Dust detection result determination method and device, storage medium and electronic device

Abstract

The invention discloses a dust detection result determination method and device, a storage medium and an electronic device, and relates to the technical field of dust detection of photovoltaic panels, and the method comprises the steps: generating a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset corresponding relation; training the initial model by taking the simulated dust-containing image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample to obtain a dust detection model; and inputting the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determining a dust detection result according to the target dust density quantitative value. The problem that an existing dust detection method is not high in dust detection efficiency of the photovoltaic panel is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of dust detection of photovoltaic panels, in particular to a dust detection result determination method and device, a storage medium and an electronic device. BACKGROUND

[0002] As a core component of solar energy conversion, the surface dust deposition of photovoltaic panels can significantly reduce light transmittance and power generation efficiency, and needs to be detected for timely cleaning. The traditional method of detecting dust by manual detection has the problems of strong subjectivity and low detection efficiency due to environmental factors such as light. The detection technology based on sensors is limited in its wide application in photovoltaic panel dust detection due to the high cost of sensors, susceptibility to environmental factors, and complexity of installation and maintenance. At present, deep learning methods are applied to the detection of dust deposition on photovoltaic panels due to their strong ability in image processing. However, deep learning models need a large amount of labeled data for training, and inaccurate labeled data can easily lead to poor model detection results, thereby affecting the dust detection results.

[0003] Therefore, in the related art, there is a technical problem of low dust detection efficiency of existing dust detection methods for photovoltaic panels.

[0004] In view of the technical problem of low dust detection efficiency of existing dust detection methods for photovoltaic panels in the related art, an effective solution has not been proposed. SUMMARY

[0005] The embodiments of the present application provide a dust detection result determination method and device, a storage medium and an electronic device to at least solve the technical problem of low dust detection efficiency of existing dust detection methods for photovoltaic panels in the related art.

[0006] According to one embodiment of the embodiments of the present application, a dust detection result determination method is provided, including: generating a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship between a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust-containing image; training an initial model with the simulated dust-containing image as an input sample and a historical dust density quantitative value corresponding to the second pixel value as an output sample to obtain a dust detection model; inputting a preprocessed standard image of the photovoltaic panel to the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determining a dust detection result according to the target dust density quantitative value.

[0007] In an example embodiment, before the pre-processed standard image of the photovoltaic panel is input into the dust detection model to obtain the target dust density quantitative value output by the dust detection model, the method comprises: performing an image correction operation on the collected original image of the photovoltaic panel to obtain a first corrected image; performing a silver line removal operation on the first image to obtain a second image; and performing an image enhancement operation on the second image after gray scale processing to obtain the standard image.

[0008] In an example embodiment, the method of generating an analog dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship comprises: the preset correspondence relationship is represented by the following formula: I(m) = J(m)t(m) + a(1-t(m)); wherein J(m) represents a first pixel value, I(m) represents a second pixel value, t(m) represents the medium transmittance, a is a global atmospheric light value, and m is a natural number; a new second pixel value corresponding to a preset medium transmittance is calculated by the formula, and an image corresponding to the new second pixel value is determined as the analog dust-containing image.

[0009] In an example embodiment, the method further comprises: determining the initial model constructed according to a preset model architecture, wherein the preset model architecture comprises a feature extraction layer, a residual layer and a fully connected layer; the feature extraction layer is used to extract feature vectors of the analog dust-containing image in parallel through multi-scale convolution kernels, and outputs a fusion result of the feature vectors to the residual layer; the residual layer is used to perform a residual operation on the fusion result, and outputs an operation result of the residual operation to the fully connected layer; and the fully connected layer is used to perform dimension reduction and linear transformation processing on the feature vectors in the operation result, and outputs the transmittance.

[0010] In an example embodiment, the method further comprises: in the process of training the initial model, using a mean square error function as a loss function, and using minimization of a function value of the loss function as a training target to iteratively train the loss function and an optimizer.

[0011] wherein the mean square error function is represented as

[0012] wherein L(Θ) is the function value of the loss function, Θ represents a learning parameter of the model, N is the total number of samples, and N is a positive integer, When Θ is given, the initial model is based on input The output prediction transmittance is is the i-th analog dust-containing image, t i is the real transmittance of the i-th historical image.

[0013] In an example embodiment, determining the dust detection result according to the target dust density quantitative value comprises: generating a transmittance distribution map corresponding to the target dust density quantitative value in a visual manner, and determining the dust detection result according to the transmittance distribution map; and in a case where the target dust density quantitative value is greater than a preset threshold, sending a cleaning instruction to a target object, wherein the cleaning instruction is used to prompt the target object to clean the photovoltaic panel.

[0014] According to another aspect of the embodiments of the present application, a dust detection result determination apparatus is further provided, comprising: a generation module configured to generate a simulation dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulation dust-containing image; a obtaining module configured to obtain a dust detection model by training an initial model with the simulation dust-containing image as an input sample and a historical dust density quantitative value corresponding to the second pixel value as an output sample; and a determination module configured to input a preprocessed standard image of the photovoltaic panel into the dust detection model, obtain a target dust density quantitative value output by the dust detection model, and determine a dust detection result according to the target dust density quantitative value.

[0015] According to still another aspect of the embodiments of the present application, a computer readable storage medium is further provided, and the computer readable storage medium stores a computer program, wherein the computer program is configured to execute the dust detection result determination method when running.

[0016] According to still another aspect of the embodiments of the present application, an electronic device is further provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the dust detection result determination method through the computer program.

[0017] According to still another aspect of the embodiments of the present application, a computer program product is further provided, comprising a computer program, and the computer program is executed by a processor to implement the dust detection result determination method.

[0018] In the embodiment of the present application, the historical image of the photovoltaic panel is generated based on a preset corresponding relationship, wherein the preset corresponding relationship represents the corresponding relationship between the first pixel value of the historical image, the medium transmittance of the photovoltaic panel, and the second pixel value of the simulated dust-containing image; the initial model is trained by taking the simulated dust-containing image as the input sample and the historical dust density quantitative value corresponding to the second pixel value as the output sample, to obtain a dust detection model; the preprocessed standard image of the photovoltaic panel is input into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and a dust detection result is determined according to the target dust density quantitative value; by using the above technical solution, the present application detects the dust on the photovoltaic panel based on the light transmittance, that is, constructs a dust detection model to realize quantitative evaluation of the dust density, solves the technical problem that the existing dust detection method has low dust detection efficiency on the photovoltaic panel, and further improves the dust detection efficiency. BRIEF DESCRIPTION OF DRAWINGS

[0019] The accompanying drawings, which are incorporated into and form a part of the specification, illustrate preferred embodiments of the present application and, together with the description, serve to explain the principles of the present application.

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the accompanying drawings required to be used in the embodiments or prior art description will be briefly introduced. Obviously, for those skilled in the art, other drawings can also be obtained based on these drawings without creative labor.

[0021] Figure 1 Fig. 1 is a hardware environment schematic diagram of a dust detection result determination method according to an embodiment of the present application;

[0022] Figure 2 Fig. 2 is a flowchart of a dust detection result determination method according to an embodiment of the present application;

[0023] Figure 3 Fig. 3 is a flowchart of a dust detection result determination method according to an embodiment of the present application;

[0024] Figure 4 Fig. 4 is an architecture schematic diagram of a DVNET detection model according to an embodiment of the present application;

[0025] Figure 5 Fig. 5 is a structural block diagram of a dust detection result determination device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In the following, the technical solutions in the embodiments of the present application will be described clearly and completely in conjunction with the drawings in the embodiments of the present application, so that those skilled in the art can better understand the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work should be within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second" and the like in the description and claims of the present application and the above drawings are used to distinguish similar objects, and do not necessarily indicate a specific order or a sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units need not be limited to those steps or units clearly listed, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] According to an aspect of the embodiments of the present application, a method for determining a dust detection result is provided. The method provided in the embodiments of the present application can be executed in a terminal device, a computer terminal or a similar computing device. Taking the case of running on a terminal device as an example, Figure 1 is a hardware structure block diagram of a terminal device of a method for determining a dust detection result according to an embodiment of the present application. As shown in Figure 1 , the terminal device can include one or more (only one is shown in Figure 1 ) processors 102 (the processor 102 can include but is not limited to a processing device such as a microprocessor or a field programmable gate array (FPGA)) and a memory 104 for storing data, wherein the terminal device can further include a transmission device 106 for communication function and an input and output device 108. Those skilled in the art can understand that Figure 1 the structure shown is only schematic, which does not limit the structure of the terminal device. For example, the terminal device can further include more or less components than those shown in Figure 1 , or have a different configuration from that shown in Figure 1 .

[0029] The memory 104 can be used to store computer programs, such as software programs of application software and modules, such as a computer program corresponding to the dust detection result determination method in the embodiments of the present application. The processor 102 performs various functional applications and data processing, that is, implements the above method, by running the computer program stored in the memory 104. The memory 104 can include a high-speed random access memory, and can also include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some examples, the memory 104 can further include a memory remotely arranged with respect to the processor 102, which can be connected to the terminal device through a network. Examples of the above network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and a combination thereof.

[0030] The transmission device 106 is used to receive or send data via a network. Specific examples of the above network can include a wireless network provided by a communication provider of the terminal device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, NIC for short), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 can be a radio frequency (Radio Frequency, RF for short) module, which is used to communicate with the Internet in a wireless manner.

[0031] In the embodiments, a dust detection result determination method is provided, which is applied to the above terminal device, Figure 2 is a flowchart of the dust detection result determination method according to the embodiments of the present application, and the flowchart includes the following steps:

[0032] In step S202, a simulated dust-containing image corresponding to a historical image of a photovoltaic panel is generated based on a preset correspondence relationship, where the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust-containing image.

[0033] In step S204, an initial model is trained by taking the simulated dust-containing image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model.

[0034] Optionally, in the training process of this step, the initial model can be further trained by taking the historical image as an input sample and taking a historical dust density quantitative value corresponding to the first pixel value as an output sample, to improve the model training effect.

[0035] In step S206, the pre-processed standard image of the photovoltaic panel is input into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and a dust detection result is determined according to the target dust density quantitative value.

[0036] Through the above steps, in the embodiment of the present application, a historical image of a photovoltaic panel is generated based on a preset corresponding relationship, wherein the preset corresponding relationship represents a corresponding relationship between a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of a simulated dust-containing image; an initial model is trained by taking the simulated dust-containing image as an input sample and a historical dust density quantitative value corresponding to the second pixel value as an output sample to obtain a dust detection model; a pre-processed standard image of the photovoltaic panel is input into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and a dust detection result is determined according to the target dust density quantitative value; by using the above technical solution, the present application detects dust on the photovoltaic panel based on light transmittance, that is, a dust detection model is constructed to realize quantitative evaluation of dust density, thereby solving the technical problem that the dust detection efficiency of the existing dust detection method on the photovoltaic panel is not high, and further improving the dust detection efficiency.

[0037] In one exemplary embodiment, before the pre-processed standard image of the photovoltaic panel is input into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, an image correction operation can also be performed on the collected original image of the photovoltaic panel to obtain a corrected first image; a silver line removal operation is performed on the first image to obtain a second image; and an image enhancement operation is performed on the grayed second image to obtain the standard image. Image preprocessing is a key step to ensure the accuracy of subsequent dust detection. It improves the quality of the image by correcting the original image, removing interference factors (such as silver lines), and enhancing the contrast, thereby facilitating the dust detection model to more accurately identify the dust area. This series of operations solves the problem that photovoltaic panel images are difficult to analyze uniformly under different lighting conditions, ensuring the consistency and reliability of model input. Using techniques including but not limited to histogram equalization, adaptive contrast enhancement, etc., the clarity and contrast of the image can be further improved, making the dust detection more accurate.

[0038] In an example embodiment, the process of generating a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship includes: the preset correspondence relationship is represented by the following formula: I(m) = J(m)t(m) + a(1-t(m)); where J(m) represents a first pixel value, I(m) represents a second pixel value, t(m) represents the medium transmittance, a is a global atmospheric light value, and m is a natural number; a new second pixel value corresponding to a preset medium transmittance is calculated by the formula, and an image corresponding to the new second pixel value is determined as the simulated dust-containing image.

[0039] Optionally, the historical image can be divided into n x n pixel blocks (for example, n is 16, and different values can be selected according to different conditions), and then the transmittance t(m) is randomly generated n times. After each time of randomly generating the transmittance t(m), a corresponding dust-containing image pixel value is obtained, and then a simulated dust-containing image is obtained, wherein the dust-containing image pixel value is used to represent the dust density, and finally a dust-containing image simulating different dust densities is obtained, and a training data set is constructed.

[0040] This formula takes into account the effect of dust blocking on light penetration, and by adjusting the value of t(m), different degrees of dust coverage effect can be simulated. Through the combination of this technical feature, the inaccuracy and inefficiency of traditional detection methods in complex environments are solved, and high-precision detection of the dust deposition degree of photovoltaic panels is realized. By changing the transmittance, a series of simulated images with different dust densities can be generated, which serve as training samples to help the model learn the influence law of dust on the light transmittance of photovoltaic panels. This method not only improves the generalization ability of the model, but also can cope with the uncertainty of dust deposition in the actual environment, thereby ensuring the accuracy and stability of dust detection. In addition, this technical feature can be extended to simulate multiple pollution sources, such as water stains and bird droppings, to comprehensively evaluate the cleaning condition of photovoltaic panels.

[0041] Further, in an example embodiment, the initial model constructed according to a preset model architecture can be determined, wherein the preset model architecture includes a feature extraction layer, a residual layer, and a fully connected layer; the feature extraction layer is configured to extract feature vectors of the simulated dust-containing image in parallel through multi-scale convolution kernels, and output a fusion result of the feature vectors to the residual layer; the residual layer is configured to perform a residual operation on the fusion result, and output an operation result of the residual operation to the fully connected layer; and the fully connected layer is configured to perform dimension reduction and linear transformation processing on the feature vectors in the operation result, and output transmittance. The innovative architecture of the DVNET model realizes deep learning processing of the simulated dust-containing image through the combination of the feature extraction layer, the residual layer, and the fully connected layer. The feature extraction layer uses multi-scale convolution kernels to capture different detailed features of dust deposition in the image, thereby enhancing the model's understanding of dust distribution. The residual layer solves the gradient vanishing problem in deep network training through the residual operation, thereby ensuring the stability and training efficiency of the model. The fully connected layer is responsible for converting the feature vectors into transmittance, which is directly related to the quantitative indicator of dust density. This architecture design not only improves the detection accuracy of the model, but also reduces the demand for computing resources, so that the model can efficiently run on edge devices.

[0042] In an example embodiment, during the training of the initial model, a mean square error function is used as a loss function, and the loss function and an optimizer are iteratively trained with the goal of minimizing the function value of the loss function.

[0043] wherein the mean square error function is expressed as

[0044] wherein L(Θ) is the function value of the loss function, Θ represents the learning parameter of the model, N is the total number of samples, and N is a positive integer. When Θ is given, the initial model is based on the input The output predicted transmittance is is the i-th simulated dust-containing image, t i is the true transmittance of the i-th historical image.

[0045] During the model training process, the mean square error function is used as the loss function, aiming to minimize the difference between the predicted transmittance of the model and the true transmittance, thereby improving the prediction accuracy of the model. Through iterative training, the model can continuously adjust its learning parameter Θ until the function value L(Θ) of the loss function is minimized. This process solves the bias and overfitting problems that may exist in the initial stage of the model, ensuring the generalization ability of the model on new data.

[0046] In an exemplary embodiment, the scheme for determining the dust detection result based on the quantitative value of the target dust density includes: generating a transmittance distribution map corresponding to the quantitative value of the target dust density in a visual manner, and determining the dust detection result based on the transmittance distribution map; and, when the quantitative value of the target dust density is determined to be greater than a preset threshold, sending a cleaning instruction to the target object, wherein the cleaning instruction is used to prompt the target object to clean the photovoltaic panel. The visual presentation of the transmittance distribution map provides photovoltaic system maintenance personnel with intuitive information on the spatial distribution of dust deposition, which helps to quickly locate areas that need cleaning. When the detected dust density exceeds the preset threshold, the system automatically issues a cleaning instruction to prompt maintenance personnel or automated cleaning devices to perform cleaning work. The introduction of this technical feature solves the problems of inefficiency of manual periodic inspections and uncertainty in judging the timing of cleaning, realizing intelligent management of photovoltaic panel cleaning.

[0047] To better understand the process of determining the dust detection results described above, the implementation flow of the method for determining the dust detection results will be further explained below with reference to optional embodiments, but this is not intended to limit the technical solutions of the embodiments of this application.

[0048] This embodiment provides a method for determining dust detection results. Figure 3 This is a flowchart illustrating a method for determining dust detection results according to an embodiment of this application, as shown below. Figure 3 As shown, the specific steps are as follows:

[0049] Step S301: Data Acquisition. Acquire clean images (i.e., raw images) of the photovoltaic panel using a visible light camera.

[0050] Step S302: Image preprocessing, performing preprocessing operations such as image correction, silver line removal, and noise filtering.

[0051] Specifically, image distortion is corrected through perspective transformation to remove interfering elements such as silver lines, resulting in a standardized image. The standardized image is then subjected to grayscale conversion and noise filtering. Next, HSV (Hue Saturation Value) is used for color space segmentation and morphological dilation to remove the silver lines. Finally, Gaussian smoothing is applied to enhance image continuity and improve the input quality for subsequent models.

[0052] The following is the process of obtaining the training set:

[0053] Historical images of photovoltaic panels are acquired using a visible light camera. Preprocessing operations such as image correction, silver line removal, and noise filtering are performed, and a dust-laden image dataset is generated based on the principle of light attenuation.

[0054] Based on the atmospheric scattering model, the dust deposition process is simulated by formula I(m) = J(m)t(m) + a(l-t(m)), wherein J(m) is the pixel value of the clean image (i.e. the first pixel value), I(m) is the pixel value of the dust-containing image (i.e. the second pixel value), t(m) is the medium transmittance, and a is the global atmospheric light value (a constant value). The normalized image is divided into n*n pixel blocks (n is 16 in this application, and different values can be selected according to different conditions), and then the transmittance t(m) is randomly generated n times. After each random generation of the transmittance t(m), the corresponding pixel value of the dust-containing image is obtained, and then the simulated dust-containing image is obtained, wherein the pixel value of the dust-containing image is used to represent the dust density, and finally the dust-containing images with different dust densities are obtained to construct the training data set. The randomly generated transmittance represents the random generation within a reasonable range of transmittance values.

[0055] For example, a high-definition camera is used to collect an RGB image of a clean photovoltaic panel with a resolution of 1920*1080 (according to the performance of the actual use equipment), the image distortion is corrected by a perspective transformation matrix, 4 groups of corresponding points are selected to solve the transformation parameters, the mapping from the image plane to the real plane is realized, the obtained image is converted to the HSV color space, the region is segmented by the threshold value, the mask range is expanded by the morphological dilation operation, and the region after the silver line is removed is filled by combining the Navier-Stokes algorithm. Then, the clean image is divided into 16*16 pixel blocks, a data set containing N samples is generated, 80% of the data is used for training, and 20% of the data is used for verification.

[0056] Step S303: model construction. The model can be constructed in combination with Figure 4 The model architecture is described. Taking the end-to-end neural network DVNET as an example, the model architecture includes a multi-scale convolution feature extraction layer, a residual layer (corresponding to Figure 4 the residual block and the SEBlock attention mechanism in the application), and a fully connected (FC) layer.

[0057] In this application, DVNET represents a deep learning network designed for photovoltaic panel dust detection for processing image data, which combines multiple advanced deep learning technologies such as multi-scale convolution, residual connection and SEBlock attention mechanism, aiming to provide an efficient, accurate and low-cost photovoltaic panel dust detection solution.

[0058] SEBlock, full name Squeeze-and-Excitation Block, is a channel attention mechanism used in deep learning models, which allows the model to adaptively calibrate the feature response of each channel, thereby enhancing the model's focus on important features while suppressing irrelevant or redundant information.

[0059] In the feature extraction layer, 3x3, 5x5, 7x7 multi-scale convolution kernels are used to extract features of different spatial resolutions in parallel, and the multi-scale feature maps are output after fusion. The convolution formula is wherein is the output after batch normalization (BN) and ReLU activation of each scale convolution, k = 2i + 1, i is a natural number.

[0060] In the residual block, residual connection is introduced to avoid gradient vanishing in deep network, and SEBlock attention mechanism is combined to adaptively calibrate channel weights using "squeeze-activation operation" to enhance the model's attention to dust areas.

[0061] In the full connection layer, the dimensionality is reduced by average pooling, and the quantitative value of dust density is output after two linear transformations to realize quantitative estimation of dust density.

[0062] Alternatively, the average transmittance and transmittance distribution map are directly output by two linear transformations after dimensionality reduction by average pooling, thereby realizing quantitative estimation and spatial distribution visualization of dust density. It should be noted that this process requires designing a model architecture capable of handling multiple outputs to simultaneously predict multiple related target variables, i.e., the model needs to predict a quantitative value of historical dust density and the distribution of the entire dust density. For example, a multi-head model or a multi-task learning model is constructed, and independent output branches are set in the last one or several layers of the model, each branch being responsible for predicting one target. Moreover, by adjusting the relative weights of the two tasks in the loss function, the performance of the model on the two output targets is balanced.

[0063] Step S304: model training and optimization. In the training process, mean square error is used as the loss function, as shown in the following formula.

[0064]

[0065] L(Θ) is the function value of the loss function, Θ represents the learning parameters of the model, N is the total number of samples, N is a positive integer, When Θ is given, the initial model is based on the input The output predicted transmittance is is the i-th simulated dust image, t i is the real transmittance of the i-th historical image.

[0066] wherein the learning parameters of the model represent the hyperparameters, weights, bias terms, and other parameters of the DVNET model.

[0067] This formula measures the transmittance The average gap between the corresponding real transmittance, the goal is to adjust Θ to make As close as possible to the real transmittance, to minimize this loss, so as to improve the prediction accuracy of the model.

[0068] Iterative training with Adam optimizer, through a number of comparative experiments to verify the performance of the model.

[0069] Step S305: model output results.

[0070] Through the above steps, through the accurate mapping of light transmittance and dust density, combined with multi-scale feature extraction and SEBlock attention mechanism, high-precision estimation of dust deposition is realized, which effectively improves the accuracy of detection. Secondly, the lightweight design of DVNET model greatly reduces the consumption of computing resources, shortens the training time, makes its application more extensive on large-scale data set, enhances the scalability and practicality of the model. Finally, the visualization of transmittance distribution map not only reflects the spatial distribution of dust intuitively, but also supports real-time monitoring and cleaning strategy optimization, greatly reducing the workload of manual maintenance.

[0071] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be realized by means of software and necessary general hardware platform, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of software product, which is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including a plurality of instructions for making a terminal device (which can be a mobile phone, computer, server, or network device, etc.) execute the method of each embodiment of the present application.

[0072] Figure 5 is a structural block diagram of a dust detection result determination device according to an embodiment of the present application; as Figure 5 shown, comprising:

[0073] The generation module 52 is configured to generate a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence, wherein the preset correspondence represents the correspondence between the first pixel value of the historical image, the medium transmittance of the photovoltaic panel, and the second pixel value of the simulated dust-containing image.

[0074] The obtaining module 54 is configured to train an initial model by taking the simulated dust-containing image as an input sample and taking the historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model.

[0075] The determining module 56 is configured to input the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determine a dust detection result according to the target dust density quantitative value.

[0076] By the above device, in the embodiment of the application, the historical image of the photovoltaic panel is generated based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust-containing image; the initial model is trained by taking the simulated dust-containing image as an input sample and a historical dust density quantitative value corresponding to the second pixel value as an output sample to obtain a dust detection model; the preprocessed standard image of the photovoltaic panel is input into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and a dust detection result is determined according to the target dust density quantitative value; by using the above technical solution, the application detects the dust on the photovoltaic panel based on the light transmittance, that is, the dust detection model is constructed to realize the quantitative evaluation of the dust density, thereby solving the technical problem that the existing dust detection method has low dust detection efficiency on the photovoltaic panel, and further improving the dust detection efficiency.

[0077] In one example embodiment, the determining module is further configured to perform an image correction operation on the collected original image of the photovoltaic panel to obtain a corrected first image before inputting the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model; perform a silver line removal operation on the first image to obtain a second image; and perform an image enhancement operation on the grayed second image to obtain the standard image.

[0078] In one example embodiment, the generating module is further configured to: the preset correspondence relationship is represented by the following formula: I(m) = J(m)t(m) + a(1-t(m)); wherein J(m) represents the first pixel value, I(m) represents the second pixel value, t(m) represents the medium transmittance, a is a global atmospheric light value, and m is a natural number; a new second pixel value corresponding to a preset medium transmittance is calculated by the formula, and an image corresponding to the new second pixel value is determined as the simulated dust-containing image.

[0079] In an example embodiment, the obtaining module is further configured to determine the initial model constructed according to a preset model architecture, wherein the preset model architecture comprises a feature extraction layer, a residual layer, and a fully connected layer; the feature extraction layer is configured to extract feature vectors of the simulated dust-containing image in parallel through multi-scale convolution kernels, output a fusion result of the feature vectors to the residual layer; the residual layer is configured to perform a residual operation on the fusion result, and output an operation result of the residual operation to the fully connected layer; and the fully connected layer is configured to perform dimension reduction and linear transformation processing on the feature vectors in the operation result, and output the transmittance.

[0080] In an example embodiment, the obtaining module is further configured to, in the process of training the initial model, perform iterative training in combination with a loss function and an optimizer, wherein the loss function is a mean square error function, and a function value of the loss function is minimized as a training target.

[0081] The mean square error function is expressed as

[0082] wherein L(Θ) is the function value of the loss function, Θ represents a learning parameter of the model, and N is a total number of samples, N being a positive integer. When Θ is given, the initial model is based on input The output predicted transmittance is is the i-th simulated dust-containing image, t i is the real transmittance of the i-th historical image.

[0083] In an example embodiment, the determining module is further configured to generate a transmittance distribution map corresponding to the target dust density quantitative value in a visual manner, and determine the dust detection result according to the transmittance distribution map; and in a case where the target dust density quantitative value is greater than a preset threshold, send a cleaning instruction to a target object, wherein the cleaning instruction is used to prompt the target object to clean the photovoltaic panel.

[0084] Embodiments of the present application also provide a storage medium comprising a stored program, wherein the program performs any of the above methods when executed.

[0085] Optionally, in the present embodiment, the storage medium can be configured to store program code for performing the following steps:

[0086] S1, generating a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust-containing image;

[0087] S2, training an initial model by taking the simulated dust image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model;

[0088] S3, inputting the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determining a dust detection result according to the target dust density quantitative value.

[0089] Embodiments of the present application also provide an electronic device including a memory and a processor, the memory storing a computer program, and the processor being configured to execute the computer program to perform the steps in any of the above method embodiments.

[0090] Optionally, the electronic device can further include a transmission device and an input / output device, wherein the transmission device is connected to the processor, and the input / output device is connected to the processor.

[0091] Optionally, in the embodiment, the processor can be configured to execute the following steps through the computer program:

[0092] S1, generating a simulated dust image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust image;

[0093] S2, training an initial model by taking the simulated dust image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model;

[0094] S3, inputting the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determining a dust detection result according to the target dust density quantitative value.

[0095] Optionally, in the embodiment, the storage medium can include but is not limited to a U disk, a read-only memory (ROM), a random access memory (RAM), a mobile hard disk, a magnetic disk or an optical disk, and various media that can store program codes.

[0096] Optionally, the embodiments of the present application also provide a computer program product, and the computer program product includes a computer program, which, when executed by a processor, implements the steps in any of the above method embodiments.

[0097] Optionally, the embodiment of the present application further provides another computer program product, comprising a nonvolatile computer readable storage medium, the nonvolatile computer readable storage medium stores a computer program, the computer program is executed by a processor to realize the steps in any one of the method embodiments.

[0098] Optionally, the embodiment of the present application further provides a computer program, the computer program comprises computer instructions stored in a computer readable storage medium; a processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions to enable the computer device to execute the steps in any one of the method embodiments.

[0099] Optionally, the specific examples in the embodiment can refer to the examples described in the above embodiments and optional implementation manners, and the embodiment will not be described here.

[0100] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present application can be realized by general computing devices, which can be concentrated on a single computing device, or distributed on a network composed of multiple computing devices, and optionally, they can be realized by program codes executable by computing devices, so that they can be stored in storage devices and executed by computing devices, and in some cases, the steps shown or described can be executed in different order, or they can be manufactured into individual integrated circuit modules, or multiple modules or steps can be manufactured into a single integrated circuit module. Thus, the present application is not limited to any specific combination of hardware and software.

[0101] The above only describes the preferred embodiments of the present application, and it should be pointed out that, for those skilled in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, and these improvements and refinements should be regarded as the protection scope of the present application.

Claims

1. A method of determining a dust detection result, characterized by, The method comprises: generating a simulated dust image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust image; training an initial model by taking the simulated dust image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model; inputting a preprocessed standard image of the photovoltaic panel into the dust detection model to obtain a target dust density quantitative value output by the dust detection model, and determining a dust detection result according to the target dust density quantitative value.

2. The method of claim 1, wherein Before inputting the preprocessed standard image of the photovoltaic panel into the dust detection model to obtain the target dust density quantitative value output by the dust detection model, the method comprises: performing an image correction operation on a collected original image of the photovoltaic panel to obtain a corrected first image; performing a silver line removal operation on the first image to obtain a second image; performing an image enhancement operation on the grayed second image to obtain the standard image.

3. The method of claim 1, wherein The method of generating a simulated dust image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship comprises: The preset correspondence relationship is represented by the following formula: I(m) = J(m)t(m) + a(1-t(m)); wherein J(m) represents a first pixel value, I(m) represents a second pixel value, t(m) represents the medium transmittance, a is a global atmospheric light value, and m is a natural number; a new second pixel value corresponding to a preset medium transmittance is calculated through the formula, and an image corresponding to the new second pixel value is determined as the simulated dust image.

4. The method of claim 1, wherein The method further comprises: determining the initial model constructed according to a preset model architecture, wherein the preset model architecture comprises a feature extraction layer, a residual layer, and a fully connected layer; the feature extraction layer is configured to extract feature vectors of the simulated dust image in parallel through multi-scale convolution kernels, and output a fusion result of the feature vectors to the residual layer; the residual layer is configured to perform a residual operation on the fusion result, and output an operation result of the residual operation to the fully connected layer; the fully connected layer is configured to perform dimension reduction and linear transformation processing on the feature vectors in the operation result, and output a transmittance.

5. The method of claim 4, wherein The method further comprises: during the training of the initial model, iteratively training the loss function and an optimizer with a mean square error function as the loss function and minimization of a function value of the loss function as a training target; wherein the mean square error function is expressed as where L(Θ) is a function value of the loss function, Θ represents a learning parameter of the model, N is a total number of samples, and N is a positive integer, The initial model is based on input The output predicted transmittance, is the i-th simulated dusty image, t i is the real transmittance of the i-th historical image.

6. The method of claim 1, wherein determining the dust detection result according to the target dust density quantitative value comprises: generating a transmittance distribution map corresponding to the target dust density quantitative value in a visual manner, and determining the dust detection result according to the transmittance distribution map; and in the case where the target dust density quantitative value is greater than a preset threshold, sending a cleaning instruction to a target object, wherein the cleaning instruction is used to prompt the target object to clean the photovoltaic panel.

7. A dust detection result determination apparatus characterized by comprising: The method comprises: The generating module is configured to generate a simulated dust-containing image corresponding to a historical image of a photovoltaic panel based on a preset correspondence relationship, wherein the preset correspondence relationship represents a correspondence relationship among a first pixel value of the historical image, a medium transmittance of the photovoltaic panel, and a second pixel value of the simulated dust-containing image. The obtaining module is configured to train an initial model by taking the simulated dust-containing image as an input sample and taking a historical dust density quantitative value corresponding to the second pixel value as an output sample, to obtain a dust detection model.

8. A computer readable storage medium, characterized in that, The computer readable storage medium comprises a stored program, wherein the program performs the method of any one of claims 1 to 6 when executed. 9.An electronic device comprising a memory and a processor, the electronic device characterized by, The memory stores a computer program, and the processor is configured to execute the method of any one of claims 1 to 6 by using the computer program.

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

Images