CN112529875A - Photovoltaic module glass burst early warning method and system based on artificial intelligence - Google Patents

Photovoltaic module glass burst early warning method and system based on artificial intelligence Download PDF

Info

Publication number
CN112529875A
CN112529875A CN202011466941.6A CN202011466941A CN112529875A CN 112529875 A CN112529875 A CN 112529875A CN 202011466941 A CN202011466941 A CN 202011466941A CN 112529875 A CN112529875 A CN 112529875A
Authority
CN
China
Prior art keywords
frame
bubble
image
area
grid line
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Withdrawn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
CN202011466941.6A
Other languages
Chinese (zh)
Inventor
孙占民
杨晓敏
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 CN202011466941.6A priority Critical patent/CN112529875A/en
Publication of CN112529875A publication Critical patent/CN112529875A/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Withdrawn legal-status Critical Current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
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/12—Edge-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/60—Analysis of geometric attributes
    • G06T7/62—Analysis of geometric attributes of area, perimeter, diameter or volume
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/90—Determination of colour characteristics
    • 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]

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • General Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Quality & Reliability (AREA)
  • Geometry (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to the technical field of artificial intelligence, in particular to a photovoltaic module glass burst early warning method and system based on artificial intelligence. The method comprises the following steps: collecting RGB images on the surface of a photovoltaic cell panel; obtaining a frame grid line image of the photovoltaic cell panel according to the RGB image; dividing an interested area image by the frame and the outermost peripheral grid line of the photovoltaic cell panel; sending the image of the region of interest into a twin network, and outputting a similarity vector with a normal battery panel region; segmenting bubbles in the region of interest to obtain bubble outlines; obtaining a plurality of frame areas according to the number of each frame to the interesting area, and dividing the bubble influence area; calculating Euclidean distances between the outermost grid line pixel point and the frame pixel point in the bubble influence area and obtaining a Euclidean distance mean square error; and obtaining a burst estimation value according to the Euclidean distance mean square error and the bubble area, and feeding back early warning information when the burst estimation value reaches an early warning value. The invention realizes the early warning of the glass burst of the photovoltaic assembly.

Description

Photovoltaic module glass burst early warning method and system based on artificial intelligence
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a photovoltaic module glass burst early warning method and system based on artificial intelligence.
Background
The photovoltaic power generation technical field has wide application range and covers various fields such as civil use, commercial use and the like. Becomes an indispensable means for utilizing clean energy. As an important technology in the photovoltaic industry, maintenance and inspection of photovoltaic panels is also concerned. The operation condition and the generated power of the photovoltaic cell panel directly influence the operation condition of a photovoltaic power plant, so that the detection and early warning of the photovoltaic cell panel are key points in maintenance, and the most effective way for avoiding the occurrence of problems of the photovoltaic module is provided.
The glass of the photovoltaic module bursts can have many reasons, for example, the frame deformation extrusion of the photovoltaic cell panel causes the burst of glass, the bubbles appear in the photovoltaic cell panel and cause the burst of glass due to nonuniform heating around the bubbles in the photovoltaic cell panel, general frame deformation can cause the photovoltaic module to be sealed poorly, the bubbles appear at the frame joint, the glass cracks can appear after a long time, and the cleaning and the power generation performance of the photovoltaic cell panel are influenced. At present, most of the defects of the photovoltaic modules in the prior art are detected, for example, the EL tester detects various hidden cracking problems, and prediction and early warning can not be carried out on the photovoltaic modules which do not have glass cracking but have cracking hidden danger.
Disclosure of Invention
In order to solve the technical problems, the invention aims to provide a photovoltaic module glass burst early warning method and system based on artificial intelligence, and the adopted technical scheme is as follows:
the invention provides an artificial intelligence-based photovoltaic module glass burst early warning method, which comprises the following steps:
collecting RGB images of a photovoltaic cell panel;
extracting an R channel image of the RGB image to obtain a first image; performing binarization processing on the first image according to a preset first pixel threshold value to obtain a frame grid line image of the photovoltaic cell panel;
dividing an interested area image by a frame and the outermost peripheral grid line of the photovoltaic cell panel according to the frame grid line image;
sending the images of the interested areas into a pre-trained twin network for processing, and outputting similarity vectors of the images of the interested areas and the areas of normal battery plates; determining an abnormal cell panel with a deformed frame according to the similarity vector;
segmenting the bubble image in the region of interest image according to a preset second pixel threshold value; performing edge extraction on the bubble image to obtain a bubble profile;
numbering the interested regions on the abnormal cell panel according to each frame to obtain a plurality of frame regions; dividing a bubble influence area by all the bubble outline coordinates and the outermost periphery grid lines in each border area;
calculating Euclidean distance between each grid line and a pixel point corresponding to the frame in the bubble influence area, and taking the mean square error of the Euclidean distance as an influence factor of frame deformation on glass burst;
summing pixel points in each bubble outline in the frame region to obtain a bubble area, and obtaining a burst estimation value of each frame grid line region according to the bubble area and the Euclidean distance mean square error; summing the burst estimated values to obtain an overall burst estimated value of the overall cell panel; and normalizing the overall burst estimation value, and feeding back early warning information when the burst estimation value reaches a set early warning value.
Further, dividing an interested area image by a frame and the outermost periphery grid line of the photovoltaic cell panel according to the frame grid line image:
taking the area between the frame and the outermost peripheral grid line as an interested area, marking the pixel in the interested area as 1, and marking other areas as 0 to obtain a mask image;
and multiplying the mask image and the first image to obtain the region-of-interest image.
Further, the following operations are also included after the abnormal battery panel with deformed frame is determined according to the similarity vector:
and if the abnormal solar panel is judged by outputting the similarity vector through the twin network but no bubble appears in the abnormal solar panel, carrying out special marking on the abnormal solar panel.
Further, the method for dividing the bubble influence area by the bubble outline coordinate in the frame area comprises the following steps:
taking the difference value between the maximum abscissa and the minimum abscissa of the bubble outline in the upper and lower frame areas as the length of the bubble influence area; the width of the upper frame area and the lower frame area is the width of the bubble influence area; and/or
Taking the difference value between the maximum ordinate and the minimum ordinate of the bubble profile in the left and right border areas as the width of the bubble influence area; the length of the left and right frame regions is the length of the bubble affected region.
Further, the method for calculating the euclidean distance between the pixel points corresponding to each grid line and the frame in the bubble influence area includes:
carrying out Hough line fitting after the binarization processing of the bubble affected area to obtain the slope of a grid line in the bubble affected area;
making grid line pixel points on the grid lines as vertical lines of the grid lines; the intersection point of the vertical line and the frame is a corresponding frame pixel point; and calculating the Euclidean distance between the grid line pixel point and the frame pixel point.
Further, the burst estimation value of each frame grid line region is obtained according to the bubble area and the euclidean distance mean square error:
calculating a burst estimate for each of said border regions by the formula:
Qj＝W(MSEj+Sj)
wherein W is the adjustment weight, j is the number of the frame area, MSEjThe Euclidean distance mean square error, S, of the bounding box region numbered jjThe bubble area of the border region numbered j.
The invention also provides a photovoltaic module glass burst early warning system based on artificial intelligence, which is characterized by comprising an image acquisition module, a frame grid line image acquisition module, an interested region image division module, a bubble outline acquisition module, a twin network module, a bubble affected region acquisition module, a frame deformation affected factor acquisition module and a burst estimation early warning module;
the image acquisition module is used for acquiring RGB images of the photovoltaic cell panel;
the frame grid line image acquisition module is used for extracting an R channel image of the RGB image to obtain a first image; performing binarization processing on the first image according to a preset first gray threshold value to obtain a frame grid line image of the photovoltaic cell panel;
the interested region image dividing module is used for dividing interested region images by using a frame and the outermost peripheral grid line of the photovoltaic cell panel according to the frame grid line images;
the twin network module is used for sending the image of the region of interest into a pre-trained twin network for processing and outputting a similarity vector with a normal battery panel region; determining an abnormal cell panel with a deformed frame according to the similarity vector;
the bubble outline acquisition module is used for segmenting a bubble image in the region of interest according to a preset second gray threshold; performing edge extraction on the bubble image to obtain a bubble outline image;
the bubble affected area acquisition module is used for numbering the interested areas on the abnormal battery panel according to each frame to acquire a plurality of frame areas; dividing a bubble influence area by all the bubble outline coordinates and the outermost periphery grid lines in each border area;
the frame deformation influence factor acquisition module is used for calculating Euclidean distances between each grid line and a pixel point corresponding to the frame in the bubble influence area, and taking the mean square error of the Euclidean distances as an influence factor of frame deformation on glass burst;
the burst estimation early warning module is used for summing pixel points in each bubble outline in the frame region to obtain a bubble area, and obtaining a burst estimation value of each frame grid line region according to the bubble area and the Euclidean distance mean square error; summing the burst estimated values to obtain an overall burst estimated value of the overall cell panel; and normalizing the overall burst estimation value, and feeding back early warning information when the burst estimation value reaches a set early warning value.
Further, the region-of-interest image dividing module takes a region between the frame and the outermost peripheral grid line as a region of interest, marks pixels in the region of interest as 1, and marks other regions as 0 to obtain a mask image; and multiplying the mask image and the first image to obtain the region-of-interest image.
Further, the twin network module outputs a similarity vector to judge an abnormal battery panel, and if no bubble occurs in the abnormal battery panel, the abnormal battery panel is specially marked.
Further, the calculating, by the frame deformation influence factor obtaining module, the euclidean distance between each grid line and the pixel point corresponding to the frame in the bubble influence area includes:
carrying out Hough line fitting after the binarization processing of the bubble affected area to obtain the slope of a grid line in the bubble affected area;
making a straight line by using the grid line pixel points on the grid line and taking the negative of the reciprocal of the slope of the grid line as the slope; the intersection point of the straight line and the frame is a corresponding frame pixel point; and calculating the Euclidean distance between the grid line pixel point and the frame pixel point.
The invention has the following beneficial effects:
1. according to the embodiment of the invention, the bursting estimation is carried out on the photovoltaic assembly glass plate through the frame deformation degree and the bubbles caused by the deformation, the bursting estimation threshold value is changed according to different conditions, and when the early warning value is reached, the early warning function of bursting is realized by timely feeding back to the working personnel.
2. According to the embodiment of the invention, the collected images of the photovoltaic cell panel are sent to the twin network for processing, and the similarity vector is output. The twin network does not need to manually extract features, and can be adaptively matched with the frame deformation difference of a normal image. And the sequence of abnormal panels can be determined by the output similarity vectors.
3. According to the embodiment of the invention, the battery plates are numbered through the frame, so that the burst point can be positioned through the numbering during detection.
4. According to the embodiment of the invention, the bubble influence area is divided by utilizing bubbles in the frame, the grid line slope is obtained by utilizing corresponding grid line pixel points in the bubble influence area through a Hough line fitting algorithm, the mean square error of the distance between the pixel points of the vertical line of the grid line and the frame of the normal grid line is calculated to evaluate the deformation degree of the frame, and the deformation evaluation of the cell panel can be more accurately obtained.
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 flow chart of a photovoltaic module glass burst warning method based on artificial intelligence according to an embodiment of the present invention;
fig. 2 is a block diagram of a photovoltaic module glass burst warning system 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 for achieving the predetermined objects, the following detailed description of the method and system for warning the glass burst of the photovoltaic module based on artificial intelligence according to the present invention with reference to the accompanying drawings and preferred embodiments will be made below. 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. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
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 specifically describes a specific scheme of the photovoltaic module glass burst early warning method and system based on artificial intelligence in combination with the accompanying drawings.
Referring to fig. 1, a flow chart of a photovoltaic module glass burst warning method based on artificial intelligence according to an embodiment of the present invention is shown. The method comprises the following steps:
step S1: and collecting the RGB image of the photovoltaic cell panel.
Aiming at the specific scene of the photovoltaic power station, the RGB camera is installed on the unmanned aerial vehicle to overlook and shoot above the photovoltaic cell panel, the height is kept consistent in the shooting process, and clear and complete RGB images are obtained.
Step S2: extracting an R channel image of the RGB image to obtain a first image; and carrying out binarization processing on the first image according to a preset first gray threshold value to obtain a frame grid line image of the photovoltaic cell panel.
The image in the R channel in the RGB image is extracted to obtain a first image, and in order to enable the subsequent image detection to be more effective, the first image needs to be subjected to preprocessing operations such as denoising.
Because the difference between the glass background color of the photovoltaic cell panel and the background color of the grid lines of the frame is large, a first pixel threshold value is set, and the frame and the grid line images are obtained through binarization processing.
Preferably, the first pixel threshold is set to 40 in the embodiment of the present invention.
Step S3: and dividing the interested area image by the frame and the outermost peripheral grid line of the photovoltaic cell panel according to the frame grid line image.
And taking the area between the frame and the outermost peripheral grid line as an interested area, marking the pixel in the interested area as 1 and marking other areas as 0, and obtaining a mask image.
And multiplying the mask image and the first image to obtain an image of the region of interest.
Step S4: and (4) sending the image of the region of interest into a pre-trained twin network for processing, and outputting a similarity vector with the normal panel region.
And (4) processing the normal panel image through the same region of interest division, taking the processed normal panel image as a positive sample of the twin network, and taking the region of interest image acquired in the step S3 as a negative sample of the twin network. And (4) outputting a similarity vector through the twin network to evaluate the deformation of the frame.
The twin network adopts a coding-full connection layer structure, and specifically comprises the following steps:
1) and sending the positive type sample and the negative type sample into the twin network for training.
2) And performing downsampling operations such as convolution pooling on the image through a deformation encoder, performing batch normalization on image data, accelerating the convergence speed of the model, and outputting through a full connection layer to obtain the characteristic vector of the frame.
Calculating Euclidean distance d of different image feature vectors through feature vectors of framesiAnd evaluating the frame difference of the collected photovoltaic cell panel image and the normal photovoltaic cell panel image by using the Euclidean distance as the standard of the similarity vector, and determining the abnormal cell panel.
Preferably, if the abnormal solar panel is judged by the similarity vector output by the twin network, but no bubble appears in the abnormal solar panel, the abnormal solar panel is specially marked, so that the detection is preferentially carried out in the later detection process.
Step S5: segmenting the bubble image in the region of interest image according to a preset second gray threshold; and performing edge extraction on the bubble image to obtain a bubble profile.
Because the gray level of the bubble area of the frame glass plate is obviously different from the grid line of the frame, the bubbles are divided according to the preset second pixel threshold value. And obtaining a bubble image through threshold segmentation on the image of the region of interest, and obtaining a bubble profile through an edge detection technology.
Step S6: numbering the interested areas according to each frame to obtain a plurality of frame areas; and demarcate the bubble affected zone.
And numbering the interested areas on the abnormal cell panel according to 4 frames. In the embodiment of the present invention, the upper frame area is set to be numbered 1, the right frame area is set to be numbered 2, the lower frame area is set to be numbered 3, and the left frame area is set to be numbered 4. These four frames are illustrated in the present embodiment.
Dividing a bubble influence area in the region of interest according to bubble contour coordinates, and specifically comprising the following steps: the difference value between the maximum abscissa and the minimum abscissa of the bubble profile in the upper and lower frame areas is taken as the length of the bubble influence area, and the width of the upper and lower frame areas is the width of the bubble influence area; taking the difference value between the maximum ordinate and the minimum ordinate of the bubble profile in the left and right border areas as the width of the bubble influence area; the length of the left and right frame regions is the length of the bubble affected zone. Thereby determining the bubble affected zone.
Step S7: and calculating the Euclidean distance between the pixel points corresponding to each grid line and the frame in the bubble influence area to obtain the mean square error of the Euclidean distance.
In order to eliminate the influence of left and right offset of the imaging of the unmanned aerial vehicle in the shooting process, the method and the device for analyzing the deformation degree of the frame part are used for analyzing the frame and the corresponding outermost grid lines, and the default grid lines cannot be offset or deformed in the analyzing process.
Carrying out binarization processing in the bubble influence area and then carrying out Hough line fitting to obtain the slope k of the grid line of the bubble influence area1. Taking grid line pixel points on the grid line as vertical lines of the grid line, wherein the slope of the vertical lines is k2I.e. k2*k1Is-1. And the intersection point of the vertical line and the corresponding frame is used as a frame pixel point. Pixel point (x) of grid linei,yi) And frame pixel point (x'i,y′i) Correspondingly connected, and calculating the Euclidean distance d between two pixel pointsiAnd mean square error of Euclidean distance MSEj：
Figure BDA0002834633920000071
Figure BDA0002834633920000072
Wherein d isiIs the Euclidean distance, xiIs the abscissa, y, of the grid line pixel pointiIs a grid line pixel point ordinate, x'iIs the abscissa, y 'of the frame pixel point'iIs the vertical coordinate of the frame pixel point.
And evaluating the deformation degree of the numbered frame region by the obtained mean square error.
Step S8: and summing pixel points in the bubble outline to obtain the bubble area, and obtaining a burst estimation value according to the bubble area and the Euclidean distance mean square error.
Poor sealing performance between the photovoltaic cell panels caused by deformation of the frame is one of the reasons for the occurrence of bubbles in the frame. Therefore, the size of the area of the bubble appearing at the frame is counted and further used as one of the indexes for evaluating the burst of the glass plate of the photovoltaic module. Summing pixel points in the obtained bubble outline to obtain the area of the bubble:
Figure BDA0002834633920000073
wherein, Ii(x, y) represents a pixel point in the bubble outline, SjIndicated as the area of the bubble corresponding to the four numbered boxes.
And obtaining burst estimated values of 4 frame areas according to the area of the bubbles and the Euclidean distance mean square error obtained in the step S7:
Qj＝W(MSEj+Sj)j＝1,2,3,4
wherein W is the adjustment weight, j is the number of the frame area, MSEjEuclidean distance mean square error, S, of the bounding box region numbered jjThe bubble area of the border region numbered j.
Preferably, in the present embodiment, W is 0.7.
And summing the burst estimated values of the 4 frame areas to obtain the overall burst estimated value of the overall solar panel. And carrying out normalization processing on the overall burst estimation value, wherein the value range of the overall burst estimation value after the normalization processing is between [0 and 1 ]. When the overall burst estimate is close to 1, the probability of burst is greater; when the overall burst estimate is close to 0, the likelihood of burst is less. And setting a burst early warning value, performing early warning when the overall burst estimated value reaches the burst early warning, and positioning a burst area through the burst estimated values of the four frames to enable a worker to timely process.
Preferably, the burst warning value is set to 0.5 in the embodiment of the present invention.
In summary, the embodiment of the present invention determines the abnormal cell panel through the preliminary evaluation of the cell panel, then performs the burst estimation on each frame region by the evaluation and detection of each frame region of the abnormal cell panel in combination with the area of the occurred bubble, and sums and normalizes the burst estimation values to obtain the overall burst estimation value. When the overall burst estimated value reaches the preset burst early warning value, early warning is carried out, and workers can rapidly position burst points and process the burst points in time according to the burst estimated values of all the frame areas.
Referring to fig. 2, a block diagram of a photovoltaic module glass burst warning system based on artificial intelligence according to an embodiment of the present invention is shown. The system comprises: the system comprises an image acquisition module 101, a frame grid line image acquisition module 102, a region-of-interest image division module 103, a bubble contour acquisition module 104, a twin network module 105, a bubble affected area acquisition module 106, a frame deformation influence factor acquisition module 107 and a burst estimation early warning module 108.
The image acquisition module 101 is used for acquiring an RGB image of the photovoltaic cell panel.
The frame grid line image acquisition module 102 is configured to extract an R channel image of an RGB image to obtain a first image; and carrying out binarization processing on the first image according to a preset first gray threshold value to obtain a frame grid line image of the photovoltaic cell panel.
The interested region image dividing module 103 is configured to divide the interested region image by the border and the outermost peripheral grid lines of the photovoltaic cell panel according to the border grid line image.
The twin network module 104 is used for sending the images of the region of interest into a pre-trained twin network for processing, and outputting similarity vectors with the normal panel region; and determining the abnormal cell panel with the deformed frame according to the similarity vector.
The bubble outline acquisition module 105 is configured to segment the bubble image according to a preset second gray threshold in the region-of-interest image; and performing edge extraction on the bubble image to obtain a bubble outline image.
The bubble affected area obtaining module 106 is configured to number the regions of interest on the abnormal cell panel according to each frame to obtain a plurality of frame areas; the bubble affected zone is divided by all bubble contour coordinates and the outermost peripheral gridlines within each border area.
The frame deformation influence factor obtaining module 107 is configured to calculate a euclidean distance between each gate line and a pixel point corresponding to the frame in the bubble influence region, and use a mean square error of the euclidean distance as an influence factor of the frame deformation on the glass burst.
The burst estimation early warning module 108 is used for summing pixel points in each bubble outline in the frame area to obtain a bubble area, and obtaining a burst estimation value of each frame grid line area according to the bubble area and the Euclidean distance mean square error; summing the burst estimated values to obtain an integral burst estimated value of the integral battery panel; and carrying out normalization processing on the overall burst estimated value, and feeding back early warning information when the burst estimated value reaches a set early warning value.
Preferably, the region-of-interest image dividing module 103 takes a region between the border and the outermost peripheral gate line as a region of interest, marks a pixel in the region of interest as 1, and marks other regions as 0, to obtain a mask image; and multiplying the mask image and the first image to obtain an image of the region of interest.
Preferably, the twin network module 104 outputs the similarity vector to judge the abnormal cell panel and the abnormal cell panel has no bubble, and then the abnormal cell panel is specially marked
Preferably, the calculating, by the frame deformation influence factor obtaining module, the euclidean distance between each grid line and the pixel point corresponding to the frame in the bubble influence region includes:
carrying out Hough line fitting after carrying out binarization processing on the bubble affected area to obtain the slope of a grid line in the bubble affected area;
making a straight line by using the negative number of the reciprocal of the slope of the grid line as the slope of grid line pixel points on the grid line; the intersection point of the straight line and the frame is a corresponding frame pixel point; and calculating the Euclidean distance between the grid line pixel point and the frame pixel point.
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 module glass burst early warning method based on artificial intelligence is characterized by comprising the following steps:
collecting RGB images of a photovoltaic cell panel;
extracting an R channel image of the RGB image to obtain a first image; performing binarization processing on a first image according to a preset first pixel threshold value to obtain a frame grid line image of the photovoltaic cell panel;
dividing an interested area image by a frame and the outermost peripheral grid line of the photovoltaic cell panel according to the frame grid line image;
sending the images of the interested areas into the twin network trained in advance for processing, and outputting similarity vectors of the images of the interested areas and the areas of normal battery plates; determining an abnormal cell panel with a deformed frame according to the similarity vector;
segmenting the bubble image in the region of interest image according to a preset second pixel threshold value; performing edge extraction on the bubble image to obtain a bubble profile;
numbering the interested regions on the abnormal cell panel according to each frame to obtain a plurality of frame regions; dividing a bubble influence area by all the bubble outline coordinates and the outermost periphery grid lines in each border area;
calculating Euclidean distance between each grid line and a pixel point corresponding to the frame in the bubble influence area, and taking the mean square error of the Euclidean distance as an influence factor of frame deformation on glass burst;
summing pixel points in each bubble outline in the frame region to obtain a bubble area, and obtaining a burst estimation value of each frame grid line region according to the bubble area and the Euclidean distance mean square error; summing the burst estimated values to obtain an overall burst estimated value of the overall cell panel; and normalizing the overall burst estimation value, and feeding back early warning information when the burst estimation value reaches a set early warning value.
2. The artificial intelligence based photovoltaic module glass burst warning method according to claim 1, wherein the region-of-interest image is divided by a frame and the outermost grid lines of the photovoltaic cell panel according to the frame grid line image:
taking the area between the frame and the outermost peripheral grid line as an interested area, marking the pixel in the interested area as 1, and marking other areas as 0 to obtain a mask image;
and multiplying the mask image and the first image to obtain the region-of-interest image.
3. The artificial intelligence based photovoltaic module glass burst early warning method according to claim 1, wherein the following operations are further included after the abnormal cell panel with deformed frame is determined according to the similarity vector:
and if the abnormal solar panel is judged by outputting the similarity vector through the twin network but no bubble appears in the abnormal solar panel, carrying out special marking on the abnormal solar panel.
4. The artificial intelligence based photovoltaic module glass burst warning method according to claim 1, wherein the method for dividing the bubble influence area in the frame area by the bubble outline coordinates comprises the following steps:
taking the difference value between the maximum abscissa and the minimum abscissa of the bubble outline in the upper and lower frame areas as the length of the bubble influence area; the width of the upper frame area and the lower frame area is the width of the bubble influence area; and/or
Taking the difference value between the maximum ordinate and the minimum ordinate of the bubble profile in the left and right border areas as the width of the bubble influence area; the length of the left and right frame regions is the length of the bubble affected region.
5. The artificial intelligence based photovoltaic module glass burst early warning method according to claim 1, wherein the method for calculating the Euclidean distance between pixel points corresponding to each grid line and the frame in the bubble influence area comprises the following steps:
carrying out Hough line fitting after the binarization processing of the bubble affected area to obtain the slope of a grid line in the bubble affected area;
making grid line pixel points on the grid lines as vertical lines of the grid lines; the intersection point of the vertical line and the frame is a corresponding frame pixel point; and calculating the Euclidean distance between the grid line pixel point and the frame pixel point.
6. The artificial intelligence based photovoltaic module glass burst early warning method according to claim 1, wherein the burst estimation value of each border grid line region is obtained according to the bubble area and the Euclidean distance mean square error:
calculating a burst estimate for each of said border regions by the formula:
Qj＝W(MSEj+Sj)
wherein W is the adjustment weight, j is the number of the frame area, MSEjThe Euclidean distance mean square error, S, of the bounding box region numbered jjThe bubble area of the border region numbered j.
7. The photovoltaic module glass burst early warning system based on artificial intelligence is characterized by comprising an image acquisition module, a frame grid line image acquisition module, an interested area image division module, a bubble contour acquisition module, a twin network module, a bubble affected area acquisition module, a frame deformation affected factor acquisition module and a burst estimation early warning module;
the image acquisition module is used for acquiring RGB images of the photovoltaic cell panel;
the frame grid line image acquisition module is used for extracting an R channel image of the RGB image to obtain a first image; performing binarization processing on the first image according to a preset first gray threshold value to obtain a frame grid line image of the photovoltaic cell panel;
the interested region image dividing module is used for dividing interested region images by using a frame and the outermost peripheral grid line of the photovoltaic cell panel according to the frame grid line images;
the twin network module is used for sending the image of the region of interest into a pre-trained twin network for processing and outputting a similarity vector with a normal battery panel region; determining an abnormal cell panel with a deformed frame according to the similarity vector;
the bubble outline acquisition module is used for segmenting a bubble image in the region of interest according to a preset second gray threshold; performing edge extraction on the bubble image to obtain a bubble outline image;
the bubble affected area acquisition module is used for numbering the interested areas on the abnormal battery panel according to each frame to acquire a plurality of frame areas; dividing a bubble influence area by all the bubble outline coordinates and the outermost periphery grid lines in each border area;
the frame deformation influence factor acquisition module is used for calculating Euclidean distances between each grid line and a pixel point corresponding to the frame in the bubble influence area, and taking the mean square error of the Euclidean distances as an influence factor of frame deformation on glass burst;
the burst estimation early warning module is used for summing pixel points in each bubble outline in the frame region to obtain a bubble area, and obtaining a burst estimation value of each frame grid line region according to the bubble area and the Euclidean distance mean square error; summing the burst estimated values to obtain an overall burst estimated value of the overall cell panel; and normalizing the overall burst estimation value, and feeding back early warning information when the burst estimation value reaches a set early warning value.
8. The artificial intelligence based photovoltaic module glass burst early warning system according to claim 7, wherein the region of interest image dividing module takes a region between the border and the outermost peripheral grid line as a region of interest, marks pixels in the region of interest as 1, and obtains a mask image with other regions as 0; and multiplying the mask image and the first image to obtain the region-of-interest image.
9. The artificial intelligence based photovoltaic module glass burst warning system according to claim 7, wherein the twin network module outputs similarity vectors to judge an abnormal panel and the abnormal panel has no bubbles, and the abnormal panel is specially marked.
10. The artificial intelligence based photovoltaic module glass burst early warning system of claim 7, wherein the calculating of Euclidean distance between each grid line and the pixel point corresponding to the frame in the bubble influence area by the frame deformation influence factor obtaining module comprises:
carrying out Hough line fitting after the binarization processing of the bubble affected area to obtain the slope of a grid line in the bubble affected area;
making a straight line by using the grid line pixel points on the grid line and taking the negative of the reciprocal of the slope of the grid line as the slope; the intersection point of the straight line and the frame is a corresponding frame pixel point; and calculating the Euclidean distance between the grid line pixel point and the frame pixel point.
CN202011466941.6A 2020-12-14 2020-12-14 Photovoltaic module glass burst early warning method and system based on artificial intelligence Withdrawn CN112529875A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011466941.6A CN112529875A (en) 2020-12-14 2020-12-14 Photovoltaic module glass burst early warning method and system based on artificial intelligence

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011466941.6A CN112529875A (en) 2020-12-14 2020-12-14 Photovoltaic module glass burst early warning method and system based on artificial intelligence

Publications (1)

Publication Number Publication Date
CN112529875A true CN112529875A (en) 2021-03-19

Family

ID=74999486

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011466941.6A Withdrawn CN112529875A (en) 2020-12-14 2020-12-14 Photovoltaic module glass burst early warning method and system based on artificial intelligence

Country Status (1)

Country Link
CN (1) CN112529875A (en)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113470014A (en) * 2021-08-31 2021-10-01 江苏裕荣光电科技有限公司 Fault detection method of solar cell panel based on artificial intelligence
CN113487563A (en) * 2021-07-02 2021-10-08 河海大学常州校区 Photovoltaic module subfissure self-adaptive detection method based on EL image
CN113610799A (en) * 2021-08-04 2021-11-05 沭阳九鼎钢铁有限公司 Artificial intelligence-based photovoltaic cell panel rainbow line detection method, device and equipment
CN113658133A (en) * 2021-08-16 2021-11-16 江苏鑫丰源机电有限公司 Gear surface defect detection method and system based on image processing
CN113673526A (en) * 2021-07-23 2021-11-19 浙江大华技术股份有限公司 Bubble detection method, terminal and computer-readable storage medium
CN116523906A (en) * 2023-06-28 2023-08-01 长沙韶光芯材科技有限公司 Method and system for detecting optical performance of glass substrate
CN117670880A (en) * 2024-01-31 2024-03-08 中成空间（深圳）智能技术有限公司 Detection and correction method and system for flexible photovoltaic cells

Cited By (11)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113487563A (en) * 2021-07-02 2021-10-08 河海大学常州校区 Photovoltaic module subfissure self-adaptive detection method based on EL image
CN113487563B (en) * 2021-07-02 2023-09-05 河海大学常州校区 EL image-based self-adaptive detection method for hidden cracks of photovoltaic module
CN113673526A (en) * 2021-07-23 2021-11-19 浙江大华技术股份有限公司 Bubble detection method, terminal and computer-readable storage medium
CN113610799A (en) * 2021-08-04 2021-11-05 沭阳九鼎钢铁有限公司 Artificial intelligence-based photovoltaic cell panel rainbow line detection method, device and equipment
CN113658133A (en) * 2021-08-16 2021-11-16 江苏鑫丰源机电有限公司 Gear surface defect detection method and system based on image processing
CN113470014A (en) * 2021-08-31 2021-10-01 江苏裕荣光电科技有限公司 Fault detection method of solar cell panel based on artificial intelligence
CN113470014B (en) * 2021-08-31 2021-12-07 江苏裕荣光电科技有限公司 Fault detection method of solar cell panel based on artificial intelligence
CN116523906A (en) * 2023-06-28 2023-08-01 长沙韶光芯材科技有限公司 Method and system for detecting optical performance of glass substrate
CN116523906B (en) * 2023-06-28 2023-09-12 长沙韶光芯材科技有限公司 Method and system for detecting optical performance of glass substrate
CN117670880A (en) * 2024-01-31 2024-03-08 中成空间（深圳）智能技术有限公司 Detection and correction method and system for flexible photovoltaic cells
CN117670880B (en) * 2024-01-31 2024-05-07 中成空间（深圳）智能技术有限公司 Detection and correction method and system for flexible photovoltaic cells

Similar Documents

Publication Publication Date Title
CN112529875A (en) Photovoltaic module glass burst early warning method and system based on artificial intelligence
CN109615653B (en) Leakage water area detection and identification method based on deep learning and visual field projection model
CN114998355B (en) Production defect identification method and device for sealing rubber ring
CN108665468B (en) Device and method for extracting tangent tower insulator string
CN109146860B (en) Full-automatic mechanical equipment installation leakage detection method and device
CN113487563B (en) EL image-based self-adaptive detection method for hidden cracks of photovoltaic module
CN106780526A (en) A kind of ferrite wafer alligatoring recognition methods
CN114170552A (en) Natural gas leakage real-time early warning method and system based on infrared thermal imaging
CN111462057A (en) Transmission line glass insulator self-explosion detection method based on deep learning
CN113538503A (en) Solar panel defect detection method based on infrared image
CN114782442B (en) Photovoltaic cell panel intelligent inspection method and system based on artificial intelligence
CN117876308A (en) Photovoltaic panel detection method based on image analysis
CN112381759A (en) Monocrystalline silicon solar wafer defect detection method based on optical flow method and confidence coefficient method
CN113591973B (en) Intelligent comparison method for appearance state change of track plate
CN109671084A (en) A kind of measurement method of workpiece shapes
CN111161264B (en) Method for segmenting TFT circuit image with defects
CN112396582A (en) Mask RCNN-based equalizing ring skew detection method
CN116739986A (en) Optical acoustic combination method for detecting typical defects of AUV in water conveyance tunnel
CN112651936B (en) Steel plate surface defect image segmentation method and system based on image local entropy
CN114821520A (en) Lane line detection method integrating spatio-temporal information
CN112288372B (en) Express bill identification method capable of simultaneously identifying one-dimensional bar code and three-segment code characters
CN114663402A (en) Cable prolapse detection method based on Hough linear detection and curve fitting
CN111079852B (en) Method for detecting icing thickness of insulator of power transmission line
CN113504244A (en) New energy automobile battery flaw detection method and device
CN117522863B (en) Integrated box body quality detection method based on image features

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: 20210319