CN113570587A - Photovoltaic cell broken grid detection method and system based on computer vision - Google Patents

Photovoltaic cell broken grid detection method and system based on computer vision Download PDF

Info

Publication number
CN113570587A
CN113570587A CN202110882422.6A CN202110882422A CN113570587A CN 113570587 A CN113570587 A CN 113570587A CN 202110882422 A CN202110882422 A CN 202110882422A CN 113570587 A CN113570587 A CN 113570587A
Authority
CN
China
Prior art keywords
grid line
grid
image
battery
panel
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Withdrawn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
CN202110882422.6A
Other languages
Chinese (zh)
Inventor
牧笛
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.)
Henan University of Animal Husbandry and Economy
Original Assignee
Henan University of Animal Husbandry and Economy
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 Henan University of Animal Husbandry and Economy filed Critical Henan University of Animal Husbandry and Economy
Priority to CN202110882422.6A priority Critical patent/CN113570587A/en
Publication of CN113570587A publication Critical patent/CN113570587A/en
Withdrawn legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/13—Edge detection
    • 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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/70—Determining position or orientation of objects or cameras
    • G06T7/73—Determining position or orientation of objects or cameras using feature-based methods
    • H—ELECTRICITY
    • H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
    • H02S—GENERATION OF ELECTRIC POWER BY CONVERSION OF INFRARED RADIATION, VISIBLE LIGHT OR ULTRAVIOLET LIGHT, e.g. USING PHOTOVOLTAIC [PV] MODULES
    • H02S50/00—Monitoring or testing of PV systems, e.g. load balancing or fault identification
    • H02S50/10—Testing of PV devices, e.g. of PV modules or single PV cells
    • 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
    • G06T2207/20024—Filtering details
    • G06T2207/20032—Median filtering
    • 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
    • G06T2207/20048—Transform domain processing
    • G06T2207/20056—Discrete and fast Fourier transform, [DFT, FFT]
    • 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
    • G06T2207/20112—Image segmentation details
    • G06T2207/20132—Image cropping
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
    • Y02E10/00—Energy generation through renewable energy sources
    • Y02E10/50—Photovoltaic [PV] energy

Abstract

The invention relates to the technical field of artificial intelligence, and provides a photovoltaic cell broken grid detection method and system based on computer vision, which comprises the following steps: converting the surface image of the single cell panel into a frequency domain; preliminarily extracting grid lines of the battery panel based on the frequency domain information to obtain an initial battery grid line image; and performing gradient calculation on the image on the surface of the single battery plate, determining the grid line reference position and the grid line reference width of the battery plate, and correcting the initial battery plate grid line image according to the grid line reference position and the grid line reference width to obtain an actual battery plate grid line image so as to realize the identification of the broken grid condition. According to the method, the grid lines are preliminarily extracted based on the frequency domain information of the image on the surface of the battery panel to obtain the initial battery panel grid lines, then the grid line reference information is determined based on the gradient condition of the image on the surface of the battery panel, the initial battery panel grid lines are corrected by utilizing the grid line reference information, the battery panel grid lines can be accurately identified, and the accuracy of broken grid identification is improved.

Description

Photovoltaic cell broken grid detection method and system based on computer vision
Technical Field
The invention relates to the technical field of photovoltaic power generation, in particular to a photovoltaic cell grid breakage detection method and system based on computer vision.
Background
Solar energy is more and more popular among people as a clean, pollution-free, convenient and easily available renewable energy source. By means of renewable green energy such as solar energy, photovoltaic power generation attracts the attention of governments and people of various countries in recent years. Solar energy resources in China are very rich, and in recent years, a photovoltaic power station is supported by the nation as a green power development energy project.
The traditional photovoltaic cell surface grid line detection is finished by adopting a manual detection method through manual naked eyes, the detection efficiency is low, the detection is not timely, and the detection result is easily influenced by artificial subjective factors, so that the detection reliability is poor. With the large-scale expansion of photovoltaic power stations, the method for monitoring the grid breaking defect of the photovoltaic power station through manual inspection obviously cannot meet the inspection requirements of timeliness, accuracy, rapidness and high efficiency.
Disclosure of Invention
In order to solve the above technical problems, the present invention aims to provide a method and a system for detecting grid breakage of a photovoltaic cell based on computer vision, and the adopted technical scheme is as follows:
the invention provides a photovoltaic cell grid breakage detection method based on computer vision, which comprises the following steps of:
acquiring a surface image of a single battery panel;
converting the surface image of the single battery panel into a frequency domain to obtain frequency domain information of the surface image of the battery panel;
extracting grid lines of the battery panel according to the frequency domain information of the surface image of the battery panel to obtain an initial battery grid line image;
performing gradient calculation on the acquired image of the surface of the single battery plate to extract the grid lines of the battery plate, and acquiring a battery plate grid line image based on the gradient;
determining a grid line reference position and a grid line reference width of a battery panel according to the gradient-based battery grid line image;
correcting grid lines in the initial battery grid line image according to the grid line reference position and the grid line reference width of the battery panel to obtain an actual battery grid line image;
and determining the grid breaking condition of the grid lines of the battery panel according to the obtained actual grid line image of the battery.
Further, the step of determining the grid line reference position and the grid line reference width of the battery panel includes:
dividing the battery grid line image based on the gradient into mu sub-regions, respectively calculating the illumination factor of each sub-region according to the grid line characteristics in each sub-region, and obtaining the illumination factor sequence { tau ] of the mu sub-regions1，τ2…τμ}；
The sequence of illumination factors { tau }1，τ2…τμTaking the sub-region corresponding to the minimum illumination factor in the data as the optimal sub-region, and determining the grid line reference width, the grid line reference interval and the grid line reference position in the optimal sub-region according to the grid line information in the optimal sub-region;
and determining the grid line reference positions in other sub-areas according to the grid line reference width, the grid line reference interval and the grid line reference position in the optimal sub-area, thereby obtaining the grid line reference position and the grid line reference width of the battery panel.
Further, the calculation formula of the illumination factor is as follows:
Figure BDA0003192834050000021
wherein, tauiLighting factor of ith sub-area; sigmali、σwiRespectively representing the grid line interval variance and the grid line width variance in the ith sub-area for the grid line characteristics in the ith sub-area; k is an illumination factor calculation parameter, and k is 5.
Further, the calculation formula for determining the grid line reference width and the grid line reference interval is as follows:
Figure BDA0003192834050000022
Figure BDA0003192834050000023
wherein, w0A grid line reference width which is transverse or longitudinal; q is the total number of transverse or longitudinal grid lines in the optimal sub-area; w is axThe width of the grid line of the x-th transverse or longitudinal grid line in the optimal sub-area is determined; d0A grid line reference interval which is transverse or longitudinal; dzThe spacing of the z-th transverse or longitudinal grid line in the optimal sub-area.
Further, according to the frequency domain information of the surface image of the battery panel, extracting the grid lines of the battery panel, wherein the step of obtaining the initial grid lines of the battery panel comprises:
calculating a grid line extraction index according to phase information in frequency domain information of the surface image of the cell panel;
comparing the grid line extraction index value with a grid line extraction index set value to determine frequency domain information of the grid line;
and converting the frequency domain information of the grid lines into a time domain to obtain the initial cell panel grid lines.
Further, the calculation formula of the grid line extraction index is as follows:
Figure BDA0003192834050000024
wherein E (t) is a grid line extraction index;
Figure BDA0003192834050000025
the phase of the nth harmonic after Fourier transform is taken as the smoothed image M0; n is the total number of harmonics.
Further, the step of determining the grid breaking condition of the grid lines of the battery panel according to the obtained actual grid line image of the battery comprises the following steps:
identifying the actual length of each grid line according to the obtained actual grid line image of the battery;
comparing the actual length of each grid line with a preset threshold value of the length of the corresponding complete grid line;
and if the actual length of the grid line is smaller than the length of the corresponding complete grid line preset threshold, judging that the grid line is broken.
Further, the method also comprises the following steps:
mapping the obtained actual battery grid line image to a three-dimensional space, wherein each grid line corresponds to a point in the three-dimensional space, the x-axis coordinate and the y-axis coordinate of the point are Hough space coordinates of the grid lines and are used for representing position information of the grid lines, and the z-axis coordinate of the point is the actual length of the grid lines;
and carrying out visual display on the three-dimensional space by utilizing the BIM of the photovoltaic power station, and alarming under the condition of judging that the grid line is broken.
Further, the step of acquiring an image of the surface of a single panel comprises:
acquiring multi-frame battery panel surface images shot continuously, and arranging the images according to the shooting sequence;
extracting edge characteristic lines of the surface images of the battery plates according to the gray value information of the pixels of the images;
extracting feature points from edge feature lines of the surface images of the two adjacent frames of battery plates, and matching the extracted feature points to realize splicing of the surface images of the continuous multi-frame battery plates;
filtering the spliced part of the images obtained after splicing to obtain an overall image of the total area of the cell panel;
carrying out binarization and morphological opening operation processing on the whole image of the total area of the battery panel to obtain an edge line image of the battery panel;
and cutting the whole image of the total area of the battery panel by taking the edge line image of the battery panel as a mask to obtain the surface image of the single battery panel.
The invention also provides a photovoltaic cell grid breakage detection system based on computer vision, which comprises a processor and a memory, wherein the processor is used for processing instructions stored in the memory to realize the photovoltaic cell grid breakage detection method based on computer vision.
The invention has the following beneficial effects: converting the image of the surface of the cell panel into a frequency domain, and preliminarily extracting grid lines based on frequency domain information corresponding to the grid lines to obtain initial cell panel grid lines; and then, grid line reference information such as grid line reference positions, grid line reference widths and the like is determined based on the gradient condition of the image on the surface of the battery panel, and the initial battery panel grid lines are corrected by using the grid line reference information, so that the battery panel grid lines can be accurately identified, and the accuracy of broken grid identification is improved.
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 method for detecting grid breakage of a photovoltaic cell based on computer vision according to the present invention;
FIG. 2 is a schematic diagram of a surface image of a battery plate obtained through shooting;
FIG. 3 is a battery panel white edge line image obtained after binarization processing;
FIG. 4 is a battery panel edge line image obtained after morphological opening operation;
fig. 5 is a schematic diagram of the gate line distribution of the optimal sub-region.
Detailed Description
To further illustrate the technical means and effects of the present invention adopted to achieve the predetermined objects, the following detailed description will be given to a method and an apparatus for detecting abnormality of a photovoltaic cell panel based on artificial intelligence according to the present invention, with reference to the accompanying drawings and preferred embodiments.
The method comprises the following steps:
the embodiment provides a photovoltaic cell grid breakage detection method based on computer vision, which converts a cell panel surface image into a frequency domain and then extracts an initial cell grid line image by utilizing phase information in frequency domain information. Because the phenomenon that part of grid lines are thick exists in the initial battery grid line image, in order to further obtain accurate grid lines, gradient calculation is carried out on the surface image of the battery panel, accurate grid line positions and widths in the battery panel are obtained, the initial battery grid line image is corrected according to the accurate grid line positions and widths, so that the accurate battery grid line image can be obtained, and further judgment on grid breaking conditions of the grid lines of the battery panel is achieved. As shown in fig. 1, the photovoltaic cell grid breakage detection method based on computer vision specifically includes the following steps:
(1) the method comprises the following steps of obtaining a surface image of a single battery panel:
and (1-1) acquiring multi-frame panel surface images shot continuously.
Wherein, in order to acquire the multiframe panel surface image of shooing in succession, this embodiment adopts unmanned aerial vehicle to carry the mode that the camera carried to take photo by plane to the panel surface and realizes. In order to obtain a clear image of the surface of each cell panel of the photovoltaic power station, the flight path, the flight height and the flight speed of the unmanned aerial vehicle need to be set in advance before aerial photography. In the process of taking photo by plane, make unmanned aerial vehicle fly along the flight line of setting for to flight in-process keeps the flight height and the speed of setting for, do not have shelters such as trees, building in the definition and the image of the panel surface image of guaranteeing to shoot in succession.
(1-2) splicing the continuous multi-frame battery plate surface images in sequence to obtain the whole image of the total area of the battery plate, wherein the method comprises the following specific steps:
and (1-2-1) extracting edge characteristic lines of the surface images of the continuous multi-frame cell plates according to the information of the pixel gray values of the images.
As shown in fig. 2, on a certain captured image of the panel surface, 1 indicates a panel edge line, 2 indicates a region other than the edge line on the panel surface, and 3 indicates a surrounding region other than the panel surface on the image. Because the pixel gray value of the edge line of the panel is greatly different from the pixel gray values of other areas on the surface of the panel and the surrounding area, the edge feature line of the panel can be extracted according to the image pixel gray value information, and the edge feature line of the panel is used as an ROI (region of interest) so as to be convenient for extracting and matching feature points in the ROI.
And (1-2-2) extracting feature points from edge feature lines of the surface images of the two continuous frames of the solar panels, and matching the extracted feature points to realize splicing of the surface images of the continuous frames of the solar panels.
The method comprises the steps of arranging continuous multi-frame panel images according to a sampling sequence, dividing ROI areas of two adjacent frame panel surface images into N small areas, namely dividing the ROI area in the first frame panel surface image into N small areas, and dividing the ROI area in the second frame panel surface image into N small areas in the same mode. After dividing ROI areas of two adjacent frames of panel surface images into N small areas, respectively extracting feature points in the same small areas of the two frames of panel surface images. After extracting the characteristic points in each small area of the ROI area of the two frames of the surface images of the battery panel, respectively matching the characteristic points in the two corresponding small areas to realize the splicing of the surface images of the continuous multi-frame battery panel.
In this embodiment, when feature point extraction is performed, an orb (organized FAST and Rotated brief) feature point detection algorithm capable of realizing rapid feature point extraction is adopted; when matching the feature points, a GMS feature matching algorithm is adopted. The specific implementation of the ORB feature point detection algorithm and the GMS feature matching algorithm belongs to the prior art, and is not described herein again. Of course, as another embodiment, when feature point extraction is performed, other feature point detection and extraction methods applicable in the prior art, such as Harris corner point detection algorithm, ORB feature point detection algorithm, SIFT key point detection algorithm, FAST feature point detection algorithm, etc., may also be adopted according to actual situations; when feature point matching is performed, other feature point matching methods applicable in the prior art, such as a RANSAC feature point matching algorithm, may also be used according to the actual situation.
The ROI area of the surface images of the two adjacent frames of the battery plates is divided into N small areas, and then the characteristic point extraction and matching mode is carried out in each small area, so that the calculation amount of the characteristic point extraction and matching is reduced, the image matching accuracy is ensured, and the characteristic point detection speed is effectively improved. Of course, in the case of not considering the calculation amount of feature point extraction and matching, as another embodiment, the feature point extraction and matching may be directly performed in the ROI areas adjacent to two adjacent frame panel surface images without dividing the ROI areas into a plurality of small areas.
And (1-2-3) filtering the spliced part of the continuous multi-frame battery panel surface images obtained by splicing to obtain an overall image of the total area of the battery panel.
Because the spliced battery panel surface images have a sudden change phenomenon of the pixel points of the overlapped parts, the median filtering method is adopted in the embodiment to filter the spliced battery panel surface images, and the points higher than a certain threshold value are removed to eliminate the sudden change of the pixel values. Since the specific implementation process of the median filtering method belongs to the prior art, it is not described here again.
The steps (1-2-1) - (1-2-3) are to perform image registration according to the edge straight line of each battery plate, that is, extracting feature points from the edge straight line of the battery plate of the image, and then performing splicing of the surface images of the continuous multi-frame battery plates based on the extracted feature points. The splicing mode can reduce the calculated amount of the image registration process to a great extent, and the generation efficiency of the whole image of the total area of the solar panel is improved.
However, it should be noted that the above-mentioned stitching method is only a method of sequentially stitching the multiple continuous frames of battery panel surface images to obtain the whole image of the total battery panel area, and as another embodiment, on the basis of being able to accurately obtain the whole image of the total battery panel area according to the multiple continuously shot frames of battery panel surface images, other applicable stitching methods may be selected from the prior art to implement the method. Since there are many splicing methods available, they will not be explained here.
In addition, after the overall image of the total area of the solar panel is obtained, the overall image is projected to a pre-constructed three-dimensional power station BIM (building information model), and real-time imaging of the solar panel area of the photovoltaic power station can be achieved. The BIM is an organic complex of a three-dimensional power station space model and power station information established on the basis of building information data, and the specific structure of the BIM belongs to the prior art and is not described herein again.
And (1-3) cutting the whole image of the total area of the battery panel by taking the edge line of the battery panel as a mask to obtain the surface image of the single battery panel.
In order to obtain a surface image of a single battery panel, firstly, the whole image of the total area of the battery panel is subjected to binarization processing to obtain a white edge line image of the battery panel, which contains white strip grid lines in the middle, as shown in fig. 3. In order to eliminate the influence of the white strip-shaped grid lines in the middle part, the binarized white edge line image of the battery panel is processed by adopting a morphological open operation method, so that an accurate edge line image of the battery panel is obtained, as shown in fig. 4. The specific implementation processes of the binarization processing and the morphological opening operation processing on the image belong to the prior art, and are not described here.
And (3) taking the obtained accurate battery plate edge line image as a mask, and multiplying the accurate battery plate edge line image with the whole battery plate total area image obtained in the step (1-2), so as to cut out a single battery plate surface image. Since the specific implementation process of performing the cropping and segmentation process on the image belongs to the prior art, it is not described here.
(2) And converting the surface image of the single battery plate into a frequency domain to obtain frequency domain information of the surface image of the battery plate.
After the single battery panel surface image is obtained, channel separation is carried out on the single battery panel surface image by considering the characteristics of the battery panel surface color, the G channel pixel value of the battery panel is extracted, so that the G channel image of the single battery panel is obtained, and the single channel image is convenient for subsequent analysis processing. Since there is a lot of noise in the G channel image of a single panel, which will interfere with the subsequent image processing, the G channel image of the panel is gaussian smoothed to remove the noise, i.e. a smoothed image M0 is obtained.
Considering that in the process of collecting the image on the surface of the battery plate, because of the influence of the angle of the camera and external factors, the illumination can not be kept uniform, and the change of the brightness of local light seriously influences the extraction of grid lines on the surface of the subsequent battery plate, therefore, in order to accurately extract the grid lines, the detection precision of broken grid lines is improved, grid line pixels are accurately distinguished under the condition of eliminating the influence of image brightness and contrast, Fourier transformation is carried out on the image M0 after smoothing, and the image on the surface of the single battery plate is converted into a frequency domain:
Figure BDA0003192834050000071
wherein F (u, v) is a frequency domain signal of the smoothed image M0 after fourier transform; f (x, y) is the amplitude of the nth harmonic of the smoothed image M0 after Fourier transformation;
Figure BDA0003192834050000072
the phase of the nth harmonic after Fourier transform is taken as the smoothed image M0; j is an imaginary unit; n is the total number of harmonics.
(3) And extracting grid lines of the battery panel according to the frequency domain information of the surface image of the battery panel to obtain an initial battery grid line image.
According to the priori knowledge, the grid lines are formed by pixel points with large gradient changes relative to the whole battery panel, and after the two-dimensional image on the surface of the battery panel is converted into a frequency domain, the phase similarity degree of each harmonic wave in the frequency domain corresponding to the pixel points with large gradient changes of the two-dimensional image is high. Therefore, based on the priori knowledge, the raster image can be extracted according to the frequency domain information of the image, and the specific implementation steps are as follows:
(3-1) calculating a grid line extraction index according to phase information in the frequency domain information of the surface image of the battery panel:
Figure BDA0003192834050000073
wherein E (t) is a grid line extraction index;
Figure BDA0003192834050000074
the phase of the nth harmonic after Fourier transform is taken as the smoothed image M0; n is the total number of harmonics.
And (3-2) comparing the grid line extraction index value with a grid line extraction index set value to determine the frequency domain information of the grid line.
When the grid line extraction index value approaches to zero, the point of the airspace image corresponding to the phase is the grid line position, and the other positions are non-grid line positions. Therefore, the set value of the grid line extraction index is set to be 0.2, the grid line extraction index value is compared with the set value of the grid line extraction index 0.2, when the grid line extraction index value is smaller than the set value of the grid line extraction index 0.2, the point on the image corresponding to the phase is considered as the grid line, otherwise, the point on the image corresponding to the phase is considered as not the grid line. Therefore, all phase information, namely frequency domain information, in the frequency domain corresponding to all grid lines in the image can be screened out.
And (3-3) converting the frequency domain information of the grid lines into a time domain to obtain the initial cell panel grid lines.
And (3) preliminarily acquiring the grid lines on the surface of the battery plate through inverse Fourier transform according to the phase information, namely frequency domain information, in the frequency domain corresponding to the grid lines acquired in the step (3-2), so as to obtain an initial battery grid line image M1.
The cell panel surface grid lines obtained based on the frequency domain phase analysis in the step (3) can effectively avoid the influence of uneven image brightness, but the cell panel surface grid lines are too thick and the grid line information is inaccurate due to the fact that the cell panel surface grid lines are easily influenced by noise in the image based on the frequency domain phase analysis and the grid line extraction index values of the pixel points near the small range of the target grid lines are close. Therefore, in order to improve the accuracy of judging the grid breaking condition of the subsequent cell panel grid line, the currently obtained initial cell panel grid line needs to be corrected subsequently, so that the more accurate cell panel grid line is obtained.
(4) And performing gradient calculation on the acquired surface image of the battery panel to extract grid lines of the battery panel, so as to obtain a gradient-based battery grid line image.
In order to obtain an accurate and reliable grid line image on the basis of the battery panel surface grid lines preliminarily obtained in the step (3), the gradient of a single battery panel surface image, namely the smoothed image M0, is calculated, and the battery panel grid lines are extracted, so that a battery grid line image M2 based on the gradient is obtained. In the embodiment, a first-order Sobel operator is adopted to calculate the gradient of the surface image of the panel, and the operator can effectively reduce the noise influence. Since the first order Sobel operator and how to calculate the image gradient by using the first order Sobel operator belong to the prior art, the details are not described here. Of course, as another embodiment, when calculating the gradient of the panel surface image, other gradient calculation methods in the prior art, such as a median difference method, an edge detection operator, and the like, may be used.
(5) And determining the grid line reference position and the grid line reference width of the battery panel according to the battery grid line image based on the gradient.
The cell panel grid lines extracted based on the gradient are susceptible to uneven illumination, and grid lines with small local gray levels affected by illumination cannot be accurately extracted, so that grid line information is inaccurate. Therefore, an illumination factor analysis model is constructed based on the gradient image, and the illumination distribution of the panel is analyzed. The purpose of illumination analysis is to extract an area with uniform illumination distribution so as to accurately extract grid line information and obtain more accurate grid line width and grid line interval, and further determine a grid line reference position and a grid line reference width on the whole battery panel so as to correct a grid line image acquired based on a phase position subsequently, and thus obtain accurate grid line information on the surface of the battery panel.
Specifically, the specific steps of analyzing the illumination distribution of the panel to determine the grid line reference position and the grid line reference width on the whole panel are as follows:
(5-1) dividing the battery grid line image based on the gradient into mu sub-regions, constructing an illumination factor analysis model based on the grid line characteristics of each sub-region, respectively calculating the illumination factor of each sub-region, and obtaining an illumination factor sequence { tau ] of the mu sub-regions1，τ2…τμ}. The purpose of constructing an illumination factor analysis model, namely calculating the illumination factor of each sub-area, is to screen the area with the most accurate grid lines in the gradient-based panel grid line image, namely to screen the area with the most balanced grid lines in the area. Specifically, the mathematical expression of the illumination factor analysis model is as follows:
Figure BDA0003192834050000091
wherein, tauiLighting factor of ith sub-area; sigmali、σwiRespectively representing the grid line interval variance and the grid line width variance in the ith sub-area for the grid line characteristics in the ith sub-area; k is an illumination factor calculation parameter, which can be selected according to the actual situation, and in this embodiment, k is 5.
Of course, the above-mentioned illumination factor analysis model is only a specific implementation for screening out the region where the grid lines are most evenly distributed in the gradient-based panel grid line image, and as another implementation, other analysis models in the prior art may be used in the case where the purpose of screening out the region where the grid lines are most evenly distributed in the gradient-based panel grid line image is achieved.
(5-2) sequence of illumination factors { tau1，τ2…τμThe sub-region corresponding to the smallest illumination factor in the wavelength division multiplexing is taken as the optimal sub-region, i.e. τj＝MIN{τ1，τ2…τμAnd (4) extracting the grid lines in the area j most accurately, and taking the area j as an optimal sub-area if the grid line information is the most standard. Determining the reference width and the reference interval of the grid line and the reference position of the grid line in the optimal sub-area according to the grid line information in the optimal sub-area。
Wherein, the width and the interval of each gate line in the calculation region j are respectively marked as { w1，w2…wQ}、{D1，D2…DQ-1And Q is the number of the grid lines in the area, and the grid line reference width and the grid line reference interval are determined as follows:
Figure BDA0003192834050000092
Figure BDA0003192834050000093
wherein, w0Is the grid line reference width; q is the total number of grid lines in the optimal sub-area, namely the area j; w is axThe width of the grid line of the x-th grid line in the optimal sub-area, namely the area j; d0A grid line reference interval; dzIs the z-th grid line spacing in the optimal sub-area, i.e. area j.
It should be noted that the grid line reference width and the grid line reference interval determined herein actually refer to the grid line reference width and the grid line reference interval in a certain direction on the panel, that is, the transverse grid line reference width and the grid line reference interval or the longitudinal grid line reference width and the grid line reference interval. The transverse direction and the longitudinal direction are only used for distinguishing grid lines in two different directions on the battery plate, and do not refer to the actual directions of the grid lines. Therefore, { w ] is involved in the calculation when the grid line reference width and the grid line reference interval in the lateral direction are calculated using the formulas (1) and (2)1，w2…wQAnd { D }1，D2…DQ-1The width and spacing of each horizontal grid line in the region j are calculated by using the formulas (1) and (2), and the reference width and spacing of the grid lines in the longitudinal direction are calculated by using the formulas (w)1，w2…wQAnd { D }1，D2…DQ-1And the width and the interval of each longitudinal grid line in the region j are multiplied.
When the grid line reference position in the optimal sub-area is determined, the position of the central axis of each grid line in the optimal sub-area is taken as the grid line reference position of the grid line. The central axis of the grating here means that the length direction of the axis is the same as the length direction of the grating, and the axis is located at the center of the width of the grating.
For ease of understanding, the grid lines of the best sub-region are distributed as shown in fig. 5. Where 4 denotes the best sub-region, 5 denotes the width of the horizontal grid lines in the best sub-region, 6 denotes the horizontal grid line spacing in the best sub-region, 7 denotes the horizontal grid line reference position in the best sub-region, 55 denotes the vertical grid line width in the best sub-region, 66 denotes the vertical grid line spacing in the best sub-region, and 77 denotes the vertical grid line reference position in the best sub-region. When the above formulas (1) and (2) are used to determine the reference width of the gate line and the reference interval of the gate line, the transverse gate line width and the longitudinal gate line width in the optimal sub-area need to be used to determine the reference width of the gate line in the transverse direction and the reference width of the gate line in the longitudinal direction respectively; and respectively determining the transverse grid line reference interval and the longitudinal grid line reference interval by utilizing the transverse grid line interval and the longitudinal grid line interval in the optimal sub-area. When the grid line reference position in the optimal sub-area is determined, the transverse grid line reference position and the longitudinal grid line reference position need to be determined.
And (5-3) determining the grid line reference position in other sub-areas according to the grid line reference width, the grid line reference interval and the grid line reference position in the optimal sub-area, thereby obtaining the grid line reference position and the grid line reference width of the whole battery panel.
And (3) because the grid lines on the solar panel are uniformly distributed, the grid line reference width and the grid line reference interval calculated in the step (5-2) are suitable for the whole solar panel. Therefore, the grid line reference width and the grid line reference interval calculated in the step (5-2) can be used as the grid line reference width and the grid line reference interval of the whole battery panel, and meanwhile, the grid line reference positions in other sub-areas can be calculated by combining the grid line reference position in the optimal sub-area so as to determine the grid line reference position of the whole battery panel. When the grid line reference positions in other sub-areas are calculated, the transverse grid line reference positions in other sub-areas are determined according to the transverse grid line reference width, the transverse grid line reference interval and the transverse grid line reference position in the optimal sub-area, and the longitudinal grid line reference positions in other sub-areas are determined according to the longitudinal grid line reference width, the longitudinal grid line reference interval and the longitudinal grid line reference position in the optimal sub-area.
In this embodiment, when determining the reference position of the gate line in the other non-optimal sub-area, the reference position of the horizontal or vertical gate line at the peripheral edge of the optimal sub-area is used as a reference, and then the reference width of the horizontal gate line and the reference interval of the gate line or the reference width of the vertical gate line and the reference interval of the gate line are translated every time in a direction perpendicular to the reference position of the gate line and away from the optimal sub-area. Of course, as another embodiment, when determining the reference positions of the gate lines in other non-optimal sub-areas, the reference positions of the gate lines in the optimal sub-areas may be used as references, and then the reference widths of the gate lines and the reference intervals of the gate lines are combined to determine the reference positions.
(6) And correcting the initial battery plate grid lines according to the grid line reference position and the grid line reference width of the battery plate to obtain an actual battery plate grid line image.
The method comprises the steps of obtaining a grid line reference position of a battery panel, comparing the grid line reference position with an initial battery panel grid line one by one under the condition that the grid line reference position of the battery panel is known, then cutting the initial battery panel grid line by taking the grid line reference position as the center of an actual grid line and taking the grid line reference width as the width of the actual grid line, wherein the central line of the obtained cut grid line is the grid line reference position, and the width of the grid line is the grid line reference width. By cutting, the part of the grid line which is affected by noise and is too thick based on phase analysis can be deleted, and the optimized grid line is obtained, so that the grid line can be accurately extracted, and an actual grid line image of the battery can be obtained.
(7) And determining the grid breaking condition of the grid lines of the battery panel according to the obtained actual grid line image of the battery.
Wherein, after the actual battery grid line image that obtains, in converting the panel grid line that draws into three-dimensional space according to the switching mode of setting for to show the testing result in the BIM system, be used for the disconnected bars condition of visual display panel, generate the suggestion task to the panel of disconnected bars, so that the staff in time maintains the change, prevent the problem that the panel generating power that arouses because of disconnected bars reduces, concrete realization process is as follows:
(7-1) after the obtained image of the actual grid lines of the battery, the actual length of each grid line can be identified in the image. Meanwhile, grid lines in the obtained actual battery grid line image are mapped to a three-dimensional space, each grid line corresponds to a point in the three-dimensional space, the x-axis coordinates and the y-axis coordinates of the point are Hough space coordinates of the grid lines and are used for representing position information of the grid lines, the z-axis coordinates of the point are the actual length of the grid lines, and the three-dimensional space is visually displayed by utilizing a photovoltaic power station BIM.
(7-2) in order to determine the grid breaking condition of each grid line and prompt workers in time under the grid breaking condition so as to inspect and maintain the grid lines of the solar panel in the area, setting a preset threshold L of the length of the grid lines, and if the numerical value of a point in the three-dimensional space corresponding to the grid line is not less than the threshold L, indicating that the grid line is complete, and displaying a green light by an alarm system in the BIM; if the numerical value of the point in the three-dimensional space corresponding to the grid line is smaller than the threshold value L, the situation that the grid line is incomplete and the grid is broken is shown, and the warning system in the BIM of the photovoltaic power station displays red light early warning.
The photovoltaic cell grid breakage detection method based on computer vision can accurately extract the grid line graph of the cell panel, has high robustness, and can accurately and quickly realize real-time detection of the grid breakage phenomenon of the photovoltaic power station.
The embodiment of the system is as follows:
the embodiment provides a photovoltaic cell broken grid detection system based on computer vision, which comprises a real-time imaging module, a grid line extraction module and a broken grid detection module, wherein the three modules are connected in pairs and used for realizing the photovoltaic cell broken grid detection method based on computer vision in the method embodiment by division of labor. The real-time imaging module is mainly used for acquiring the whole image of the battery panel and inputting the whole image of the battery panel and a grid breakage detection result into the BIM system for grid breakage display; the grid line extraction module is mainly used for acquiring a surface image of a single battery panel and accurately extracting grid lines of the battery panel; and the broken grid detection module is mainly used for judging the broken grid condition based on the extracted grid lines of the cell panel.
In essence, the real-time imaging module, the grid line extraction module and the grid break detection module may be regarded as a processor and a memory, and the processor is configured to process instructions stored in the memory to implement the method for detecting grid break of a photovoltaic cell based on computer vision in the above method embodiment.
It should be noted that: the precedence order of the above embodiments of the present invention is only for description, and does not represent the merits of the embodiments. And specific embodiments thereof have been described above. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.

Claims (10)

1. A photovoltaic cell grid breakage detection method based on computer vision is characterized by comprising the following steps:
acquiring a surface image of a single battery panel;
converting the surface image of the single battery panel into a frequency domain to obtain frequency domain information of the surface image of the battery panel;
extracting grid lines of the battery panel according to the frequency domain information of the surface image of the battery panel to obtain an initial battery grid line image;
performing gradient calculation on the acquired image of the surface of the single battery plate to extract the grid lines of the battery plate, and acquiring a battery plate grid line image based on the gradient;
determining a grid line reference position and a grid line reference width of a battery panel according to the gradient-based battery grid line image;
correcting grid lines in the initial battery grid line image according to the grid line reference position and the grid line reference width of the battery panel to obtain an actual battery grid line image;
and determining the grid breaking condition of the grid lines of the battery panel according to the obtained actual grid line image of the battery.
2. The method for detecting broken grid of photovoltaic cell based on computer vision as claimed in claim 1, wherein the step of determining the reference position and reference width of grid line of the panel comprises:
dividing the battery grid line image based on the gradient into mu sub-regions, respectively calculating the illumination factor of each sub-region according to the grid line characteristics in each sub-region, and obtaining the illumination factor sequence { tau ] of the mu sub-regions1，τ2…τμ}；
The sequence of illumination factors { tau }1，τ2…τμTaking the sub-region corresponding to the minimum illumination factor in the data as the optimal sub-region, and determining the grid line reference width, the grid line reference interval and the grid line reference position in the optimal sub-region according to the grid line information in the optimal sub-region;
and determining the grid line reference positions in other sub-areas according to the grid line reference width, the grid line reference interval and the grid line reference position in the optimal sub-area, thereby obtaining the grid line reference position and the grid line reference width of the battery panel.
3. The method for detecting grid breakage of a photovoltaic cell based on computer vision according to claim 2, wherein the calculation formula of the illumination factor is as follows:
Figure FDA0003192834040000011
wherein, tauiLighting factor of ith sub-area; sigmali、σwiRespectively representing the grid line interval variance and the grid line width variance in the ith sub-area for the grid line characteristics in the ith sub-area; k is an illumination factor calculation parameter, and k is 5.
4. The method for detecting broken grid of photovoltaic cell based on computer vision according to claim 2 or 3, characterized in that the calculation formula for determining the grid line reference width and the grid line reference interval is as follows:
Figure FDA0003192834040000012
Figure FDA0003192834040000021
wherein, w0A grid line reference width which is transverse or longitudinal; q is the total number of transverse or longitudinal grid lines in the optimal sub-area; w is axThe width of the grid line of the x-th transverse or longitudinal grid line in the optimal sub-area is determined; d0A grid line reference interval which is transverse or longitudinal; dzThe spacing of the z-th transverse or longitudinal grid line in the optimal sub-area.
5. The method for detecting broken grid of photovoltaic cell based on computer vision as claimed in any one of claims 1-3, wherein the step of extracting grid lines of the cell panel according to the frequency domain information of the image on the surface of the cell panel to obtain initial grid lines of the cell panel comprises:
calculating a grid line extraction index according to phase information in frequency domain information of the surface image of the cell panel;
comparing the grid line extraction index value with a grid line extraction index set value to determine frequency domain information of the grid line;
and converting the frequency domain information of the grid lines into a time domain to obtain the initial cell panel grid lines.
6. The method for detecting broken grid of photovoltaic cell based on computer vision of claim 5, wherein the grid line extraction index is calculated by the formula:
Figure FDA0003192834040000022
wherein E (t) is a grid line extraction index;
Figure FDA0003192834040000023
the phase of the nth harmonic after Fourier transform is taken as the smoothed image M0; n is the total number of harmonics.
7. The method for detecting broken grid of photovoltaic cell based on computer vision as claimed in any one of claims 1-3, wherein the step of determining broken grid condition of grid lines of the cell panel according to the obtained actual image of grid lines of the cell panel comprises:
identifying the actual length of each grid line according to the obtained actual grid line image of the battery;
comparing the actual length of each grid line with a preset threshold value of the length of the corresponding complete grid line;
and if the actual length of the grid line is smaller than the length of the corresponding complete grid line preset threshold, judging that the grid line is broken.
8. The method for detecting grid breakage of a photovoltaic cell based on computer vision according to claim 7, further comprising:
mapping the obtained actual battery grid line image to a three-dimensional space, wherein each grid line corresponds to a point in the three-dimensional space, the x-axis coordinate and the y-axis coordinate of the point are Hough space coordinates of the grid lines and are used for representing position information of the grid lines, and the z-axis coordinate of the point is the actual length of the grid lines;
and carrying out visual display on the three-dimensional space by utilizing the BIM of the photovoltaic power station, and alarming under the condition of judging that the grid line is broken.
9. The computer vision-based photovoltaic cell grid breakage detection method according to any one of claims 1-3, wherein the step of acquiring an image of the surface of a single panel comprises:
acquiring multi-frame battery panel surface images shot continuously, and arranging the images according to the shooting sequence;
extracting edge characteristic lines of the surface images of the battery plates according to the gray value information of the pixels of the images;
extracting feature points from edge feature lines of the surface images of the two adjacent frames of battery plates, and matching the extracted feature points to realize splicing of the surface images of the continuous multi-frame battery plates;
filtering the spliced part of the images obtained after splicing to obtain an overall image of the total area of the cell panel;
carrying out binarization and morphological opening operation processing on the whole image of the total area of the battery panel to obtain an edge line image of the battery panel;
and cutting the whole image of the total area of the battery panel by taking the edge line image of the battery panel as a mask to obtain the surface image of the single battery panel.
10. A computer vision based photovoltaic cell grid break detection system comprising a processor and a memory, the processor being configured to process instructions stored in the memory to implement the computer vision based photovoltaic cell grid break detection method of any one of claims 1-9.
CN202110882422.6A 2021-08-02 2021-08-02 Photovoltaic cell broken grid detection method and system based on computer vision Withdrawn CN113570587A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110882422.6A CN113570587A (en) 2021-08-02 2021-08-02 Photovoltaic cell broken grid detection method and system based on computer vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110882422.6A CN113570587A (en) 2021-08-02 2021-08-02 Photovoltaic cell broken grid detection method and system based on computer vision

Publications (1)

Publication Number Publication Date
CN113570587A true CN113570587A (en) 2021-10-29

Family

ID=78170003

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110882422.6A Withdrawn CN113570587A (en) 2021-08-02 2021-08-02 Photovoltaic cell broken grid detection method and system based on computer vision

Country Status (1)

Country Link
CN (1) CN113570587A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272306A (en) * 2022-09-26 2022-11-01 太阳谷再生资源(江苏)有限公司 Solar cell panel grid line enhancement method utilizing gradient operation

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115272306A (en) * 2022-09-26 2022-11-01 太阳谷再生资源(江苏)有限公司 Solar cell panel grid line enhancement method utilizing gradient operation

Similar Documents

Publication Publication Date Title
WO2018028103A1 (en) Unmanned aerial vehicle power line inspection method based on characteristics of human vision
GB2569751A (en) Static infrared thermal image processing-based underground pipe leakage detection method
CN110648364B (en) Multi-dimensional space solid waste visual detection positioning and identification method and system
CN110910350B (en) Nut loosening detection method for wind power tower cylinder
CN112149543B (en) Building dust recognition system and method based on computer vision
Mao et al. Development of power transmission line defects diagnosis system for UAV inspection based on binocular depth imaging technology
CN109253722A (en) Merge monocular range-measurement system, method, equipment and the storage medium of semantic segmentation
CN110991452B (en) Parking space frame detection method, device, equipment and readable storage medium
CN110189375A (en) A kind of images steganalysis method based on monocular vision measurement
CN103674975A (en) Visual LCD (Liquid Crystal Display) Mura defect machine inspection method based on Gabor filtering and C-V (Chan-Vese) model segmentation
CN105096305A (en) Method and device for analyzing state of insulators
CN109684914A (en) Based on unmanned plane image intelligent identification Method
CN103091330A (en) Copper strip surface defect identifying device for simulating human visual perception mechanism and method thereof
CN115222884A (en) Space object analysis and modeling optimization method based on artificial intelligence
CN113570587A (en) Photovoltaic cell broken grid detection method and system based on computer vision
CN108022245A (en) Photovoltaic panel template automatic generation method based on upper thread primitive correlation model
CN113705564B (en) Pointer type instrument identification reading method
CN113469938A (en) Pipe gallery video analysis method and system based on embedded front-end processing server
CN117351472A (en) Tobacco leaf information detection method and device and electronic equipment
CN111626104A (en) Cable hidden danger point detection method and device based on unmanned aerial vehicle infrared thermal imagery
CN111079752A (en) Method and device for identifying circuit breaker in infrared image and readable storage medium
CN115690077A (en) Bubble plate back defect detection method and system
CN114187269A (en) Method for rapidly detecting surface defect edge of small-sized device
CN113869407A (en) Monocular vision-based vehicle length measuring method and device
CN105447496B (en) A kind of docking aircraft model identification verification method and system based on machine vision

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: 20211029