Photovoltaic Image Denoising Method, Computer-Readable Storage Medium, Electronic Device

Abstract

The present invention discloses a photovoltaic image denoising method, a computer-readable storage medium, and an electronic device, relating to the technical field of photovoltaics. The method includes: decomposing a photovoltaic image to be denoised into a plurality of latent variables, where the latent variables follow a normal distribution, and the number of the latent variables is determined according to the photovoltaic image to be denoised; removing a target latent variable from the plurality of latent variables, and converting the remaining latent variables into a target image to obtain a photovoltaic image after removing shadow noise, where the target latent variable is the latent variable corresponding to the shadow noise. Thus, photovoltaic image denoising can be achieved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic technology, and in particular, to a photovoltaic image denoising method, a computer-readable storage medium, and an electronic device. Background Art

[0002] Due to the many advantages of solar energy, such as pollution-free, wide distribution, and renewable, it is widely regarded as one of the most promising clean energy sources. However, in a photovoltaic power station, cables and wires with a diameter of a few centimeters often block the photovoltaic modules, forming linear shadows. Since the diameter of overhead cables and the like is relatively thin and the distance from the photovoltaic modules is relatively far, the linear shadows have little impact on power generation itself, but they will interfere with the identification of other photovoltaic faults such as hot spots. Especially when using automated means such as AI image recognition for component fault identification, the linear shadows can be regarded as a kind of image noise in image processing. If it cannot be effectively eliminated, it will directly affect the accuracy of subsequent AI image recognition. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems in the related art to some extent. To this end, the first object of the present invention is to propose a photovoltaic image denoising method to achieve photovoltaic image denoising.

[0004] The second object of the present invention is to propose a computer-readable storage medium.

[0005] The third object of the present invention is to propose an electronic device.

[0006] To achieve the above object, an embodiment of the first aspect of the present invention proposes a photovoltaic image denoising method, including: decomposing a photovoltaic image to be denoised into a plurality of latent variables, where the latent variables follow a normal distribution, and the number of the latent variables is determined according to the photovoltaic image to be denoised; removing a target latent variable from the plurality of latent variables, and converting the remaining latent variables into a target image to obtain a photovoltaic image after removing shadow noise, where the target latent variable is the latent variable corresponding to the shadow noise.

[0007] In addition, the photovoltaic image denoising method according to the embodiment of the present invention may further have the following additional technical features:

[0008] In an embodiment of the present invention, the process of determining the number of the latent variables includes: dividing the photovoltaic image to be denoised into a plurality of image blocks; classifying the plurality of image blocks to obtain a classification result; and obtaining the number of the latent variables according to the classification result and the size of the image to be denoised.

[0009] In one embodiment of the present invention, the classifying of the multiple image blocks includes: calculating the similarity between the image blocks; and classifying the multiple image blocks according to the similarity.

[0010] In one embodiment of the present invention, the classification results include shadowless photovoltaic panels, shaded photovoltaic panels, and backgrounds; the number of latent variables is obtained by the following formula:

[0011] ,

[0012] where N represents the number of latent variables, W represents the width of the image to be denoised, H represents the height of the image to be denoised, and P represents the proportion of the image blocks of the shaded photovoltaic panels among the multiple image blocks.

[0013] In one embodiment of the present invention, a pre-trained encoder is used to decompose the photovoltaic image to be denoised into multiple latent variables, wherein the encoder adopts the reparameterization method to implement backpropagation.

[0014] In one embodiment of the present invention, the reparameterization method includes: obtaining the latent variables by the following formula:

[0015] z = μ + e σ

[0016] where z represents the latent variable, e represents Gaussian noise satisfying the unit normal distribution, and μ and σ are the mean and variance of the variables generated by the encoder respectively. represents multiplication.

[0017] In one embodiment of the present invention, a pre-trained decoder is used to remove the target latent variable from the multiple latent variables and convert the remaining latent variables into a target image.

[0018] In one embodiment of the present invention, the output end of the encoder is connected to the input end of the decoder, and the training process of the encoder and the decoder includes: obtaining multiple shadowless photovoltaic images, and respectively superimposing shadow data on each of the shadowless photovoltaic images to obtain multiple shadowed photovoltaic images; for each of the shadowed photovoltaic images, inputting the shadowed photovoltaic image into the encoder, and outputting a denoised photovoltaic image through the decoder; obtaining a loss function according to the output variable distribution of the encoder and the standard normal distribution, and the denoised photovoltaic image and the corresponding shadowless photovoltaic image; and updating the parameters of the encoder and the decoder according to the loss function.

[0019] To achieve the above object, a second aspect embodiment of the present invention proposes a computer-readable storage medium storing a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0020] To achieve the above object, an embodiment of the third aspect of the present invention provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0021] According to the photovoltaic image denoising method, computer-readable storage medium, and electronic device of the embodiments of the present invention, a photovoltaic image to be denoised is decomposed into a plurality of latent variables, where the latent variables follow a normal distribution, and the number of the latent variables is determined according to the photovoltaic image to be denoised; a target latent variable among the plurality of latent variables is removed, and the remaining latent variables are converted into a target image to obtain a photovoltaic image after removing shadow noise, where the target latent variable is the latent variable corresponding to the shadow noise. Thus, photovoltaic image denoising can be achieved.

[0022] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 is a flowchart of the photovoltaic image denoising method according to an embodiment of the present invention;

[0024] Figure 2 is an image of a photovoltaic module without shadow as an example;

[0025] Figure 3 is an image of a photovoltaic module with shadow as an example;

[0026] Figure 4 is a schematic diagram of the photovoltaic image denoising method according to an example of the present invention;

[0027] Figure 5 is a schematic diagram of a similar block as an example;

[0028] Figure 6 is a photovoltaic image to be denoised according to an example of the present invention;

[0029] Figure 7 is a schematic diagram of dividing the photovoltaic image to be denoised into blocks according to an example of the present invention;

[0030] Figure 8 is a schematic diagram of reparameterization according to an embodiment of the present invention;

[0031] Figure 9 is a schematic diagram of the photovoltaic image denoising method according to another example of the present invention;

[0032] Figure 10 is a schematic diagram of the photovoltaic image denoising method according to still another example of the present invention;

[0033] Figure 11 is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed implementation manners

[0034] The following describes a photovoltaic image denoising method, a computer-readable storage medium, and an electronic device according to an embodiment of the present invention, where the same or similar reference numerals represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described with reference to the accompanying drawings are exemplary and should not be construed as limiting the present invention.

[0035] Figure 1 is a flowchart of a photovoltaic image denoising method according to an embodiment of the present invention.

[0036] As Figure 1 shown, the photovoltaic image denoising method includes:

[0037] S11, decomposing a photovoltaic image to be denoised into a plurality of latent variables, where the latent variables follow a normal distribution, and the number of latent variables is determined according to the photovoltaic image to be denoised.

[0038] Specifically, when performing aerial inspection on a photovoltaic module by a drone, it is necessary to take aerial photos with two cameras, infrared and visible light, and automatically identify faults such as hot spots through an AI image recognition algorithm. Among them, when using an infrared camera to take an image of a photovoltaic panel, it is first necessary to separate the frame of the photovoltaic module.

[0039] At this time, if there is no shadow in the image, the obtained frame of the photovoltaic module can be seen in the example shown in Figure 2 If there is a shadow in the image, the obtained frame of the photovoltaic module can be seen in the example shown in Figure 3 It can be seen that the shadow in the image will have a serious impact on subsequent segmentation, and then lead to misjudgment or missed judgment of faults such as photovoltaic hot spots. Therefore, after obtaining the photovoltaic image to be denoised as shown in Figure 3 it can be considered that the photovoltaic image to be denoised is a combination of several Gaussian distributions, that is, the photovoltaic image to be denoised is regarded as a mixture Gaussian distribution, and then the photovoltaic image to be denoised is decomposed into a plurality of latent variables, the latent variables follow a normal distribution, and the number of latent variables is determined according to the photovoltaic image to be denoised. Specifically, see the example shown in Figure 4 In the example shown in Figure 4 the GMN curve is the curve corresponding to the photovoltaic image to be denoised. This curve is a combination of the first Gaussian and the second Gaussian curves. The image corresponding to the first Gaussian is the photovoltaic, and the image corresponding to the second Gaussian is the shadow. Therefore, the photovoltaic image to be denoised can be decomposed into two Gaussian curves, and each Gaussian curve is a latent variable. Only by removing the second Gaussian, the remaining first Gaussian is the photovoltaic image after removing the shadow noise.

[0040] S12. Remove the target latent variable from multiple latent variables and convert the remaining latent variables into a target image to obtain a photovoltaic image after removing shadow noise, where the target latent variable is the latent variable corresponding to the shadow noise.

[0041] Specifically, after obtaining multiple latent variables, the latent variable corresponding to the shadow noise can be found from the multiple latent variables. After removing this latent variable, and then combining the remaining latent variables, a photovoltaic image after removing shadow noise can be obtained.

[0042] Thus, denoising processing of the photovoltaic image can be achieved.

[0043] In some embodiments of the present invention, a pre-trained encoder is used to decompose a photovoltaic image to be denoised into multiple latent variables, where the encoder implements backpropagation using the reparameterization method.

[0044] In some embodiments of the present invention, the process of determining the number of latent variables includes: dividing the photovoltaic image to be denoised into multiple image blocks; classifying the multiple image blocks to obtain a classification result; and obtaining the number of latent variables according to the classification result and the size of the photovoltaic image to be denoised.

[0045] Taking the above example of using a pre-trained encoder to decompose a photovoltaic image to be denoised into multiple latent variables, in order to decompose the photovoltaic image to be denoised into multiple latent variables, the encoder is responsible for decomposing the input photovoltaic image to be denoised into latent variables , where N is the number of latent variables, which depends on the specific data distribution. These N distributions, each of which is a normal distribution, and a part of them constitutes the shaded part of the image, denoted as S, and the other part constitutes the non-shaded part, denoted as S'.

[0046] To implement this process, the specific value of N needs to be known. For the specific value of N, it can be confirmed based on experience. For example, for a 28×28 picture, the usual value is between 2 and 4. Most of the photovoltaic images are taken by drones, and their pixels are generally around 640×512. Relying solely on experience to determine N will result in a large error.

[0047] Therefore, in order to determine the specific value of N with low error, it is set to divide the photovoltaic image to be denoised into multiple image blocks; classify the multiple image blocks to obtain a classification result; and obtain the number of latent variables according to the classification result and the size of the photovoltaic image to be denoised.

[0048] Specifically, natural images have non-local self-similarity, that is, a certain image block in a natural image can find many similar blocks in this image, and some textures or structures in natural images tend to repeat in the image. For details, seeFigure 5 The specific example shown Figure 5 The image blocks within the square frames are all similar blocks of the image block numbered R.

[0049] Therefore, after obtaining the photovoltaic image to be denoised, first divide the photovoltaic image to be denoised into several image blocks. Taking Figure 6 the photovoltaic image to be denoised shown as an example, the result of dividing the photovoltaic image to be denoised into several image blocks can be seen in Figure 7 .

[0050] After obtaining multiple image blocks, the multiple image blocks can be classified to obtain a classification result, and then the number of latent variables can be obtained according to the classification result and the size of the photovoltaic image to be denoised.

[0051] Among them, in order to classify multiple image blocks, the similarity between image blocks can be calculated; the multiple image blocks can be classified according to the similarity.

[0052] Moreover, the classification result can also be set to include a non-shadow photovoltaic panel, a shadowed photovoltaic panel, and a background.

[0053] Specifically, calculate the similarity of each image block, and divide the image blocks into the following three categories, C1: non-shadow photovoltaic panel, C2: shadowed photovoltaic panel, C3: others. Among them, the proportion of the number of C2 in the total number of image blocks is denoted as P, and the resolution of the picture is denoted as W×H. Then, for the shadowed image blocks, the number of pixels contained is . The above calculation of similarity can be based on cosine similarity. The above C3 is mainly the background around the photovoltaic panel, W is the width of the photovoltaic image to be denoised, and H is the height of the photovoltaic image to be denoised.

[0054] Since the area generated by the shadow only accounts for a part of the image block, assuming that the width and height of the image block are respectively , , generally speaking, (such as 640×512), then the area of the shadow is approximately , so the proportion of the shadow area is . Therefore, the number of latent variables is obtained by the following formula:

[0055] ,

[0056] Among them, N represents the number of latent variables, W represents the width of the photovoltaic image to be denoised, H represents the height of the photovoltaic image to be denoised, and P represents the proportion of the image blocks of the shadowed photovoltaic panel in the multiple image blocks.

[0057] Thus, the number of latent variables can be determined with low error.

[0058] In some embodiments of the present invention, the reparameterization method includes:

[0059] The latent variable is obtained by the following formula:

[0060] z = μ + e σ

[0061] where z represents the latent variable, e represents Gaussian noise that satisfies the unit normal distribution, and μ and σ are the mean and variance of the variables generated by the encoder respectively. represents multiplication.

[0062] See Figure 8 , in Figure 8 , on the left side before reparameterization, z cannot be analytically calculated using the variance σ and the mean μ, so it cannot meet the requirements of neural network backpropagation. On the right side, reparameterization is performed. By adding a parameter e that satisfies the unit normal distribution N(0,1), multiplying e by the variance σ and adding it to the mean μ, the requirements of backpropagation are met while not affecting the calculation result.

[0063] In some embodiments of the present invention, a pre-trained decoder is used to remove the target latent variable from multiple latent variables and convert the remaining latent variables into a target image.

[0064] In some embodiments of the present invention, the output end of the encoder is connected to the input end of the decoder. The training process of the encoder and the decoder includes: obtaining a plurality of shadowless photovoltaic images, and respectively superimposing shadow data on each shadowless photovoltaic image to obtain a plurality of shadowed photovoltaic images; for each shadowed photovoltaic image, inputting the shadowed photovoltaic image into the encoder and outputting a denoised photovoltaic image through the decoder; obtaining a loss function according to the output variable distribution of the encoder and the standard normal distribution, as well as the denoised photovoltaic image and the corresponding shadowless photovoltaic image; updating the parameters of the encoder and the decoder according to the loss function.

[0065] The following is illustrated with a specific embodiment.

[0066] In this specific embodiment, see Figure 9 , an encoder, a decoder, and a latent space (latent vector space) are set. The encoder is used to implement the Encoder (encoding) function, and the decoder is used to implement the Decoder (decoding) function. The input is the above-mentioned photovoltaic image to be denoised, and the output is the above-mentioned photovoltaic image after removing shadow noise.

[0067] The encoder is set to a two-class Gaussian distribution. Its mean and variance are calculated respectively by the encoder, enter the latent space, and then the decoder decodes and outputs the separated one-class Gaussian distribution. Among them, both the encoder and the decoder are neural networks. For details, see Figure 10 . In Figure 10Among them, x1, x2, x3, x4, x5, x6 represent the inputs of the encoder, a1, a2, a3, a4 represent the intermediate data in the operation, μ1, μ2 represent the means of the variables generated by the encoder, σ1, σ2 represent the variances of the variables generated by the encoder, z1, z2 represent the latent variables output by the encoder, 1. 2. 3. 4. 5. 6 represents the output of the decoder.

[0068] In this neural network, the loss function consists of the following two parts:

[0069] ,

[0070] .

[0071] Among them, the first term is the in the decoder, denoted as the reconstruction loss, that is, the difference between the reconstructed output and the input, usually using the mean squared error. In the above calculation formula, x is the input image, z is the latent variable, p(z) is the prior distribution of x, q(z|x) is the approximate distribution of p(z), and E q（z|x） is the logarithmic expected value of the p(z) distribution.

[0072] The second term is the in the encoder, denoted as the KL divergence between the true distribution p(z) and the selected approximate distribution q(z|x). Among them, (q(z|x)||p(z)) is the divergence calculation formula of q(z|x) and p(z). Here, q(z|x) is usually a normal distribution N(0,1) with a mean and unit variance of zero, so that the distribution q(z|x) approaches the true distribution p(z) during training.

[0073] Specifically, the following process can be adopted.

[0074] Data preparation: From the photovoltaic panel pictures, a dataset of a certain scale (about 1000 pictures) is selected, denoted as C0, and there is no shadow in C0. Using data augmentation technology, shadow data is randomly superimposed on the picture data in C0, denoted as C1. In the subsequent neural network training, C1 is the input data and C0 is the output target data.

[0075] Model construction: Construct a model as shown in Figure 9 , which is divided into two parts: an encoder and a decoder.

[0076] Encoder part: The encoder is responsible for decomposing each picture X1 in C1 into latent variables , where N depends on the specific data distribution. Each of the N distributions is a normal distribution, a part of which forms the shaded part of the image, denoted as S, and the other part forms the non-shaded part, denoted as S'. During the training process, its loss function is mainly , and the reparameterization method is adopted.

[0077] Decoder part: The decoder is responsible for converting into the picture X0'. The decoder is also composed of a neural network, and its goal is to make X0' as close as possible to the corresponding X0 in C0. The loss function of the decoder is , that is, the error between X0' and X0. Since a part of the normal distribution in forms the shaded part of the image S, while X0 has no shaded part, so during the training process of the decoder, the relevant Z1 distribution corresponding to S is automatically masked by the training of the neural network.

[0078] Model training: Use C0 as the model input and C1 as the target data of the model output, and use an optimization algorithm (such as stochastic gradient descent) to minimize the loss function , . After multiple rounds of batch iteration, the network parameters of the encoder and decoder are learned.

[0079] Image denoising: After the model training is completed, input a new shaded photovoltaic panel picture at the input end of the model, and a picture with shadow noise removed can be obtained at the output side of the model.

[0080] In summary, for the photovoltaic image denoising method according to the embodiments of the present invention, the photovoltaic image to be denoised is decomposed into multiple latent variables, where the latent variables follow a normal distribution, and the number of the latent variables is determined according to the photovoltaic image to be denoised; the target latent variables among the multiple latent variables are removed, and the remaining latent variables are converted into a target image to obtain a photovoltaic image with shadow noise removed, where the target latent variables are the latent variables corresponding to the shadow noise. Thus, photovoltaic image denoising can be achieved. Moreover, the determination process of setting the number of latent variables includes: dividing the photovoltaic image to be denoised into multiple image blocks, classifying the multiple image blocks to obtain a classification result, and obtaining the number of latent variables according to the classification result and the size of the image to be denoised; and classifying the multiple image blocks includes: calculating the similarity between the image blocks, and classifying the multiple image blocks according to the similarity; and the specific calculation formula for the number of latent variables. Through the calculation of the number of latent variables, the number of latent variables can be determined with low error, thereby improving the accuracy of photovoltaic image denoising.

[0081] Furthermore, the present invention proposes a computer-readable storage medium.

[0082] In an embodiment of the present invention, a computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the above-mentioned photovoltaic image denoising method are implemented.

[0083] The computer-readable storage medium according to the embodiment of the present invention can achieve high-precision photovoltaic image denoising by implementing the photovoltaic image denoising method in the above embodiment.

[0084] Furthermore, the present invention proposes an electronic device.

[0085] Figure 11 It is a structural block diagram of the electronic device according to the embodiment of the present invention.

[0086] As Figure 11 shown, the electronic device 500 includes: a processor 501 and a memory 503. Among them, the processor 501 and the memory 503 are connected, such as connected through a bus 502. Optionally, the electronic device 500 may further include a transceiver 504. It should be noted that in practical applications, the transceiver 504 is not limited to one, and the structure of the electronic device 500 does not constitute a limitation to the embodiment of the present invention.

[0087] The processor 501 may be a CPU (Central Processing Unit, central processor), a general-purpose processor, a DSP (Digital Signal Processor, digital signal processor), an ASIC (Application Specific Integrated Circuit, application-specific integrated circuit), an FPGA (Field Programmable Gate Array, field programmable gate array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It can implement or execute various exemplary logic blocks, modules and circuits described in combination with the disclosure of the present invention. The processor 501 may also be a combination that realizes a computing function, such as a combination including one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0088] The bus 502 may include a path for transmitting information between the above components. The bus 502 may be a PCI (Peripheral Component Interconnect, peripheral component interconnect standard) bus or an EISA (Extended Industry Standard Architecture, extended industry standard architecture) bus, etc. The bus 502 may be divided into an address bus, a data bus, a control bus, etc. For the sake of representation, Figure 11 only a thick line is used in

[0089] The memory 503 is used to store a computer program corresponding to the photovoltaic image denoising method in the above embodiments of the present invention, and the execution of this computer program is controlled by the processor 501. The processor 501 is used to execute the computer program stored in the memory 503 to implement the content shown in the foregoing method embodiments.

[0090] Among them, Figure 11 the illustrated electronic device 500 is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0091] The electronic device according to the embodiments of the present invention can achieve high-precision photovoltaic image denoising.

[0092] It should be noted that the logic and / or steps represented in the flowchart or otherwise described herein can be considered as a defined sequence list of executable instructions for implementing logical functions, and can be specifically implemented in any computer-readable medium for use by an instruction execution system, apparatus, or device (such as a computer-based system, a system including a processor, or other systems that can fetch and execute instructions from the instruction execution system, apparatus, or device), or used in combination with these instruction execution systems, apparatus, or devices. For the purposes of this specification, a "computer-readable medium" can be any device that can contain, store, communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. More specific examples (non-exhaustive list) of computer-readable media include the following: an electrical connection portion with one or more wirings (electronic device), a portable computer diskette (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber device, and a portable compact disc read-only memory (CDROM). Additionally, the computer-readable medium can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other media, followed by editing, interpretation, or otherwise processing as appropriate, and then stored in a computer memory.

[0093] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, any one or a combination of the following well-known technologies in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0094] In the description of this specification, the descriptions referring to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. mean that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0095] In the description of this specification, the orientation or positional relationship indicated by terms such as "center", "longitudinal", "transverse", "length", "width", "thickness", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc. is based on the orientation or positional relationship shown in the drawings, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and should not be construed as a limitation to the present invention.

[0096] In addition, the terms "first" and "second" are only used for descriptive purposes, and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include at least one of such features. In the description of the present invention, the meaning of "a plurality" is at least two, such as two, three, etc., unless otherwise specifically and clearly defined.

[0097] In the description of this specification, unless otherwise stated, terms such as "installation", "connection", "attachment", "fixation", etc. should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or integrated; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and can be the communication inside two elements or the interaction relationship between two elements, unless otherwise clearly defined. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0098] In the present invention, unless otherwise clearly specified or limited, the first feature being "on" or "under" the second feature may mean that the first and second features are in direct contact, or the first and second features are indirectly in contact through an intermediate medium. Further, the first feature being "above", "over" and "on top of" the second feature may mean that the first feature is directly above or obliquely above the second feature, or merely indicates that the horizontal height of the first feature is higher than that of the second feature. The first feature being "under", "beneath" and "underneath" the second feature may mean that the first feature is directly below or obliquely below the second feature, or merely indicates that the horizontal height of the first feature is less than that of the second feature.

[0099] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A photovoltaic image denoising method, characterized in that, Including: Decompose the photovoltaic image to be denoised into multiple latent variables, where the latent variables follow a normal distribution, and the number of the latent variables is determined according to the photovoltaic image to be denoised; Remove the target latent variable from the multiple latent variables, and convert the remaining latent variables into a target image to obtain a photovoltaic image after removing shadow noise, where the target latent variable is the latent variable corresponding to the shadow noise; The determination process of the number of the latent variables includes: Divide the photovoltaic image to be denoised into multiple image blocks; Classify the multiple image blocks to obtain a classification result; Obtain the number of the latent variables according to the classification result and the size of the photovoltaic image to be denoised; The classification result includes a shadowless photovoltaic panel, a shaded photovoltaic panel, and a background; the number of the latent variables is obtained by the following formula: ， Where N represents the number of the latent variables, W represents the width of the photovoltaic image to be denoised, H represents the height of the photovoltaic image to be denoised, and P represents the proportion of the image blocks of the shaded photovoltaic panel in the multiple image blocks.

2. The photovoltaic image denoising method according to claim 1, wherein The classification of the multiple image blocks includes: Calculate the similarity between the image blocks; Classify the multiple image blocks according to the similarity.

3. The photovoltaic image denoising method according to claim 1, wherein Use a pre-trained encoder to decompose the photovoltaic image to be denoised into multiple latent variables, where the encoder uses the reparameterization method to implement backpropagation.

4. The photovoltaic image denoising method according to claim 3, wherein The reparameterization method includes: Obtain the latent variable through the following formula: z = μ + e σ where z represents the latent variable, e represents Gaussian noise satisfying a unit normal distribution, and μ and σ are the mean and variance of the variables generated by the encoder, respectively. denotes multiplication.

5. The photovoltaic image denoising method according to claim 3, characterized in that, Use a pre-trained decoder to remove the target latent variable from the multiple latent variables, and convert the remaining latent variables into a target image.

6. The photovoltaic image denoising method according to claim 5, characterized in that, The output end of the encoder is connected to the input end of the decoder, and the training process of the encoder and the decoder includes: Obtain multiple shadowless photovoltaic images, and respectively superimpose shadow data on each of the shadowless photovoltaic images to obtain multiple shaded photovoltaic images; For each of the shaded photovoltaic images, input the shaded photovoltaic image into the encoder, and output a denoised photovoltaic image through the decoder; Obtain a loss function according to the variable distribution output by the encoder and the standard normal distribution, and the denoised photovoltaic image and the corresponding shadowless photovoltaic image; Update the parameters of the encoder and the decoder according to the loss function.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the method according to any one of claims 1-6 are implemented.

8. An electronic device, characterized in that, The electronic device includes a memory and a processor, the memory stores a computer program, and when the processor executes the computer program, the steps of the method according to any one of claims 1-6 are implemented.

Images