Photovoltaic panel image processing method, device, equipment and medium

Abstract

The invention provides a photovoltaic panel image processing method and device, equipment and a medium, and relates to the photovoltaic field. Performing segmentation operation on the initial image of the photovoltaic power station to obtain a plurality of sub-images; according to the photovoltaic panel identification method and device, due to the fact that the overlapping areas exist between the adjacent sub-images obtained through segmentation, for one photovoltaic panel, the complete photovoltaic panel image can exist in at least one sub-image, and then the accuracy of subsequent photovoltaic panel identification is improved. After the sub-image is obtained, the initial photovoltaic panel area and the initial photovoltaic array area in the sub-image are identified, and the abnormal detection operation is performed on the initial photovoltaic panel area to obtain the target area, so that the accuracy of determining the target area can be improved, and the error identification operation is avoided. And subsequently, after the to-be-processed image of the photovoltaic power station is obtained, the initial photovoltaic array area is utilized to perform photovoltaic panel identification operation on the leak detection area in the to-be-processed image to obtain the target image of the photovoltaic power station, so that the problem of image leak detection can be avoided, and the identification accuracy of the photovoltaic panel in the remote sensing image is improved.

Description

Technical Field

[0001] The present application relates to the photovoltaic field, and more specifically, to a photovoltaic panel image processing method, device, equipment and medium. Background Art

[0002] With the rapid development of photovoltaic power generation technology, the demand for monitoring of large-scale photovoltaic power stations is increasing.

[0003] When monitoring a photovoltaic power station, an image of the photovoltaic power station is first acquired, and photovoltaic panels in the image are identified, so that monitoring operations are performed using the identified photovoltaic panels.

[0004] Currently, when photovoltaic panels in an image are identified, the recognition accuracy of the photovoltaic panels is low, which in turn results in low accuracy in monitoring a photovoltaic power station using the identified photovoltaic panels. Summary of the Invention

[0005] In view of this, the present application provides a photovoltaic panel image processing method, device, equipment and medium to solve the problem of low photovoltaic panel recognition accuracy when identifying photovoltaic panels in an image.

[0006] To solve the above technical problems, this application adopts the following technical solutions:

[0007] A photovoltaic panel image processing method, comprising:

[0008] The initial image of the photovoltaic power station is segmented to obtain multiple sub-images; there are overlapping areas between adjacent sub-images;

[0009] Identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image, performing an anomaly detection operation on the initial photovoltaic panel area to obtain a target area; the target area includes the photovoltaic panel area and the initial photovoltaic array area;

[0010] performing a clustering operation based on the attribute information of the photovoltaic panel area to determine a photovoltaic array image in the initial image;

[0011] performing deduplication and splicing operations on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station;

[0012] The initial photovoltaic array area is used to perform a photovoltaic panel recognition operation on the missed detection area in the image to be processed to obtain a target image of the photovoltaic power station.

[0013] Optionally, the initial image of the photovoltaic power station is segmented to obtain multiple sub-images, including:

[0014] Acquire a remote sensing image of the photovoltaic power station and use the remote sensing image as an initial image;

[0015] Obtaining segmentation parameters; the segmentation parameters include image overlap size and slice size;

[0016] The initial image is segmented according to the segmentation parameters from different preset starting positions to obtain a plurality of sub-images corresponding to each preset starting position.

[0017] Optionally, identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image includes:

[0018] Identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image using an image recognition model;

[0019] A contour recognition operation is performed on the initial photovoltaic panel area to obtain edge information of the initial photovoltaic panel area, and a contour recognition operation is performed on the initial photovoltaic array area to obtain edge information of the initial photovoltaic array area.

[0020] Optionally, performing an anomaly detection operation on the initial photovoltaic panel area to obtain a target area includes:

[0021] performing an anomaly detection operation on edge information of the initial photovoltaic panel area to screen out a photovoltaic panel area from the initial photovoltaic panel area;

[0022] The photovoltaic panel area and the initial photovoltaic array area are used as target areas.

[0023] Optionally, performing a clustering operation based on the attribute information of the photovoltaic panel region to determine the photovoltaic array image in the initial image includes:

[0024] Determining attribute information of the photovoltaic panel area, the attribute information including center point coordinates and vertex coordinates;

[0025] performing a sorting operation on the photovoltaic panel areas by using the attribute information of the photovoltaic panel areas;

[0026] Based on the spatial relationship of the photovoltaic panels and the coordinates of the center points of the photovoltaic panel areas, clustering the sorted photovoltaic panel areas to obtain an initial clustering result; the initial clustering result includes at least one photovoltaic array image; the photovoltaic array image includes the clustered photovoltaic panel areas;

[0027] The initial clustering result is corrected using the vertex coordinates of the photovoltaic panel area to obtain a target clustering result.

[0028] Optionally, performing a deduplication operation and a splicing operation on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station includes:

[0029] performing a labeling operation on the photovoltaic array image in the target clustering result to obtain labeling information of the photovoltaic array image;

[0030] Acquire multiple photovoltaic panel areas corresponding to the same photovoltaic panel in the photovoltaic array image with the same marking information;

[0031] Taking the photovoltaic panel region with the largest area as the actual image of the photovoltaic panel;

[0032] A stitching operation is performed on the actual images of the photovoltaic panels in each of the photovoltaic array images to obtain an image to be processed of the photovoltaic power station.

[0033] Optionally, using the initial photovoltaic array area, performing a photovoltaic panel recognition operation on a missed detection area in the image to be processed to obtain a target image of the photovoltaic power station includes:

[0034] constructing a reference image of the photovoltaic array image using an actual image of the photovoltaic panel in the photovoltaic array image;

[0035] performing a comparison operation on an initial photovoltaic array area corresponding to the photovoltaic array image and a reference image to obtain a missed detection area in the image to be processed;

[0036] Determining a photovoltaic panel reference image from the image to be processed;

[0037] Comparing the photovoltaic panel reference image with the missed detection area to obtain a comparison result;

[0038] When the comparison result shows that the missed-detection area is a photovoltaic panel image, determining edge information of the photovoltaic panel in the missed-detection area;

[0039] A marking operation is performed on the photovoltaic panels in the missed detection area to obtain a target image of the photovoltaic power station.

[0040] A photovoltaic panel image processing device, comprising:

[0041] A segmentation module is used to segment the initial image of the photovoltaic power station to obtain multiple sub-images; there are overlapping areas between adjacent sub-images;

[0042] a region recognition module, configured to recognize an initial photovoltaic panel region and an initial photovoltaic array region in the sub-image, perform an anomaly detection operation on the initial photovoltaic panel region, and obtain a target region; the target region includes the photovoltaic panel region and the initial photovoltaic array region;

[0043] an image determination module, configured to perform a clustering operation based on attribute information of the photovoltaic panel region to determine a photovoltaic array image in the initial image;

[0044] a splicing module, configured to perform a deduplication operation and a splicing operation on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station;

[0045] The missed detection analysis module is used to use the initial photovoltaic array area to perform a photovoltaic panel recognition operation on the missed detection area in the image to be processed to obtain a target image of the photovoltaic power station.

[0046] An electronic device comprising at least one processor and a memory connected to the processor, wherein:

[0047] The memory is used to store computer programs;

[0048] The processor is used to execute the computer program so that the electronic device can implement the above-mentioned photovoltaic panel image processing method.

[0049] A computer storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement the above-mentioned photovoltaic panel image processing method.

[0050] The present application provides a photovoltaic panel image processing method, apparatus, device, and medium. In this application, an initial image of a photovoltaic power station is segmented to obtain multiple sub-images. Because there are overlapping areas between adjacent sub-images obtained by segmentation, a complete photovoltaic panel image can be present in at least one sub-image for a photovoltaic panel, thereby improving the accuracy of subsequent photovoltaic panel identification. After obtaining the sub-images, an initial photovoltaic panel region and an initial photovoltaic array region in the sub-images are identified, and an anomaly detection operation is performed on the initial photovoltaic panel region to obtain a target region. This improves the accuracy of target region determination and avoids misidentification. Subsequently, a clustering operation is performed based on attribute information of the photovoltaic panel region to determine the photovoltaic array image in the initial image. Deduplication and splicing operations are performed on the photovoltaic panel region in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station. Using the initial photovoltaic array region, photovoltaic panel identification operations are performed on missed regions in the to-be-processed image to obtain a target image of the photovoltaic power station. This avoids image missed detection and improves the accuracy of photovoltaic panel identification in remote sensing images. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are merely embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without any creative work.

[0052] Figure 1 This is a flow chart of a photovoltaic panel image processing method according to an embodiment of the present application;

[0053] Figure 2 This is a schematic diagram of image processing of a photovoltaic power station in an embodiment of the present application;

[0054] Figure 3 This is a flow chart of a method for region identification in an embodiment of the present application;

[0055] Figure 4 This is a flow chart of a method for determining clustering results in an embodiment of the present application;

[0056] Figure 5 This is a flowchart of an image processing method in an embodiment of the present application;

[0057] Figure 6 This is a flowchart of another image processing method in an embodiment of the present application;

[0058] Figure 7 Schematic diagram of the structure of a photovoltaic panel image processing device in an embodiment of the present application. DETAILED DESCRIPTION

[0059] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0060] With the rapid development of photovoltaic power generation technology, the demand for monitoring, management and maintenance of large-scale photovoltaic power stations is increasing.

[0061] When monitoring, managing and maintaining photovoltaic power stations, drone mapping, as a non-contact monitoring method, is widely used in the series management and maintenance of photovoltaic panel arrays.

[0062] In specific implementation, remote sensing images of photovoltaic power stations are obtained through drone mapping, photovoltaic panels in the remote sensing images are identified, and photovoltaic panel identification results are obtained. The photovoltaic panel identification results are used to manage and maintain the photovoltaic power station.

[0063] Currently, when identifying photovoltaic panels in remote sensing images, the image false detection rate and missed detection rate are high.

[0064] This is due to:

[0065] 1. Photovoltaic panels located at the edge of the image are easily cut into multiple pieces, which may cause the photovoltaic panels to be missed due to incompleteness.

[0066] 2. The actual installation heights of different photovoltaic panels are different. Photovoltaic panels at different installation heights are easily identified as the same photovoltaic panel due to shooting angles and environmental influences (such as color, noise, light noise, etc.), resulting in false detection and missed detection.

[0067] 3. Most of the photovoltaic power station images taken by drones are spliced. In actual applications, the images are distorted and have errors in the splicing, which makes the detection of areas with splicing distortion and errors invalid, thus causing false detection and missed detection.

[0068] Due to the above reasons, when identifying photovoltaic panels in remote sensing images, the image false detection rate and missed detection rate are high, which in turn makes the management and maintenance of photovoltaic power stations using photovoltaic panel identification results low in accuracy.

[0069] To this end, in an embodiment of the present application, an initial image of a photovoltaic power station is segmented to obtain multiple sub-images. Because adjacent sub-images obtained by segmentation have overlapping areas, a complete image of a photovoltaic panel can be included in at least one sub-image, thereby improving the accuracy of subsequent photovoltaic panel recognition.

[0070] In addition, after obtaining the sub-image, the target area in the sub-image is identified, and the target area includes the photovoltaic panel area. The photovoltaic panel area with abnormal detection is subsequently eliminated to improve the accuracy of photovoltaic panel recognition.

[0071] In addition, after obtaining the photovoltaic panel area, the photovoltaic array image in the initial image is determined based on the attribute information of the photovoltaic panel area, and the photovoltaic panel area in the photovoltaic array image is stitched to obtain the image to be processed of the photovoltaic power station. The photovoltaic panel recognition operation is performed on the missed detection area in the image to be processed, which can avoid the problem of image missed detection and improve the accuracy of photovoltaic panel recognition in remote sensing images.

[0072] An embodiment of the present application provides a photovoltaic panel image processing method, the execution subject can be a processor, controller and other devices of a photovoltaic power station, or a cloud device.

[0073] Reference Figure 1 , a photovoltaic panel image processing method may include:

[0074] S11 . Segment the initial image of the photovoltaic power station to obtain multiple sub-images.

[0075] In this embodiment, drone mapping is a non-contact monitoring method with high reliability. Therefore, in this embodiment of the application, drone mapping can be used to collect images of the photovoltaic power station to obtain an initial image of the photovoltaic power station.

[0076] Generally speaking, remote sensing images obtained using drone mapping often have extremely high resolution, with sizes reaching tens of thousands by tens of thousands of pixels. Directly processing these images requires a large amount of computing resources, resulting in slow processing speeds, low efficiency in processing large-scale remote sensing images, and high memory usage. To this end, in the embodiments of the present application, an initial image of a photovoltaic power station is segmented to obtain multiple sub-images, each with a resolution lower than that of the initial image. Therefore, processing the sub-images is more efficient than processing the initial image, and memory usage is reduced.

[0077] In addition, in the present application, in order to avoid the situation where a photovoltaic panel image is divided into multiple pieces due to environmental factors such as lighting or being located at the edge of the image, in an embodiment of the present application, there is an overlapping area between adjacent sub-images. In this way, for a photovoltaic panel, its complete photovoltaic panel image can exist in at least one sub-image, and thus the photovoltaic panel can be accurately identified, thereby improving the accuracy of photovoltaic panel identification.

[0078] S12: Identify the initial photovoltaic panel area and the initial photovoltaic array area in the sub-image, perform an anomaly detection operation on the initial photovoltaic panel area, and obtain a target area.

[0079] The target area includes the photovoltaic panel area and the initial photovoltaic array area.

[0080] The initial photovoltaic panel area refers to the imaging area of each photovoltaic panel in the sub-image, and the initial photovoltaic array area refers to the imaging area of each photovoltaic array in the sub-image.

[0081] The model can be used to identify the initial photovoltaic panel and array regions within the sub-images. Because the model is trained with a large amount of training data, using the model to identify the target region can improve recognition accuracy. Furthermore, to mitigate inaccurate model recognition results, anomaly analysis can be performed on the target regions identified by the model. Initial photovoltaic panel regions incorrectly identified by the model are eliminated, retaining only those correctly identified.

[0082] S13. Perform a clustering operation based on the attribute information of the photovoltaic panel area to determine the photovoltaic array image in the initial image.

[0083] In this embodiment, after the photovoltaic panel area in the sub-image is identified, corresponding attribute information is provided for each photovoltaic panel, and the attribute information can be coordinate information. Generally, the shape of the photovoltaic panel is rectangular, so the attribute information in the embodiment of the present application can be center point coordinates and vertex coordinates. Among them, vertex coordinates refer to the coordinates of at least one vertex of the photovoltaic panel area. In general embodiments, the coordinates of the four vertices of the photovoltaic panel area can be included. The center point coordinates refer to the coordinates of the center point of the photovoltaic panel area, and the center point coordinates can be determined based on the vertex coordinates.

[0084] In this embodiment, the coordinate system origin of the center point coordinates and vertex coordinates of each photovoltaic panel area is the same coordinate system origin, such as the upper left corner, lower left corner, upper right corner, lower right corner, etc. of the initial image.

[0085] Once the attributes of the photovoltaic panel regions are known, they can be used to identify the individual photovoltaic array images within the initial image. Generally, the center coordinates of photovoltaic panel regions within the same photovoltaic array image tend to cluster, while the vertex coordinates of photovoltaic panel regions in different photovoltaic array images tend to differ significantly. Therefore, this characteristic can be exploited to identify the individual photovoltaic array images within the initial image through clustering.

[0086] S14: performing deduplication and splicing operations on the photovoltaic panel areas in the photovoltaic array image to obtain an image to be processed of the photovoltaic power station.

[0087] After the photovoltaic array image in the initial image is known, it is necessary to determine the actual image of each photovoltaic panel in the photovoltaic array image. The actual image of the photovoltaic panel can be determined using the above-mentioned photovoltaic panel area. After the actual image of the photovoltaic panel is known, the actual image of the photovoltaic panel in the photovoltaic array image is deduplicated and spliced to obtain the image to be processed of the photovoltaic power station.

[0088] S15 , using the initial photovoltaic array area, performing a photovoltaic panel recognition operation on the missed detection area in the image to be processed, and obtaining a target image of the photovoltaic power station.

[0089] In this embodiment, there may be missed detection areas in the image to be processed of the photovoltaic power station. The missed detection areas may be areas in the photovoltaic array image that are not identified as photovoltaic panels. For example, for a certain photovoltaic array image, there is an area in the lower right corner that is not identified as a photovoltaic panel. This area may be a photovoltaic panel or may not be a photovoltaic panel. Therefore, it is necessary to perform photovoltaic panel identification on this area again to determine whether the area is a photovoltaic panel. If it is a photovoltaic panel, it means that the photovoltaic panel has been missed. This step can identify the missed photovoltaic panels and improve the accuracy of photovoltaic panel identification.

[0090] In this embodiment, an initial image of a photovoltaic power station is segmented to obtain multiple sub-images. Since adjacent sub-images obtained by segmentation have overlapping areas, a complete photovoltaic panel image can be present in at least one sub-image for a photovoltaic panel, thereby improving the accuracy of subsequent photovoltaic panel identification. After obtaining the sub-images, the initial photovoltaic panel region and the initial photovoltaic array region in the sub-images are identified, and an anomaly detection operation is performed on the initial photovoltaic panel region to obtain the target region. This can improve the accuracy of target region determination and avoid misidentification operations. Subsequently, a clustering operation is performed based on the attribute information of the photovoltaic panel region to determine the photovoltaic array image in the initial image. The photovoltaic panel region in the photovoltaic array image is deduplicated and spliced to obtain a to-be-processed image of the photovoltaic power station. The initial photovoltaic array region is used to perform photovoltaic panel identification on missed areas in the to-be-processed image to obtain the target image of the photovoltaic power station. This can avoid image missed detection issues and improve the accuracy of photovoltaic panel identification in remote sensing images.

[0091] In addition, the present application automates the entire process from image slicing to obtaining the final image, which can improve processing efficiency.

[0092] Based on the above embodiments, Figure 2 , the initial image of the photovoltaic power station is segmented to obtain multiple sub-images, which may include:

[0093] 1) Obtain a remote sensing image of the photovoltaic power station and use the remote sensing image as the initial image.

[0094] In this embodiment, the image of the photovoltaic power station can be collected by drone mapping, and the collected images can be spliced to obtain a remote sensing image of the photovoltaic power station. The remote sensing image is the initial image in the embodiment of the present application.

[0095] 2) Get the segmentation parameters.

[0096] Generally, the photovoltaic area in a photovoltaic power station is large. Therefore, the initial image obtained by image acquisition of this photovoltaic area is usually very large. If this image is directly processed, it will consume too many computing resources, and there is a possibility that the image cannot be processed due to insufficient computing resources. To this end, in the embodiments of the present application, redundant dynamic overlapping slicing technology is used to cut the large initial image into small image blocks, achieving complete slicing of the large remote sensing image. At the same time, overlapping areas are introduced to solve the problem of edge objects being truncated.

[0097] When segmenting the initial image, segmentation parameters are used, including image overlap size and slice size.

[0098] The image overlap size refers to the size of the overlapping portion between adjacent images, and the slice size refers to the size of each sub-image obtained by segmentation.

[0099] In this embodiment, the slice size can be determined based on the actual size of the initial image, and the image overlap size can be determined based on the slice size and the size of a single photovoltaic array in the initial image. Providing overlapping areas between adjacent slices ensures that each photovoltaic panel is fully imaged in at least one sub-image. Even if a photovoltaic panel located at the edge of the initial image is only partially visible in one sub-image, it will still be fully imaged in at least one other sub-image.

[0100] After the initial image size, slice size and image overlap size are known, the number of slices and slice positions are dynamically calculated to perform the slicing operation.

[0101] 3) The initial image is segmented according to the segmentation parameters from different preset starting positions to obtain a plurality of sub-images corresponding to each preset starting position.

[0102] In this embodiment, in order to accurately identify each photovoltaic panel and to obtain an actual image of the photovoltaic power station by splicing the identified photovoltaic panels, an image redundancy segmentation method is adopted to segment the initial image from multiple preset starting positions.

[0103] In one implementation, the preset starting position may be one or more of the upper left corner, lower left corner, upper right corner, and lower right corner of the initial image.

[0104] In practical applications, after determining the preset starting position, the initial image is segmented from different preset starting positions according to the segmentation parameters to obtain multiple sub-images corresponding to each preset starting position. This ensures that each photovoltaic panel has multiple redundant images, and then the optimal image can be selected from the multiple redundant images as the image of the photovoltaic panel.

[0105] It should be noted that when the image is sliced to the edge, if the remaining image area is less than the size of a slice, blank areas can be added to ensure that a complete slice can be obtained. For example, if the slice size is 10×10, the edge of the initial image, after obtaining the previous sub-image, has a 3×10 portion remaining. At this time, 7×10 blank columns can be added (see Figure 2 The image extension area caused by the slice in the image is removed) to obtain a complete 10×10 slice. When each subsequent row is sliced, the supplementary blank column is used as the initial part of the slice operation. Figure 2 As shown, the width of the slice image finally obtained is greater than the actual image width, and the total height of the slice image is greater than the actual image height.

[0106] In addition, when performing the slicing operation, taking the first row as an example of slicing from left to right, after slicing the first row from left to right, the second row can be switched from right to left, and the third row can be sliced from left to right..., this can ensure the continuity of image segmentation, and each slice has a corresponding slice name, such as slice 1, slice 2, etc.

[0107] In addition, the slicing operation can also be performed only from a certain preset starting position, and the specific number of preset starting positions selected can be based on actual configuration.

[0108] In one implementation, the initial image is segmented from the upper left corner. The schematic diagram of the segmented sub-image can be referred to Figure 2 There is an overlapping area between the adjacent left and right sub-images, and there is also an overlapping area between the upper and lower adjacent sub-images. This can ensure that a photovoltaic panel has a complete image in at least one sub-image as much as possible, and thus can accurately identify the photovoltaic panel.

[0109] In this embodiment, when slicing, in order to ensure that each photovoltaic panel can be detected subsequently and the actual image of the photovoltaic power station can be obtained by splicing the detected photovoltaic panels, multiple of the four corners of the initial image are selected for redundant slicing, so that each photovoltaic panel has a redundant and complete image, avoiding the loss of segmentation detection targets during subsequent visual deep learning inference, ensuring that each photovoltaic panel can be detected subsequently, and improving the accuracy of photovoltaic panel detection.

[0110] In addition, in this application, a unified segmentation parameter is used to segment the initial image to ensure that the obtained sub-images have a uniform size, preparing for the subsequent segmentation of visual deep learning.

[0111] Based on any of the above embodiments, Figure 3 , identifying the initial photovoltaic panel area and the initial photovoltaic array area in the sub-image, which may include:

[0112] S21. Using an image recognition model, identify an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image.

[0113] In this embodiment, since deep learning has a robustness that exceeds that of traditional visual processing algorithms, it can accurately identify photovoltaic panels at different angles in different images while having strong migration capabilities, and can achieve segmentation detection of different photovoltaic panel images. Therefore, in an embodiment of the present application, the image recognition model is a deep learning model, that is, this embodiment uses a deep learning model to perform recognition operations on photovoltaic panel areas. In one implementation, the deep learning model can be a model based on a deep learning segmentation detection algorithm called YOLO (You Only Look Once). Using the YOLO model and the above-mentioned overlapping slicing processing, it can be ensured that the photovoltaic panels in the edge area are fully detected.

[0114] The training samples for the deep learning model in this embodiment are samples with the same image size as the aforementioned sub-images. These training samples include photovoltaic panel images taken at different angles, and the panel regions are labeled for each training sample. Therefore, after training the deep learning model using these training samples, the model can be used to identify photovoltaic panel regions at different angles within an image, adapting to complex scenarios. It should be noted that the deep learning model possesses detection robustness, and therefore can be trained on stacked datasets to cover all future photovoltaic panel types.

[0115] In addition, when using traditional image processing algorithms to identify photovoltaic panels, since the installation spacing of photovoltaic panels is usually only a few centimeters, after being photographed by drones at high altitudes, the images presented are affected by color, noise, lighting, etc., and it is often necessary to manually adjust the parameters to extract the photovoltaic panel area, resulting in low accuracy and efficiency in photovoltaic panel area recognition. To this end, the deep learning model in the embodiment of the present application can be trained using images with different color, noise, lighting conditions, etc., which can improve the accuracy of photovoltaic panel recognition under the influence of color, noise, lighting, etc.

[0116] After the model training is completed, the model is configured into a device that executes the photovoltaic panel image processing method in this application, the model and the sub-images obtained by the above-mentioned slicing operation are loaded, and then a deep learning model is used to perform image inference operations to detect the area of interest in each sub-image, which is each photovoltaic panel area.

[0117] In this embodiment, the photovoltaic panel area obtained through model detection can be referred to as the initial photovoltaic panel area, and the data output by the model is the mask data of the initial photovoltaic panel area.

[0118] It should be noted that, since there are overlapping areas between adjacent sub-images during the above-mentioned slicing operation, there may be multiple initial photovoltaic panel areas for the photovoltaic panels located in the overlapping areas, so that the photovoltaic panels can be accurately identified subsequently, ensuring the detection rate of the initial photovoltaic panel areas of the photovoltaic panels and reducing the missed detection rate.

[0119] In another implementation of the present application, the deep learning model can identify not only the photovoltaic panel area but also the photovoltaic array area. In a specific implementation, an image recognition model can also be used to identify the initial photovoltaic array area in the sub-image.

[0120] In this embodiment, to enable the deep learning model to identify the PV array area, the deep learning model's training samples should include PV array image samples. For the PV array image samples, the training samples are labeled with the PV panel area and the PV array area, allowing the deep learning model to simultaneously identify the PV panel area and the PV array area.

[0121] The sub-image is then input into a deep learning model, and the output of the model includes both the initial photovoltaic panel area and the initial photovoltaic array area.

[0122] S22 , performing a contour recognition operation on the initial photovoltaic panel area to obtain edge information of the initial photovoltaic panel area, and performing a contour recognition operation on the initial photovoltaic array area to obtain edge information of the initial photovoltaic array area.

[0123] In this embodiment, after obtaining the initial photovoltaic panel area, the four vertices of the initial photovoltaic panel area are extracted, and the extracted vertices are used to perform contour approximation processing on the mask data of the detected initial photovoltaic panel area to obtain the circumscribed rectangular frame of the initial photovoltaic panel area. The circumscribed rectangular frame is the edge information of the initial photovoltaic panel area in the embodiment of the present application.

[0124] Similarly, a contour recognition operation is performed on the initial photovoltaic array area to obtain edge information of the initial photovoltaic array area.

[0125] Similar to the above-mentioned initial photovoltaic panel region operation, in this embodiment, a circumscribed polygon of the initial photovoltaic array region is obtained through contour approximation to obtain the photovoltaic array region. The circumscribed polygon in this embodiment represents the edge information of the initial photovoltaic array region, which is used in subsequent missed detection operations.

[0126] Based on this embodiment, after identifying the initial photovoltaic panel area and the initial photovoltaic array area in the sub-image, an abnormality detection operation is performed on the initial photovoltaic panel area to obtain the target area.

[0127] Specifically, an abnormality detection operation is performed on edge information of the initial photovoltaic panel area to screen out the photovoltaic panel area from the initial photovoltaic panel area.

[0128] In this embodiment, since visual detection may have more or less false detection, areas that are not photovoltaic panels may be identified as photovoltaic panels. Therefore, in order to ensure the accuracy of the photovoltaic panel area detected in the embodiment of the present application, the initial photovoltaic panel area output by the model will be verified.

[0129] During verification, you can interactively select any image containing photovoltaic panel data through graphical interaction, manually select the four vertices of one of the photovoltaic panels, determine the size information within the manually selected reference photovoltaic panel area, and then use this size information to filter the initial photovoltaic panel area output by the model. If the size information of the initial photovoltaic panel area is the same or similar to the size information of the user-selected reference photovoltaic panel area, it means that the initial photovoltaic panel area is the correct photovoltaic panel area. If the size information of the initial photovoltaic panel area and the user-selected reference photovoltaic panel area is significantly different, it means that the initial photovoltaic panel area is not a photovoltaic panel area. In this case, the initial photovoltaic panel area that is not a photovoltaic panel area should be eliminated. After the above steps, the true photovoltaic panel area can be filtered out from the initial photovoltaic panel area.

[0130] In addition to the above-mentioned method of manually selecting a reference photovoltaic panel area from any image containing photovoltaic panel data and then determining the size information of the reference photovoltaic panel area, the size information of the photovoltaic panel area can also be directly input manually for subsequent screening operations.

[0131] Subsequently, the photovoltaic panel area and the initial photovoltaic array area are used as target areas.

[0132] In this embodiment, after the photovoltaic panel area and the initial photovoltaic array area are obtained, the photovoltaic panel area and the initial photovoltaic array area are used as target areas.

[0133] In this embodiment, the detection result of the target area of each sub-image is output and saved in the format of a JSON (JavaScript Object Notation) file for subsequent processing.

[0134] In this embodiment, the size information of the reference photovoltaic panel area is obtained through a graphical interactive method or manually input. The size information is used to eliminate falsely detected initial photovoltaic panel areas, thereby improving the accuracy of the detected photovoltaic panel areas. In addition, in this application, this method can also eliminate false detection of photovoltaic panel areas caused by a small number of model training samples or missing samples.

[0135] On the basis of the above embodiment, in another implementation of the present application, referring to Figure 4 , performing a clustering operation based on the attribute information of the photovoltaic panel area to determine the photovoltaic array image in the initial image, which may include:

[0136] S31. Determine attribute information of the photovoltaic panel area.

[0137] The attribute information includes the center point coordinates and the vertex coordinates. For the specific implementation of determining the center point coordinates and the vertex coordinates, please refer to the corresponding instructions above.

[0138] It should be noted that, in this embodiment, the attribute information of each photovoltaic panel region should be calculated and obtained. Even if two photovoltaic panel regions overlap, the attribute information of each photovoltaic panel region should still be obtained.

[0139] S32. Using the attribute information of the photovoltaic panel areas, sort the photovoltaic panel areas.

[0140] In this embodiment, when performing the slicing operation, multiple corners in the initial image are used as preset starting positions, and the initial image is sliced from the preset starting positions to obtain the detection results of the target area corresponding to each preset starting position, that is, for each preset starting position, there is a corresponding photovoltaic panel area and an initial photovoltaic array area.

[0141] In one example, if each of the four corners of the initial image is used as a preset starting position, recognition results of four groups of photovoltaic panel areas and the initial photovoltaic array area will be obtained.

[0142] In this embodiment, the recognition results for all photovoltaic panel regions and the initial photovoltaic array region are collected and sorted. The sorting can be performed based on the relative position of the coordinates of a vertex of the photovoltaic panel region and the starting point, or based on the relative position of the coordinates of the center point of the photovoltaic panel region and the starting point, using a corner of the initial image as the starting point.

[0143] If the same photovoltaic panel appears in multiple recognition results at the same time, theoretically, the actual sorting positions of the multiple photovoltaic panel areas corresponding to the photovoltaic panel should overlap or have slight position differences.

[0144] S33. Based on the spatial relationship of the photovoltaic panels and the coordinates of the center points of the photovoltaic panel areas, a clustering operation is performed on the sorted photovoltaic panel areas to obtain an initial clustering result.

[0145] In this embodiment, the initial clustering result includes at least one photovoltaic array image, and the photovoltaic array image includes the clustered photovoltaic panel area.

[0146] In a specific implementation, the relative positions of photovoltaic panels in the same photovoltaic array are relatively close. Therefore, in this embodiment, the photovoltaic panels can be clustered together through a clustering operation to identify the individual photovoltaic array images in the initial image.

[0147] The clustering algorithm in this embodiment can be a density clustering algorithm. Using the density clustering algorithm, according to the actual spatial relationship between the photovoltaic panels and based on the exponential distance of the center point coordinates of the photovoltaic panel area, the photovoltaic array is automatically identified and grouped through the clustering algorithm to obtain the photovoltaic array image in the initial image.

[0148] S34. Using the vertex coordinates of the photovoltaic panel area, the initial clustering result is corrected to obtain the target clustering result.

[0149] In actual scenarios, two or more photovoltaic arrays may be identified as one photovoltaic array. To avoid this problem, the vertex coordinates of the photovoltaic panel area can be used to perform photovoltaic array correction operations.

[0150] In specific implementations, the vertex coordinates of adjacent photovoltaic panels in the same photovoltaic array are similar or identical. For example, the coordinates of the two right vertices of the photovoltaic panel on the left are the same or similar to the coordinates of the two left vertices of the photovoltaic panel on the right. For adjacent photovoltaic panels in different photovoltaic arrays, the coordinates of the two right vertices of the photovoltaic panel on the left are not similar to the coordinates of the two left vertices of the photovoltaic panel on the right. Therefore, this principle can be used to design a rectangular vertex proximity filtering algorithm, which can split photovoltaic array images that actually belong to two photovoltaic arrays but are identified as the same photovoltaic array in the initial clustering results, so that the obtained target clustering results are closer to the actual photovoltaic array design scenario.

[0151] In this embodiment, the photovoltaic array image is identified by the center point coordinates and vertex coordinates of the photovoltaic panel area, which can avoid identifying adjacent photovoltaic arrays as the same photovoltaic array and improve the accuracy of photovoltaic array image recognition.

[0152] Based on the above embodiments, Figure 5 , performing deduplication and stitching operations on the photovoltaic panel area in the photovoltaic array image to obtain the image to be processed of the photovoltaic power station, which may include:

[0153] S41 . Perform a labeling operation on the photovoltaic array image in the target clustering result to obtain labeling information of the photovoltaic array image.

[0154] In this embodiment, after accurately obtaining the photovoltaic array image in the initial image, the next step is to determine the image of each photovoltaic panel in the photovoltaic array image.

[0155] At this time, the photovoltaic array images in the target clustering results are first labeled in order. For example, each identified photovoltaic array image can be assigned a unique number, which is the label information of the photovoltaic array image.

[0156] S42 , acquiring multiple photovoltaic panel regions corresponding to the same photovoltaic panel in the photovoltaic array image with the same marking information.

[0157] In this embodiment, when processing photovoltaic panels in a photovoltaic array image, the photovoltaic panels located in the same photovoltaic array image should be screened out first. Due to the above-mentioned slicing operation, the same photovoltaic panel may correspond to multiple photovoltaic panel areas. In this embodiment, all photovoltaic panel areas in the photovoltaic array image with the same number are screened out. At this time, one photovoltaic panel generally corresponds to multiple photovoltaic panel areas.

[0158] S43. Taking the photovoltaic panel region with the largest area as the actual image of the photovoltaic panel.

[0159] In this embodiment, since one photovoltaic panel generally corresponds to multiple photovoltaic panel areas, it is necessary to select one photovoltaic panel area from the multiple photovoltaic panel areas as the actual image of the photovoltaic panel.

[0160] When performing the specific screening operation, the photovoltaic panel region with the largest area may be used as the actual image of the photovoltaic panel, so as to ensure that the photovoltaic panel occupies the largest area.

[0161] For the same photovoltaic panel, when screening the photovoltaic panel area with the largest area, all photovoltaic panel areas in the photovoltaic array image with the same number can be sorted according to the coordinates of the four vertices of the rectangle, and then multiple photovoltaic panel areas with similar center point coordinates are screened out. These photovoltaic panel areas are the photovoltaic panel areas corresponding to the same photovoltaic panel. For these photovoltaic panel areas, the photovoltaic panel area with the largest area rectangle is retained, and other photovoltaic panel areas are filtered out to obtain the actual image of the photovoltaic panel.

[0162] S44 , performing a splicing operation on the actual images of the photovoltaic panels in each photovoltaic array image to obtain an image to be processed of the photovoltaic power station.

[0163] In this embodiment, for each photovoltaic array image, the actual images of the photovoltaic panels are spliced according to their original positions to obtain a processed image of the photovoltaic power station. The processed image can be saved in JSON format.

[0164] In this embodiment, duplicate rectangles are filtered out using the center points of the photovoltaic panel areas. Multiple overlapping photovoltaic panel areas corresponding to the same photovoltaic panel are then deduplicated based on their area. This results in an actual image of the photovoltaic panel with the largest area. This avoids inaccurate image capture of the photovoltaic panel area due to occlusions and other factors during actual image acquisition, making the actual image of the photovoltaic panel more accurate for the actual scene design. Furthermore, the actual images of the photovoltaic panels in each photovoltaic array image are spliced together to obtain the processed image of the photovoltaic power station, achieving a transition from a single photovoltaic panel to a photovoltaic array.

[0165] Based on the above embodiments, Figure 6 , using the initial photovoltaic array area, the photovoltaic panel recognition operation is performed on the missed detection area in the image to be processed to obtain the target image of the photovoltaic power station, which may include:

[0166] S51 . Construct a reference image of the photovoltaic array image using the actual image of the photovoltaic panel in the photovoltaic array image.

[0167] In this embodiment, for a photovoltaic array area, after obtaining the actual images of each photovoltaic panel in the photovoltaic array area, the stitching result of the actual images of the photovoltaic panels in each photovoltaic array image in the photovoltaic array area is the reference image of the photovoltaic array image. The shape of the reference image can be a polygon, and the reference image can be subsequently displayed graphically.

[0168] S52 : performing a comparison operation on the initial photovoltaic array area corresponding to the photovoltaic array image and the reference image to obtain a missed detection area in the image to be processed.

[0169] In this embodiment, the initial photovoltaic array region corresponding to the photovoltaic array image is identified using the aforementioned deep learning model, and the reference image corresponding to the photovoltaic array image is obtained by stitching the actual images of the photovoltaic panels within the photovoltaic array image. In theory, the two should overlap or nearly overlap. However, in real-world scenarios, the photovoltaic panels may not be accurately identified, resulting in a low degree of overlap. In this case, this indicates that there are areas of missed detection in the processed image.

[0170] To this end, in the embodiment of the present application, a contour search algorithm based on OpenCV (Open Source Computer Vision Library) is used to compare the initial photovoltaic array area corresponding to the photovoltaic array image with the overlapping area of the reference image to obtain the missed detection area that exists in the initial photovoltaic array area but does not exist in the image to be processed. The missed detection area can be referred to Figure 2 The area where the missed photovoltaic panels are located.

[0171] S53: Determine a photovoltaic panel reference image from the image to be processed.

[0172] In this embodiment, a photovoltaic panel can be randomly selected from the image to be processed, and the image of the photovoltaic panel is the photovoltaic panel reference image in the embodiment of the present application. The selected photovoltaic panel reference image, such as the RGB image, can be cropped for subsequent comparison operations.

[0173] S54: performing a comparison operation on the photovoltaic panel reference image and the missed detection area to obtain a comparison result.

[0174] In this embodiment, a template matching algorithm based on OpenCV is used to map the cropped photovoltaic panel reference image to the missed detection area, and to find out whether there is a single photovoltaic panel in the missed detection area to obtain a comparison result.

[0175] S55 . When the comparison result shows that the missed-detection area is a photovoltaic panel image, determine edge information of the photovoltaic panel in the missed-detection area.

[0176] In this embodiment, when the comparison result shows that the missed detection area is a photovoltaic panel image, it means that at least one photovoltaic panel is missed here. Based on the contour approximation algorithm, the circumscribed rectangle of each photovoltaic panel in the missed detection area is obtained to obtain the edge information of the photovoltaic panel.

[0177] S56 , marking the photovoltaic panels in the missed inspection area to obtain a target image of the photovoltaic power station.

[0178] In this embodiment, the photovoltaic panels in the photovoltaic array image have photovoltaic panel numbers, which can be 1, 2, 3, etc. In this embodiment, it is necessary to supplement the numbers of the missed areas and renumber each photovoltaic panel in the photovoltaic array image according to the arrangement order of the photovoltaic panels. The specific numbering results can be referred to Figure 2 The photovoltaic panel numbers in an array area shown are shown. The array area is a photovoltaic array image.

[0179] After the numbering is completed, the interface obtains the target image of the photovoltaic power station and interactively displays the target image to present the final processing result of the photovoltaic power station image to the operator. The target image can then be used for operations such as photovoltaic panel maintenance.

[0180] In this embodiment, by performing a missed detection operation on the image to be processed, the detection rate of the photovoltaic panels in the image can be guaranteed, thereby improving the recognition accuracy of the photovoltaic panels.

[0181] In summary, this application utilizes redundant dynamic overlapping slicing technology to achieve complete slicing of large remote sensing images. Deep learning segmentation detection technology is used to identify photovoltaic panel and array regions. Based on the spatial relationship between photovoltaic panels and the exponential distance of the center point coordinates of the photovoltaic panels, a density clustering algorithm and a rectangular vertex proximity filtering algorithm are used to group and identify photovoltaic arrays. Traditional image contour extraction and template matching techniques are used to detect missed areas, thereby achieving overlapping and complete extraction of individual photovoltaic panels in the photovoltaic array and obtaining the final image. The synergy between these technologies achieves optimal detection results.

[0182] Based on the embodiment of the photovoltaic panel image processing method described above, another embodiment of the present application provides a photovoltaic panel image processing device, referring to Figure 7 ,include:

[0183] A segmentation module 11 is used to segment the initial image of the photovoltaic power station to obtain multiple sub-images; there are overlapping areas between adjacent sub-images;

[0184] A region recognition module 12 is configured to recognize an initial photovoltaic panel region and an initial photovoltaic array region in the sub-image, perform an anomaly detection operation on the initial photovoltaic panel region, and obtain a target region; the target region includes the photovoltaic panel region and the initial photovoltaic array region;

[0185] An image determination module 13 is configured to perform a clustering operation based on the attribute information of the photovoltaic panel area to determine the photovoltaic array image in the initial image;

[0186] A stitching module 14 is used to perform a deduplication operation and a stitching operation on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station;

[0187] The missed detection analysis module 15 is configured to use the initial photovoltaic array area to perform photovoltaic panel recognition operations on the missed detection areas in the image to be processed, thereby obtaining a target image of the photovoltaic power station.

[0188] In one implementation, the segmentation module 11 may include:

[0189] An image acquisition submodule is used to acquire a remote sensing image of the photovoltaic power station and use the remote sensing image as an initial image;

[0190] The parameter acquisition submodule is used to obtain segmentation parameters; segmentation parameters include image overlap size and slice size;

[0191] The sub-slicing module is used to perform a segmentation operation on the initial image from different preset starting positions according to the segmentation parameters to obtain multiple sub-images corresponding to each preset starting position.

[0192] In one implementation, the region identification module 12 includes:

[0193] A first identification submodule is used to identify an initial photovoltaic panel area and an initial photovoltaic array area in a sub-image using an image recognition model;

[0194] The second recognition submodule is configured to perform contour recognition on the initial photovoltaic panel area to obtain edge information of the initial photovoltaic panel area, and to perform contour recognition on the initial photovoltaic array area to obtain edge information of the initial photovoltaic array area.

[0195] In one implementation, the region identification module 12 includes:

[0196] An anomaly detection submodule is used to perform an anomaly detection operation on the edge information of the initial photovoltaic panel area to screen out the photovoltaic panel area from the initial photovoltaic panel area;

[0197] The region determination submodule is used to take the photovoltaic panel region and the initial photovoltaic array region as the target region.

[0198] In one implementation, the image determination module 13 includes:

[0199] An information determination submodule is used to determine the attribute information of the photovoltaic panel area, the attribute information including the center point coordinates and the vertex coordinates;

[0200] The sorting submodule is used to sort the photovoltaic panel areas using the attribute information of the photovoltaic panel areas;

[0201] A clustering submodule is configured to perform a clustering operation on the sorted photovoltaic panel areas based on the spatial relationship of the photovoltaic panels and the coordinates of the center points of the photovoltaic panel areas to obtain an initial clustering result; the initial clustering result includes at least one photovoltaic array image; the photovoltaic array image includes the clustered photovoltaic panel areas;

[0202] The correction submodule is used to use the vertex coordinates of the photovoltaic panel area to perform correction operations on the initial clustering results to obtain the target clustering results.

[0203] In one implementation, the splicing module 14 includes:

[0204] The marking submodule is used to mark the photovoltaic array image in the target clustering result to obtain the marking information of the photovoltaic array image;

[0205] The region acquisition submodule is used to acquire multiple photovoltaic panel regions corresponding to the same photovoltaic panel in the photovoltaic array image with the same marking information;

[0206] A first image determination submodule is configured to use the photovoltaic panel region with the largest area as the actual image of the photovoltaic panel;

[0207] The stitching submodule is used to stitch the actual images of the photovoltaic panels in each photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station.

[0208] In one implementation, the missed detection analysis module 15 includes:

[0209] An image construction submodule is used to construct a reference image of the photovoltaic array image using an actual image of the photovoltaic panel in the photovoltaic array image;

[0210] A first comparison submodule is used to compare the initial photovoltaic array area corresponding to the photovoltaic array image with the reference image to obtain a missed detection area in the image to be processed;

[0211] A second image determination submodule, configured to determine a photovoltaic panel reference image from the image to be processed;

[0212] The second comparison submodule is used to compare the photovoltaic panel reference image with the missed detection area to obtain a comparison result;

[0213] an edge information determination submodule, for determining edge information of the photovoltaic panels in the missed detection area when the comparison result shows that the missed detection area is a photovoltaic panel image;

[0214] The third image determination submodule is used to determine to perform a marking operation on the photovoltaic panels in the missed inspection area to obtain a target image of the photovoltaic power station.

[0215] In this embodiment, in this application, an initial image of a photovoltaic power station is segmented to obtain multiple sub-images. Since there are overlapping areas between adjacent sub-images obtained by segmentation, a complete photovoltaic panel image can be present in at least one sub-image for a photovoltaic panel, thereby improving the accuracy of subsequent photovoltaic panel recognition. After obtaining the sub-images, the initial photovoltaic panel region and the initial photovoltaic array region in the sub-image are identified, and an anomaly detection operation is performed on the initial photovoltaic panel region to obtain the target region. This can improve the accuracy of target region determination and avoid misidentification operations. Subsequently, a clustering operation is performed based on the attribute information of the photovoltaic panel region to determine the photovoltaic array image in the initial image. The photovoltaic panel region in the photovoltaic array image is deduplicated and spliced to obtain a to-be-processed image of the photovoltaic power station. The initial photovoltaic array region is used to perform photovoltaic panel recognition operations on missed areas in the to-be-processed image to obtain a target image of the photovoltaic power station. This can avoid image missed detection problems and improve the accuracy of photovoltaic panel recognition in remote sensing images.

[0216] It should be noted that, for the working process of each module and sub-module in this embodiment, please refer to the corresponding description in the above embodiment, which will not be repeated here.

[0217] An embodiment of the present application further provides an electronic device, including at least one processor and a memory connected to the processor, wherein:

[0218] Memory is used to store computer programs;

[0219] The processor is used to execute the computer program so that the electronic device can implement the above-mentioned photovoltaic panel image processing method.

[0220] An embodiment of the present application also provides a computer program product including computer-readable instructions. When the computer-readable instructions are executed on an electronic device, the electronic device implements any photovoltaic panel image processing method provided in the embodiment of the present application.

[0221] A computer-readable storage medium is also provided in an embodiment of the present application. The storage medium carries one or more computer programs. When the one or more computer programs are executed by an electronic device, the electronic device can implement any photovoltaic panel image processing method provided in the embodiment of the present application.

[0222] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A photovoltaic panel image processing method, characterized in that: include: The initial image of the photovoltaic power station is segmented to obtain multiple sub-images; there are overlapping areas between adjacent sub-images; Identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image, performing an anomaly detection operation on the initial photovoltaic panel area to obtain a target area; the target area includes the photovoltaic panel area and the initial photovoltaic array area; performing a clustering operation based on the attribute information of the photovoltaic panel area to determine a photovoltaic array image in the initial image; performing deduplication and splicing operations on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station; The initial photovoltaic array area is used to perform a photovoltaic panel recognition operation on the missed detection area in the image to be processed to obtain a target image of the photovoltaic power station.

2. The photovoltaic panel image processing method according to claim 1, characterized in that: The initial image of the photovoltaic power station is segmented to obtain multiple sub-images, including: Acquire a remote sensing image of the photovoltaic power station and use the remote sensing image as an initial image; Obtaining segmentation parameters; the segmentation parameters include image overlap size and slice size; The initial image is segmented according to the segmentation parameters from different preset starting positions to obtain a plurality of sub-images corresponding to each preset starting position.

3. The photovoltaic panel image processing method according to claim 1, characterized in that: Identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image includes: Identifying an initial photovoltaic panel area and an initial photovoltaic array area in the sub-image using an image recognition model; A contour recognition operation is performed on the initial photovoltaic panel area to obtain edge information of the initial photovoltaic panel area, and a contour recognition operation is performed on the initial photovoltaic array area to obtain edge information of the initial photovoltaic array area.

4. The photovoltaic panel image processing method according to claim 3, characterized in that: An anomaly detection operation is performed on the initial photovoltaic panel area to obtain a target area, including: performing an anomaly detection operation on edge information of the initial photovoltaic panel area to screen out a photovoltaic panel area from the initial photovoltaic panel area; The photovoltaic panel area and the initial photovoltaic array area are used as target areas.

5. The photovoltaic panel image processing method according to claim 4, characterized in that: Performing a clustering operation based on the attribute information of the photovoltaic panel area to determine a photovoltaic array image in the initial image includes: Determining attribute information of the photovoltaic panel area, the attribute information including center point coordinates and vertex coordinates; performing a sorting operation on the photovoltaic panel areas by using the attribute information of the photovoltaic panel areas; Based on the spatial relationship of the photovoltaic panels and the coordinates of the center points of the photovoltaic panel areas, clustering the sorted photovoltaic panel areas to obtain an initial clustering result; the initial clustering result includes at least one photovoltaic array image; the photovoltaic array image includes the clustered photovoltaic panel areas; The initial clustering result is corrected using the vertex coordinates of the photovoltaic panel area to obtain a target clustering result.

6. The photovoltaic panel image processing method according to claim 5, characterized in that: Performing a deduplication operation and a splicing operation on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station, including: performing a labeling operation on the photovoltaic array image in the target clustering result to obtain labeling information of the photovoltaic array image; Acquire multiple photovoltaic panel areas corresponding to the same photovoltaic panel in the photovoltaic array image with the same marking information; Taking the photovoltaic panel region with the largest area as the actual image of the photovoltaic panel; A stitching operation is performed on the actual images of the photovoltaic panels in each of the photovoltaic array images to obtain an image to be processed of the photovoltaic power station.

7. The photovoltaic panel image processing method according to claim 6, characterized in that: Using the initial photovoltaic array area, performing a photovoltaic panel recognition operation on the missed detection area in the image to be processed to obtain a target image of the photovoltaic power station, including: constructing a reference image of the photovoltaic array image using an actual image of the photovoltaic panel in the photovoltaic array image; performing a comparison operation on an initial photovoltaic array area corresponding to the photovoltaic array image and a reference image to obtain a missed detection area in the image to be processed; Determining a photovoltaic panel reference image from the image to be processed; Comparing the photovoltaic panel reference image with the missed detection area to obtain a comparison result; When the comparison result shows that the missed-detection area is a photovoltaic panel image, determining edge information of the photovoltaic panel in the missed-detection area; A marking operation is performed on the photovoltaic panels in the missed detection area to obtain a target image of the photovoltaic power station.

8. A photovoltaic panel image processing device, characterized in that: include: A segmentation module is used to segment the initial image of the photovoltaic power station to obtain multiple sub-images; there are overlapping areas between adjacent sub-images; a region recognition module, configured to recognize an initial photovoltaic panel region and an initial photovoltaic array region in the sub-image, perform an anomaly detection operation on the initial photovoltaic panel region, and obtain a target region; the target region includes the photovoltaic panel region and the initial photovoltaic array region; an image determination module, configured to perform a clustering operation based on attribute information of the photovoltaic panel region to determine a photovoltaic array image in the initial image; a splicing module, configured to perform a deduplication operation and a splicing operation on the photovoltaic panel area in the photovoltaic array image to obtain a to-be-processed image of the photovoltaic power station; The missed detection analysis module is used to use the initial photovoltaic array area to perform a photovoltaic panel recognition operation on the missed detection area in the image to be processed to obtain a target image of the photovoltaic power station.

9. An electronic device, characterized in that: comprising at least one processor and a memory connected to the processor, wherein: The memory is used to store computer programs; The processor is configured to execute the computer program so that the electronic device can implement the photovoltaic panel image processing method according to any one of claims 1 to 7.

10. A computer storage medium, characterized in that The storage medium carries one or more computer programs, and when the one or more computer programs are executed by an electronic device, the electronic device can implement the photovoltaic panel image processing method according to any one of claims 1 to 7.

Images