pixels, W, of the segmented image of the battery piece defectxAnd representing the weight of the x-th type shape defect type of the battery piece.
8. The artificial intelligence based photovoltaic cell panel quality assessment system according to claim 7, wherein said texture defect assessment module further comprises a texture defect assessment index calculation module for obtaining a cell texture defect assessment index value Td：
Figure FDA0002743176530000032
Wherein S' represents the number of texture defect pixels, WyAnd representing the weight of the y-th texture defect type of the battery piece.
9. The artificial intelligence based photovoltaic panel quality assessment system according to claim 6, wherein said color defect assessment module further comprises:
the image conversion module is used for obtaining a single battery piece color image of the battery piece plate color image and converting the battery piece color image into an HSV space;
the hue analysis evaluation module is used for performing hue analysis on the HSV image:
Figure FDA0002743176530000033
wherein P represents a hue analysis evaluation value, O represents the pixel number of an image in a specified interval of an H channel, and N represents the total pixel number of the HSV image;
the boundary color difference evaluation module is used for analyzing the boundary color difference of the battery pieces:
Figure FDA0002743176530000034
wherein, BdRepresenting a cell boundary color difference evaluation value, Hi、HjRespectively represent the ith and jth ROI H-channel images, Si、SjRespectively represent the ith and jth ROI area S channel images, Vi、VjRespectively representing an ith ROI area and a jth ROI area V channel image, wherein the ROI areas are interested areas divided based on four corner points of the cell, and mean is an average value of all pixels of the image;
a color defect evaluation index calculation module for calculating color defect evaluation index value C of the whole celld：
Cd＝P+Bd
10. The artificial intelligence based photovoltaic cell panel quality assessment system according to claim 6, wherein said panel quality assessment module further comprises a segment assessment module for dividing said panel quality assessment value M' into at least two segment intervals according to a preset threshold value, such that each segment interval corresponds to a corresponding quality level.
CN202011157344.5A 2020-10-26 2020-10-26 Photovoltaic cell panel quality evaluation method and system based on artificial intelligence Withdrawn CN112233101A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011157344.5A CN112233101A (en) 2020-10-26 2020-10-26 Photovoltaic cell panel quality evaluation method and system based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011157344.5A CN112233101A (en) 2020-10-26 2020-10-26 Photovoltaic cell panel quality evaluation method and system based on artificial intelligence

Publications (1)

Publication Number Publication Date
CN112233101A true CN112233101A (en) 2021-01-15

Family

ID=74110797

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011157344.5A Withdrawn CN112233101A (en) 2020-10-26 2020-10-26 Photovoltaic cell panel quality evaluation method and system based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN112233101A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112517473A (en) * 2020-11-11 2021-03-19 曾忠英 Photovoltaic cleaning robot stable operation method and system based on artificial intelligence
CN113674215A (en) * 2021-07-26 2021-11-19 浙江大华技术股份有限公司 Light spot identification method and device of photovoltaic panel and computer readable storage medium
CN115049645A (en) * 2022-08-12 2022-09-13 瀚能太阳能(山东)集团有限公司 Solar cell panel surface defect detection method

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112517473A (en) * 2020-11-11 2021-03-19 曾忠英 Photovoltaic cleaning robot stable operation method and system based on artificial intelligence
CN113674215A (en) * 2021-07-26 2021-11-19 浙江大华技术股份有限公司 Light spot identification method and device of photovoltaic panel and computer readable storage medium
CN113674215B (en) * 2021-07-26 2024-05-31 浙江大华技术股份有限公司 Light spot identification method and device for photovoltaic panel and computer readable storage medium
CN115049645A (en) * 2022-08-12 2022-09-13 瀚能太阳能(山东)集团有限公司 Solar cell panel surface defect detection method
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

Similar Documents

Publication Publication Date Title
CN112233101A (en) Photovoltaic cell panel quality evaluation method and system based on artificial intelligence
CN108961235B (en) Defective insulator identification method based on YOLOv3 network and particle filter algorithm
CN111598861B (en) Improved Faster R-CNN model-based non-uniform texture small defect detection method
CN109614996B (en) Weak visible light and infrared image fusion identification method based on generation countermeasure network
CN111667455B (en) AI detection method for brushing multiple defects
CN109376792B (en) Photovoltaic cell appearance defect classification method based on multi-channel residual error neural network
CN112241699B (en) Object defect type identification method, object defect type identification device, computer equipment and storage medium
WO2019214268A1 (en) Photovoltaic array fault diagnosis method based on composite information
CN110610475B (en) Visual defect detection method of deep convolutional neural network
CN110689011A (en) Solar cell panel defect detection method of multi-scale combined convolution neural network
CN110766681B (en) Bamboo strip surface defect detection method based on triple loss network
CN112588764A (en) Multi-robot collaborative cleaning method and device on photovoltaic module based on artificial intelligence
CN111291814B (en) Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy
CN111178177A (en) Cucumber disease identification method based on convolutional neural network
CN115082744B (en) Artificial intelligence-based solar heat collection efficiency analysis method and system
CN112200121A (en) Hyperspectral unknown target detection method based on EVM and deep learning
CN112215821A (en) Photovoltaic cell panel grid line defect detection method and system based on artificial intelligence
CN113837994B (en) Photovoltaic panel defect diagnosis method based on edge detection convolutional neural network
CN112258490A (en) Low-emissivity coating intelligent damage detection method based on optical and infrared image fusion
CN108711150B (en) End-to-end pavement crack detection and identification method based on PCA
CN111695436B (en) High spatial resolution remote sensing image scene classification method based on target enhancement
CN117218101A (en) Composite wind power blade defect detection method based on semantic segmentation
CN113962929A (en) Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line
CN117392065A (en) Cloud edge cooperative solar panel ash covering condition autonomous assessment method
CN114581407B (en) Self-adaptive defect detection method for photovoltaic module

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
WW01 Invention patent application withdrawn after publication
WW01 Invention patent application withdrawn after publication

Application publication date: 20210115