A method for automatically detecting anti-reflection film on the surface of large-scale photovoltaic modules

Abstract

The present invention relates to an automatic detection method for anti-reflection film on the surface of large-scale photovoltaic modules, including: commanding a drone to carry out aerial photography under specific parameters; forming a digital orthophoto of a photovoltaic power station; identifying the longitude and latitude information of each group string in a defined area; marking the group string area; extracting the group string area from the corner point coordinates to convert it into a regular group string image; identifying the component frame of the regular group string image, extracting the component center area in the regular group string image, and calculating the grayscale value of each component center point area; calculating the relationship function of the component reflectivity; calculating the reflectivity values ​​of all other photovoltaic modules, and comprehensively planning the failure of the anti-reflection film layer of the module in the defined area. The beneficial effect of the present invention is that the present invention can complete the accurate detection of the reflectivity of the anti-reflection film layer on the surface of the photovoltaic power station module in a short time, comprehensively evaluate the film performance of the photovoltaic power station module, and improve the detection efficiency and detection accuracy.

Description

Technical Field

[0001] The invention relates to the field of detection technology, and more specifically, to an automatic detection method for an anti-reflection film layer on the surface of a large-scale photovoltaic module. Background Art

[0002] With the continuous advancement of photovoltaic power generation technology and the continuous decline in power generation costs, photovoltaic power generation has become a new favorite investment in the energy industry. Photovoltaic modules account for about 40% of the initial cost of ground photovoltaic systems, and defects in photovoltaic modules can directly affect power generation revenue. The peeling of the anti-reflection film on the surface of the module is a relatively common module defect. After the anti-reflection film-coated modules have been subjected to long-term outdoor tests, dust, rain, organic pollutants, etc. will gradually erode the hollow SiO2 anti-reflection film on the surface of the glass, causing the gradual loss of anti-reflection effect, and the power generation rate may drop by more than 3%. The working life of photovoltaic modules can be more than 25 years, but the anti-reflection film prepared by high-temperature sintering with glass cannot match the life cycle of photovoltaic modules. A high proportion of photovoltaic modules installed for more than 5 years have already had the problem of film shedding. Therefore, an efficient full-station module film evaluation method is needed to accurately evaluate the module film condition to determine whether re-coating is required and the area that needs re-coating.

[0003] The detection of anti-reflection film on the surface of photovoltaic power station components is usually carried out by contact measurement using a portable spectrophotometer. However, there are difficulties such as the large number of components to be tested, the inability to take samples for testing due to the diverse installation heights and terrains. Summary of the invention

[0004] The purpose of the present invention is to overcome the deficiencies in the prior art and provide a method for automatically detecting the anti-reflection film layer on the surface of a large-scale photovoltaic module. The technical solution is as follows:

[0005] Step 1: Use a drone equipped with a visible light gimbal camera to carry out aerial photography under specific parameters within a designated area;

[0006] Step 2: Based on the collected aerial images, a digital orthophoto of the photovoltaic power station is formed;

[0007] Step 3: Based on the digital orthophoto and image processing algorithm, identify the longitude and latitude information of each string in the defined area;

[0008] Step 4: Mark the string area;

[0009] Step 5: Extract the corner points of the marked string areas in turn, and convert the string areas into regular string images according to the corner point coordinates.

[0010] Step 6: Identify the component borders of the regular string image, extract the component center area in the regular string image, and calculate the grayscale value of each component center point area;

[0011] Step 7: Select a marking component according to the gray value and distribution of the component, detect the reflectivity of the marking component, and calculate the relationship function of the component reflectivity according to the gray value of the marking component, the component distribution, the image shooting parameters and the reflectivity value;

[0012] Step 8: Calculate the reflectivity values ​​of all other photovoltaic modules according to the relationship function, and comprehensively calculate the failure conditions of the anti-reflection film layers of the modules in the specified area.

[0013] Preferably, in step 1, the designated area is the entire photovoltaic power station or any part of the photovoltaic power station; the specific parameters include: the flight altitude, heading overlap rate and lateral overlap rate of the UAV; the flight altitude is comprehensively determined by the focal length of the optical camera, the sensor size, and the size of the photovoltaic string; the heading overlap rate and the lateral overlap rate have a value range of 50% to 90%.

[0014] Preferably, step 3 comprises:

[0015] Step 3.1, obtain the latitude and longitude coordinates of several feature points in the digital orthophoto;

[0016] Step 3.2, combining the latitude and longitude coordinates of the feature point with the XY model coordinates, and obtaining the conversion formula between the XY model coordinates and the latitude and longitude coordinates;

[0017] Step 3.3, obtain the longitude and latitude coordinates of each string in the designated area.

[0018] Preferably, in step 3, the image processing algorithm includes a threshold segmentation algorithm, a color space conversion algorithm, a connected region analysis algorithm and an image morphology processing algorithm; the image processing algorithm is used to separate the photovoltaic string area from the background area, and read the XY model coordinates of the center point of each string.

[0019] Preferably, step 4 comprises:

[0020] Step 4.1, determine whether to conduct further flight tests, if not, execute step 4.2, if further flight tests are to be conducted, execute step 4.3;

[0021] Step 4.2, performing image processing on the visible light image according to the latitude and longitude coordinates, flight altitude, yaw angle and other parameters of the visible light image, and marking the area where the string is to be extracted;

[0022] Step 4.3: Set the longitude and latitude coordinates of the center point of each string as the path point, adjust the flight parameters, and take an image of a single complete string at each waypoint according to the optimized path, and mark the string area.

[0023] Preferably, in step 5, the corner points of the group strings are extracted manually or by an automatic extraction algorithm, and the automatic extraction algorithm includes: a threshold segmentation algorithm, an edge detection algorithm and an edge morphology monitoring algorithm.

[0024] Preferably, in step 6, the component border is identified by a manual or automatic identification algorithm, and the automatic identification algorithm includes: a threshold segmentation algorithm and an image morphology processing algorithm; the component border is identified to segment the string image into a plurality of component areas.

[0025] Preferably, in step 6, the grayscale value of the center point area of ​​each component is the average grayscale value of all pixels in the center area of ​​each component except the gate line area.

[0026] Preferably, in step 7, the marking components are selected based on the average grayscale value and distribution of each component in the demarcated area, the selected marking components are evenly distributed in the demarcated area and the average grayscale value is evenly distributed in the grayscale interval, and the grayscale interval includes a high grayscale interval, a medium grayscale interval and a low grayscale interval.

[0027] Preferably, in step 8, the relationship function is determined by numerical fitting or fuzzy logic; the failure conditions of the anti-reflection film layer of the components in the designated area are systematically planned, including: summarizing the reflectivity value information of all components in the designated area, setting a failure threshold T; when the component reflectivity is greater than or equal to T, the anti-reflection film layer of the component is considered to have failed; counting the failure conditions of the anti-reflection film layer of the components in the region, and calculating the failure rate of the anti-reflection film of all photovoltaic components in the designated area.

[0028] The beneficial effects of the present invention are as follows: the present invention can complete the accurate detection of the reflectivity of the anti-reflection film layer on the surface of the photovoltaic power station component in a short time, comprehensively evaluate the film layer performance of the photovoltaic power station component, and improve the detection efficiency and detection accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0029] Figure 1 A flow chart of a method for automatically detecting anti-reflection film on the surface of a large-scale photovoltaic module provided in this application;

[0030] Figure 2 A schematic diagram of regular string extraction provided for this application;

[0031] Figure 3 Schematic diagram of the central area extraction of the photovoltaic module provided for this application. DETAILED DESCRIPTION

[0032] The present invention is further described below in conjunction with embodiments. The description of the following embodiments is only used to help understand the present invention. It should be noted that for ordinary persons in the art, without departing from the principle of the present invention, the present invention can also be modified in some ways, and these improvements and modifications also fall within the scope of protection of the claims of the present invention.

[0033] Embodiment 1:

[0034] A method for automatically detecting an anti-reflection film layer on the surface of a large-scale photovoltaic module, comprising:

[0035] Step 1: Use a drone equipped with a visible light gimbal camera to carry out aerial photography under specific parameters within a designated area;

[0036] Step 2: Based on the collected aerial images, a digital orthophoto of the photovoltaic power station is formed;

[0037] Step 3: Based on the digital orthophoto and image processing algorithm, identify the longitude and latitude information of each string in the defined area;

[0038] Step 4: Mark the string area;

[0039] Step 5: Extract the corner points of the marked string areas in turn, and convert the string areas into regular string images according to the corner point coordinates.

[0040] Step 6: Identify the component borders of the regular string image, extract the component center area in the regular string image, and calculate the grayscale value of each component center point area;

[0041] Step 7: Select a marking component according to the gray value and distribution of the component, detect the reflectivity of the marking component, and calculate the relationship function of the component reflectivity according to the gray value of the marking component, the component distribution, the image shooting parameters and the reflectivity value;

[0042] Step 8: Calculate the reflectivity values ​​of all other photovoltaic modules based on the relationship function, and comprehensively calculate the failure status of the anti-reflection film layer of the modules in the specified area.

[0043] In step 1, the area is defined as the entire photovoltaic power station or any part of the photovoltaic power station; the specific parameters include: the flight altitude, heading overlap rate and lateral overlap rate of the drone; the flight altitude is determined by the focal length of the optical camera, the sensor size, and the size of the photovoltaic string, so that each photo can fully capture the bracket; the heading overlap rate and the lateral overlap rate range from 50% to 90%.

[0044] In step 2, digital orthophotos can be generated with the help of 3D / 2D model building software, such as Smart3D, Pix4D, DJI Zhitu, etc.

[0045] Step 3 includes:

[0046] Step 3.1, obtain the latitude and longitude coordinates of several feature points in the digital orthophoto;

[0047] Step 3.2, combining the latitude and longitude coordinates of the feature point with the XY model coordinates, and obtaining the conversion formula between the XY model coordinates and the latitude and longitude coordinates;

[0048] Step 3.3, obtain the longitude and latitude coordinates of each string in the designated area.

[0049] In step 3, the image processing algorithm includes a threshold segmentation algorithm, a color space conversion algorithm, a connected region analysis algorithm and an image morphology processing algorithm; the image processing algorithm is used to separate the photovoltaic string area from the background area and read the XY model coordinates of the center point of each string.

[0050] It should be noted that after step 3, two extraction methods are used depending on whether further flight tests are conducted. If further tests cannot be conducted on the designated area due to experimental conditions, the closest visible light image is matched according to the longitude and latitude of the group string to be extracted, and the visible light image is further processed. The area where the group string is to be extracted is marked in the matched visible light image by combining the longitude and latitude of the group string, the longitude and latitude of the center point of the visible light image, the flight altitude, the yaw angle and other parameters; if the experimental conditions allow repeated tests, the longitude and latitude coordinates of the center point of each group string are set as path points, and the path is optimized according to the optimized path planning method, so that the UAV can fly to the waypoints according to the optimized path and the set yaw angle and flight altitude, and hover and take pictures at each waypoint. The path planning method includes ant colony algorithm, genetic algorithm, particle swarm algorithm, discrete ecological symbiosis algorithm, etc.; the set yaw angle and flight altitude must ensure that the image of the complete string is captured at each waypoint, and the string area is marked through the image processing algorithm; the image processing algorithm includes threshold segmentation method, color space conversion, connected region analysis method, semantic segmentation method, etc.; therefore, step 4 includes:

[0051] Step 4.1, determine whether to conduct further flight tests, if not, execute step 4.2, if further flight tests are to be conducted, execute step 4.3;

[0052] Step 4.2, perform image processing on the visible light image according to the latitude and longitude coordinates, flight altitude, yaw angle and other parameters of the visible light image, and mark the area where the string is to be extracted;

[0053] Step 4.3: Set the longitude and latitude coordinates of the center point of each string as the path point, adjust the flight parameters, and take an image of a single complete string at each waypoint according to the optimized path, and mark the string area.

[0054] In step 5, the corner points of the group strings are extracted manually or by an automatic extraction algorithm. The automatic extraction algorithm includes: a threshold segmentation algorithm, an edge detection algorithm, and an edge morphology monitoring algorithm.

[0055] In step 6, the component border is identified by manual or automatic recognition algorithm, and the automatic recognition algorithm includes: threshold segmentation algorithm and image morphology processing algorithm; the component border is identified to divide the string image into several component areas, such as m×n component areas. For example, the horizontal coordinate of the identified string border is Sx(i), i=1,2,m,m+1, and the vertical coordinate of the string border is Sy(j), i=1,2,n,n+1; the vertex coordinates of the extracted center area of ​​each component are as follows, LT=[Sx(i)+α,Sy(j)+α], RT=[Sx(i+1)-α,Sy(j)+α], LB=[Sx(i)+α,Sy(j+1)-α], RB=[Sx(i+1)-α,Sy(j+1)-α], where i=1,2,…m, j=1,2,…n, α is the pixel value of the component border to the inside of the component, which can be selected according to actual conditions.

[0056] In step 6, the grayscale value of the center point area of ​​each component is the average grayscale value of all pixels in the center area of ​​each component except the gate line area.

[0057] In step 7, the marking components are selected based on the average grayscale value and distribution of each component in the designated area. The selected marking components are evenly distributed in the designated area and the average grayscale value is evenly distributed in the grayscale interval, which includes a high grayscale interval, a medium grayscale interval and a low grayscale interval.

[0058] In step 8, the relationship function is determined by numerical fitting or fuzzy logic; the failure conditions of the anti-reflection film layer of the components in the designated area are comprehensively planned, including: summarizing the reflectivity value information of all components in the designated area, setting a failure threshold T; when the reflectivity of the component is greater than or equal to T, it is considered that the anti-reflection film layer of the component has failed; the failure conditions of the anti-reflection film layer of the components in the region are counted, and the failure ratio of the anti-reflection film of all photovoltaic components in the designated area is calculated.

[0059] Embodiment 2:

[0060] A method for automatically detecting an anti-reflection film layer on the surface of a large-scale photovoltaic module, comprising:

[0061] Step 1: Use the M300RTK DJI drone equipped with the H20T visible light gimbal camera to conduct aerial photography in a round-trip manner at one of the plots of a photovoltaic power station.

[0062] In step 1, the gimbal camera is 35mm, the focal length is 47mm, the resolution is 3888×5184, the length and width of the string are 12m and 4m respectively, and the drone flight altitude is selected to be 60m. At this time, the string can be completely photographed, and the image clarity meets the subsequent recognition requirements. Set the heading overlap rate and the lateral overlap rate to 80%.

[0063] Step 2: Collect all the aerial images collected and use DJI mapping software to reconstruct the model and generate a digital orthophoto (DOM) of the area.

[0064] Step 3: Based on the digital orthophoto and image processing algorithm, identify the longitude and latitude information of each string in the demarcated area. The specific steps include:

[0065] Step 3.1: Extract the model coordinates of the center point of the string, including:

[0066] Step 3.1.1, convert the original image IMG (size 3888×5184×3) into a black and white logo image BW. First, convert IMG into HSV color space, and the resulting single-dimensional images are H, S, and V. Use the OTSU algorithm to determine the grayscale thresholds of H, S, and V images as TH, respectively. H , TH S and TH V , traverse the pixel points (i, j) in IMG, each pixel point 1≤i≤ 3888,1≤j≤5184, if H(i,j)>TH H And V(i,j)>TH V , if IMG(i,j,1), IMG(i,j,2), and IMG(i,j,3) are not zero at the same time, then BW(i,j)=1, otherwise BW(i,j)=0.

[0067] Step 3.1.2: Remove the areas in the BW image that are smaller than a certain threshold TH region The connected graph, here we take TH region =600.

[0068] Step 3.1.3: Fill the holes in the BW image.

[0069] Step 3.1.4: The boundary of each connected image of the BW image is eroded inwards into a specific shape, here a 3×3 rectangle is selected.

[0070] Step 3.1.5: Determine the area range of each string in the image based on the BW image features.

[0071] Step 3.1.6: Determine the area range of each string in the image based on the BW image features.

[0072] Step 3.1.7, record the model coordinates of the center point of the connected area selected as the string area.

[0073] Step 3.2: Confirm the conversion formula between latitude and longitude coordinates and model coordinates.

[0074] In step 3.2, several feature points are selected in the model, and the longitude and latitude coordinates of the feature points are combined with the model coordinates to obtain the linear conversion formula between the model coordinates and the longitude and latitude coordinates.

[0075] Step 3.3, further obtain the longitude and latitude coordinates of each string group in the demarcated area Stringlatlng(k)=[latitude(k),longitude(k)], k=1,2,...StringNum, StringNum is the total number of string groups in the demarcated area.

[0076] Step 4: extract the visible light image of each string respectively, mark the string area to be extracted in the visible light image, and perform image extraction and single string area image extraction according to the longitude and latitude of the photo, including:

[0077] Step 4.1: Match the original image of each extracted group string according to the longitude and latitude coordinates. For the kth group string, find the visible light original image img with the closest longitude and latitude coordinates. k , image processing is performed on visible light images, including:

[0078] Step 4.1.1, read the yaw angle yaw(k) of the visible light image, and convert img k Image rotation - yaw(k) angle.

[0079] Step 4.1.2, img k (size is h×w×3) is converted into HSV color space, and the grayscale thresholds of H, S, and V dimensions are determined as TH H , TH S and TH V .

[0080] Step 4.1.3, traverse img k Pixel (i, j), 1≤i≤h, 1≤j≤w, if H(i, j)>TH H And S(i,j)>TH S , or V(i,j)>TH V , at the same time img k (i,j,1),img k (i,j,2),img k (i,j,3), then BW k (i,j)=1, otherwise BW k (i,j)=0.

[0081] Step 4.1.4: Remove BW k The area in the image is smaller than a certain threshold TH region The connected graph, here we take TH region = 1000.

[0082] Step 4.1.5, BW k The boundary of each connected image of the image expands outward in a certain shape, and a 3*3 rectangle is selected here.

[0083] Step 4.1.6: Fill BW k Holes in the image.

[0084] Step 4.1.7, BW k The boundary of each connected image of the image is eroded inwards to a certain shape, here a 80*80 rectangle is selected.

[0085] Step 4.1.8: Fill BW k Holes in the image.

[0086] Step 4.1.9: According to BW k Image features,determine the area range of each group of strings in the image.

[0087] Step 4.1.10, record the center point coordinates Stringxy of each string area k (t), t=1,2,…StringNum k , StringNum k , for img k The number of clusters in the image.

[0088] Step 4.2: Calculate the location of the proposed string in img according to the longitude and latitude of the proposed string, the distance between the drone and the string, and the longitude and latitude of the center point of the original image. k The approximate coordinates xy in k .

[0089] Step 4.3, calculate img k The coordinates of the center points of each string and xy k The string with the smallest distance is the string to be extracted. k .

[0090] Step 4.4: Label the covered string k Rectangular area of ​​the string Part k .

[0091] Step 5: For each marked rectangular area Part k, k = 1, 2, ... StringNum, respectively, carry out the corner point extraction of the group string, and extract the group string area according to the corner point coordinates and convert it into a regular group string image. The steps of the group string corner point extraction algorithm include:

[0092] Step 5.1: Rectangular area Part k For edge recognition, for example, the "canny" operator is used for edge detection, and the convex hull of the detected edge is identified. The convex hull consists of Bound(l), l = 1, 2, ... pointno, where pointno is the number of points in the convex hull. The angle between adjacent points of Bound(l) is detected, and the vertices with an angle greater than a certain set threshold are identified as vertices. In this example, the threshold is selected as 20°. The identified vertices are LT, RT, RB, and LB.

[0093] Step 5.2: Map the four points LT, RT, RB, and LB to the pixel points [1,1], [1,1100], [400,1100], and [400,1], and use this mapping relationship to form a regular string image Panel. k ,See Figure 2 shown.

[0094] Step 6: Extract each regular string image Panel k (size is k1×k2, k=1,2,…StringNum), and identify the component borders respectively. The identification process includes:

[0095] Step 6.1: Panel k Image converted to grayscale image PanelG k (k1×k2), traverse PanelG k For each pixel above, if the gray value is greater than Gray TH , then the value of the pixel is set to 1, otherwise it is set to 0. Here Gray TH =200.

[0096] Step 6.2: Count the values ​​of pixels in each column and row that are 1, which are TX(i), TY(j), i=1,2,…k1, j=1,2,…k2.

[0097] Step 6.3: If TX(i)>ThX, and the difference between i and the previous i value that meets the conditions is greater than β, then i is confirmed as the horizontal coordinate of the string frame; if TY(j)>ThY, and the difference between j and the previous j value that meets the conditions is greater than β, then j is confirmed as the vertical coordinate of the string frame. Here ThX=0.2k1, ThY=0.2k2, β=20. The horizontal coordinate of the string frame thus identified is Sx(i), i=1,2,m,m+1, and the vertical coordinate of the string frame is Sy(j), i=1,2,n,n+1. The arrangement of the components in the string is m×n.

[0098] Step 6.4: Extract the center area of ​​each component in the string as follows: LT = [Sx(i) + α, Sy(j) + α], RT = [Sx(i+1) - α, Sy(j) + α], LB = [Sx(i) + α, Sy(j+1) - α], RB = [Sx(i+1) - α, Sy(j+1) - α], where i = 1, 2, ... m, j = 1, 2, ... n, α is the pixel value from the component border to the inside of the component. The extracted center area of ​​the component is as follows: Figure 3 As shown. Eliminate the grid line area in the center area of ​​the component, that is, larger than GL Th Pixels, here we take GL Th = 160. Calculate the average grayscale value of the pixels in the center area of ​​each component except the grid line area

[0099] Step 7: Since the distribution of components in this embodiment is highly consistent, the selection of marked components is performed only based on the grayscale value of the components; all photovoltaic components in the designated area are marked according to the average grayscale value. (where k = 1, 2, ... StringNum, i = 1, 2, ... m, j = 1, 2, ... n) is divided into high / medium / low grayscale intervals, and equal components are selected in each grayscale interval as marking components. The average grayscale values ​​of the marking components are Where p = 1, 2, ... MarkNo, MarkNo is the number of marking components.

[0100] Step 8: Detect the reflectivity curve of the marking component and record its reflectivity value Rp (p = 1, 2, ... MarkNo) in the visible light region. Since the shooting parameters of each image in this embodiment are basically the same, only the average gray value and the gray value of the reflectivity are calculated, that is, Rp and The values ​​are numerically fitted to obtain the corresponding polynomial equation.

[0101] Step 9: Place all remaining photovoltaic modules Substitute the value into the polynomial equation to calculate the reflectivity value R_k of the remaining photovoltaic components i,j Calculated value.

[0102] Step 10: Summarize the reflectivity R_k of all components in the defined area i,j Value information, set the failure threshold T, and R_k i,j ≥T means failure. The failure of the anti-reflection film layer of the modules in the area is counted, and the failure ratio of the anti-reflection film of all photovoltaic modules in the designated area is calculated.

Claims

1. A method for automatically detecting anti-reflection film on the surface of large-scale photovoltaic modules, characterized in that: include: Step 1: Use a drone equipped with a visible light gimbal camera to carry out aerial photography under specific parameters within a designated area; Step 2: Based on the collected aerial images, a digital orthophoto of the photovoltaic power station is formed; Step 3: Based on the digital orthophoto and image processing algorithm, identify the longitude and latitude information of each string in the defined area; Step 4: Mark the string area; Step 5: Extract the corner points of the marked string areas in turn, and convert the string areas into regular string images according to the corner point coordinates. Step 6: Identify the component borders of the regular string image, extract the component center area in the regular string image, and calculate the grayscale value of each component center point area; Step 7: Select a marking component according to the gray value and distribution of the component, detect the reflectivity of the marking component, and calculate the relationship function of the component reflectivity according to the gray value of the marking component, the component distribution, the image shooting parameters and the reflectivity value; Step 8: Calculate the reflectivity values ​​of all other photovoltaic modules according to the relationship function, and comprehensively calculate the failure conditions of the anti-reflection film layers of the modules in the specified area.

2. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1 is characterized in that: In step 1, the designated area is the entire photovoltaic power station or any part of the photovoltaic power station; the specific parameters include: the flight altitude, heading overlap rate and lateral overlap rate of the drone; the flight altitude is determined by the focal length of the optical camera, the sensor size, and the size of the photovoltaic string; the heading overlap rate and the lateral overlap rate range from 50% to 90%.

3. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1 is characterized in that: Step 3 includes: Step 3.1, obtain the latitude and longitude coordinates of several feature points in the digital orthophoto; Step 3.2, combining the latitude and longitude coordinates of the feature point with the XY model coordinates, and obtaining the conversion formula between the XY model coordinates and the latitude and longitude coordinates; Step 3.3, obtain the longitude and latitude coordinates of each string in the designated area.

4. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 3 is characterized in that: In step 3, the image processing algorithm includes a threshold segmentation algorithm, a color space conversion algorithm, a connected region analysis algorithm and an image morphology processing algorithm; the image processing algorithm is used to separate the photovoltaic string area from the background area, and read the XY model coordinates of the center point of each string.

5. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1 is characterized in that: Step 4 includes: Step 4.1, determine whether to conduct further flight tests, if not, execute step 4.2, if further flight tests are to be conducted, execute step 4.3; Step 4.2, performing image processing on the visible light image according to the latitude and longitude coordinates, flight altitude, and yaw angle parameters of the visible light image, and marking the area where the strings are to be extracted; Step 4.3: Set the longitude and latitude coordinates of the center point of each string as the path point, adjust the flight parameters, and take an image of a single complete string at each waypoint according to the optimized path, and mark the string area.

6. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1 is characterized in that: In step 5, the corner points of the group strings are extracted manually or by an automatic extraction algorithm, and the automatic extraction algorithm includes: a threshold segmentation algorithm, an edge detection algorithm, and an edge morphology monitoring algorithm.

7. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1, characterized in that: In step 6, the component frame is identified by manual or automatic identification algorithm, and the automatic identification algorithm includes: a threshold segmentation algorithm and an image morphology processing algorithm; the component frame is identified to segment the string image into a plurality of component areas.

8. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1, characterized in that: In step 6, the grayscale value of the center point area of ​​each component is the average grayscale value of all pixels in the center area of ​​each component except the gate line area.

9. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1, characterized in that: In step 7, the marking components are selected based on the average grayscale value and distribution of each component in the designated area. The selected marking components are evenly distributed in the designated area and the average grayscale value is evenly distributed in the grayscale interval, and the grayscale interval includes a high grayscale interval, a medium grayscale interval and a low grayscale interval.

10. The automatic detection method for anti-reflection film layer on the surface of large-scale photovoltaic modules according to claim 1, characterized in that: In step 8, the relationship function is determined by numerical fitting or fuzzy logic; the failure conditions of the anti-reflection film layer of the components in the designated area are systematically planned, including: summarizing the reflectivity value information of all components in the designated area, setting a failure threshold T; if the component reflectivity is greater than or equal to T, it is considered that the anti-reflection film layer of the component has failed; counting the failure conditions of the anti-reflection film layer of the components in the area, and calculating the failure ratio of the anti-reflection film of all photovoltaic components in the designated area.

Images