CN113962929A - Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line - Google Patents
Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line Download PDFInfo
- Publication number
- CN113962929A CN113962929A CN202111043149.4A CN202111043149A CN113962929A CN 113962929 A CN113962929 A CN 113962929A CN 202111043149 A CN202111043149 A CN 202111043149A CN 113962929 A CN113962929 A CN 113962929A
- Authority
- CN
- China
- Prior art keywords
- photovoltaic cell
- defect
- grade
- percent
- cell assembly
- 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
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
- G06T7/0006—Industrial image inspection using a design-rule based approach
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Quality & Reliability (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
- Photovoltaic Devices (AREA)
Abstract
The invention discloses a method and a system for detecting defects of a photovoltaic cell assembly and a photovoltaic cell assembly production line, wherein the method comprises the steps of dividing an EL image of the photovoltaic cell assembly into photovoltaic cell images; marking the defect type and the defect position on the image with the defect by using a labeling tool; constructing a sample set; constructing a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set; detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell; the defect identification is carried out on the photovoltaic cell image, so that internal defects and micro defects can be identified, and the defect position can be identified; model training is carried out before target recognition model recognition is utilized, an artificial intelligence method is adopted for defect recognition, recognition efficiency is high, recognition accuracy is high, and multiple defects can be recognized.
Description
Technical Field
The invention belongs to the technical field of solar cell detection, and particularly relates to a method and a system for detecting defects of a photovoltaic cell module and a photovoltaic cell module production line.
Background
At present, methods for detecting defects of a photovoltaic battery assembly on a production line include:
1. manual visual inspection: the method has the advantages that the defects of the photovoltaic cell assembly on the production line are detected by artificial vision, so that the quality of the photovoltaic cell assembly is controlled, the detection efficiency is low, the detection cost is high, and the detection precision is low due to visual fatigue.
2. Visible light image method: the method can only detect apparent defects of the photovoltaic cell assembly and is not suitable for detecting micro defects and defects inside the photovoltaic cell assembly.
3. And (3) near-infrared detection method: the method has the advantages that EL imaging is carried out on the photovoltaic cell assembly by utilizing the near-infrared camera, and the defect detection is carried out by combining the V-I characteristic curve of the photovoltaic cell assembly, so that the defect position is difficult to determine, the detected defect types are few, and the defect detection effect on the appearance is not obvious.
4. Laser diode array method: the method is characterized in that photoluminescence imaging is carried out by means of laser scanning of a laser diode array or direct irradiation of an ultraviolet light emitting diode, so that defect detection of the photovoltaic cell assembly is realized.
Disclosure of Invention
The invention aims to provide a method and a system for detecting defects of a photovoltaic cell assembly and a photovoltaic cell assembly production line, and aims to solve the problems that the detection efficiency is low, the detection precision is low, the detection defect types are few, the defect positions are difficult to determine, and internal defects and micro defects cannot be detected in the traditional technology.
The invention solves the technical problems through the following technical scheme: a method for detecting defects of a photovoltaic cell assembly comprises the following steps:
step 1: acquiring an EL image of the photovoltaic cell assembly, and dividing the EL image of the photovoltaic cell assembly into photovoltaic cell images;
step 2: screening out images with defects in the photovoltaic cell images, and marking the defect types and defect positions on the images with the defects by using a labeling tool, wherein the defect positions are the positions of a marking frame;
and step 3: constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities;
and 4, step 4: constructing a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set to obtain a trained target recognition model;
and 5: and detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
According to the invention, the EL image is segmented to obtain a photovoltaic cell image, and the defect identification is carried out on the photovoltaic cell to identify the internal defect and the micro defect and identify the defect position; the model is trained before the target recognition model is used for recognition, and an artificial intelligence method is used for defect recognition, so that the recognition efficiency is high, the recognition precision is high, and various defects can be recognized; compared with a manual visual detection method, the method has the advantages of low detection cost, higher accuracy, higher speed and the like, and can adapt to various severe working environments.
Further, in the step 1, the EL image of the photovoltaic cell module is divided into photovoltaic cell images by a template matching method, and the specific implementation process is as follows:
step 1.1: cutting off information at the edge of the EL image;
step 1.2: performing SURF feature extraction on the cut EL image to obtain feature point information;
step 1.3: selecting a point which accords with the coordinate of the expected point in the characteristic point information as a template center to generate a template;
step 1.4: identifying a non-luminous black diamond-like area in the EL image by adopting the template in the step 1.3;
step 1.5: calculating a central point of the black diamond-like region, wherein the central point is a vertex of the photovoltaic cell;
step 1.6: and connecting four adjacent central points to form a quadrangle, wherein the external rectangle of the quadrangle is a photovoltaic cell image.
Further, the target recognition model adopts a YOLO v5s model.
Further, in the step 2, the defect types include grid breakage, scratch, hidden crack, black spot, cold solder joint, pollution, crack, corner collapse, failure and over-soldering.
Further, the detection method further comprises:
step 6: dividing the defect grade of each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell;
and 7: and determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
And determining the defect grade of the photovoltaic cell assembly according to the defect grades of all the photovoltaic cells on the single photovoltaic cell assembly so as to control the quality according to the defect grade of the photovoltaic cell assembly.
Further, in the step 6, the defect grades of the photovoltaic cell sheet comprise a grade a, a grade b, a grade c and a grade d;
when only scratch exists on a single photovoltaic cell sheet, l is more than or equal to 02≤L/5，0≤n2Less than or equal to 2; or no defect type is present; the defect grade of the photovoltaic cell is a grade; wherein L is the side length of the photovoltaic cell piece, L2To length of scratch, n2The number of scratches on the photovoltaic cell sheet, and n2Is an integer;
when there is a broken gate on a single photovoltaic cell, and 0 < ∑ s1Less than or equal to 5 percent of S; or has scratches, and L/5 < L2≤2L/3，2＜n2Less than or equal to 4; or with subfissure and 0 < l3≤L/12，0＜n3Less than or equal to 1; or black spots are present, and 0 < l4≤L/4，0＜∑s4Less than or equal to 5 percent of S; or in the presence of contamination, and 0 < ∑ s6Less than or equal to 5 percent of S; the defect grade of the photovoltaic cell is b grade; wherein S is the area of the photovoltaic cell sheet, S1Is the area of a single broken gate,/3Is subfissure length, n3The number of subfissure on the photovoltaic cell sheet, /)4Is the length of the black spot, s4Area of a single black spot, s6As an area of individual contamination;
when a broken grid exists on a single photovoltaic cell, and 5% S < Sigmas1Less than or equal to 10 percent of S; or has scratches, and L/5 < L2Less than or equal to 2L/3; or having subfissure and L/12 < L3≤L/5，1＜n3Less than or equal to 2; or black spots are present and 5% S < Sigmas4Less than or equal to 10 percent of S; or a cold solder joint is present and 0 < l5Less than or equal to L/4; or in the presence of contamination, and 0 < ∑ s6Less than or equal to 10 percent of S; or cracks are present and 0 < l7≤L/5，0＜n7Less than or equal to 1; or a breakout angle of 0 < s8≤4％S，0＜n8Less than or equal to 1; or a failure exists and 0 < s9≤10％S，0＜n9Less than or equal to 1; the defect grade of the photovoltaic cell is grade c; wherein s is7Area of a single crack, /)7Is the crack length, n8Is the number of corner breakouts on the photovoltaic cell sheet, s8Is the area of a single corner break, n9Number of failures on the photovoltaic cell sheet, s9Area of single failure,/10Is the overweld length;
when a broken grid exists on a single photovoltaic cell, and 10% S < Sigmas1(ii) a Or has scratch, and 2L/3 < L2(ii) a Or having subfissure and L/5 < L3，2＜n3(ii) a Or black spots are present and 10% S < Sigmas4(ii) a Or cold solder joints are present and L/4 < L5(ii) a Or contamination, and 10% S < Sigmas6(ii) a Or cracks are present and L/5 < L7，1＜n7(ii) a Or a breakout angle is present and 4% S < S8，1＜n8(ii) a Or failure, and 10% S < S9，1＜n9(ii) a Or over-welding exists and n is more than or equal to 110If so, the defect grade of the photovoltaic cell is d grade; wherein n is10The number of the over-welding on the photovoltaic cell slice.
Further, in the step 7, the defect grades of the photovoltaic cell assembly comprise grade A, grade B, grade C and grade D;
when the number of photovoltaic cell pieces with the defect grade of 94.4 percent to a grade a is in proportion, the number of photovoltaic cell pieces with the defect grade of 0 to b is in proportion to 5.6 percent, and the number of photovoltaic cell pieces with the defect grade of c and d is in proportion to 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is in a grade A;
when the number of the photovoltaic cell pieces with the defect grade of 88.9 percent to a grade a is less than 94.4 percent, the number of the photovoltaic cell pieces with the defect grade of 5.6 percent to a grade B is less than 11.1 percent, the number of the photovoltaic cell pieces with the defect grade of 0 to a grade c is less than 5.6 percent, and the number of the photovoltaic cell pieces with the defect grade of d is 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is B grade;
when the number of the photovoltaic cell pieces with the defect grade of 83.3 percent to more than or equal to a grade a is less than 88.9 percent, the number of the photovoltaic cell pieces with the defect grade of 11.1 percent to more than or equal to a grade b is less than 16.7 percent, the number of the photovoltaic cell pieces with the defect grade of 5.6 percent to more than a grade C is less than 16.7 percent, and the number of the photovoltaic cell pieces with the defect grade of d is 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is grade C;
when the number of the photovoltaic cell pieces with the defect grade of a grade is less than 83.3 percent, the number of the photovoltaic cell pieces with the defect grade of 16.7 percent is less than the number of the photovoltaic cell pieces with the defect grade of b grade, the number of the photovoltaic cell pieces with the defect grade of 16.7 percent is less than the number of the photovoltaic cell pieces with the defect grade of c grade and the number of the photovoltaic cell pieces with the defect grade of D grade are more than 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is D grade.
The invention also provides a photovoltaic cell module defect detection system, which comprises:
the image acquisition and segmentation unit is used for acquiring an EL image of the photovoltaic cell assembly and segmenting the EL image of the photovoltaic cell assembly into photovoltaic cell images;
the marking unit is used for screening out images with defects in the photovoltaic cell images and marking the defect types and the defect positions on the images with the defects by using a labeling tool, wherein the defect positions refer to the positions of the marking frames;
the sample construction unit is used for constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities;
the model building and training unit is used for building a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set to obtain a trained target recognition model;
and the defect detection unit is used for detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
Further, the system further comprises:
the first grade confirming unit is used for carrying out defect grade division on each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell;
and the second grade confirmation unit is used for determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
The invention also provides a photovoltaic cell assembly production line which comprises the photovoltaic cell assembly defect detection system.
Advantageous effects
Compared with the prior art, the invention has the advantages that:
according to the method and the system for detecting the defects of the photovoltaic cell assembly, provided by the invention, the EL image is segmented to obtain the photovoltaic cell image, the defects on the photovoltaic cell can be identified to identify the internal defects and the micro defects, and the defect positions can be identified; the model is trained before the target recognition model is used for recognition, and an artificial intelligence method is used for defect recognition, so that the recognition efficiency is high, the recognition precision is high, and various defects can be recognized; compared with a manual visual detection method, the method has the advantages of low detection cost, higher accuracy, higher speed and the like, and can adapt to various severe working environments.
Drawings
In order to more clearly illustrate the technical solution of the present invention, the drawings needed to be used in the description of the embodiments are briefly introduced below, and it is obvious that the drawings in the following description are only one embodiment of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on the drawings without creative efforts.
FIG. 1 is a flow chart of a method for detecting defects in a photovoltaic cell assembly according to an embodiment of the present invention;
FIG. 2 is a schematic view of an image of a photovoltaic cell assembly EL according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of the image segmentation of a photovoltaic cell assembly EL into photovoltaic cells in an embodiment of the present invention;
FIG. 4 is a diagram of a black diamond-like area in an embodiment of the present invention;
FIG. 5 is a diagram illustrating conversion of defect label information into defect levels according to an embodiment of the present invention;
FIG. 6 is a schematic defect level diagram of an entire photovoltaic cell assembly in an embodiment of the invention;
FIG. 7 is a schematic diagram of various defects of a photovoltaic cell sheet according to an embodiment of the present invention;
fig. 8 is a schematic diagram of various defects of a photovoltaic cell assembly according to an embodiment of the present invention.
The photovoltaic cell module comprises a 1-photovoltaic cell module and a 2-photovoltaic cell piece.
Detailed Description
The technical solutions in the present invention are clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The technical solution of the present application will be described in detail below with specific examples. The following several specific embodiments may be combined with each other, and details of the same or similar concepts or processes may not be repeated in some embodiments.
As shown in fig. 1, the method for detecting defects of a photovoltaic cell module provided in this embodiment includes the following steps:
step 1: and acquiring an EL image of the photovoltaic cell assembly, and dividing the EL image of the photovoltaic cell assembly into photovoltaic cell slice images.
In this embodiment, the number of the near-infrared cameras is 3, the 3 near-infrared cameras are stacked, lenses of the near-infrared cameras are aligned with the photovoltaic cell assemblies, and EL imaging is performed in a scanning shooting manner, as shown in fig. 2, for the photovoltaic cell assembly EL image, pixels of each near-infrared camera are 1360 × 1024, 13756 EL images are obtained in total, and the resolution of each image is 5400 × 2700.
In order to improve the detection accuracy of internal defects and micro defects, the EL image is segmented into photovoltaic cell images, as shown in fig. 3, the specific segmentation process is as follows:
step 1.1: and cutting off information at the edge of the EL image, and eliminating the influence of the edge information on the feature point extraction.
Step 1.2: performing SURF feature extraction on the cut EL image to obtain feature point information; the SURF feature extraction method has the advantages of high precision and unchanged scale, can filter unstable or uninteresting points, and can accurately extract a sufficient number of feature points.
Step 1.3: and selecting a point which accords with the coordinate of the expected point from the characteristic point information as a template center to generate a template.
Step 1.4: non-luminescent black diamond-like regions were identified in the EL image using the template of step 1.3, as shown in fig. 4.
Step 1.5: and calculating the central point of each black diamond-like area, wherein each central point is a vertex of a certain photovoltaic cell.
Step 1.6: and connecting four adjacent central points to form a quadrangle, wherein the circumscribed rectangle of the quadrangle is a photovoltaic cell image.
And constructing a template for each black diamond-like area. The same as the conventional template matching method, the constructed template is the same area as the black diamond-like area. Since the template matching method is known in the art, the detailed process is not described in detail herein.
Step 2: screening out images with defects in the images of the photovoltaic cells, and marking the defect types and the defect positions on the images with the defects by using a labeling tool, wherein the defect positions refer to the positions of the marking frames.
In order to obtain a photovoltaic cell image with known defect types and defect positions, preparation is made for constructing a sample set, an image with defects in the photovoltaic cell image is screened out manually, and the defect types and the defect positions are marked on the image with the defects by a labeling tool. In this embodiment, the labeling tool is label, the label is a visual image calibration tool, the defect frame is directly framed on the photovoltaic cell image by the labeling frame, and an xml file of the defect position corresponding to the defect type can be generated, where the defect position is the position of the labeling frame. In this embodiment, the position of the labeling frame includes the coordinates of the upper left corner and the lower right corner of the labeling frame, the width and the height, the length or the area of the defect framed by the labeling frame can be calculated according to the coordinates of the upper left corner and the lower right corner, the width and the height, the length of the defect is the length of the diagonal line of the labeling frame, and the area of the defect is the area of the labeling frame. And coordinates of the upper left corner coordinate and the lower right corner coordinate in a coordinate system of the marking frame, wherein the coordinate system takes the upper left corner as the origin of coordinates, the horizontal direction is an x axis, and the vertical direction is a y axis.
In this embodiment, the defect types include, but are not limited to, grid breakage, scratching, hidden cracking, black spots, insufficient solder, contamination, cracking, corner collapse, failure, and overwelding.
And step 3: and constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities.
In this embodiment, there are 30304 labeled photovoltaic cell images, wherein the training set is 16224, the verification set is 4048, and the test set is 10032.
And 4, step 4: and constructing a target recognition model, and training, verifying and testing the target recognition model by adopting a sample set to obtain the trained target recognition model.
In this embodiment, the target recognition model adopts a YOLO v5s model, which can be converged quickly on multiple data sets and has strong customizability. The YOLO v5s model is an existing model, and the construction process thereof is not described in detail here.
And 5: and detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
Similar to the sample set acquisition process, the defect detection of the photovoltaic cell assembly by using the target identification model actually comprises the steps of acquiring a photovoltaic cell assembly EL image acquired by a near-infrared camera in real time, and dividing the photovoltaic cell assembly EL image into photovoltaic cell images; and inputting the photovoltaic cell image into a target identification model to obtain the defect label information of the photovoltaic cell image.
The richer the sample set, the higher the recognition accuracy of the target recognition model is, the more the defect label information of each photovoltaic cell is recognized in real time by adopting the target recognition model, the training process is actually realized, when the trained target recognition model is adopted to detect the defects of the photovoltaic cells in real time, manual participation is not needed, the recognition efficiency is high, and the recognition accuracy is high.
Step 6: and performing defect grade classification on each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell.
The defect label information comprises defect types and defect positions in the photovoltaic cell pieces, the defect area and/or defect length can be calculated according to the defect positions, and the defect types, the defect areas and/or the defect length are main factors influencing the output power of the photovoltaic cell pieces, so that the defect grades of the photovoltaic cell pieces are classified according to the defect types, the defect areas and/or the defect lengths, and then the defect grades of the single photovoltaic cell assemblies are determined according to the defect grades of the photovoltaic cell pieces so as to control the quality.
The defect label information comprises defect types, defect positions and defect confidence coefficients, the defect positions refer to the positions of the marking frames, the positions of the marking frames comprise the coordinates of the upper left corner and the lower right corner of the marking frames, the width and the height, and the defect length and the defect area can be calculated according to the coordinates of the upper left corner and the lower right corner, the width and the height. The defect grades corresponding to different defect types, defect lengths and/or defect areas and defect numbers on a single photovoltaic cell sheet are divided, the defect grades of the photovoltaic cell sheet comprise a grade a, a grade b, a grade c and a grade d, and specific division rules are shown in table 1.
TABLE 1 Defect rating Scale for Individual photovoltaic cells
In Table 1, the horizontal header indicates the type of defect, the vertical header indicates the defect level, and the x indicates that the defect is not allowedThe defects are represented by L, the side length of the photovoltaic cell piece2To length of scratch, n2The number of scratches on the photovoltaic cell sheet, and n2Is an integer, S is the area of the photovoltaic cell, S1Is the area of a single broken gate,/3Is subfissure length, n3The number of subfissure on the photovoltaic cell sheet, /)4Is the length of the black spot, s4Area of a single black spot, s6Is the area of a single contamination, s7Area of a single crack, /)7Is the crack length, n8Is the number of corner breakouts on the photovoltaic cell sheet, s8Is the area of a single corner break, n9Number of failures on the photovoltaic cell sheet, s9Area of single failure,/10Is an over-welding length, n10The number of the over-welding on the photovoltaic cell slice. In this embodiment, the defect length is the diagonal length of the marking frame.
As shown in fig. 5, the defect label information is converted into a defect level label, fig. 6 is the defect level information of each photovoltaic cell on a single photovoltaic cell assembly, in fig. 6, a single photovoltaic cell, for example, a first photovoltaic cell, a box of the upper left corner a represents a classification result of manual detection, a box of the upper right corner a represents a result of detection by using the object recognition model of the present embodiment and conversion into a defect level label, and a lower box represents a classification result and confidence of the classification network DenseNet 121.
Fig. 7 shows different defects of the cell, in which in the graph (a), the cell side length L is 602 (indicated by the number of pixels), and the scratch defect length L is2The defect grade of the battery piece can be judged to be grade a according to the table 1, namely 61.5; in the graph (b), the cell side length L is 602 (in number of pixels), and the crack defect length L is shown3The defect grade of the battery piece can be judged to be b grade according to the table 1, wherein 40.45 is obtained; in the graph (c), the cell area S is 204744, the side length L is 456 (in pixel number), the contamination area and Σ S6528, black spot area and ∑ s4645, the length of the diagonal line of the two black spot labeling boxes is l4＝29，l429.15, the defect of the battery piece can be judged to belong to the b grade according to the table 1; in the graph (d), the cell side length L is 602 (number of pixels)Expressed), subfissure defect length l363.65, judging the defect of the battery piece to belong to the class c according to the table 1; in the graph (e), the cell side length L is 456 (in pixel number), and the crack defect length L is7195.3, the defect of the battery piece can be judged to belong to d class according to the table 1. And 7: and determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
The defect grade of a single photovoltaic cell piece cannot reflect the defect of the whole photovoltaic cell assembly, and the quality control cannot be performed, so that the defect grade of the whole photovoltaic cell assembly needs to be determined according to the defect grade of each photovoltaic cell piece, the defect grade of the single photovoltaic cell assembly comprises a grade A, a grade B, a grade C and a grade D, and the specific division rule is shown in table 2.
TABLE 2 determination of Defect grade of photovoltaic cell Assembly
In table 2, the horizontal direction header indicates the defect grade of the photovoltaic cell module, the vertical direction header indicates the defect grade of the photovoltaic cell, and the percentage in the table indicates the number of the photovoltaic cell corresponding to the defect grade. Taking 72 photovoltaic cells on a certain photovoltaic cell assembly as an example, 94.4% of the second row and the second column of the second photovoltaic cell assembly represents the number of the photovoltaic cells with the defect level of a grade to the ratio paThe number of the photovoltaic cells with the defect grade of a grade accounts for p when the defect grade is more than or equal to 94.4 percentaThe number of the photovoltaic cell pieces with the defect grade of a grade is divided by the total number of the photovoltaic cell pieces in the photovoltaic cell assembly, namely 68/72. 11.1% of the third row and the third column represents the proportion p of the number of the photovoltaic cells with the defect grade of b gradebLess than 11.1 percent, and the number of the photovoltaic cells with the defect grade of b grade accounts for pbEqual to the number of photovoltaic cells with defect grade b divided by the total number of photovoltaic cells in the photovoltaic cell assembly, 8/72.
Fig. 8 shows a photovoltaic cell module with different defect levels, wherein in the diagram (a), the module comprises 72 cells, the number of a-level cells: 69, number of b-stage battery pieces: 3, the number of c-grade cells: 0, d-grade cell number: 0; judging that the component belongs to the A-level component according to the table 2; in fig. b, the number of battery cells included in the assembly is 72, the number of a-stage battery cells: 68, number of b-grade cells: 3, the number of c-grade cells: 1, d-grade cell number: 0; judging that the component belongs to a B-level component according to the table 2; in fig. (c), the number of battery pieces contained in the assembly is 72, the number of a-stage battery pieces: 62, number of b-grade cells: 10, c-grade cell number: 0, d-grade cell number: 0; judging that the component belongs to the C-level component according to the table 2; in fig. (d), the number of cells included in the assembly is 72, the number of a-stage cells: 69, number of b-stage battery pieces: 1, c-grade cell number: 0, d-grade cell number: 2; the component is judged to belong to the D-level component according to the table 2. When the defect grade of the photovoltaic cell assembly is A grade, judging the photovoltaic cell assembly to be a normal assembly; when the defect grade of the photovoltaic cell assembly is B grade, judging the photovoltaic cell assembly to be a defective product; and when the defect grade of the photovoltaic cell assembly is grade C, judging that the photovoltaic cell assembly is a management and control product, and when the defect grade of the photovoltaic cell assembly is grade D, judging that the photovoltaic cell assembly is a waste product.
Next, the effect of the target recognition model on defect detection of the photovoltaic cell is evaluated through accuracy and reliability, for the target recognition model, AP (IoU ═ 0.5) is used as a parameter for evaluating the target recognition model, and table 3 lists APs with various defects in the order of the training set, the verification set, and the test set. The defect label information is converted into defect grade labels, a confusion matrix of each model is calculated, and the accuracy, the F1, the omission factor and the false positive are used as evaluation indexes, as shown in the table 4.
TABLE 3 AP and mAP of various types of defects
In table 3, the horizontal header represents different defect types of the photovoltaic cell, the vertical header represents different target recognition models, and the data in the table sequentially represents the defect AP corresponding to the training set, the defect AP corresponding to the verification set, and the defect AP corresponding to the test set, for example, 96|88|100 in the second row and the second column sequentially represents that the AP of the defect identified by inputting the training set into the YOLO v5s model is 96%, the AP of the defect identified by inputting the verification set into the YOLO v5s model is 88%, and the AP of the defect identified by inputting the test set into the YOLO v5s model is 100%. The AP is an average accuracy and is an evaluation index of the accuracy of the target recognition model. The last column of mAP is the average of the AP values of all defect types, for example 84.55|77.26|62.07 sequentially indicates that the average of ten defect AP values identified by inputting the training set into the YOLO v5s model is 84.55%, the average of ten defect AP values identified by inputting the verification set into the YOLO v5s model is 77.26%, and the average of ten defect AP values identified by inputting the test set into the YOLO v5s model is 62.07%. As shown in Table 3, the YOLO v5 model has high accuracy in determining various defects.
TABLE 4 accuracy of object recognition model and time of inference
| Metrics(％) | YOLO v5s | DenseNet121 | Human visual inspection |
| Accuracy | 90.45 | 87.27 | - |
| F1 score | 94.71 | 89.09 | - |
| Miss rate | 2.55 | 10.91 | - |
| False positive | 7.88 | 2.97 | - |
| Time spent in single sheet detection | 9.06ms | 15.37ms | 18s |
In table 4, Accuracy represents Accuracy, F1 score represents harmonic mean of P-R (Precision-reduce), misrate represents False positive rate, False positive represents FP False positive rate, that is, probability that a counter example is erroneously determined as a positive example, and single detection time is time required for target recognition of a picture by using a model and reflects real-time performance.
According to the data analysis in table 4, the target recognition model YOLO v5s performs best, the accuracy reaches 90.45%, and the detection speed of each photovoltaic cell reaches 9.06ms, so the YOLO v5s is selected to detect defects of the photovoltaic cell assembly on the production line. As shown in fig. 2, the undetected photovoltaic cell module shows, as shown in fig. 6, the results of manual inspection and defect detection and classification using the object recognition model and the classification network, each photovoltaic cell is classified by the three methods, the upper left box of each photovoltaic cell represents the classification result of the manual inspection, the upper right box represents the classification result after defect detection and defect class conversion using the object recognition model YOLO v5s, and the lower box represents the classification result and confidence using the classification network DenseNet 121.
The rules in tables 1 and 2 can be updated according to the actual quality control requirements of the production line, so that the produced components meet the actual quality requirements.
The detection method is suitable for quality control of the photovoltaic cell assembly on a production line, is improved on the basis of electroluminescence detection, can ensure that the photovoltaic cell assembly is subjected to non-contact real-time high-precision defect detection under the condition of uninterrupted production, and therefore intelligent quality control of the photovoltaic cell assembly on the production line is achieved.
The embodiment also provides a photovoltaic cell module defect detection system, which comprises an image acquisition and segmentation unit, an annotation unit, a sample construction unit, a model construction and training unit, a defect detection unit, a first grade confirmation unit and a second grade confirmation unit.
And the image acquisition and division unit is used for acquiring the EL image of the photovoltaic cell assembly and dividing the EL image of the photovoltaic cell assembly into photovoltaic cell images.
In this embodiment, the number of the near-infrared cameras is 3, the 3 near-infrared cameras are stacked, lenses of the near-infrared cameras are aligned with the photovoltaic cell assemblies, and EL imaging is performed in a scanning shooting manner, as shown in fig. 2, for the photovoltaic cell assembly EL image, pixels of each near-infrared camera are 1360 × 1024, 13756 EL images are obtained in total, and the resolution of each image is 5400 × 2700.
In order to improve the detection accuracy of internal defects and micro-defects, the EL image is segmented into photovoltaic cell images, and as shown in fig. 3, the image acquisition and segmentation unit is further configured to:
cutting off information at the edge of the EL image, and eliminating the influence of the edge information on the feature point extraction; performing SURF feature extraction on the cut EL image to obtain feature point information; selecting a point which accords with the coordinate of the expected point from the characteristic point information as a template center to generate a template; identifying non-luminous black diamond-like regions in the EL image using a template, as shown in fig. 4; calculating the central point of each black diamond-like area, wherein each central point is a vertex of a certain photovoltaic cell slice; and connecting four adjacent central points to form a quadrangle, wherein the circumscribed rectangle of the quadrangle is a photovoltaic cell image.
And the marking unit is used for screening out the images with the defects in the photovoltaic cell images and marking the defect types and the defect positions on the images with the defects by using a labeling tool, wherein the defect positions refer to the positions of the marking frames.
In order to obtain a photovoltaic cell image with known defect types and defect positions, preparation is made for constructing a sample set, an image with defects in the photovoltaic cell image is screened out manually, and the defect types and the defect positions are marked on the image with the defects by a labeling tool. In this embodiment, the labeling tool is label, the label is a visual image calibration tool, the defect frame is directly framed on the photovoltaic cell image by the labeling frame, and an xml file of the defect position corresponding to the defect type can be generated, where the defect position is the position of the labeling frame. In this embodiment, the position of the labeling frame includes the coordinates of the upper left corner and the lower right corner of the labeling frame, the width and the height, the length or the area of the defect framed by the labeling frame can be calculated according to the coordinates of the upper left corner and the lower right corner, the width and the height, the length of the defect is the length of the diagonal line of the labeling frame, and the area of the defect is the area of the labeling frame.
In this embodiment, the defect types include, but are not limited to, grid breakage, scratching, hidden cracking, black spots, insufficient solder, contamination, cracking, corner collapse, failure, and overwelding.
And the sample construction unit is used for constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities.
In this embodiment, there are 30304 labeled photovoltaic cell images, wherein the training set is 16224, the verification set is 4048, and the test set is 10032.
And the model building and training unit is used for building a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set to obtain the trained target recognition model.
In this embodiment, the target recognition model adopts a YOLO v5s model, which can be converged quickly on multiple data sets and has strong customizability. The YOLO v5s model is an existing model, and the construction process thereof is not described in detail here.
And the defect detection unit is used for detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
Similar to the sample set acquisition process, the defect detection of the photovoltaic cell assembly by using the target identification model actually comprises the steps of acquiring a photovoltaic cell assembly EL image acquired by a near-infrared camera in real time, and dividing the photovoltaic cell assembly EL image into photovoltaic cell images; and inputting the photovoltaic cell image into a target identification model to obtain the defect label information of the photovoltaic cell image.
And the first grade confirming unit is used for classifying the defect grade of each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell.
The defect label information comprises defect types and defect positions in the photovoltaic cell pieces, the defect area and/or defect length can be calculated according to the defect positions, and the defect types, the defect areas and/or the defect length are main factors influencing the output power of the photovoltaic cell pieces, so that the defect grades of the photovoltaic cell pieces are classified according to the defect types, the defect areas and/or the defect lengths, and then the defect grades of the single photovoltaic cell assemblies are determined according to the defect grades of the photovoltaic cell pieces so as to control the quality.
The defect label information comprises defect types, defect positions and defect confidence coefficients, the defect positions refer to the positions of the marking frames, the positions of the marking frames comprise the coordinates of the upper left corner and the lower right corner of the marking frames, the width and the height, and the defect length and the defect area can be calculated according to the coordinates of the upper left corner and the lower right corner, the width and the height. The defect grades corresponding to different defect types, defect lengths and/or defect areas and defect numbers on a single photovoltaic cell sheet are divided, the defect grades of the photovoltaic cell sheet comprise a grade a, a grade b, a grade c and a grade d, and specific division rules are shown in table 1.
And the second grade confirmation unit is used for determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
The defect grade of a single photovoltaic cell piece cannot reflect the defect of the whole photovoltaic cell assembly, and the quality control cannot be performed, so that the defect grade of the whole photovoltaic cell assembly needs to be determined according to the defect grade of each photovoltaic cell piece, the defect grade of the single photovoltaic cell assembly comprises a grade A, a grade B, a grade C and a grade D, and the specific division rule is shown in table 2.
When the defect grade of the photovoltaic cell assembly is A grade, judging the photovoltaic cell assembly to be a normal assembly; when the defect grade of the photovoltaic cell assembly is B grade, judging the photovoltaic cell assembly to be a defective product; and when the defect grade of the photovoltaic cell assembly is grade C, judging that the photovoltaic cell assembly is a management and control product, and when the defect grade of the photovoltaic cell assembly is grade D, judging that the photovoltaic cell assembly is a waste product.
The above disclosure is only for the specific embodiments of the present invention, but the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive of changes or modifications within the technical scope of the present invention, and shall be covered by the scope of the present invention.
Claims (10)
1. A defect detection method for a photovoltaic cell assembly is characterized by comprising the following steps:
step 1: acquiring an EL image of the photovoltaic cell assembly, and dividing the EL image of the photovoltaic cell assembly into photovoltaic cell images;
step 2: screening out images with defects in the photovoltaic cell images, and marking the defect types and defect positions on the images with the defects by using a labeling tool, wherein the defect positions are the positions of a marking frame;
and step 3: constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities;
and 4, step 4: constructing a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set to obtain a trained target recognition model;
and 5: and detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
2. The method for detecting the defects of the photovoltaic cell assembly according to claim 1, wherein in the step 1, the EL image of the photovoltaic cell assembly is segmented into the images of the photovoltaic cells by adopting a template matching method, and the specific implementation process is as follows:
step 1.1: cutting off information at the edge of the EL image;
step 1.2: performing SURF feature extraction on the cut EL image to obtain feature point information;
step 1.3: selecting a point which accords with the coordinate of the expected point in the characteristic point information as a template center to generate a template;
step 1.4: identifying a non-luminous black diamond-like area in the EL image by adopting the template in the step 1.3;
step 1.5: calculating a central point of the black diamond-like region, wherein the central point is a vertex of the photovoltaic cell;
step 1.6: and connecting four adjacent central points to form a quadrangle, wherein the external rectangle of the quadrangle is a photovoltaic cell image.
3. The method for detecting defects in a photovoltaic cell assembly according to claim 1, wherein the target identification model is a YOLO v5s model.
4. The method for detecting the defects of the photovoltaic cell assembly according to claim 1, wherein in the step 2, the defect types comprise grid breakage, scratch, subfissure, black spot, cold solder joint, pollution, crack, corner breakage, failure and over solder joint.
5. The method for detecting defects of a photovoltaic cell assembly according to any one of claims 1 to 4, further comprising:
step 6: dividing the defect grade of each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell;
and 7: and determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
6. The method for detecting the defects of the photovoltaic cell assembly according to claim 5, wherein in the step 6, the defect grades of the photovoltaic cell pieces comprise a grade a, a grade b, a grade c and a grade d;
when only scratch exists on a single photovoltaic cell sheet, l is more than or equal to 02≤L/5，0≤n2Less than or equal to 2; or no defect type is present; the defect grade of the photovoltaic cell is a grade; wherein L is the side length of the photovoltaic cell piece, L2To length of scratch, n2The number of scratches on the photovoltaic cell sheet, and n2Is an integer;
when there is a broken gate on a single photovoltaic cell, and 0 < ∑ s1Less than or equal to 5 percent of S; or has scratches, and L/5 < L2≤2L/3，2＜n2Less than or equal to 4; or with subfissure and 0 < l3≤L/12，0＜n3Less than or equal to 1; or black spots are present, and 0 < l4≤L/4，0＜∑s4Less than or equal to 5 percent of S; or in the presence of contamination, and 0 < ∑ s6Less than or equal to 5 percent of S; the defect grade of the photovoltaic cell is b grade; wherein S is a photovoltaic cellArea of sheet, s1Is the area of a single broken gate,/3Is subfissure length, n3The number of subfissure on the photovoltaic cell sheet, /)4Is the length of the black spot, s4Area of a single black spot, s6As an area of individual contamination;
when a broken grid exists on a single photovoltaic cell, and 5% S < Sigmas1Less than or equal to 10 percent of S; or has scratches, and L/5 < L2Less than or equal to 2L/3; or having subfissure and L/12 < L3≤L/5，1＜n3Less than or equal to 2; or black spots are present and 5% S < Sigmas4Less than or equal to 10 percent of S; or a cold solder joint is present and 0 < l5Less than or equal to L/4; or in the presence of contamination, and 0 < ∑ s6Less than or equal to 10 percent of S; or cracks are present and 0 < l7≤L/5，0＜n7Less than or equal to 1; or a breakout angle of 0 < s8≤4％S，0＜n8Less than or equal to 1; or a failure exists and 0 < s9≤10％S，0＜n9Less than or equal to 1; the defect grade of the photovoltaic cell is grade c; wherein s is7Area of a single crack, /)7Is the crack length, n8Is the number of corner breakouts on the photovoltaic cell sheet, s8Is the area of a single corner break, n9Number of failures on the photovoltaic cell sheet, s9Area of single failure,/10Is the overweld length;
when a broken grid exists on a single photovoltaic cell, and 10% S < Sigmas1(ii) a Or has scratch, and 2L/3 < L2(ii) a Or having subfissure and L/5 < L3，2＜n3(ii) a Or black spots are present and 10% S < Sigmas4(ii) a Or cold solder joints are present and L/4 < L5(ii) a Or contamination, and 10% S < Sigmas6(ii) a Or cracks are present and L/5 < L7，1＜n7(ii) a Or a breakout angle is present and 4% S < S8，1＜n8(ii) a Or failure, and 10% S < S9，1＜n9(ii) a Or over-welding exists and n is more than or equal to 110If so, the defect grade of the photovoltaic cell is d grade; wherein n is10The number of the over-welding on the photovoltaic cell slice.
7. The method for detecting defects of a photovoltaic cell assembly according to claim 5, wherein in the step 7, the defect levels of the photovoltaic cell assembly include a level A, a level B, a level C and a level D;
when the number of photovoltaic cell pieces with the defect grade of 94.4 percent to a grade a is in proportion, the number of photovoltaic cell pieces with the defect grade of 0 to b is in proportion to 5.6 percent, and the number of photovoltaic cell pieces with the defect grade of c and d is in proportion to 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is in a grade A;
when the number of the photovoltaic cell pieces with the defect grade of 88.9 percent to a grade a is less than 94.4 percent, the number of the photovoltaic cell pieces with the defect grade of 5.6 percent to a grade B is less than 11.1 percent, the number of the photovoltaic cell pieces with the defect grade of 0 to a grade c is less than 5.6 percent, and the number of the photovoltaic cell pieces with the defect grade of d is 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is B grade;
when the number of the photovoltaic cell pieces with the defect grade of 83.3 percent to more than or equal to a grade a is less than 88.9 percent, the number of the photovoltaic cell pieces with the defect grade of 11.1 percent to more than or equal to a grade b is less than 16.7 percent, the number of the photovoltaic cell pieces with the defect grade of 5.6 percent to more than a grade C is less than 16.7 percent, and the number of the photovoltaic cell pieces with the defect grade of d is 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is grade C;
when the number of the photovoltaic cell pieces with the defect grade of a grade is less than 83.3 percent, the number of the photovoltaic cell pieces with the defect grade of 16.7 percent is less than the number of the photovoltaic cell pieces with the defect grade of b grade, the number of the photovoltaic cell pieces with the defect grade of 16.7 percent is less than the number of the photovoltaic cell pieces with the defect grade of c grade and the number of the photovoltaic cell pieces with the defect grade of D grade are more than 0 percent in a single photovoltaic cell assembly, the defect grade of the photovoltaic cell assembly is D grade.
8. A photovoltaic cell assembly defect detection system, comprising:
the image acquisition and segmentation unit is used for acquiring an EL image of the photovoltaic cell assembly and segmenting the EL image of the photovoltaic cell assembly into photovoltaic cell images;
the marking unit is used for screening out images with defects in the photovoltaic cell images and marking the defect types and the defect positions on the images with the defects by using a labeling tool, wherein the defect positions refer to the positions of the marking frames;
the sample construction unit is used for constructing a sample set by taking the marked image as an input quantity and taking the defect type and the defect position corresponding to the image as output quantities;
the model building and training unit is used for building a target recognition model, and training, verifying and testing the target recognition model by adopting the sample set to obtain a trained target recognition model;
and the defect detection unit is used for detecting the defects of the photovoltaic cell assembly by using the trained target identification model to obtain the defect label information of each photovoltaic cell, wherein the defect label information comprises defect types, defect positions and defect confidence coefficients.
9. The photovoltaic cell assembly defect detection system of claim 8, further comprising:
the first grade confirming unit is used for carrying out defect grade division on each photovoltaic cell according to the defect label information of the photovoltaic cell to obtain the defect grade of the photovoltaic cell;
and the second grade confirmation unit is used for determining the defect grade of the whole photovoltaic cell assembly according to the defect grade of each photovoltaic cell.
10. A photovoltaic cell module production line comprising the photovoltaic cell module defect detection system according to claim 8 or 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111043149.4A CN113962929A (en) | 2021-09-07 | 2021-09-07 | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111043149.4A CN113962929A (en) | 2021-09-07 | 2021-09-07 | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113962929A true CN113962929A (en) | 2022-01-21 |
Family
ID=79460992
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111043149.4A Pending CN113962929A (en) | 2021-09-07 | 2021-09-07 | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113962929A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627122A (en) * | 2022-05-16 | 2022-06-14 | 北京东方国信科技股份有限公司 | Defect detection method and device |
| CN115797270A (en) * | 2022-11-15 | 2023-03-14 | 正泰集团研发中心(上海)有限公司 | Training method, detection method and equipment of light leakage detection model and storage medium |
-
2021
- 2021-09-07 CN CN202111043149.4A patent/CN113962929A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114627122A (en) * | 2022-05-16 | 2022-06-14 | 北京东方国信科技股份有限公司 | Defect detection method and device |
| CN115797270A (en) * | 2022-11-15 | 2023-03-14 | 正泰集团研发中心(上海)有限公司 | Training method, detection method and equipment of light leakage detection model and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111223088B (en) | Casting surface defect identification method based on deep convolutional neural network | |
| CN109993094A (en) | Fault in material intelligent checking system and method based on machine vision | |
| CN111223093A (en) | AOI defect detection method | |
| CN108711148B (en) | Tire defect intelligent detection method based on deep learning | |
| CN113962929A (en) | Photovoltaic cell assembly defect detection method and system and photovoltaic cell assembly production line | |
| CN115601355A (en) | Method and device for detecting and classifying product surface defects and storage medium | |
| CN113393426A (en) | Method for detecting surface defects of rolled steel plate | |
| CN117132844B (en) | Classifying method for cracks and scratches of photovoltaic panel based on image processing | |
| CN116091505B (en) | Automatic defect detection and classification method and system for sapphire substrate | |
| CN114119591A (en) | Display screen picture quality detection method | |
| CN113313107A (en) | Intelligent detection and identification method for multiple types of diseases on cable surface of cable-stayed bridge | |
| CN115995056A (en) | Automatic bridge disease identification method based on deep learning | |
| CN114359235A (en) | Wood surface defect detection method based on improved YOLOv5l network | |
| CN115866502A (en) | Microphone part surface defect online detection process | |
| CN114881987A (en) | Improved YOLOv 5-based hot-pressing light guide plate defect visual detection method | |
| CN114663393A (en) | Convolutional neural network-based detection method for defect of poor spacing between photovoltaic modules | |
| CN117455917B (en) | Establishment of false alarm library of etched lead frame and false alarm on-line judging and screening method | |
| CN114529536A (en) | Solid wood quality detection method | |
| CN114387261A (en) | Automatic detection method suitable for railway steel bridge bolt diseases | |
| CN111598851B (en) | Solar cell broken piece detection method based on morphological image processing | |
| CN116052110B (en) | Intelligent positioning method and system for pavement marking defects | |
| CN115330768B (en) | Quality grading method for solar cell panel | |
| CN115953387A (en) | Radiographic image weld defect detection method based on deep learning | |
| CN110889418A (en) | Gas contour identification method | |
| CN114943738A (en) | Sensor packaging curing adhesive defect identification method based on visual identification |
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 |