Method, device, equipment, medium and program product for dust detection of a photovoltaic module

Abstract

The application relates to a dust detection method, device, equipment, medium and program product of a photovoltaic module, and relates to the technical field of photovoltaics. The method comprises the following steps: in response to a dust detection instruction for a photovoltaic module, obtaining a module image of the photovoltaic module and environment features of an environment in which the photovoltaic module is located; performing feature extraction on the module image to obtain a plurality of image features for representing dust coverage influence; fusing the plurality of image features and the environment features to obtain fused features; and performing dust coverage detection on the photovoltaic module based on the fused features to obtain a dust coverage detection result of the photovoltaic module. The method can improve the dust detection accuracy of the photovoltaic module.

Description

Technical Field

[0001] This application relates to the field of photovoltaic technology, and in particular to a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for dust detection of photovoltaic modules. Background Technology

[0002] As a crucial component of clean energy, photovoltaic (PV) power generation's efficiency directly impacts the economic benefits of power plants. Dust accumulation on PV module surfaces significantly reduces light transmittance, leading to a decrease in module output power. Therefore, timely and effective monitoring of dust levels on PV module surfaces is essential for ensuring the efficient operation of PV power plants.

[0003] Currently, common dust detection methods mainly involve directly collecting and weighing dust using contact physical means, or evaluating based on the attenuation of optical transmittance. However, these methods are either destructive, discontinuous, or susceptible to interference from stray light in the environment, resulting in relatively low accuracy in dust detection for photovoltaic modules. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, apparatus, computer equipment, computer-readable storage medium, and computer program product for detecting dust in photovoltaic modules, which can improve the accuracy of dust detection in photovoltaic modules, in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for dust detection in photovoltaic modules, comprising:

[0006] In response to a dust detection command for photovoltaic modules, the system acquires images of the photovoltaic modules and the environmental characteristics of the environment in which the photovoltaic modules are located.

[0007] Feature extraction is performed on the component images to obtain multiple image features that characterize the impact of dust coverage;

[0008] Multiple image features and environmental features are fused to obtain fused features;

[0009] Based on the fusion characteristics, dust coverage detection is performed on photovoltaic modules to obtain the dust coverage detection results of photovoltaic modules.

[0010] In one embodiment, the plurality of image features include a first image feature for characterizing the global texture distribution of the photovoltaic module;

[0011] Feature extraction is performed on the component images to obtain multiple image features characterizing the impact of dust cover, including:

[0012] Obtain the spatial location information of each pixel in the component image;

[0013] Based on the spatial location information, multiple pixel pairs that satisfy the spatial location conditions are determined from each pixel point;

[0014] Obtain the gray value pair corresponding to each pixel pair, and construct the gray value matrix of the component image based on each gray value pair;

[0015] Based on the grayscale matrix, a first image feature is determined to characterize the global texture distribution of the photovoltaic module.

[0016] In one embodiment, the multiple image features include a second image feature for characterizing the local texture distribution of the photovoltaic module;

[0017] Feature extraction is performed on the component images to obtain multiple image features characterizing the impact of dust cover, including:

[0018] For each pixel in the component image, search for multiple neighboring pixels corresponding to the pixel according to a preset pixel search range;

[0019] The gray value of each neighboring pixel is compared with the gray value of the pixel to obtain multiple gray value comparison results;

[0020] Based on the comparison results of each gray value, the first gray value parameter that characterizes the local gray value distribution of the pixel is determined.

[0021] Based on the first grayscale parameter corresponding to each pixel, a second image feature representing the local texture distribution of the photovoltaic module is determined.

[0022] In one embodiment, the plurality of image features include a third image feature for characterizing the edge texture distribution of the photovoltaic module;

[0023] Feature extraction is performed on the component images to obtain multiple image features characterizing the impact of dust cover, including:

[0024] Multiple edge pixels are identified from each pixel in the component image;

[0025] For each edge pixel, extract a second grayscale parameter that represents the degree of drastic grayscale change of the edge pixel;

[0026] Based on each of the second grayscale parameters, a third image feature is determined to characterize the edge texture distribution of the photovoltaic module.

[0027] In one embodiment, based on fusion features, dust coverage detection is performed on the photovoltaic module to obtain the dust coverage detection result of the photovoltaic module, including:

[0028] The dust cover detection model is invoked. This model is pre-trained based on multiple sample image features extracted from sample images of photovoltaic modules, corresponding sample environment features, and labeled true dust cover.

[0029] By analyzing the fusion characteristics using a dust coverage detection model, the dust coverage detection results of photovoltaic modules are obtained.

[0030] In one embodiment, multiple image features and environmental features are fused to obtain fused features, including:

[0031] Importance analysis is performed on each image feature to obtain the importance analysis results for each image feature;

[0032] Determine the fusion weights that match the results of each importance analysis;

[0033] Based on each fusion weight, the image features are fused to obtain preliminary features;

[0034] The preliminary features are fused with the environmental features to obtain the fused features.

[0035] Secondly, this application also provides a dust detection device for photovoltaic modules, comprising:

[0036] The information acquisition module is used to acquire images of the photovoltaic modules and environmental characteristics of the environment in which the photovoltaic modules are located in response to dust detection commands for the photovoltaic modules.

[0037] The feature extraction module is used to extract features from the component images to obtain multiple image features that characterize the impact of dust coverage.

[0038] The feature fusion module is used to fuse multiple image features and environmental features to obtain fused features;

[0039] The dust detection module is used to detect dust coverage on photovoltaic modules based on fusion features, and obtain the dust coverage detection results of photovoltaic modules.

[0040] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0041] In response to a dust detection command for photovoltaic modules, the system acquires images of the photovoltaic modules and the environmental characteristics of the environment in which the photovoltaic modules are located.

[0042] Feature extraction is performed on the component images to obtain multiple image features that characterize the impact of dust coverage;

[0043] Multiple image features and environmental features are fused to obtain fused features;

[0044] Based on the fusion characteristics, dust coverage detection is performed on photovoltaic modules to obtain the dust coverage detection results of photovoltaic modules.

[0045] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, performs the following steps:

[0046] In response to a dust detection command for photovoltaic modules, the system acquires images of the photovoltaic modules and the environmental characteristics of the environment in which the photovoltaic modules are located.

[0047] Feature extraction is performed on the component images to obtain multiple image features that characterize the impact of dust coverage;

[0048] Multiple image features and environmental features are fused to obtain fused features;

[0049] Based on the fusion characteristics, dust coverage detection is performed on photovoltaic modules to obtain the dust coverage detection results of photovoltaic modules.

[0050] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, performs the following steps:

[0051] In response to a dust detection command for photovoltaic modules, the system acquires images of the photovoltaic modules and the environmental characteristics of the environment in which the photovoltaic modules are located.

[0052] Feature extraction is performed on the component images to obtain multiple image features that characterize the impact of dust coverage;

[0053] Multiple image features and environmental features are fused to obtain fused features;

[0054] Based on the fusion characteristics, dust coverage detection is performed on photovoltaic modules to obtain the dust coverage detection results of photovoltaic modules.

[0055] The aforementioned dust detection method, apparatus, computer equipment, computer-readable storage medium, and computer program product for photovoltaic modules, in response to a dust detection command for the photovoltaic module, first acquires an image of the photovoltaic module and the environmental features of the environment in which the photovoltaic module is located. Next, feature extraction is performed on the module image to obtain multiple image features characterizing the impact of dust coverage. These multiple image features and environmental features are then fused to obtain fused features. Finally, based on the fused features, dust coverage detection is performed on the photovoltaic module to obtain the dust coverage detection result. Thus, this solution, by fusing multi-scale visual features corresponding to the photovoltaic module with real-time environmental parameters, eliminates the need for contact with the module or collection of dust samples during the detection process. It also effectively overcomes interference caused by changes in ambient light, thereby significantly improving the accuracy of dust detection for photovoltaic modules. Attached Figure Description

[0056] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is an application environment diagram of a dust detection method for photovoltaic modules in one embodiment;

[0058] Figure 2 This is a flowchart illustrating a dust detection method for a photovoltaic module in one embodiment;

[0059] Figure 3 This is a flowchart illustrating the process of determining a first image feature in one embodiment;

[0060] Figure 4 This is a flowchart illustrating the process of determining a second image feature in one embodiment;

[0061] Figure 5 This is a flowchart illustrating the process of determining a third image feature in one embodiment;

[0062] Figure 6 This is a schematic diagram of the architecture of a dust detection system for a photovoltaic module in a specific embodiment;

[0063] Figure 7 This is a flowchart illustrating a dust detection method for a photovoltaic module in a specific embodiment.

[0064] Figure 8 This is a schematic diagram of multi-scale image feature extraction in a specific embodiment;

[0065] Figure 9This is a structural block diagram of a dust detection device for a photovoltaic module in one embodiment;

[0066] Figure 10 This is an internal structural diagram of a computer device in one embodiment;

[0067] Figure 11 This is the internal structure of a computer device in another embodiment. Detailed Implementation

[0068] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0069] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0070] The dust detection method for photovoltaic modules provided in this application embodiment can be applied to, for example... Figure 1 In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on other network servers. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services.

[0071] In one exemplary embodiment, such as Figure 2 As shown, a method for dust detection in photovoltaic modules is provided. This method can be performed by... Figure 1 The method can be executed independently by terminal 102 or server 104, or it can be executed interactively by terminal 102 and server 104. The following describes how this method is applied to... Figure 1 Taking server 104 as an example, the following steps are included:

[0072] Step S202: In response to the dust detection command for the photovoltaic module, obtain the module image of the photovoltaic module and the environmental characteristics of the environment in which the photovoltaic module is located.

[0073] In this context, a photovoltaic module refers to a solar panel unit to be inspected, typically covered by glass and containing regularly spaced metal grids and silicon wafers. A dust detection command is a signal or command that triggers the entire inspection process; it can be automatically issued by a timed task, manually triggered by maintenance personnel, or automatically generated by the power plant monitoring system based on power anomaly warnings. A module image is an image containing visual information about the photovoltaic module's surface, captured by an image acquisition device. Environmental characteristics refer to quantitative parameters recorded simultaneously during module image acquisition, used to describe the physical conditions of the shooting scene. In this embodiment, the core environmental characteristic is primarily ambient light intensity, which directly affects the image's brightness and contrast. Additionally, environmental characteristics may include auxiliary parameters such as solar incidence angle, temperature, and humidity; this embodiment does not impose limitations on these.

[0074] For example, when the server receives a dust detection command for the photovoltaic modules, the image acquisition module (such as a fixed high-definition industrial camera or a gimbal camera mounted on a drone) will take pictures of the photovoltaic modules to obtain raw RGB or grayscale images. Simultaneously, the environmental information acquisition module (such as at least one high-precision light sensor or solar position sensor) will record at least one piece of information about the photovoltaic modules, including the current light intensity, solar azimuth angle, and altitude angle. This data is transmitted to the server via a communication link.

[0075] In some embodiments, the server can dynamically adjust the contrast and brightness of the component image based on environmental characteristics. Specifically, after obtaining the component image Iraw and environmental characteristics such as the ambient light intensity value Lenv, the server first normalizes the light intensity, that is, maps the ambient light intensity value Lenv to a standard range (such as between 0 and 1). The normalization expression is as follows:

[0076]

[0077] in, , These are the preset maximum and minimum possible light intensities for the environment in which the photovoltaic modules are located. This is the normalized ambient light intensity value, which reflects the relative strength of the current light.

[0078] After normalizing the ambient light intensity, the server then sets the target brightness of the component image Iraw. Understandably, in low light (such as dawn or dusk), the target brightness should be set slightly higher to fully enhance shadow details while suppressing noise amplification. In strong light (such as midday), the target brightness should be set slightly lower to avoid overall overexposure and preserve highlight details. The expression for setting the target brightness is as follows:

[0079]

[0080] in, Target brightness. This is the reference target brightness, such as 128 or 255. It is an adjustment coefficient, when In low light conditions, For positive, Increase. Conversely, when During periods of intense (strong) light, Negative, reduce.

[0081] The server then calculates the average grayscale of the component image Iraw. If the component image Iraw is in color, it is first converted to grayscale, resulting in a single-channel matrix. A filtering algorithm, such as Gaussian filtering or median filtering, is applied to this grayscale image to reduce noise, smoothing out random noise generated during the shooting process while preserving the main edges and textures representing dust and component structure. The processed component image is defined as Ipre. The global average grayscale value is calculated based on the component image Ipre, using the following expression:

[0082]

[0083] in, This is the global average grayscale value. This represents the total number of pixels in the image, where i and j represent individual pixels.

[0084] By comparing the global average grayscale with the target brightness, the brightness gain coefficient can be calculated.

[0085]

[0086] in, Indicates the brightness gain coefficient. It is a very small constant to prevent division by zero. When this happens, it indicates that the area needs to be highlighted; When the time is right, it indicates that the image needs to be darkened.

[0087] Finally, the server uses a pixel-level transformation function to adjust both the brightness and contrast of the image. For example, a variant of linear stretching maps the original pixel value p to a new value through a linear transformation with controlled slope and intercept. The expression is as follows:

[0088]

[0089] Where k represents contrast and b represents brightness offset. At the same time, the server will increase both k and b, which will increase the overall brightness while appropriately stretching the contrast of the midtones to avoid the image appearing grayish. When this happens, both k and b will be reduced simultaneously, darkening the overall brightness while preventing the loss of detail in the shadows. Among these, The simplified mapping relationship between k and b is: , The value is usually 0.7, but other values ​​can also be used, depending on the specific circumstances. .

[0090] Step S204: Extract features from the component image to obtain multiple image features used to characterize the impact of dust coverage.

[0091] Among them, multiple image features are a set of quantitative indicators that describe image attributes from different angles and scales. In this embodiment, the following three types of features constitute multi-scale features: a first image feature for characterizing the global texture distribution of the photovoltaic module, a second image feature for characterizing the local texture distribution of the photovoltaic module, and a third image feature for characterizing the edge texture distribution of the photovoltaic module.

[0092] For example, after the server completes the initial adjustment of the brightness and contrast of the component image, it can execute three feature extraction pipelines in parallel: First, it calculates a grayscale matrix on the adjusted entire image or a defined region of interest, thereby determining a first image feature representing the global texture distribution of the photovoltaic module based on the grayscale matrix. Second, it traverses each pixel in the image, calculates the grayscale value comparison results between pixels, and determines a second image feature representing the local texture distribution of the photovoltaic module based on the grayscale value comparison results. Third, it identifies edge pixels in the image, and determines a third image feature representing the edge texture distribution of the photovoltaic module based on the grayscale information of the edge pixels.

[0093] Step S206: Fuse multiple image features and environmental features to obtain fused features.

[0094] Among them, the fusion feature refers to the final feature vector obtained by fusing multiple image features and environmental features. It simultaneously contains the visual state information of the photovoltaic module surface (carried by multi-scale image features) and the physical environmental context information that generates the visual state (carried by environmental features).

[0095] For example, after obtaining multiple image features, the server can first normalize these features to standardize them to the same scale, as shown in the following expression:

[0096]

[0097] in, It is the original feature value of the i-th image feature. and These are the mean and standard deviation of the i-th class feature, calculated from a large number of training samples.

[0098] After feature normalization, the server will concatenate multiple image features into a single visual feature vector in a preset order. This is then combined with the normalized environmental feature E to form the final, comprehensive fused feature. Among them, environmental characteristic E includes ambient light intensity. In one example, features are fused. The expression is as follows: Concat is a feature fusion operation in machine learning. Its core meaning is to concatenate two or more vectors (or matrices) end to end according to a specific dimension to form a new vector (or matrix) with a larger dimension, without changing the values ​​and order of the elements inside the original vector.

[0099] For example: = [0.85, 1.20, 0.45];E1= (Ambient light intensity) = 0.95, E2 (cosine of solar incidence angle) = 0.80, E = [E1, E2] = [0.95, 0.80]; =Concat([0.85, 1.20,0.45], [0.95, 0.80]) = [0.85, 1.20, 0.45, 0.95, 0.80].

[0100] Step S208: Based on the fusion characteristics, perform dust coverage detection on the photovoltaic module to obtain the dust coverage detection results of the photovoltaic module.

[0101] The dust coverage detection result is the final output, typically a quantified dust coverage level, such as a percentage or grade, used to objectively measure the severity of dust accumulation. In some embodiments, the server can also provide corresponding component cleaning suggestions based on the dust coverage monitoring results, such as "clean immediately," "clean recommended," or "keep an eye on it."

[0102] In some embodiments, dust coverage detection of photovoltaic modules is performed based on fusion features to obtain dust coverage detection results of photovoltaic modules, including: calling a dust coverage detection model, wherein the dust coverage detection model is pre-trained based on multiple sample image features extracted from sample images of photovoltaic modules, corresponding sample environment features, and labeled real dust coverage; and analyzing fusion features through the dust coverage detection model to obtain dust coverage detection results of photovoltaic modules.

[0103] The dust coverage detection model is a mathematical function or computational framework built using machine learning algorithms to establish a mapping relationship between fused features and dust coverage. Essentially, it's a predictor that takes fused features as input and outputs a quantified dust coverage value. In specific implementations, this model can be a support vector regression model, a gradient boosting decision tree model, a lightweight convolutional neural network (CNN), or a fully connected neural network, etc. Sample image features and sample environment features refer to the historical data used during model training. Sample image features are multiple image features extracted from a large number of historical photovoltaic module images. These can include first sample image features characterizing the global texture distribution of the photovoltaic module, second sample image features characterizing the local texture distribution, and third sample image features characterizing the edge texture distribution. Sample environment features are environmental features recorded synchronously when collecting historical photovoltaic module images, such as at least one of ambient light intensity, solar incidence angle, temperature, and humidity. The labeled true dust coverage is the true value (label) corresponding to each training sample (i.e., a set of image features and environmental features) used to supervise model learning. This value is usually obtained through high-precision offline calibration methods, such as manual calibration or instrument calibration.

[0104] For example, during the offline training phase, the server first constructs a high-quality training dataset. This involves collecting a large amount of sample data covering photovoltaic modules across different seasons, weather conditions, time of day, dust levels, and geographical locations. For each sample dataset, the module image and ambient lighting information are recorded simultaneously. Subsequently, using a feature extraction method identical to the online detection process, multi-scale image features are extracted from the sample images and concatenated with the sample environmental features to form fused features for training. Simultaneously, the actual dust coverage of each sample is labeled and used as the training label. Next, select an initial model structure, and then ( , The model takes a sample as input and iteratively trains it using a specific loss function (such as mean squared error) and optimization algorithm (such as gradient descent). During training, the data is typically divided into training and validation sets. The model's hyperparameters (such as tree depth, learning rate, or at least one other) are adjusted based on the performance on the validation set to prevent overfitting. The final result is a stable and highly generalizable dust cover detection model file.

[0105] During the online inference phase, the server only needs to load the trained dust cover detection model file and the real-time generated fusion features. Input the data, and the model can quickly complete the calculation and output the result, namely the predicted dust coverage.

[0106] In some embodiments, multiple image features and environmental features are fused to obtain fused features, including: performing importance analysis on each image feature to obtain the importance analysis result corresponding to each image feature; determining the fusion weight that matches each importance analysis result; fusing each image feature based on each fusion weight to obtain preliminary features; and fusing the preliminary features with environmental features to obtain fused features.

[0107] Importance analysis aims to quantify the contribution or discriminative power of each image feature to the final dust cover detection. The results of importance analysis are typically a set of numerical values ​​or ranks. For example, it could be a feature importance score vector [s1, s2, s3], where s1, s2, and s3 represent the scores of the first, second, and third image features, respectively. Higher scores indicate a stronger correlation between the feature and dust cover, or a greater contribution to the model. Fusion weights are multiplicative coefficients corresponding one-to-one with each image feature, used to adjust the contribution ratio of the feature during fusion. The magnitude of the weights is directly determined by the importance analysis results; higher importance results in higher weights. Preliminary features refer to a new feature representation generated by linearly weighting multiple image features according to their corresponding fusion weights. Fusion with environmental features refers to the operation of combining the obtained preliminary features with environmental features. Here, fusion usually refers to concatenation, where the environmental feature vector is directly appended to the preliminary feature vector to form a final feature vector that simultaneously contains weighted visual information and environmental context. In practical applications, environmental features can also have corresponding fusion weights.

[0108] For example, the server can pre-calculate the Pearson correlation coefficient or mutual information between each image feature and dust coverage. The absolute value of the correlation coefficient or mutual information is used as the importance score of that feature. After obtaining the importance score, the score is converted into a fusion weight that sums to 1 using Softmax (normalized exponential function). For example, the fusion weight for the first image feature is w1, the fusion weight for the second image feature is w2, and the fusion weight for the third image feature is w3. Then, the image feature vectors extracted in real time (such as the first image feature v1, the second image feature v2, and the third image feature v3) are weighted and fused. V visual =[w1 * v1,w2 * v2,w3 * v3] . This constitutes the initial features. Subsequently, the environmental feature vector E is compared with... To splice: The final fusion features are obtained. .

[0109] In this embodiment, in response to a dust detection command for the photovoltaic module, the module image and the environmental features of the environment in which the photovoltaic module is located are first acquired. Next, feature extraction is performed on the module image to obtain multiple image features characterizing the impact of dust coverage. These multiple image features and environmental features are then fused to obtain a fused feature. Finally, based on the fused feature, dust coverage detection is performed on the photovoltaic module to obtain the dust coverage detection result. Thus, this embodiment, by fusing multi-scale visual features corresponding to the photovoltaic module with real-time environmental parameters, eliminates the need for contact with the module or collection of dust samples during the detection process. It also effectively overcomes interference caused by changes in ambient light, thereby significantly improving the accuracy of dust detection for photovoltaic modules.

[0110] In one exemplary embodiment, such as Figure 3 As shown, feature extraction is performed on the component image to obtain multiple image features used to characterize the impact of dust coverage, including:

[0111] Step S302: Obtain the spatial location information corresponding to each pixel in the component image.

[0112] In this context, a pixel is the smallest unit that makes up the component image, and each pixel stores a grayscale value representing the brightness information at that location. Spatial location information refers to the coordinates of each pixel in the image matrix, usually represented by an ordered pair (i,j), where i represents the row index of the pixel and j represents the column index of the pixel. These coordinates uniquely determine the position of the pixel on the two-dimensional image plane.

[0113] For example, the component image is represented as a two-dimensional array in memory or a computing unit. The server automatically associates the index (i,j) of each element, i.e., each pixel, with this array by traversing it. For instance, the pixel located in the 3rd row and 5th column has the spatial location information (3,5).

[0114] Step S304: Based on the spatial location information, determine multiple pixel pairs that satisfy the spatial location conditions from each pixel point.

[0115] The spatial location condition is a predefined rule used to determine whether two pixels constitute a valid pixel pair for texture analysis. This condition is defined by two core parameters: an orientation parameter and a distance parameter. The orientation parameter specifies the relative orientation between the two pixels; common orientations include 0° (horizontal to the right), 45° (upper right), 90° (vertical upward), and 135° (upper left). The distance parameter specifies the number of pixels between the two pixels; the distance parameter is usually a positive integer, such as a distance parameter of 1 indicating adjacent pixels. A pixel pair is a combination of two pixels whose spatial relationship satisfies the above spatial location condition. For a given reference pixel P(i,j) and a set of (θ,d) parameters, where θ represents the orientation parameter and d represents the distance parameter, the position of the other target pixel Q in its corresponding pixel pair is uniquely determined. For example, when (θ=0°, d=1), for pixel P(i,j), its pixel pair is (P,Q), where the coordinates of pixel Q are (i,j+1).

[0116] For example, the server pre-sets spatial location conditions, i.e., configures one or more sets of (θ, d) parameters. Then, it iterates through each pixel P in the image. For each pixel P, based on the currently used (θ, d) parameters, the coordinates of the target pixel Q are directly calculated using the coordinates (i, j) of pixel P. (P, Q) thus constitute a pixel pair that satisfies the spatial location conditions.

[0117] It can be understood that the essence of component texture is the spatial repetition pattern of gray levels. By systematically and regularly examining a large number of pixel pairs with fixed relative positions in an image, the statistical characteristics of this spatial repetition pattern can be quantified. Different (θ,d) parameters can capture texture information from different directions and scales.

[0118] Step S306: Obtain the gray value pair corresponding to each pixel pair, and construct the gray value matrix of the component image based on each gray value pair.

[0119] In this context, a grayscale value pair is an unordered or ordered combination of the grayscale values ​​of two pixels in a pixel pair, represented as (g1, g2), where g1 is the grayscale value of pixel P and g2 is the grayscale value of pixel Q. The grayscale value matrix is ​​a two-dimensional statistical matrix, where the row and column indices correspond to possible grayscale values ​​(e.g., 0 to 255). The value of the element in the g1-th row and g2-th column represents the frequency of occurrence of pixel pairs with grayscale values ​​(g1, g2) among all pixel pairs satisfying the current spatial location conditions in the entire component image. Essentially, this matrix is ​​a statistical table of grayscale value co-occurrence frequencies.

[0120] For example, the server first initializes a zero matrix GLCM (Gray-Level Co-occurrence Matrix) of size L×L, where L is the number of gray levels, such as 256. Then, it iterates through all the previously determined pixel pairs. For each pixel pair (P, Q), the server reads the gray value g1 of pixel P and the gray value g2 of pixel Q. Subsequently, it increments the element value in the g1-th row and g2-th column of the GLCM matrix by 1. After iterating through all pixel pairs, the GLCM matrix is ​​constructed.

[0121] Step S308: Determine the first image feature to characterize the global texture distribution of the photovoltaic module based on the grayscale value matrix.

[0122] The first image feature includes one or more statistics calculated from the grayscale matrix that quantify the global texture characteristics of the image, such as contrast and entropy. Contrast reflects the sharpness of the texture and the intensity of local changes. A higher contrast value indicates a sharper transition between bright and dark areas in the image, and a coarser texture. Dust cover typically leads to microscopic unevenness on the surface, increasing local contrast and thus raising the contrast value. Entropy reflects the complexity, randomness, or disorder of the image texture. A higher entropy value indicates a more random grayscale distribution and a more complex texture. Irregular dust deposition disrupts the uniformity of the component surface, increasing the randomness of the texture and thus raising the entropy value.

[0123] For example, the server calculates values ​​such as contrast and entropy based on the constructed grayscale matrix, and these values ​​together constitute the first image feature.

[0124] In one example, the expression for calculating contrast is as follows:

[0125]

[0126] in, The row index represents the gray value of the first pixel in the pixel pair. The column index represents the grayscale value of the second pixel (adjacent pixel) in a pixel pair. It is located at the line, number The element values ​​of a column represent combinations of grayscale values ​​given a spatial relationship. The probability of it appearing in the entire image. This represents the square of the grayscale difference. This weighting factor causes pixel pairs with large grayscale differences to contribute to the contrast value by a square factor.

[0127] In one example, the expression for calculating entropy is as follows:

[0128]

[0129] in, It is a logarithmic function.

[0130] In this embodiment, by constructing a grayscale matrix and extracting statistics such as contrast and entropy, the macroscopic texture features of the photovoltaic module surface are transformed into stable and quantifiable digital indicators. This effectively captures the surface roughness and disorder changes caused by dust coverage, providing a basis for subsequent judgment and significantly improving the objectivity and accuracy of dust detection.

[0131] In one exemplary embodiment, such as Figure 4 As shown, feature extraction is performed on the component image to obtain multiple image features used to characterize the impact of dust coverage, including:

[0132] Step S402: For each pixel in the component image, search for multiple neighboring pixels corresponding to the pixel according to a preset pixel search range.

[0133] The preset pixel search range is a predefined rule used to determine which pixels around the center pixel will be considered its neighborhood. It is typically defined by a neighborhood shape (such as a square, circle, etc.) and a radius or side length (i.e., the search distance). For example, a common search range is a square area centered on the center pixel with a side length of 3 pixels. This means searching for 8 adjacent pixels within a 1-pixel range above, below, to the left, and to the right of the center pixel. Neighboring pixels are all other pixels around the center pixel that meet the criteria, found according to the preset search range. These pixels have spatial proximity to the center pixel, and their grayscale values, together with those of the center pixel, describe the texture details of the local area.

[0134] For example, the server pre-defines the search range, i.e., the shape and size of the neighborhood, such as a 3x3 square neighborhood. Then, it traverses every pixel in the image and uses it as the center pixel. For each center pixel, based on its coordinates and the pre-defined search range, it calculates the coordinates of all its neighboring pixels and uses the pixels corresponding to these coordinates as the set of neighboring pixels of the current center pixel.

[0135] Step S404: Compare the gray value of each neighboring pixel with the gray value of the pixel to obtain multiple gray value comparison results.

[0136] The grayscale value comparison result is a binary result, with a value of either 0 or 1. Generally, if the grayscale value of a neighboring pixel is greater than or equal to the grayscale value of the center pixel, the grayscale value comparison result is 1; otherwise, it is 0. Therefore, for a search range with P neighboring pixels, a sequence of P binary bits will be obtained.

[0137] For example, for each center pixel and its set of neighboring pixels, the server sequentially reads the grayscale value of each neighboring pixel and compares it with the grayscale value of the center pixel. A binary digit is generated according to a preset comparison rule. After traversing all P neighboring pixels, a binary sequence of length P is obtained. For example, for a 3x3 neighborhood (P=8), the sequence [1,0,1,1,0,1,0,0] might be obtained. This binary sequence represents the comparison results of multiple grayscale values ​​of the center pixel.

[0138] Step S406: Based on the comparison results of each gray value, determine the first gray value parameter that represents the local gray value distribution of the pixel.

[0139] The first grayscale parameter is a value derived from the comparison of grayscale values, specifically the binary sequence described above, that uniquely represents the local texture pattern encoded by the sequence. Typically, the first grayscale parameter is obtained by treating the binary sequence as a P-bit binary number and then converting it to a decimal number. For example, the decimal number corresponding to the binary sequence [1,0,1,1,0,1,0,0] is (10110100)^2 = 180. This decimal number is the first grayscale parameter of the center pixel, or the local pattern encoded value.

[0140] For example, after obtaining the P-bit binary sequence corresponding to each center pixel, the server arranges the sequence in a predetermined order, such as starting from the neighboring point at the top left corner, in a clockwise or counterclockwise order, forming a binary string. Then, using a standard binary-to-decimal algorithm, this string is converted into an integer. This integer is the first grayscale parameter of the center pixel. Different local grayscale change patterns produce unique binary sequences, which are then mapped to a unique integer. This integer encodes the microstructural features of the local region, such as flat areas, edges, corners, etc. By calculating the first grayscale parameter for each center pixel in the image, the original component grayscale image can be converted into a local pattern coding map, where the value of each pixel represents the local texture pattern coding at that location.

[0141] Step S408: Based on the first grayscale parameter corresponding to each pixel, determine the second image feature that characterizes the local texture distribution of the photovoltaic module.

[0142] The second image feature includes one or more statistical quantities calculated based on the first gray-level parameter corresponding to each pixel, which quantifies the overall characteristics of the local texture of the image. These include, for example, at least one of the following: the entropy of the local pattern histogram, the energy of the local pattern histogram, and the mean of the local region gray-level variance. The local pattern histogram is the statistical distribution of the encoded local texture patterns in the local pattern coding diagram. The entropy of the local pattern histogram reflects the uniformity and complexity of the distribution of different local texture patterns in the image. Dust cover tends to homogenize the texture, which may lead to a decrease in entropy. The energy of the local pattern histogram reflects the dominance of one or more local texture patterns; dust cover may increase the energy value. The mean of the local region gray-level variance is calculated by first calculating the gray-level variance of each pixel in its local neighborhood before calculating the first gray-level parameter, and then averaging the variances of all neighborhoods in the entire image. This value directly measures the average level of local contrast; dust cover usually leads to a decrease in local contrast, thus lowering this mean.

[0143] For example, the server counts the first grayscale parameter values ​​of all pixels and generates a local pattern histogram. Then, the histogram is normalized to obtain the probability of each local texture pattern occurring. Next, the histogram's entropy, energy, and other statistics are calculated. In addition, the server simultaneously calculates the grayscale variance of pixels within each local neighborhood and calculates the arithmetic mean of these variances.

[0144] In this embodiment, by comparing the grayscale of each pixel with its neighborhood and encoding it as a local mode parameter, the uniformity and complexity of the local texture are quantified from a global statistical perspective. This can capture the blurring of microscopic details and the decrease in local contrast caused by dust coverage, providing a reliable microscale feature basis for dust detection and making up for the deficiencies of macroscopic texture features.

[0145] In one exemplary embodiment, such as Figure 5 As shown, feature extraction is performed on the component image to obtain multiple image features used to characterize the impact of dust coverage, including:

[0146] Step S502: Identify multiple edge pixels from each pixel in the component image.

[0147] Among them, edge pixels are pixels in the component image whose grayscale values ​​change drastically or significantly in local space. Visually, these pixels form the outline of clear structures such as component grid lines and borders.

[0148] For example, servers typically identify edge pixels using edge detection operators, the core of which is measuring the rate of grayscale change (gradient) at each pixel location. Taking the classic Sobel operator as an example, firstly, the original grayscale image is convolved using convolution kernels in two directions (horizontal and vertical), yielding approximate gradient values ​​Gx and Gy for each pixel in the horizontal and vertical directions, respectively. These two values ​​together describe the intensity and direction of the grayscale change at that point. Subsequently, the gradient magnitude of each pixel is calculated, typically... or , The gradient magnitude represents the square root. The gradient magnitude quantifies the degree of grayscale change at that point. Finally, a preset magnitude threshold is set. Iterating through all pixels, if the gradient magnitude of a point is greater than the threshold, that point is determined to be an edge pixel; otherwise, it is determined to be a non-edge pixel.

[0149] Step S504: For each edge pixel, extract a second grayscale parameter that represents the degree of grayscale change of the edge pixel.

[0150] The second grayscale parameter refers to the gradient magnitude corresponding to the edge pixel. Its value directly and quantitatively reflects the degree of grayscale change at the location of a specific edge pixel, that is, the local sharpness of the edge.

[0151] For example, the server reads the gradient magnitude of pixels identified as edge pixels. For each edge pixel, its gradient magnitude is the second grayscale parameter of that point.

[0152] Step S506: Based on each of the second grayscale parameters, determine the third image features used to characterize the edge texture distribution of the photovoltaic module.

[0153] The third image feature is one or more statistics derived from the second set of grayscale parameters of all edge pixels, used to quantify the sharpness of edge structures in the component image as a whole. Core features typically include, but are not limited to, at least one of the following: mean gradient magnitude and variance of gradient magnitude. The mean gradient magnitude is the arithmetic mean of the gradient magnitudes of all edge pixels, representing the average sharpness of the edges overall. The variance of gradient magnitude reflects the dispersion of the gradient magnitudes of all edge pixels relative to their mean, representing the non-uniformity or consistency of edge sharpness.

[0154] For example, the server aggregates the gradient magnitudes of all edge pixels to form a dataset. Next, the mean and variance, or standard deviation, of this dataset are calculated. The mean is calculated by summing all magnitudes and dividing by the total number of edge pixels. The variance is calculated by summing the squares of the differences between each magnitude and the mean, and then dividing by the total number of pixels. These two statistics together constitute the third image feature. It can be understood that dust coverage acts like a semi-transparent film, causing sharp edges to diffuse and blur, directly leading to a general decrease in the gradient magnitude of edge pixels, thus significantly reducing the mean gradient magnitude. Simultaneously, the distribution of dust may be uneven, resulting in some edges being severely blurred while others are relatively clear; this difference manifests as an increase in the gradient magnitude variance. By monitoring the decrease in the mean and the increase in the variance, the edge blurring effect caused by dust on the critical structure (gate lines) of the component can be accurately and quantitatively detected.

[0155] In this embodiment, by quantifying the gradient intensity of edge pixels, the edge blurring effect caused by dust is transformed into a precisely calculable mean and variance index, thereby directly capturing the clarity degradation of key structures such as photovoltaic module grid lines. This edge feature effectively complements macroscopic texture and microscopic detail features, jointly enhancing the dust detection system's immunity to interference from complex environments and significantly improving the accuracy and reliability of dust detection.

[0156] In some specific embodiments, Figure 6 A schematic diagram of the architecture of a dust detection system for photovoltaic modules is shown. The system includes an image acquisition unit, an image preprocessing unit, a feature extraction unit, a dust coverage detection unit, an environmental information acquisition unit, a result output unit, and a model optimization unit. The image acquisition unit acquires images of the photovoltaic modules. The image preprocessing unit processes the module images for contrast, brightness, etc. The feature extraction unit extracts features from the module images to obtain multiple image features characterizing the impact of dust coverage. The dust coverage detection unit deploys a dust coverage detection model, which fuses environmental features with multiple image features and performs dust coverage detection based on the fused features. The environmental information acquisition unit collects environmental information about the environment in which the photovoltaic modules are located and extracts environmental features from this information. The result output unit is connected to the dust coverage detection unit and outputs the dust coverage result (dust coverage degree) from the dust coverage detection unit, sending the result to a terminal or monitoring center. The model optimization unit is connected to the dust coverage detection unit and optimizes and updates the dust coverage detection model based on the dust coverage result; the updated model continues to participate in dust coverage detection.

[0157] Figure 7The flowchart illustrates a dust detection method for photovoltaic modules. The first stage is offline training. In this stage, the server extracts multiple sample image features and corresponding environmental features from sample images of the photovoltaic modules. Next, data annotation is performed, specifically labeling the actual dust coverage. The labeled data is then used to train the model until the training stop condition is met. If the training iterations are reached, the dust coverage detection model is output.

[0158] The dust cover detection model has entered the online detection phase. That is, in response to dust detection commands for photovoltaic modules, the server acquires images of the photovoltaic modules and the environmental features of the environment in which the modules are located. After preprocessing the module images, multi-scale image features are extracted. Figure 8 A schematic diagram of multi-scale image feature extraction is shown. Next, the multi-scale image features and environmental features are fused to obtain fused features. These fused features serve as input to the dust coverage detection model. Based on the pre-learned relationship between the fused features and dust coverage, the model outputs the current dust coverage of the photovoltaic modules. Further, it determines whether the dust coverage exceeds a preset threshold. If so, a cleaning suggestion is generated and an alarm is issued. If not, a normal detection report is output. This output information is sent to the monitoring center and the terminal.

[0159] The output information mentioned above will also be used in the model optimization phase as a basis for model optimization. Specifically, the server will store the dust coverage detection results and collect the actual cleaning results. Based on these dust coverage detection and cleaning results, the model's performance will be evaluated. Then, based on the model performance evaluation results, the model parameters will be updated, thereby completing the continuous optimization of the model.

[0160] This embodiment integrates multi-scale visual features corresponding to photovoltaic modules with real-time environmental parameters, enabling the detection process to proceed without contact with the modules or collecting dust samples. It also effectively overcomes interference caused by changes in ambient light, thereby significantly improving the accuracy of dust detection in photovoltaic modules.

[0161] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0162] Based on the same inventive concept, this application also provides a dust detection device for photovoltaic modules to implement the dust detection method for photovoltaic modules described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations in one or more embodiments of the dust detection device for photovoltaic modules provided below can be found in the limitations of the dust detection method for photovoltaic modules described above, and will not be repeated here.

[0163] In one exemplary embodiment, such as Figure 9 As shown, a dust detection device for photovoltaic modules is provided, comprising:

[0164] The information acquisition module 902 is used to acquire the module image of the photovoltaic module and the environmental characteristics of the environment in which the photovoltaic module is located in response to the dust detection command for the photovoltaic module;

[0165] The feature extraction module 904 is used to extract features from the component image to obtain multiple image features that characterize the impact of dust coverage;

[0166] The feature fusion module 906 is used to fuse multiple image features and environmental features to obtain fused features;

[0167] The dust detection module 908 is used to detect dust coverage on photovoltaic modules based on fusion features and obtain the dust coverage detection results of photovoltaic modules.

[0168] In one embodiment, the plurality of image features include a first image feature for characterizing the global texture distribution of the photovoltaic module; the feature extraction module 904 further includes a first extraction unit for:

[0169] Obtain the spatial location information of each pixel in the component image;

[0170] Based on the spatial location information, multiple pixel pairs that satisfy the spatial location conditions are determined from each pixel point;

[0171] Obtain the gray value pair corresponding to each pixel pair, and construct the gray value matrix of the component image based on each gray value pair;

[0172] Based on the grayscale matrix, a first image feature is determined to characterize the global texture distribution of the photovoltaic module.

[0173] In one embodiment, the plurality of image features include a second image feature for characterizing the local texture distribution of the photovoltaic module; the feature extraction module 904 further includes a second extraction unit for:

[0174] For each pixel in the component image, search for multiple neighboring pixels corresponding to the pixel according to a preset pixel search range;

[0175] The gray value of each neighboring pixel is compared with the gray value of the pixel to obtain multiple gray value comparison results;

[0176] Based on the comparison results of each gray value, the first gray value parameter that characterizes the local gray value distribution of the pixel is determined.

[0177] Based on the first grayscale parameter corresponding to each pixel, a second image feature representing the local texture distribution of the photovoltaic module is determined.

[0178] In one embodiment, the plurality of image features include a third image feature for characterizing the edge texture distribution of the photovoltaic module; the feature extraction module 904 further includes a third extraction unit for:

[0179] Multiple edge pixels are identified from each pixel in the component image;

[0180] For each edge pixel, extract a second grayscale parameter that represents the degree of drastic grayscale change of the edge pixel;

[0181] Based on each of the second grayscale parameters, a third image feature is determined to characterize the edge texture distribution of the photovoltaic module.

[0182] In one embodiment, the dust detection module 908 is further configured to:

[0183] The dust cover detection model is invoked. This model is pre-trained based on multiple sample image features extracted from sample images of photovoltaic modules, corresponding sample environment features, and labeled true dust cover.

[0184] By analyzing the fusion characteristics using a dust coverage detection model, the dust coverage detection results of photovoltaic modules are obtained.

[0185] In one embodiment, the feature fusion module 906 is further configured to:

[0186] Importance analysis is performed on each image feature to obtain the importance analysis results for each image feature;

[0187] Determine the fusion weights that match the results of each importance analysis;

[0188] Based on each fusion weight, the image features are fused to obtain preliminary features;

[0189] The preliminary features are fused with the environmental features to obtain the fused features.

[0190] Each module in the aforementioned dust detection device for photovoltaic modules can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0191] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 10 As shown, the computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores dust detection data for photovoltaic modules. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a dust detection method for photovoltaic modules.

[0192] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 11As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When the computer program is executed by the processor, it implements a dust detection method for photovoltaic modules.

[0193] Those skilled in the art will understand that Figure 10 or Figure 11 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0194] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0195] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0196] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0197] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0198] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0199] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0200] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for detecting dust in photovoltaic modules, characterized in that, The method includes: In response to a dust detection command for a photovoltaic module, an image of the photovoltaic module and environmental characteristics of the environment in which the photovoltaic module is located are acquired. Feature extraction is performed on the component image to obtain multiple image features used to characterize the impact of dust coverage; The multiple image features and the environmental features are fused to obtain fused features; Based on the fusion characteristics, dust coverage detection is performed on the photovoltaic module to obtain the dust coverage detection result of the photovoltaic module.

2. The method according to claim 1, characterized in that, The plurality of image features include a first image feature used to characterize the global texture distribution of the photovoltaic module; The feature extraction of the component image yields multiple image features characterizing the impact of dust coverage, including: Obtain the spatial location information of each pixel in the component image; Based on the spatial location information, multiple pixel pairs that satisfy the spatial location conditions are determined from each of the pixel points; Obtain the grayscale value pair corresponding to each pixel pair, and construct the grayscale value matrix of the component image based on each grayscale value pair; Based on the grayscale matrix, a first image feature is determined to characterize the global texture distribution of the photovoltaic module.

3. The method according to claim 1, characterized in that, The plurality of image features include a second image feature used to characterize the local texture distribution of the photovoltaic module; The feature extraction of the component image yields multiple image features characterizing the impact of dust coverage, including: For each pixel in the component image, search for multiple neighboring pixels corresponding to the pixel according to a preset pixel search range; The gray value of each neighboring pixel is compared with the gray value of the pixel to obtain multiple gray value comparison results; Based on the comparison results of each gray value, a first gray value parameter is determined to characterize the local gray value distribution of the pixel. Based on the first grayscale parameter corresponding to each pixel, a second image feature representing the local texture distribution of the photovoltaic module is determined.

4. The method according to claim 1, characterized in that, The plurality of image features include a third image feature used to characterize the edge texture distribution of the photovoltaic module; The feature extraction of the component image yields multiple image features characterizing the impact of dust coverage, including: Multiple edge pixels are identified from each pixel of the component image; For each edge pixel, a second grayscale parameter is extracted to characterize the degree of drastic grayscale change of the edge pixel; Based on each of the second grayscale parameters, a third image feature is determined to characterize the edge texture distribution of the photovoltaic module.

5. The method according to claim 1, characterized in that, The step of detecting dust coverage on the photovoltaic module based on the fusion features to obtain the dust coverage detection result of the photovoltaic module includes: The dust coverage detection model is invoked, wherein the dust coverage detection model is pre-trained based on multiple sample image features extracted from sample images of the photovoltaic module, corresponding sample environment features, and labeled true dust coverage. The dust coverage detection results of the photovoltaic module are obtained by analyzing the fusion characteristics using the dust coverage detection model.

6. The method according to claim 1, characterized in that, The process of fusing the multiple image features and the environmental features to obtain fused features includes: An importance analysis is performed on each of the image features to obtain the importance analysis results for each image feature. Determine the fusion weights that match the results of each importance analysis. Based on the fusion weights, the image features are fused to obtain preliminary features; The preliminary features are fused with the environmental features to obtain the fused features.

7. A dust detection device for photovoltaic modules, characterized in that, The device includes: The information acquisition module is used to acquire, in response to a dust detection command for the photovoltaic module, an image of the photovoltaic module and the environmental characteristics of the environment in which the photovoltaic module is located; The feature extraction module is used to extract features from the component image to obtain multiple image features that characterize the impact of dust coverage; The feature fusion module is used to fuse the multiple image features and the environmental features to obtain fused features; A dust detection module is used to detect dust coverage on the photovoltaic module based on the fusion features, and obtain the dust coverage detection result of the photovoltaic module.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images