A method and device for detecting a light spot for centralized photovoltaics

Abstract

The application discloses a kind of spot detection method and device for centralized photovoltaic, and the initial image of centralized photovoltaic is obtained by carrying out fish eye correction, rotation fine adjustment and size processing to the image to be handled of centralized photovoltaic area received from unmanned aerial vehicle collection, effective photovoltaic panel detection and photovoltaic panel cutting extraction are carried out to initial image in turn by sectional detection method and fixed-width detection method, while the corresponding of flight point is carried out to photovoltaic panel detection image after detection by flight point relationship table, output several photovoltaic panel images, and spot recognition is carried out to the output photovoltaic panel image to determine spot photovoltaic panel.The application carries out the corresponding of flight point to effective photovoltaic panel in photovoltaic panel detection image after removing invalid area, so that the specific position of spot photovoltaic panel in centralized photovoltaic is determined by system, the workload and work pressure of photovoltaic panel maintenance engineer are reduced, the work efficiency of engineer is also improved, and the operation efficiency of centralized photovoltaic is further improved.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and apparatus for detecting light spots in centralized photovoltaic systems. Background Technology

[0002] Due to the unique nature of the photovoltaic (PV) power generation industry, PV power plants are often located in remote areas, have numerous pieces of equipment, and are distributed over a wide geographical area. As the PV industry and technology continue to advance, effectively testing the quality and functionality of centralized PV systems within a power plant has always been a challenging problem for PV engineers.

[0003] A major problem with centralized photovoltaic (PV) spot detection is the high density of PV panels and the lack of distinctive location features. Current techniques involve using drones to photograph PV panels from high altitudes, then using software to filter images, classifying those with temperatures exceeding a threshold as suspicious, followed by manual screening and severity grading. Finally, based on GPS information recorded in the images, personnel manually locate the specific positions of the PV panels with spots using handheld infrared detectors, and replace them after confirmation.

[0004] However, the above-mentioned methods for detecting light spots on photovoltaic panels are complicated to implement, and because the system cannot provide accurate locations, photovoltaic panel repair personnel have to spend a lot of extra time and manpower to identify and confirm the specific locations of the light spots on the photovoltaic panels. Summary of the Invention

[0005] This invention provides a method and apparatus for detecting light spots in centralized photovoltaic systems, thereby improving the accuracy and efficiency of locating light spots on photovoltaic panels in centralized photovoltaic systems.

[0006] To address the aforementioned technical problems, this invention provides a method for detecting light spots in centralized photovoltaic systems, comprising the following steps:

[0007] The preset first flight path is sent to the target drone so that the target drone can collect images of the centralized photovoltaic area to be identified while flying along the first flight path, and return the collected images to be processed.

[0008] The system receives and sequentially performs fisheye correction, rotation fine-tuning, and size processing on the image to be processed, outputs an initial image, calls the segmented detection method to perform effective photovoltaic panel detection on the initial image, outputs a photovoltaic panel detection image, and performs waypoint mapping on the photovoltaic panel detection image through a preset waypoint relationship table.

[0009] The fixed-width detection method is used to cut, extract, and locate the photovoltaic panel in the detected image, outputting several photovoltaic panel images. Then, spot recognition is performed on the several photovoltaic panel images to identify the spot photovoltaic panel.

[0010] The spot detection method provided by this invention, after sending the first flight path to the target UAV and receiving the centralized photovoltaic image transmitted back by the target UAV, first performs fisheye correction, rotation fine-tuning, and size processing on the image. Since images captured by the UAV at high altitudes have fisheye protrusion and rotation errors, it is necessary to first perform fisheye correction and rotation fine-tuning on the acquired raw image to improve the accuracy and feasibility of subsequent spot detection by the system. Simultaneously, to facilitate subsequent detection of effective photovoltaic panel areas, the image also needs to be size-processed to reduce the impact of invalid areas in the image on subsequent photovoltaic panel detection and further improve the accuracy of spot recognition.

[0011] After outputting the initial image, the system uses a segmented detection method to detect valid photovoltaic panels in the initial image, further eliminating invalid areas in the image. This ensures that the output photovoltaic panel detection image contains a large proportion of valid photovoltaic panels. Simultaneously, the system uses a pre-stored waypoint relationship table to match waypoints with the valid photovoltaic panels in the image after removing invalid areas. This allows the system to further determine the specific location of the spotted photovoltaic panel in the centralized photovoltaic system after identifying it, reducing the workload and pressure on photovoltaic panel maintenance engineers. It also improves the efficiency of engineers in maintaining spotted photovoltaic panels, thereby improving the operational efficiency of the centralized photovoltaic system.

[0012] After completing the effective photovoltaic panel detection and outputting the corresponding photovoltaic panel detection image, the system performs photovoltaic panel cutting, extraction, and positioning on the photovoltaic panel detection image. Each photovoltaic panel in the image is cut and extracted, and then spot detection is performed on each cut photovoltaic panel to identify the photovoltaic panels with spots. Once identified, the photovoltaic panels with spots to be repaired can be determined, and the specific location of the photovoltaic panels with spots can be determined based on the previously determined waypoints. This reduces the workload of photovoltaic panel maintenance engineers and reduces the human and time resources required for photovoltaic power plants to perform spot detection and maintenance on centralized photovoltaics, thereby improving the working efficiency of centralized photovoltaics in photovoltaic power plants.

[0013] As a preferred example, before sending the preset first route to the target drone, the method further includes:

[0014] Centralized photovoltaic image acquisition is performed on the centralized photovoltaic area to be identified to obtain a basic photovoltaic image, so that the basic photovoltaic image includes area markers;

[0015] The photovoltaic panels in the basic photovoltaic image are divided and marked according to the regional markers, and the marked images are associated with UAV waypoints and the flight routes are divided to generate the waypoint relationship table.

[0016] The waypoint relationship table includes the first route and multiple waypoints along the first route.

[0017] To facilitate the subsequent system's marking of the specific locations of the detected effective photovoltaic panels, the spot detection method provided in this invention first uses the drone to acquire images of the centralized photovoltaic area before sending the first flight path to the target drone. This obtains the corresponding basic images, and the photovoltaic panels in the basic images are divided into regions and marked with waypoints. This includes the photovoltaic panel corresponding to each waypoint and the specific location of the photographed photovoltaic panel in the centralized photovoltaic area. In this way, waypoints are marked on the photovoltaic panels in the images, and flight paths are divided for each waypoint, generating multiple drone image acquisition flight paths. Then, the waypoints, flight paths, and the specific locations of the photovoltaic panels are correlated and integrated to generate the waypoint relationship table, providing data support for the subsequent system to determine the waypoints for effective photovoltaic panel detection.

[0018] As a preferred example, the step of calling the segmented detection method to perform effective photovoltaic panel detection on the initial image, outputting a photovoltaic panel detection image, and then performing waypoint mapping on the photovoltaic panel detection image using a preset waypoint relationship table, specifically:

[0019] RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output.

[0020] The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output.

[0021] The pixel height of the interval region between two adjacent marked regions is compared with a preset height threshold. Two marked regions whose pixel height is less than or equal to the height threshold are merged into one marked region. All marked regions are filtered out and the photovoltaic panel detection image is output.

[0022] The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table.

[0023] To improve the accuracy of the system in detecting valid photovoltaic panels, the spot detection method provided in this invention, after obtaining the initial image, first extracts RGB data from the image to confirm the RGB value of each pixel block, i.e., the first image RGB data. Then, it compares the determined RGB value of each pixel block with the RGB threshold data pre-stored in the system to obtain the corresponding first comparison result. Since the image of a centralized photovoltaic system mainly consists of grasses besides the photovoltaic panels, the RGB values ​​corresponding to the grasses and photovoltaic panels in the image are not the same. Therefore, the system compares the RGB values ​​of each pixel block in the image to identify invalid areas and further determine the valid photovoltaic panel areas.

[0024] The identified invalid areas are defined as marked areas. Multiple marked areas are merged, and the intervals between adjacent marked areas are removed to generate a merged marked area. All the merged marked areas are then removed to obtain and output the corresponding photovoltaic panel detection image.

[0025] After output, the system will also retrieve the waypoint relationship table and match and filter the output photovoltaic panel detection image to determine the specific waypoint in the first route corresponding to the photovoltaic panel in the photovoltaic panel detection image. Based on the specific position of the photovoltaic panel determined by each waypoint in the waypoint relationship table, the system will locate and mark each photovoltaic panel in the photovoltaic panel detection image. This provides data reference for the subsequent system to send the specific location of the photovoltaic panel to the maintenance personnel after determining the spot photovoltaic panel, thereby improving the work efficiency of photovoltaic engineers and thus improving the operation and maintenance efficiency of the photovoltaic power station.

[0026] As a preferred example, the step of using the fixed-width detection method to perform photovoltaic panel segmentation, extraction, and positioning on the photovoltaic panel detection image, and outputting several photovoltaic panel images, specifically involves:

[0027] RGB data is extracted from the photovoltaic panel detection image to obtain corresponding second image RGB data. The second image RGB data is compared with RGB threshold data, and the corresponding second comparison result is output.

[0028] The portion of the RGB data of the second image in the second comparison result that is greater than or equal to the RGB threshold data is taken as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked RGB data of the second image to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is cut and extracted according to the complete cutting line, and the several photovoltaic panel images are output.

[0029] At the same time, the waypoint associations of the several photovoltaic panel images are confirmed according to the corresponding waypoints of the corresponding routes in the waypoint relationship table.

[0030] To further improve the detection accuracy of spot detection in the subsequent system when performing spot detection on photovoltaic panel images, the system in this invention, after outputting the photovoltaic panel detection image, will also perform targeted segmentation and extraction of each photovoltaic panel image in the photovoltaic panel detection image using a fixed-width detection method. This will divide a single photovoltaic panel detection image into multiple photovoltaic panel images corresponding to individual photovoltaic panels, so that the subsequent system can perform spot detection on them. This improves the system's computational efficiency and reduces the computational space required for the subsequent system to perform spot detection on the photovoltaic panels.

[0031] As a preferred example, the step of identifying the photovoltaic panel by spot recognition in the plurality of photovoltaic panel images specifically includes:

[0032] RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold.

[0033] If it is determined that there exists RGB data that is greater than or equal to the spot RGB threshold, then the pixel blocks corresponding to the RGB data that are greater than or equal to the spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks whose calculated area is greater than or equal to a preset area threshold are determined as spots.

[0034] The photovoltaic panel with light spots is identified as the light spot photovoltaic panel. The images of the photovoltaic panels are traversed and integrated to obtain the light spot photovoltaic panel. The waypoint corresponding to the light spot photovoltaic panel is sent to the first terminal.

[0035] After obtaining several images of photovoltaic panels, the system can extract RGB data from each image. Then, based on a comparison of the extracted RGB data with a preset data threshold, it determines whether a light spot exists on the corresponding photovoltaic panel in the image. If it exists, the area of ​​the light spot is calculated to further determine whether it is a real light spot. This eliminates the possibility of incorrect light spot identification due to errors in the RGB data of a single pixel block. Only light spots with an area reaching a certain value are identified as real light spots, and only photovoltaic panels corresponding to real light spots are identified as light-spot photovoltaic panels. This information is then sent to the terminal of the photovoltaic panel maintenance personnel, prompting them to inspect the light-spot photovoltaic panels, thus improving the maintenance efficiency of photovoltaic panel maintenance personnel.

[0036] Accordingly, the present invention also provides a spot detection device for centralized photovoltaic systems, the spot detection device comprising an image acquisition module, a photovoltaic panel detection module, and a spot recognition module;

[0037] The image acquisition module is used to send a preset first flight path to the target UAV, so that the target UAV can acquire images of the centralized photovoltaic area to be identified while flying along the first flight path, and feed back the acquired images to be processed.

[0038] The photovoltaic panel detection module is used to receive and sequentially perform fisheye correction, rotation fine-tuning and size processing on the image to be processed, output an initial image, call the segmented detection method to perform effective photovoltaic panel detection on the initial image, output a photovoltaic panel detection image, and perform waypoint mapping on the photovoltaic panel detection image through a preset waypoint relationship table;

[0039] The spot recognition module is used to call the fixed width detection method to cut, extract and locate the photovoltaic panel in the photovoltaic panel detection image, output several photovoltaic panel images, and perform spot recognition on the several photovoltaic panel images to determine the photovoltaic panel with the spot.

[0040] As a preferred example, the spot detection device further includes a waypoint confirmation module;

[0041] The waypoint confirmation module is used to collect centralized photovoltaic images of the centralized photovoltaic area to be identified, and obtain a basic photovoltaic image so that the basic photovoltaic image includes area markers.

[0042] The photovoltaic panels in the basic photovoltaic image are divided and marked according to the regional markers, and the marked images are associated with UAV waypoints and the flight routes are divided to generate the waypoint relationship table.

[0043] The waypoint relationship table includes the first route and multiple waypoints along the first route.

[0044] As a preferred example, the photovoltaic panel detection module uses a segmented detection method to perform effective photovoltaic panel detection on the initial image, outputs a photovoltaic panel detection image, and performs waypoint mapping on the photovoltaic panel detection image using a preset waypoint relationship table, specifically:

[0045] RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output.

[0046] The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output.

[0047] The pixel height of the interval region between two adjacent marked regions is compared with a preset height threshold. Two marked regions whose pixel height is less than or equal to the height threshold are merged into one marked region. All marked regions are filtered out and the photovoltaic panel detection image is output.

[0048] The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table.

[0049] As a preferred example, the spot recognition module uses a fixed-width detection method to perform photovoltaic panel segmentation, extraction, and localization on the photovoltaic panel detection image, and outputs several photovoltaic panel images, specifically:

[0050] RGB data is extracted from the photovoltaic panel detection image to obtain corresponding second image RGB data. The second image RGB data is compared with RGB threshold data, and the corresponding second comparison result is output.

[0051] The portion of the RGB data of the second image in the second comparison result that is greater than or equal to the RGB threshold data is taken as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked RGB data of the second image to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is cut and extracted according to the complete cutting line, and the several photovoltaic panel images are output.

[0052] At the same time, the waypoint associations of the several photovoltaic panel images are confirmed according to the corresponding waypoints of the corresponding routes in the waypoint relationship table.

[0053] As a preferred example, the spot recognition module performs spot recognition on the plurality of photovoltaic panel images to determine the photovoltaic panel with the spot, specifically including:

[0054] RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold.

[0055] If it is determined that there exists RGB data that is greater than or equal to the spot RGB threshold, then the pixel blocks corresponding to the RGB data that are greater than or equal to the spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks whose calculated area is greater than or equal to a preset area threshold are determined as spots.

[0056] The photovoltaic panel with light spots is identified as the light spot photovoltaic panel. The images of the photovoltaic panels are traversed and integrated to obtain the light spot photovoltaic panel. The waypoint corresponding to the light spot photovoltaic panel is sent to the first terminal. Attached Figure Description

[0057] Figure 1 : This is a flowchart illustrating an embodiment of the light spot detection method for centralized photovoltaic systems provided by the present invention;

[0058] Figure 2 : This is a schematic diagram of an embodiment of the light spot detection device for centralized photovoltaic systems provided by the present invention;

[0059] Figure 3 : A schematic diagram of an embodiment of centralized photovoltaic area division provided by the present invention;

[0060] Figure 4 This is a schematic diagram of the image to be processed provided by the present invention before fisheye processing;

[0061] Figure 5 : A schematic diagram of the image to be processed provided by the present invention after fisheye processing;

[0062] Figure 6 This is a schematic diagram illustrating the effective photovoltaic panel detection of an initial image provided by the present invention.

[0063] Figure 7 : A schematic diagram of an embodiment of the photovoltaic panel detection image provided by the present invention. Detailed Implementation

[0064] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0065] Example 1

[0066] Please refer to Figure 1 This is a flowchart illustrating an embodiment of the light spot detection method for centralized photovoltaic systems provided by the present invention, including steps 101 to 103, each step of which is detailed below:

[0067] Step 101: Send the preset first flight path to the target UAV, so that the target UAV can collect images of the centralized photovoltaic area to be identified while flying along the first flight path, and return the collected images to be processed.

[0068] To address the issue of precise location during inspections of suspicious photovoltaic (PV) panels, this invention first requires dividing the PV panels in a centralized PV area into zones and setting drone flight paths. Since centralized PV systems are typically neatly installed, the system in this embodiment can utilize this characteristic to calculate and form flight paths, and then capture images of the centralized PV panels using predetermined actions, obtaining infrared and visible light images. After obtaining the images, the system creates a database corresponding to the centralized PV panels based on the flight path names and physical location identifiers. The specific location of the PV panels in the images is then determined according to the set flight paths and the divided areas, thereby improving the efficiency of inspecting PV panels with visible spots.

[0069] For example, before sending the preset first route to the target drone in this embodiment, the method further includes:

[0070] Centralized photovoltaic image acquisition is performed on the centralized photovoltaic area to be identified to obtain a basic photovoltaic image, so that the basic photovoltaic image includes area markers;

[0071] The photovoltaic panels in the basic photovoltaic image are divided and marked according to the regional markers, and the marked images are associated with UAV waypoints and the flight routes are divided to generate the waypoint relationship table.

[0072] The waypoint relationship table includes the first route and multiple waypoints along the first route.

[0073] To facilitate subsequent system marking of the specific locations of the detected effective photovoltaic panels, the spot detection method provided in this embodiment of the invention first acquires images of the centralized photovoltaic area using the drone before sending the first flight path to the target drone. This obtains corresponding basic images, and the photovoltaic panels in the basic images are divided into regions and marked with waypoints. This includes the photovoltaic panel corresponding to each waypoint and the specific location of the photographed photovoltaic panel in the centralized photovoltaic area. In this way, waypoints are marked on the photovoltaic panels in the images, and flight paths are divided for each waypoint, generating multiple drone image acquisition flight paths. Then, the waypoints, flight paths, and the specific locations of the photovoltaic panels are correlated and integrated to generate the waypoint relationship table, providing data support for the subsequent system to determine the waypoints for effective photovoltaic panel detection.

[0074] In this embodiment, see Figure 3 , Figure 3 This is a schematic diagram of one embodiment of the centralized photovoltaic area division provided by the present invention. Figure 3 As shown, the entire centralized photovoltaic area can be divided into sections according to the pre-set area markers. Figure 3 The various regions shown, among which, Figure 3A1-1 in the diagram represents the first branch line in the first row. Each branch line represents a section of a photovoltaic panel in a centralized photovoltaic area.

[0075] The image acquisition device used in this embodiment is the "DJI Mavic 3T" drone. This embodiment does not further limit the specific model of the drone. The drone only needs to support infrared photography and have a resolution of not less than 640x512.

[0076] For images captured by drones, each image must have at least one spaced marker on the left or right side, i.e., the area marker described in this embodiment. The photovoltaic panel branch line should be slightly greater than 50% of the image captured. If possible, the image should be as balanced as possible, with the heading angle being a key parameter. Based on the principle of infrared thermal imaging, the images must be taken on a sunny day with sunlight of a certain intensity. The inspection and shooting time should ideally be between 11:00 and 15:00 to fully reflect the temperature of the photovoltaic panel, thereby obtaining the light spots and improving the system's accuracy in detecting light-spotted photovoltaic panels.

[0077] After obtaining the basic images from the drone, the system can plan flight routes based on these images. Specifically, in regular areas, the system uses the waypoint recording function to capture three points. Based on the obtained GPS, altitude, ellipsoidal altitude, pitch angle, and heading angle, and by retrieving the number of rows to be calculated and the number of images captured in each row provided by the user, the system calculates all the intermediate points to form a flight route. For irregular areas, the user manually records the routes. Finally, the regular and irregular routes are linked together to form a waypoint relationship table.

[0078] Step 102: Receive and sequentially perform fisheye correction, rotation fine-tuning, and size processing on the image to be processed, output the initial image, call the segmented detection method to perform effective photovoltaic panel detection on the initial image, output the photovoltaic panel detection image, and perform waypoint mapping on the photovoltaic panel detection image through a preset waypoint relationship table.

[0079] The spot detection method provided in this embodiment of the invention, after sending the first flight path to the target UAV and receiving the centralized photovoltaic image transmitted back by the target UAV, first performs fisheye correction, rotation fine-tuning, and size processing on the image. Since images captured by the UAV at high altitudes have fisheye protrusion and rotation errors, it is necessary to first perform fisheye correction and rotation fine-tuning on the acquired original image to improve the accuracy and feasibility of subsequent spot detection by the system. Simultaneously, to facilitate subsequent detection of effective photovoltaic panel areas, the image also needs to be size-processed to reduce the impact of invalid areas in the image on subsequent photovoltaic panel detection and further improve the accuracy of spot recognition.

[0080] In this embodiment, before performing effective photovoltaic panel detection on the image using the segmented detection method, the system will preprocess the image to be processed, i.e., the original image, through fisheye correction, rotation fine-tuning, and size processing, thereby improving the accuracy of the subsequent photovoltaic detection by the system. The image before the system performs fisheye correction on the original captured image is shown below. Figure 4 As shown, and Figure 5 This is a schematic diagram of the image to be processed after fisheye processing provided by the present invention. After fisheye processing by the system, the slight fisheye protrusion in the original captured image due to drone shooting can be corrected. In this embodiment, the specific method of fisheye correction is to use the OpenCV cv2.getOptimalNewCameraMatrix method to correct the image, so that the objects in the image are restored to their original balance. The distortion correction method is as follows:

[0081] #Set the parameters required for distortion correction

[0082] camera_matrix=np.array([[1000,0,320],[0,1000,240],[0,0,1]],dtype=np.float32)

[0083] The main problem with infrared images is the "fisheye" bulge; simply correcting it slightly will solve the issue.

[0084] distortion_coefficient = np.array([-0.5, 0.01, -0.001, 0.002, 0], dtype = np.float32) # Distortion correction

[0085] h,w = gray.shape[:2]

[0086] #cv2.getOptimalNewCameraMatrix(original intrinsic parameter, distortion correction coefficient, original image width and height, 0, / / no black border)

[0087] new_camera_matrix,roi=cv2.getOptimalNewCameraMatrix(camera_matrix,distortion_coefficient,(w,h),0,(w,h))

[0088] However, for the image after fisheye correction processing by the system, there may be errors in both the balance error of the photovoltaic panel installation and the heading angle of the UAV shooting, and there may also be slight rotation errors (within about 5 degrees deviation). Therefore, the system needs to perform "picture rotation micro-correction" as much as possible, that is, the rotation fine-tuning described in this embodiment. The rotation fine-tuning specifically includes: First, perform binaryzation processing on the image, and then use the "cv2.Canny()" edge detection algorithm in OpenCV to identify the edges of the objects in the image, and transmit the results to the "HoughLinesP(image,rho,theta,threshold,minLineLength,maxLineGap)" probabilistic Hough transform line detection algorithm, that is, the Hough line detection algorithm described in this embodiment. Detect all possible lines in the picture through the Hough line detection algorithm, and then screen and obtain all the lines that meet the range of the approximate horizontal line (-10 < x < 10), and calculate the average value of all the screened lines. Further, use the "cv2.getRotationMatrix2D() and cv2.warpAffine()" methods to rotate the image according to the calculated average value, so that the objects in the image are kept as horizontal as possible, so as to achieve the rotation fine-tuning of the image.

[0089] For the size processing of the image, it is because according to the recorded data information, it is determined that there are some rows in the image that only take a single row. Therefore, the system needs to intercept 60% of the upper or lower part of the image for this type of image, so as to reduce the impact of the invalid area on the subsequent operations of the system.

[0090] After outputting the initial image, the system performs effective photovoltaic panel detection on the initial image through the segmented detection method, and further excludes the invalid areas in the image, so that the effective photovoltaic panel part included in the output photovoltaic panel detection image accounts for the vast majority. At the same time, the system pre-stores the waypoint relationship table to perform waypoint correspondence on the effective photovoltaic panels in the image after removing the invalid areas, so that the system can further determine the specific position of the spot photovoltaic panel in the centralized photovoltaic after determining the spot photovoltaic panel, reducing the workload and work pressure of the photovoltaic panel maintenance engineer, and at the same time improving the work efficiency of the engineer, that is, the maintenance efficiency for the spot photovoltaic panel, and further improving the operation efficiency of the centralized photovoltaic.

[0091] Exemplarily, the call to the segmented detection method described in this embodiment to perform effective photovoltaic panel detection on the initial image, output a photovoltaic panel detection image, and perform waypoint correspondence on the photovoltaic panel detection image through a preset waypoint relationship table is specifically as follows:

[0092] RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output.

[0093] The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output.

[0094] The pixel height of the interval region between two adjacent marked regions is compared with a preset height threshold. Two marked regions whose pixel height is less than or equal to the height threshold are merged into one marked region. All marked regions are filtered out and the photovoltaic panel detection image is output.

[0095] The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table.

[0096] To improve the accuracy of the system in detecting valid photovoltaic panels, the spot detection method provided in this embodiment of the invention first extracts RGB data from the image after obtaining the initial image. This confirms the RGB value of each pixel block in the image, i.e., the first image RGB data. Then, the determined RGB value of each pixel block is compared with the RGB threshold data pre-stored in the system to obtain the corresponding first comparison result. Since the image of centralized photovoltaic systems mainly consists of grasses besides the photovoltaic panels, the RGB values ​​corresponding to the grasses and photovoltaic panels in the image are not the same. Therefore, the system compares and detects the RGB values ​​of each pixel block in the image to identify invalid areas and further determine the valid photovoltaic panel areas.

[0097] The identified invalid areas are defined as marked areas. Multiple marked areas are merged, and the intervals between adjacent marked areas are removed to generate a merged marked area. All the merged marked areas are then removed to obtain and output the corresponding photovoltaic panel detection image.

[0098] In this embodiment, since high temperatures are rarely present near centralized photovoltaic systems, and the area is mostly covered with grass and other vegetation besides the photovoltaic panels, these objects appear as low-temperature objects in infrared images, typically appearing as dark red, purple, or black. Therefore, the system in this embodiment employs a "segmented detection method" for effective photovoltaic panel detection in the image, specifically including:

[0099] Assuming the image is divided into 16 segments, the first step is to determine that the middle 10 segments of the area with photovoltaic panels will be bright orange, yellow, or close to white in the infrared image. As for the passageway, i.e., the irrelevant area, if a dark segment appears and the RGB value of the dark segment is greater than or equal to the RGB threshold data described in this embodiment (e.g., R<=180 and G<=80), the system will determine whether to mark it as an irrelevant area, i.e., the marked area described in this embodiment, based on whether the proportion of that segment is higher than a set threshold. At the same time, the dividing lines between photovoltaic panels will also be marked when determining the marked area.

[0100] After identification, the approximate height of a photovoltaic panel in the image is determined and set in pixels, i.e., the height threshold described in this embodiment is set. If the distance between two identification areas, i.e. the pixel height described in the embodiment, is less than or equal to the height threshold, they are merged into a new identification area.

[0101] In addition, the system also has a "first / last label area width too large anomaly detection" function. Specifically, if the upper and lower parts of the photovoltaic panel happen to be photovoltaic panels in other invalid rows, then the width of the label area at the beginning and end of the image must be greater than the width of the gap between the photovoltaic panels. In this case, the system determines that the upper and lower areas of the image are invalid areas, i.e., label areas.

[0102] If the second marked area (or the second to last marked area) is a passageway, and the upper part of the image is not tall enough, it means that it is not a complete photovoltaic panel. Therefore, the system can determine that the part above (or below) this line is an invalid area.

[0103] After detecting the image using the above detection rules, the following results can be obtained: Figure 6 The diagram shown illustrates the effective photovoltaic panel detection of an initial image provided by the present invention, wherein... Figure 6 The light-colored areas are the invalid regions, or marked areas, identified by the system. After filtering them out, the system can obtain the desired result. Figure 7 The diagram shown is a schematic representation of an embodiment of the photovoltaic panel detection image provided by the present invention.

[0104] After output, the system will also retrieve the waypoint relationship table and match and filter the output photovoltaic panel detection image to determine the specific waypoint in the first route corresponding to the photovoltaic panel in the photovoltaic panel detection image. Based on the specific position of the photovoltaic panel determined by each waypoint in the waypoint relationship table, the system will locate and mark each photovoltaic panel in the photovoltaic panel detection image. This provides data reference for the subsequent system to send the specific location of the photovoltaic panel to the maintenance personnel after determining the spot photovoltaic panel, thereby improving the work efficiency of photovoltaic engineers and thus improving the operation and maintenance efficiency of the photovoltaic power station.

[0105] In this embodiment, the route confirmation for the photovoltaic panel branch line specifically involves determining the width and number of the marked areas based on the continuity of the returned marked areas, matching them with the data stored in the previous database, i.e., the waypoint relationship table, extracting the corresponding photovoltaic panel branch line row from the waypoint relationship table, and saving it according to certain rules (such as "time__waypoint_down.JPG").

[0106] Step 103: Use the fixed-width detection method to cut, extract, and locate the photovoltaic panel in the detected image, output several photovoltaic panel images, and perform spot recognition on the several photovoltaic panel images to determine the spot photovoltaic panel.

[0107] After completing the effective photovoltaic panel detection and outputting the corresponding photovoltaic panel detection image, the system performs photovoltaic panel cutting, extraction, and positioning on the photovoltaic panel detection image. Each photovoltaic panel in the image is cut and extracted, and then spot detection is performed on each cut photovoltaic panel to identify the photovoltaic panels with spots. Once identified, the photovoltaic panels with spots to be repaired can be determined, and the specific location of the photovoltaic panels with spots can be determined based on the previously determined waypoints. This reduces the workload of photovoltaic panel maintenance engineers and reduces the human and time resources required for photovoltaic power plants to perform spot detection and maintenance on centralized photovoltaics, thereby improving the working efficiency of centralized photovoltaics in photovoltaic power plants.

[0108] Furthermore, in this embodiment, the fixed-width detection method is used to perform photovoltaic panel cutting, extraction, and positioning on the photovoltaic panel detection image, outputting several photovoltaic panel images, specifically as follows:

[0109] RGB data is extracted from the photovoltaic panel detection image to obtain corresponding second image RGB data. The second image RGB data is compared with RGB threshold data, and the corresponding second comparison result is output.

[0110] The portion of the RGB data of the second image in the second comparison result that is greater than or equal to the RGB threshold data is taken as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked RGB data of the second image to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is cut and extracted according to the complete cutting line, and the several photovoltaic panel images are output.

[0111] At the same time, the waypoint associations of the several photovoltaic panel images are confirmed according to the corresponding waypoints of the corresponding routes in the waypoint relationship table.

[0112] To further improve the detection accuracy of the subsequent system when performing spot detection on photovoltaic panel images, the system in this embodiment of the invention will, after outputting the photovoltaic panel detection image, perform targeted segmentation and extraction of each photovoltaic panel image in the photovoltaic panel detection image using a fixed-width detection method. This will divide a single photovoltaic panel detection image into multiple photovoltaic panel images corresponding to individual photovoltaic panels, so that the subsequent system can perform spot detection on them. This improves the system's computational efficiency and reduces the computational space required for the subsequent system to perform spot detection on the photovoltaic panels.

[0113] In this embodiment, the "fixed width detection method" is used to detect continuous quantitative vertical lines. Specifically, it involves determining whether to mark a dark area (with a threshold of R<=180 and G<=80) in the obtained image RGB data as an irrelevant area, i.e., the identification area described in this embodiment, based on whether the proportion of the dark area in the dark area is higher than a set threshold. This area is then used by the system to extract and cut individual photovoltaic panels.

[0114] After completing the initial segmentation of the image (identifying photovoltaic panels within the same partition), the system uses the `cv2.HoughLinesP()` method again to identify the vertical lines between photovoltaic panels. This identifies the intervals between individual photovoltaic panels in the image. Vertical lines with similar distances are merged into a single line based on their average value. Since there may be a few unidentified vertical lines, the system uses the intervals between these lines and calculates and supplements them using a similar interval to form complete lines. These complete lines are then used to segment and extract individual photovoltaic panel images, outputting several corresponding photovoltaic panel images.

[0115] After cutting, the system names and stores the photovoltaic panels according to their original locations. This includes sorting the panels by region identifiers (ls indicates starting from left, rs indicates starting from right), and naming them as "Time__Waypoint_down_ls_ur_01.JPG" (ur indicates top row, lr indicates bottom row). Based on these identifiers, the specific location of each individual photovoltaic panel can be determined, providing data support for subsequent spot-based photovoltaic panel positioning.

[0116] Furthermore, the method described in this embodiment for identifying light spots in the plurality of photovoltaic panel images to determine the photovoltaic panel with light spots specifically includes:

[0117] RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold.

[0118] If it is determined that there exists RGB data that is greater than or equal to the spot RGB threshold, then the pixel blocks corresponding to the RGB data that are greater than or equal to the spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks whose calculated area is greater than or equal to a preset area threshold are determined as spots.

[0119] The photovoltaic panel with light spots is identified as the light spot photovoltaic panel. The images of the photovoltaic panels are traversed and integrated to obtain the light spot photovoltaic panel. The waypoint corresponding to the light spot photovoltaic panel is sent to the first terminal.

[0120] After obtaining several images of photovoltaic panels, the system can extract RGB data from each image. Then, based on a comparison of the extracted RGB data with a preset data threshold, it determines whether a light spot exists on the corresponding photovoltaic panel in the image. If it exists, the area of ​​the light spot is calculated to further determine whether it is a real light spot. This eliminates the possibility of incorrect light spot identification due to errors in the RGB data of a single pixel block. Only light spots with an area reaching a certain value are identified as real light spots, and only photovoltaic panels corresponding to real light spots are identified as light-spot photovoltaic panels. This information is then sent to the terminal of the photovoltaic panel maintenance personnel, prompting them to inspect the light-spot photovoltaic panels, thus improving the maintenance efficiency of photovoltaic panel maintenance personnel.

[0121] Meanwhile, the expected lifespan of photovoltaic panels can be predicted based on the historical changes of each panel stored in the database, thus improving the operation and maintenance efficiency of centralized photovoltaic systems.

[0122] To better illustrate the working principle and steps of the method and apparatus for detecting light spots in centralized photovoltaic systems according to the present invention, please refer to the relevant descriptions above, but not limited to those provided.

[0123] Accordingly, see Figure 2 , Figure 2 This is a schematic diagram of an embodiment of the light spot detection device for centralized photovoltaic systems provided by the present invention. Figure 7 As shown, the spot detection device includes a waypoint confirmation module 201, an image acquisition module 202, a photovoltaic panel detection module 203, and a spot recognition module 204.

[0124] The waypoint confirmation module 201 is used to collect centralized photovoltaic images of the centralized photovoltaic area to be identified, obtain a basic photovoltaic image, and make the basic photovoltaic image include regional markers; divide and mark the photovoltaic panels in the basic photovoltaic image according to the regional markers, and associate the marked image with UAV waypoints and divide the route to generate the waypoint relationship table; wherein the waypoint relationship table includes the first route and multiple waypoints in the first route.

[0125] The image acquisition module 202 is used to send a preset first flight path to the target UAV, so that the target UAV can acquire images of the centralized photovoltaic area to be identified while flying along the first flight path, and feed back the acquired images to be processed.

[0126] The photovoltaic panel detection module 203 is used to receive and sequentially perform fisheye correction, rotation fine-tuning and size processing on the image to be processed, output an initial image, call the segmented detection method to perform effective photovoltaic panel detection on the initial image, output a photovoltaic panel detection image, and perform waypoint mapping on the photovoltaic panel detection image through a preset waypoint relationship table.

[0127] Furthermore, the photovoltaic panel detection module 203 uses a segmented detection method to perform effective photovoltaic panel detection on the initial image, outputs a photovoltaic panel detection image, and performs waypoint mapping on the photovoltaic panel detection image using a preset waypoint relationship table, specifically:

[0128] RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output.

[0129] The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output. The pixel height of the interval region between two adjacent identification regions is compared with a preset height threshold. The two identification regions corresponding to the pixel height being less than or equal to the height threshold are merged into one identification region. All identification regions are filtered out, and the photovoltaic panel detection image is then output. The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table.

[0130] The light spot recognition module 204 is used to call the fixed width detection method to perform photovoltaic panel cutting extraction and positioning on the photovoltaic panel detection image, output several photovoltaic panel images, and perform light spot recognition on the several photovoltaic panel images to determine the photovoltaic panel with light spots.

[0131] Furthermore, the light spot recognition module 204 uses a fixed-width detection method to perform photovoltaic panel cutting, extraction, and positioning on the photovoltaic panel detection image, and outputs several photovoltaic panel images, specifically:

[0132] RGB data is extracted from the photovoltaic panel detection image to obtain corresponding second image RGB data. The second image RGB data is compared with RGB threshold data, and the corresponding second comparison result is output.

[0133] The portion of the second image RGB data in the second comparison result that is greater than or equal to the RGB threshold data is used as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked second image RGB data to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is then cut and extracted according to the complete cutting line to output the plurality of photovoltaic panel images. At the same time, the plurality of photovoltaic panel images are associated and confirmed according to the waypoints corresponding to the routes in the waypoint relationship table.

[0134] Furthermore, the light spot recognition module 204 performs light spot recognition on each of the plurality of photovoltaic panel images to determine the photovoltaic panel with the light spot, specifically including:

[0135] RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold.

[0136] If it is determined that there exists RGB data greater than or equal to the light spot RGB threshold, then the pixel blocks corresponding to the RGB data greater than or equal to the light spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks with the calculated area greater than or equal to the preset area threshold are determined as light spots; and the photovoltaic panels with light spots are determined as light spot photovoltaic panels, the several photovoltaic panel images are traversed and integrated to obtain the light spot photovoltaic panels, and the waypoints corresponding to the light spot photovoltaic panels are sent to the first terminal.

[0137] In summary, this invention provides a method and apparatus for detecting light spots in centralized photovoltaic (PV) systems. It obtains an initial image of the centralized PV area by performing fisheye correction, rotation fine-tuning, and size processing on an image of the PV area received from a UAV. The initial image is then sequentially processed using segmented detection and fixed-width detection methods to detect and extract valid PV panels. Simultaneously, a waypoint relationship table is used to map waypoints to the detected PV panel images, outputting several PV panel images. Light spot recognition is then performed on the output PV panel images to identify the affected PV panels. This invention uses a waypoint relationship table to map waypoints to the valid PV panels in the detected images after removing invalid areas, facilitating the system's determination of the specific location of the affected PV panels within the centralized PV system. This reduces the workload and stress on PV panel maintenance engineers, improves their efficiency, and ultimately enhances the operational efficiency of the centralized PV system.

[0138] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for detecting light spots in centralized photovoltaic systems, characterized in that, Includes the following steps: The preset first flight path is sent to the target drone so that the target drone can collect images of the centralized photovoltaic area to be identified while flying along the first flight path, and return the collected images to be processed. The system receives and sequentially performs fisheye correction, rotation fine-tuning, and size processing on the image to be processed, outputs an initial image, calls a segmented detection method to perform effective photovoltaic panel detection on the initial image, outputs a photovoltaic panel detection image, and performs waypoint mapping on the photovoltaic panel detection image using a preset waypoint relationship table, specifically: RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output. The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output. The pixel height of the interval region between two adjacent marked regions is compared with a preset height threshold. Two marked regions whose pixel height is less than or equal to the height threshold are merged into one marked region. All marked regions are filtered out and the photovoltaic panel detection image is output. The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch line in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table. The photovoltaic panel detection image is cut, extracted, and located using a fixed-width detection method, and several photovoltaic panel images are output. Specifically, RGB data is extracted from the photovoltaic panel detection image to obtain corresponding second image RGB data. The second image RGB data is compared with RGB threshold data, and corresponding second comparison results are output. The portion of the RGB data of the second image in the second comparison result that is greater than or equal to the RGB threshold data is taken as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked RGB data of the second image to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is cut and extracted according to the complete cutting line, and the several photovoltaic panel images are output. Simultaneously, the waypoint association of the several photovoltaic panel images is confirmed according to the corresponding waypoints of the corresponding routes in the waypoint relationship table. The photovoltaic panels with light spots are identified by performing light spot recognition on the several photovoltaic panel images respectively.

2. The method for detecting light spots in centralized photovoltaic systems as described in claim 1, characterized in that, Before sending the preset first flight path to the target drone, the method further includes: Centralized photovoltaic image acquisition is performed on the centralized photovoltaic area to be identified to obtain a basic photovoltaic image, so that the basic photovoltaic image includes area markers; The photovoltaic panels in the basic photovoltaic image are divided and marked according to the regional markers, and the marked images are associated with UAV waypoints and the flight routes are divided to generate the waypoint relationship table. The waypoint relationship table includes the first route and multiple waypoints along the first route.

3. The method for detecting light spots in centralized photovoltaic systems as described in claim 1, characterized in that, The step of identifying the photovoltaic panel by spot recognition of the plurality of photovoltaic panel images specifically includes: RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold. If it is determined that there exists RGB data that is greater than or equal to the spot RGB threshold, then the pixel blocks corresponding to the RGB data that are greater than or equal to the spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks whose calculated area is greater than or equal to a preset area threshold are determined as spots. The photovoltaic panel with light spots is identified as the light spot photovoltaic panel. The images of the photovoltaic panels are traversed and integrated to obtain the light spot photovoltaic panel. The waypoint corresponding to the light spot photovoltaic panel is sent to the first terminal.

4. A spot detection device for centralized photovoltaic systems, characterized in that, The spot detection device includes an image acquisition module, a photovoltaic panel detection module, and a spot recognition module; The image acquisition module is used to send a preset first flight path to the target UAV, so that the target UAV can acquire images of the centralized photovoltaic area to be identified while flying along the first flight path, and feed back the acquired images to be processed. The photovoltaic panel detection module receives and sequentially performs fisheye correction, rotation fine-tuning, and size processing on the image to be processed, outputs an initial image, calls a segmented detection method to perform effective photovoltaic panel detection on the initial image, outputs a photovoltaic panel detection image, and performs waypoint mapping on the photovoltaic panel detection image through a preset waypoint relationship table, specifically: RGB data is extracted from the initial image to obtain the corresponding first image RGB data. The first image RGB data is compared with preset RGB threshold data, and the corresponding first comparison result is output. The portion of the first image RGB data in the first comparison result that is greater than or equal to the RGB threshold data is taken as the identification region, and several corresponding identification regions are output. The pixel height of the interval region between two adjacent marked regions is compared with a preset height threshold. Two marked regions whose pixel height is less than or equal to the height threshold are merged into one marked region. All marked regions are filtered out and the photovoltaic panel detection image is output. The photovoltaic panel detection image is matched and filtered with the waypoint relationship table so that each photovoltaic panel branch line in the photovoltaic panel detection image corresponds to the route in the waypoint relationship table. The spot recognition module is used to call the fixed width detection method to perform photovoltaic panel cutting extraction and positioning on the photovoltaic panel detection image and output several photovoltaic panel images. Specifically, it extracts RGB data from the photovoltaic panel detection image to obtain corresponding second image RGB data, compares the second image RGB data with RGB threshold data, and outputs corresponding second comparison results. The portion of the RGB data of the second image in the second comparison result that is greater than or equal to the RGB threshold data is taken as the marker line, and the Hough line detection algorithm is called to perform line detection on the marked RGB data of the second image to determine the complete cutting line of the photovoltaic panel detection image. The photovoltaic panel detection image is cut and extracted according to the complete cutting line, and the several photovoltaic panel images are output. Simultaneously, the waypoints of the corresponding routes in the waypoint relationship table are used to confirm the waypoint association of the photovoltaic panel images, and the spot recognition is performed on the photovoltaic panel images to identify the spot photovoltaic panels.

5. The light spot detection device for centralized photovoltaic systems as described in claim 4, characterized in that, The spot detection device also includes a waypoint confirmation module; The waypoint confirmation module is used to collect centralized photovoltaic images of the centralized photovoltaic area to be identified, and obtain a basic photovoltaic image so that the basic photovoltaic image includes area markers. The photovoltaic panels in the basic photovoltaic image are divided and marked according to the regional markers, and the marked images are associated with UAV waypoints and the flight routes are divided to generate the waypoint relationship table. The waypoint relationship table includes the first route and multiple waypoints along the first route.

6. The light spot detection device for centralized photovoltaic systems as described in claim 4, characterized in that, The spot recognition module performs spot recognition on the plurality of photovoltaic panel images to determine the photovoltaic panel with the spot, specifically including: RGB data is extracted from the several photovoltaic panel images in sequence, and the extracted RGB data is compared with a preset spot RGB threshold to determine whether the RGB data is greater than or equal to the spot RGB threshold. If it is determined that there exists RGB data that is greater than or equal to the spot RGB threshold, then the pixel blocks corresponding to the RGB data that are greater than or equal to the spot RGB threshold are selected, the area of ​​several pixel blocks is calculated, and the pixel blocks whose calculated area is greater than or equal to a preset area threshold are determined as spots. The photovoltaic panel with light spots is identified as the light spot photovoltaic panel. The images of the photovoltaic panels are traversed and integrated to obtain the light spot photovoltaic panel. The waypoint corresponding to the light spot photovoltaic panel is sent to the first terminal.

Images