Method, device and equipment for detecting dust coverage degree of photovoltaic panel and medium

Abstract

The invention provides a photovoltaic panel dust coverage degree detection method, device, equipment and medium, and the method comprises the steps: collecting a photovoltaic panel image, carrying out the preprocessing, and dividing the image into a training set, a verification set and a test set; an improved YOLOv11 segmentation model is utilized to segment the photovoltaic panel image to obtain a photovoltaic panel area image, and SCConv convolution is utilized to construct the improved YOLOv11 segmentation model; and classifying the photovoltaic panel area images by using a YOLOv11 classification model to obtain dust coverage degrees of different photovoltaic panels so as to solve the problem that a photovoltaic power generation device is damaged due to low detection precision of the dust coverage degrees on the photovoltaic panels and untimely dust cleaning.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic panel cleaning and inspection, and in particular to a method, device, equipment and medium for detecting the degree of dust coverage of a photovoltaic panel. Background Art

[0002] Dust covering photovoltaic panels will lead to a series of chain problems. The first is to reduce the photovoltaic panel's photoelectric conversion efficiency. The dust layer blocks part of the sunlight, making it impossible for the photovoltaic panel to fully absorb solar energy, thereby reducing the generation of electricity. Dust adhering to photovoltaic panels will cause uneven temperature of photovoltaic panels, increase the thermal load of photovoltaic panels, and thus accelerate aging. In addition, long-term dust accumulation may also cause surface scratches or damage, further reducing the service life of photovoltaic panels. Therefore, how to monitor the dust on the surface of photovoltaic panels has always been a key issue in the photovoltaic power generation industry.

[0003] Chinese Patent 202210243083.1 (A photovoltaic panel dust detection method based on color analysis) uses a dust-developing lamp to illuminate the photovoltaic panel and obtain an image. Since the dust-covered area in the image has a significant color difference with the photovoltaic panel, the dust and its covering thickness can be quickly identified in the HSV (Hue, Saturation, Value) color space. Although this method is low-cost, it also has some disadvantages. First, changes in illumination still have a significant impact on color analysis. Especially in the case of uneven illumination or strong reflection, the color difference between dust and photovoltaic panels may become less obvious, resulting in recognition errors. Secondly, when the HSV color space judges areas with similar colors, if the background and dust colors are close, it may lead to misjudgment. Furthermore, although the deletion of background noise can remove interference, in a complex environment, noise and dust may have similar color characteristics, resulting in mistaken deletion or omission of dust areas. Finally, the judgment of dust thickness depends only on the brightness and color area of ​​the developing lamp, and cannot be accurately evaluated when the dust is unevenly distributed or thin.

[0004] Chinese patent 202410760868.5 (Photovoltaic panel dust detection method, device, equipment and storage medium) detects dust on photovoltaic panels based on an improved YOLOv8 model. Although this method has good robustness, since dust generally covers the entire photovoltaic panel, when using Labelimg software to mark dust targets, the entire photovoltaic panel needs to be framed, which will cause the dust detection accuracy to be interfered by other foreign objects. In addition, YOLOv8 target detection has poor detection effect on the thickness of dust coverage on photovoltaic panels and cannot accurately identify the degree of dust coverage on photovoltaic panels.

[0005] Therefore, there is an urgent need to propose a method for detecting the dust coverage degree of photovoltaic panels to solve the technical problem that the photovoltaic power generation device is damaged due to the low detection accuracy of the dust coverage degree on the photovoltaic panels and the untimely cleaning of dust. Summary of the invention

[0006] In order to overcome the problems existing in the related art, the present disclosure provides a method, device, equipment and medium for detecting the degree of dust coverage on photovoltaic panels, so as to solve the technical problem in the related art that the photovoltaic power generation device is damaged due to the low detection accuracy of the degree of dust coverage on the photovoltaic panels and the untimely cleaning of dust.

[0007] One or more embodiments of this specification provide a method for detecting dust coverage of a photovoltaic panel, comprising the following steps:

[0008] Collect photovoltaic panel images and divide them into training set, validation set and test set after preprocessing;

[0009] The photovoltaic panel image is segmented using an improved YOLOv11 segmentation model to obtain a photovoltaic panel region image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution;

[0010] The photovoltaic panel area image is classified using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels.

[0011] Preferably, the method further comprises the following steps:

[0012] A hierarchical complementary attention mechanism is embedded in the feature fusion network layer of the YOLOv11 classification model.

[0013] Preferably, the method of using the YOLOv11 classification model to classify the photovoltaic panel area image to obtain the dust coverage degree of different photovoltaic panels specifically includes the following steps:

[0014] Performing front view correction on the photovoltaic panel area image by using a Blob analysis algorithm and a perspective transformation algorithm;

[0015] The rectified photovoltaic panel area images are classified according to the dust coverage to obtain a classification data set;

[0016] The classification data set is used to train a YOLOv11 classification model;

[0017] According to the YOLOv11 classification model, the dust coverage degree of different photovoltaic panels can be detected.

[0018] Preferably, the step of segmenting the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel area image specifically includes the following steps:

[0019] Utilize SCConv convolution combined with C3k2 module to construct C3k2_SCConv module, replace C3k2 module in YOLOv11 segmentation model with C3k2_SCConv module, and construct improved YOLOv11 segmentation model;

[0020] Using the training set to train the improved YOLOv11 segmentation model and perform weight optimization;

[0021] Segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel mask;

[0022] The photovoltaic panel mask is operated on the photovoltaic panel image to obtain a photovoltaic panel area image.

[0023] One or more embodiments of this specification provide a photovoltaic panel dust coverage degree detection device, including a collection module, a segmentation module and a classification module;

[0024] The acquisition module is used to acquire photovoltaic panel images and divide them into a training set, a validation set and a test set after preprocessing;

[0025] The segmentation module is used to segment the photovoltaic panel image using an improved YOLOv11 segmentation model to obtain a photovoltaic panel area image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution;

[0026] The classification module is used to classify the photovoltaic panel area image using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels.

[0027] Preferably, the classification module comprises an embedding unit for embedding a hierarchical complementary attention mechanism in a feature fusion network layer of the YOLOv11 classification model.

[0028] Preferably, the classification module further includes a correction unit, a classification unit, a first training unit and a detection unit;

[0029] The correction unit is used to perform front view correction on the photovoltaic panel area image by using a Blob analysis algorithm and a perspective transformation algorithm;

[0030] The classification unit is used to classify the corrected photovoltaic panel area image according to the dust coverage degree to obtain a classification data set;

[0031] The first training unit is used to train the classification data set to obtain a YOLOv11 classification model;

[0032] The detection unit is used to detect the dust coverage degree of different photovoltaic panels according to the YOLOv11 classification model.

[0033] Preferably, the segmentation module includes an improvement unit, a second training unit, a segmentation unit and a calculation unit;

[0034] The improvement unit is used to use SCConv convolution in combination with the C3k2 module to construct a C3k2_SCConv module, replace the C3k2 module in the YOLOv11 segmentation model with the C3k2_SCConv module, and construct an improved YOLOv11 segmentation model;

[0035] The second training unit is used to train the improved YOLOv11 segmentation model using the training set and perform weight optimization;

[0036] The segmentation unit is used to segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel mask;

[0037] The operation unit is used to operate the photovoltaic panel mask and the photovoltaic panel image to obtain a photovoltaic panel area image.

[0038] One or more embodiments of the present specification provide a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned method for detecting the degree of dust coverage of photovoltaic panels when executing the computer program.

[0039] One or more embodiments of the present specification provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the above-mentioned method for detecting the degree of dust coverage of photovoltaic panels.

[0040] The present invention provides a method, device, equipment and medium for detecting the dust coverage degree of photovoltaic panels. The advantages are that by collecting photovoltaic panel images and dividing them into training sets, verification sets and test sets after preprocessing, it is helpful for effective training and evaluation of the model; the photovoltaic panel images are segmented by an improved YOLOv11 segmentation model to obtain photovoltaic panel area images, wherein the improved YOLOv11 segmentation model is constructed by using SCConv convolution, which can enhance the model's ability to extract photovoltaic panel image features, accurately obtain photovoltaic panel area images, and provide high-quality data for subsequent classification; the photovoltaic panel area images are classified by using the YOLOv11 classification model to obtain the dust coverage degrees of different photovoltaic panels, and the operation and maintenance personnel can reasonably arrange cleaning work according to the dust coverage degrees, thereby improving the overall power generation efficiency of the photovoltaic power station, increasing power generation, and improving economic benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate one or more embodiments of this specification or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this specification. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0042] Figure 1 A schematic flow chart of a method for detecting dust coverage of a photovoltaic panel provided in one or more embodiments of this specification;

[0043] Figure 2 A hardware schematic diagram of a photovoltaic panel dust coverage degree detection method provided in one or more embodiments of this specification;

[0044] Figure 3 A schematic diagram of annotated images provided for one or more embodiments of this specification;

[0045] Figure 4 A schematic diagram of the structure of the HRAMi attention mechanism provided for one or more embodiments of this specification;

[0046] Figure 5 A schematic diagram of detection results obtained by using an improved YOLOv11 classification model provided in one or more embodiments of this specification;

[0047] Figure 6 A schematic diagram of the structure of a C3k2 module provided for one or more embodiments of this specification;

[0048] Figure 7 A schematic diagram of the structure of the C3k2 SCConv module provided for one or more embodiments of this specification; Figure 8 A network structure diagram of the SCConv module provided for one or more embodiments of this specification;

[0049] Fig. 9 A schematic diagram of photovoltaic panel image segmentation and conversion results provided by one or more embodiments of this specification;

[0050] Fig.10 A general implementation flow chart of photovoltaic panel dust coverage detection based on the YOLOv11 segmentation and classification model provided for one or more embodiments of this specification;

[0051] Fig.11 A schematic diagram of the structure of a photovoltaic panel dust coverage degree detection device provided in one or more embodiments of this specification;

[0052] Fig.12A schematic diagram of the structure of a computer device provided for one or more embodiments of this specification. DETAILED DESCRIPTION

[0053] In order to enable those skilled in the art to better understand the technical solutions in one or more embodiments of this specification, the following will be combined with the drawings in one or more embodiments of this specification to clearly and completely describe the technical solutions in one or more embodiments of this specification. Obviously, the described embodiments are only part of the embodiments of this specification, not all of the embodiments. Based on one or more embodiments of this specification, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the protection scope of the present invention document.

[0054] The present invention is described in detail below in conjunction with specific implementation methods and the accompanying drawings.

[0055] Method Embodiment

[0056] According to an embodiment of the present invention, a method for detecting dust coverage of a photovoltaic panel is provided. Figure 1 As shown in FIG. 1 , it is a flow chart of the photovoltaic panel dust coverage degree detection method provided in this embodiment, which includes two parts: hardware and algorithm. The hardware components are as follows: Figure 2 As shown, the system collects on-site working images of the photovoltaic panel cleaning robot through three monitoring cameras installed above the photovoltaic panel cleaning robot, and transmits the image data to the background server through the AP (Access Point, wireless access point) device. After the image data undergoes instance segmentation, front view correction and classification, dust and coverage can be detected, and finally the detection results are displayed on the software interface. The photovoltaic panel dust coverage detection method according to an embodiment of the present invention includes the following steps:

[0057] S110, collecting photovoltaic panel images, and dividing them into a training set, a validation set, and a test set after preprocessing.

[0058] Specifically, the photovoltaic panel cleaning robot is started, and the monitoring camera is controlled by the background server to capture images according to the set shooting interval during the movement of the photovoltaic panel cleaning robot, so as to collect solar photovoltaic panel images. If the length of each photovoltaic panel at the collection site is S, there are N panels in total, the forward movement speed of the cleaning robot is V, and the set capture time is T, then the number of images obtained is NS / TV. 800 photovoltaic panel images under different weather conditions and viewing angles are collected through the camera on the photovoltaic cleaning robot. In order to improve the generality of the segmentation model, the data set is expanded by flipping, rotating, cropping, scaling, and translating. The final data set contains 1500 images.

[0059] Use Labelme to annotate 1500 images. The annotated images are as follows: Figure 3 As shown in the figure, the photovoltaic panel area is enclosed by multi-coordinate point annotation. After the annotation is completed, the 1500 images are divided into training set, validation set and test set according to 7:2:1. The training set is used to train the YOLOv11 segmentation model, and the validation and test sets are used to verify and test the segmentation performance of the YOLOv11 model.

[0060] S120, improving and training the YOLOv11 segmentation model, and using the improved YOLOv11 segmentation model to segment the photovoltaic panel image to obtain a photovoltaic panel area image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution.

[0061] S130, using the YOLOv11 classification model to classify the photovoltaic panel area image to obtain dust coverage degrees of different photovoltaic panels.

[0062] The method provided in this embodiment collects photovoltaic panel images and divides them into training sets, validation sets and test sets after preprocessing, which is conducive to the effective training and evaluation of the model; the photovoltaic panel images are segmented using the improved YOLOv11 segmentation model to obtain photovoltaic panel area images, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution, which can enhance the model's ability to extract photovoltaic panel image features, accurately obtain photovoltaic panel area images, and provide high-quality data for subsequent classification; the photovoltaic panel area images are classified using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels, and the operation and maintenance personnel can reasonably arrange the cleaning work according to the dust coverage degree, thereby improving the overall power generation efficiency of the photovoltaic power station, increasing the power generation, and improving the economic benefits.

[0063] In one embodiment, the following steps are also included:

[0064] A hierarchical complementary attention mechanism is embedded in the feature fusion network layer of the YOLOv11 classification model.

[0065] Since the image features of photovoltaic panels with different dust coverage are not very different, the classification effect of the YOLOv11 benchmark classification model is not ideal, and it needs to be improved to improve the classification accuracy of the model. By introducing the Hierarchical Complementary Attention Mechanism (HRAMi), the feature selection and expression ability of the model are improved, so that photovoltaic panels with different dust coverage can be more accurately distinguished. The structure of HRAMi usually consists of multiple levels of attention modules, each of which extracts and fuses information for different feature levels. Its basic architecture includes the following key parts: First, the input data will be processed through multiple levels, and each layer focuses on different levels of details, such as local features, global information, contextual relationships, etc. The attention modules of each layer adopt different strategies, such as self-attention or cross-attention, to capture semantic information of different granularity. Then, the layers interact with each other through complementary mechanisms, that is, the local features of the lower layer and the global information of the higher layer can complement each other, enhancing information flow and feature integration. Finally, these levels of attention modules integrate the attention information from different levels into a global feature representation through a weighted fusion strategy, further improving the representation ability of the model when processing complex tasks. Figure 4 The figure shows the structural diagram of the HRAMi attention mechanism.

[0066] The HRAMi attention mechanism is embedded into the feature fusion network of the YOLOv11 classification model. Figure 4 The input of Stage 1-3 in the figure is the feature image after Concat in the feature fusion network, which has three different sizes. After the HRAMi attention mechanism, a feature map with richer semantic information can be obtained, which helps the model to better obtain the feature difference information of different dust coverage levels in the image and improve the classification ability of the model. The classification data set prepared in (1) is put into the improved YOLOv11 classification model for training to obtain the corresponding weight file, which has a high accuracy for inference detection. The detection results are shown in Figure 5 shown.

[0067] The method provided in this embodiment can compensate for the pixel-level information loss caused by downsampling of feature maps in the backbone network by embedding a hierarchical complementary attention mechanism, while making full use of semantic-level information, maintaining an efficient hierarchical structure, improving classification accuracy and stability, and improving computational efficiency.

[0068] In one embodiment, the photovoltaic panel area image is classified using a YOLOv11 classification model to obtain dust coverage of different photovoltaic panels, specifically including the following steps:

[0069] The photovoltaic panel area image is corrected in front view by using a Blob analysis algorithm and a perspective transformation algorithm.

[0070] The corrected photovoltaic panel area images are classified according to the dust coverage to obtain a classification data set.

[0071] The classification data set is used for training to obtain a YOLOv11 classification model.

[0072] According to the YOLOv11 classification model, the dust coverage degree of different photovoltaic panels can be detected.

[0073] The method provided in this embodiment accurately corrects the photovoltaic panel area image into a front view through Blob analysis and perspective transformation algorithm, eliminates the influence of viewing angle deviation, classifies the image according to the degree of dust coverage to form a data set, provides effective data for model training, and uses the classified data set to train the YOLOv11 classification model to generate a precise weight file. Based on the weight file, efficient and accurate detection of dust coverage degrees of different photovoltaic panels can be achieved.

[0074] In one embodiment, the photovoltaic panel image is segmented using the improved YOLOv11 segmentation model to obtain a photovoltaic panel area image, which specifically includes the following steps:

[0075] The C3k2_SCConv module is constructed by combining SCConv convolution with C3k2 module to reduce redundant information in feature maps and improve the segmentation accuracy of the model. Due to the complex environment of photovoltaic power generation devices, when segmenting photovoltaic panel images, not only weather changes are faced, but also the reflection problem caused by strong sunlight exposure of photovoltaic panels needs to be considered. In order to improve the accuracy of photovoltaic panel area segmentation by YOLOv11 model, the C3k2 module in YOLOv11 model improves feature extraction efficiency and computing performance by combining channel separation convolution (C3) and small convolution kernel (k2). The C3k2 module consists of three convolution layers, among which C3 effectively extracts local features by optimizing the convolution operation between channels, while reducing the computational complexity and adapting to features of different scales; k2 uses a small 2x2 convolution kernel, which further accelerates the calculation process by reducing the number of parameters of convolution operation. Its advantages are that it can improve the detection accuracy of small objects, reduce the model inference time, and reduce the demand for computing resources while maintaining high performance, making YOLOv11 more suitable for real-time applications. However, the disadvantage of the C3k2 module is that it relies too much on small convolution kernels, which may sacrifice the recognition accuracy of large objects in some cases, because smaller convolution kernels may find it difficult to capture a wider range of contextual information. In addition, although channel-separated convolution reduces the amount of computation, it may cause information loss or insufficient feature fusion in certain scenarios, affecting detection accuracy. The C3k2 module structure is as follows: Figure 6 shown.

[0076] The C3k2 module in the YOLOv11 segmentation model is replaced with the C3k2_SCConv module to construct an improved YOLOv11 segmentation model, so that the feature expression capability of the C3k2 module is further enhanced. The structure diagram of the improved C3k2 module (C3k2_SCConv) is shown in Figure 7 As shown in the figure, replacing all C3k2 modules in the YOLOv11 segmentation model with C3k2_SCConv modules can greatly improve the segmentation accuracy of the photovoltaic panel area of ​​the YOLOv11 segmentation model, providing a solid foundation for subsequent detection steps.

[0077] Among them, SCConv is an efficient convolution module, whose function is to optimize the feature extraction process, reduce computing resource consumption and improve network performance. The network structure is as follows: Figure 8 As shown. This module mainly includes a spatial reconstruction unit (SRU) and a channel reconstruction unit (CRU). The SRU combines traditional convolution operations with a spatial attention mechanism by spatially reconstructing the input feature map to selectively strengthen key information areas and suppress irrelevant areas. Specifically, the SRU first weights each spatial position in an adaptive manner to generate a spatial attention map, and then adjusts the weights of different areas in the input feature map based on the map. In this way, the SRU can focus on more meaningful feature areas, thereby improving the model's perception of local spatial relationships and overall performance. By introducing a channel attention mechanism, the CRU automatically learns and selects the most relevant channel features so that key information is retained first and unimportant features are suppressed in the convolution operation. By performing a channel-by-channel weighted operation on the input feature map, a channel attention map is generated, and then the activation strength of each channel is adjusted, so that the model can focus more on key channels and improve the ability to transmit information and extract features.

[0078] The improved YOLOv11 segmentation model is trained using the training set and weight optimization is performed.

[0079] The photovoltaic panel image is segmented using the improved YOLOv11 segmentation model to obtain the photovoltaic panel mask. Since there are multiple photovoltaic panels in the camera field of view, the photovoltaic panel area at the far end of the field of view in the image is small in size and difficult to detect. To avoid this interference, before performing photovoltaic panel mask mapping, the photovoltaic panel mask in the main field of view needs to be screened out according to the size of the mask area. The photovoltaic panel area mask is a binary image, and the photovoltaic panel area mask needs to be converted into the corresponding RGB original image.

[0080] The photovoltaic panel mask is operated on the photovoltaic panel image to obtain a photovoltaic panel area image, that is, an RGB image containing only the photovoltaic panel area is obtained, such as Fig. 9The figure shows the schematic diagram of photovoltaic panel image segmentation and conversion results.

[0081] like Fig.10 As shown, it is an overall implementation flow chart of photovoltaic panel dust coverage degree detection based on the YOLOv11 segmentation and classification model provided in this embodiment.

[0082] The method provided in this embodiment can utilize the YOLOv11 segmentation model's ability to extract and express features of photovoltaic panel images, making the model structure more suitable for photovoltaic panel image segmentation tasks, improving the accuracy and stability of the model in photovoltaic panel image segmentation tasks, effectively extracting photovoltaic panel area images, and ensuring the reliability and effectiveness of subsequent tasks.

[0083] Device Embodiment

[0084] According to an embodiment of the present invention, a photovoltaic panel dust coverage degree detection device is provided. Fig.11 , which is a schematic diagram of the structure of the photovoltaic panel dust coverage degree detection device provided in this embodiment. The photovoltaic panel dust coverage degree detection device according to the embodiment of the present invention includes a collection module 111, a segmentation module 112 and a classification module 113.

[0085] The acquisition module 111 is used to acquire photovoltaic panel images and divide them into a training set, a verification set and a test set after preprocessing.

[0086] The segmentation module 112 is used to segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel area image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution.

[0087] The classification module 113 is used to classify the photovoltaic panel area image using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels.

[0088] The device provided in this embodiment collects photovoltaic panel images through the acquisition module 111, and divides them into training set, verification set and test set after preprocessing, which is conducive to the effective training and evaluation of the model; the segmentation module 112 uses the improved YOLOv11 segmentation model to segment the photovoltaic panel image to obtain the photovoltaic panel area image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution, which can enhance the model's ability to extract photovoltaic panel image features, accurately obtain the photovoltaic panel area image, and provide high-quality data for subsequent classification; the classification module 113 uses the YOLOv11 classification model to classify the photovoltaic panel area image to obtain the dust coverage degree of different photovoltaic panels. The operation and maintenance personnel can reasonably arrange the cleaning work according to the dust coverage degree, improve the overall power generation efficiency of the photovoltaic power station, increase the power generation, and improve the economic benefits.

[0089] In one embodiment, the classification module 113 includes an embedding unit for embedding a hierarchical complementary attention mechanism in a feature fusion network layer of the YOLOv11 classification model.

[0090] The device provided in this embodiment can compensate for the pixel-level information loss caused by downsampling of feature maps in the backbone network by embedding a hierarchical complementary attention mechanism, while making full use of semantic-level information, maintaining an efficient hierarchical structure, improving classification accuracy and stability, and improving computing efficiency.

[0091] In one embodiment, the classification module 113 further includes a correction unit, a classification unit, a first training unit and a detection unit.

[0092] The correction unit is used to perform front view correction on the photovoltaic panel area image through a Blob analysis algorithm and a perspective transformation algorithm.

[0093] The classification unit is used to classify the corrected photovoltaic panel area image according to the dust coverage degree to obtain a classification data set.

[0094] The first training unit is used to use the classification data set to perform training to obtain a YOLOv11 classification model.

[0095] The detection unit is used to detect the dust coverage degree of different photovoltaic panels according to the YOLOv11 classification model.

[0096] The device provided in this embodiment accurately corrects the photovoltaic panel area image into a front view through Blob analysis and perspective transformation algorithm, eliminates the influence of viewing angle deviation, classifies the image according to the degree of dust coverage to form a data set, provides effective data for model training, and uses the classified data set to train the YOLOv11 classification model to generate a precise weight file. Based on the weight file, efficient and accurate detection of dust coverage degrees of different photovoltaic panels can be achieved.

[0097] In one embodiment, the segmentation module includes an improvement unit, a second training unit, a segmentation unit and a calculation unit.

[0098] The improvement unit is used to use SCConv convolution in combination with the C3k2 module to construct a C3k2_SCConv module, replace the C3k2 module in the YOLOv11 segmentation model with the C3k2_SCConv module, and construct an improved YOLOv11 segmentation model.

[0099] The second training unit is used to train the improved YOLOv11 segmentation model using the training set and perform weight optimization.

[0100] The segmentation unit is used to segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel mask.

[0101] The operation unit is used to operate the photovoltaic panel mask and the photovoltaic panel image to obtain a photovoltaic panel area image.

[0102] The device provided in this embodiment can utilize the YOLOv11 segmentation model's ability to extract and express features of photovoltaic panel images, making the model structure more suitable for photovoltaic panel image segmentation tasks, improving the accuracy and stability of the model in photovoltaic panel image segmentation tasks, effectively extracting photovoltaic panel area images, and ensuring the reliability and effectiveness of subsequent tasks.

[0103] The embodiment of the present invention is an apparatus embodiment corresponding to the above-mentioned method embodiment. The specific operations of the processing steps of each module can be understood by referring to the description of the method embodiment, which will not be repeated here.

[0104] like Fig.12 As shown, the present invention also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for detecting the degree of dust coverage of photovoltaic panels in the above-mentioned embodiment, or, when executed by a processor, implements the method for detecting the degree of dust coverage of photovoltaic panels in the above-mentioned embodiment.

[0105] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0106] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device or system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment. The device and system embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and the contents not described in detail in the specification of the present invention belong to the common knowledge of those skilled in the art.

Claims

1. A method for detecting dust coverage of photovoltaic panels, characterized in that: The following steps are involved: Collect photovoltaic panel images and divide them into training set, validation set and test set after preprocessing; The photovoltaic panel image is segmented using an improved YOLOv11 segmentation model to obtain a photovoltaic panel region image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution; The photovoltaic panel area image is classified using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels.

2. The method for detecting the degree of dust coverage of photovoltaic panels according to claim 1, characterized in that: The following steps are also included: A hierarchical complementary attention mechanism is embedded in the feature fusion network layer of the YOLOv11 classification model.

3. The method for detecting the degree of dust coverage of photovoltaic panels according to claim 1 or 2, characterized in that: The photovoltaic panel area image is classified by using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels, which specifically includes the following steps: Performing front view correction on the photovoltaic panel area image by using a Blob analysis algorithm and a perspective transformation algorithm; The rectified photovoltaic panel area images are classified according to the dust coverage to obtain a classification data set; The classification data set is used to train a YOLOv11 classification model; According to the YOLOv11 classification model, the dust coverage degree of different photovoltaic panels can be detected.

4. The method for detecting the degree of dust coverage of photovoltaic panels according to claim 1, characterized in that: The photovoltaic panel image is segmented by using the improved YOLOv11 segmentation model to obtain a photovoltaic panel area image, which specifically includes the following steps: Utilize SCConv convolution combined with C3k2 module to construct C3k2_SCConv module, replace C3k2 module in YOLOv11 segmentation model with C3k2_SCConv module, and construct improved YOLOv11 segmentation model; Using the training set to train the improved YOLOv11 segmentation model and perform weight optimization; Segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel mask; The photovoltaic panel mask is operated on the photovoltaic panel image to obtain a photovoltaic panel area image.

5. A photovoltaic panel dust coverage detection device, characterized in that: It includes acquisition module, segmentation module and classification module; The acquisition module is used to acquire photovoltaic panel images and divide them into a training set, a validation set and a test set after preprocessing; The segmentation module is used to segment the photovoltaic panel image using an improved YOLOv11 segmentation model to obtain a photovoltaic panel area image, wherein the improved YOLOv11 segmentation model is constructed using SCConv convolution; The classification module is used to classify the photovoltaic panel area image using the YOLOv11 classification model to obtain the dust coverage degree of different photovoltaic panels.

6. The photovoltaic panel dust coverage detection device according to claim 5, characterized in that: The classification module includes an embedding unit for embedding a hierarchical complementary attention mechanism in a feature fusion network layer of the YOLOv11 classification model.

7. The photovoltaic panel dust coverage detection device according to claim 5 or 6, characterized in that: The classification module also includes a correction unit, a classification unit, a first training unit and a detection unit; The correction unit is used to perform front view correction on the photovoltaic panel area image by using a Blob analysis algorithm and a perspective transformation algorithm; The classification unit is used to classify the corrected photovoltaic panel area image according to the dust coverage degree to obtain a classification data set; The first training unit is used to obtain a YOLOv11 classification model by training using the classification data set; The detection unit is used to detect the dust coverage degree of different photovoltaic panels according to the YOLOv11 classification model.

8. The photovoltaic panel dust coverage detection device according to claim 5, characterized in that: The segmentation module includes an improvement unit, a second training unit, a segmentation unit and a calculation unit; The improvement unit is used to use SCConv convolution in combination with the C3k2 module to construct a C3k2_SCConv module, replace the C3k2 module in the YOLOv11 segmentation model with the C3k2_SCConv module, and construct an improved YOLOv11 segmentation model; The second training unit is used to train the improved YOLOv11 segmentation model using the training set and perform weight optimization; The segmentation unit is used to segment the photovoltaic panel image using the improved YOLOv11 segmentation model to obtain a photovoltaic panel mask; The operation unit is used to operate the photovoltaic panel mask and the photovoltaic panel image to obtain a photovoltaic panel area image.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method for detecting the degree of dust coverage of a photovoltaic panel as claimed in any one of claims 1 to 4 is implemented.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for detecting the degree of dust coverage of a photovoltaic panel as claimed in any one of claims 1 to 4 are implemented.

Images