Image compensation method, system and medium based on electroluminescence detection

Abstract

The present application provides an image compensation method, system and medium based on electroluminescence detection, the method comprising: applying a reverse voltage on a photovoltaic panel, using an electroluminescence camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescence image and at least two background images; obtaining the intensity of the background light according to the at least two background images; removing the intensity of the background light from the electroluminescence image to obtain a corrected image without background light interference; calculating the local variance of pixels in the corrected image to obtain the noise level in the corrected image, and constructing a filter according to the noise level; applying a generative adversarial network to process the corrected image to generate an enhanced image. The method can significantly improve the accuracy and efficiency of photovoltaic panel defect detection. By removing background light interference and noise, the obtained corrected image significantly improves the clarity and contrast of the image, making the identification of tiny cracks and hidden defects more accurate.

Description

Technical Field

[0001] The present application relates to the field of photovoltaic panel detection technology, and in particular to an image compensation method, system and medium based on electroluminescence detection. Background Art

[0002] Electroluminescence (EL) detection technology plays an important role in the quality control of photovoltaic panels. It can effectively identify material defects such as cracks and hidden cracks. However, traditional EL detection needs to be carried out in a dark room, which limits its scope of application. Although daylight EL detection can be used in outdoor environments, it is affected by ambient light, and the image quality and defect recognition accuracy are low. Ordinary EL detection cannot be directly applied to installed photovoltaic systems due to its strict requirements on environmental conditions.

[0003] At present, due to the strong background noise in sunlight, ambient light will interfere with EL signals, resulting in a decrease in image quality. Although the influence of ambient light can be weakened to a certain extent through signal processing, filtering and synchronization technology, there is still a large gap in the clarity, contrast and defect recognition accuracy of sunlight EL images compared with ordinary EL inspection. Summary of the invention

[0004] In view of this, it is necessary to provide an image compensation method, system and storage medium based on electroluminescence detection, which can overcome at least one of the above defects.

[0005] In a first aspect, an embodiment of the present application provides an image compensation method based on electroluminescence detection, which is applied to detect defects of photovoltaic panels. The method includes:

[0006] Applying a reverse voltage to the photovoltaic panel, and using an electroluminescent camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images;

[0007] The intensity of the background light is obtained according to at least two of the background images, and the calculation formula of the intensity is:

[0008]

[0009] Where N is the number of background images, I i (x,y) is the light intensity of the pixel (x,y) in the i-th image;

[0010] removing the intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference;

[0011] The local variance of the pixels in the corrected image is calculated to obtain the noise level in the corrected image, and a filter is constructed according to the noise level, the filter is used to remove the noise in the corrected image, and the calculation formula of the local variance is:

[0012]

[0013] in, To correct the local variance of the pixel (x, y) in the image, M is the size of the local window, m is the radius of the window, and i and j are the relative position offsets of the pixel points;

[0014] A generative adversarial network is applied to process the corrected image to generate an enhanced image, wherein the enhanced image is used to identify defects of the photovoltaic panel.

[0015] In one embodiment, the adversarial network includes a generator and a discriminator, the generator is used to reduce the adversarial network of the enhanced image, and the loss function of the adversarial network is:

[0016]

[0017] Among them, G(z) represents the enhanced image output by the generator, D(G(z)) is the judgment result of the discriminator on the generator output image, G is the generator, D is the discriminator, and z is the random noise input. Represents the probability distribution p z (z) is the expected value of the random noise z extracted from

[0018] In one embodiment, applying a generative adversarial network to process the modified image to generate an enhanced image includes:

[0019] The enhanced image is optimized by applying a perceptual loss function to improve the quality of the enhanced image compensation. The formula of the perceptual loss function is:

[0020]

[0021] Among them, φ i is the pre-trained convolutional neural network feature extraction layer, I real Represents the light intensity value of the corresponding pixel point (x, y) in the real image.

[0022] In one embodiment, the applying a generative adversarial network to process the modified image to generate an enhanced image further includes:

[0023] Gradient calculation is applied to introduce structural information into the enhanced image, and the structural information is used to improve the details of the enhanced image to improve the accuracy of identifying defects of the photovoltaic panel. The formula for the gradient calculation is:

[0024]

[0025] Among them, G(x,y) represents the gradient amplitude of image I at point (x,y), and are the partial derivatives of the image in the x and y directions respectively.

[0026] In one embodiment, removing the intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference includes:

[0027] The intensity of the background light is removed by a linear correction algorithm to obtain the corrected image. The formula of the linear correction algorithm is:

[0028] I correct (x,y)=I EL (x,y)-I bg (x,y)

[0029] Among them, I correct (x, y) is the corrected image, I EL (x,y) is the electroluminescence image.

[0030] In one embodiment, the method further comprises:

[0031] Obtaining the material type of the photovoltaic panel;

[0032] Acquiring a linear adjustment parameter of the photovoltaic panel according to a material type of the photovoltaic panel;

[0033] The corrected image is corrected according to the linear adjustment parameter to form an adapted image. The calculation formula of the adapted image is:

[0034] I adapted (x,y)=k m I correct (x,y)+b m

[0035] Among them, k m and b m is a linear adjustment parameter.

[0036] In one embodiment, the filter is an adaptive filter, and the output of the adaptive filter is:

[0037]

[0038] Where W is the size of the adaptive filter,

[0039] In one embodiment, the method further comprises:

[0040] Acquiring shape features, texture features and color features in the enhanced image;

[0041] The shape feature, the texture feature and the color feature are input into a classifier to identify defects of the photovoltaic panel.

[0042] In a second aspect, an embodiment of the present application provides an image compensation system based on electroluminescence detection, which is applied to perform the image compensation method based on electroluminescence detection as described in the first aspect, and the system includes:

[0043] An electroluminescent camera module, used for applying a reverse voltage to the photovoltaic panel, and using an electroluminescent camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images;

[0044] The light intensity acquisition module is used to acquire the light intensity of the background light according to at least two background images. The calculation formula of the light intensity is:

[0045]

[0046] Where N is the number of background images, I i (x,y) is the light intensity of the pixel (x,y) in the i-th image;

[0047] A light intensity removal module, used for removing the light intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference;

[0048] A filter construction module is used to calculate the local variance of pixels in the modified image to obtain the noise level in the modified image, and to construct a filter according to the noise level. The filter is used to remove the noise in the modified image. The calculation formula of the local variance is:

[0049]

[0050] in, To correct the local variance of the pixel (x, y) in the image, M is the size of the local window, m is the radius of the window, and i and j are the relative position offsets of the pixel points;

[0051] The enhanced image generation module applies a generative adversarial network to process the corrected image to generate an enhanced image, and the enhanced image is used to identify defects of the photovoltaic panel.

[0052] In a third aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions, wherein the instructions instruct a device to execute the image compensation method based on electroluminescence detection as described in the first aspect.

[0053] The image compensation method, system and storage medium for electroluminescence detection provided in the embodiments of the present application can significantly improve the accuracy and efficiency of photovoltaic panel defect detection. By removing background light interference and noise, the corrected image obtained significantly improves the image clarity and contrast, making the identification of tiny cracks and hidden defects more accurate. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 A schematic flow chart of an image compensation method for electroluminescence detection provided in one embodiment of the present application.

[0055] Figure 2 A schematic diagram of an image compensation system module based on electroluminescence detection provided in one embodiment of the present application.

[0056] Figure 3 A schematic diagram of an electronic device provided in one embodiment of the present application.

[0057] Main component symbols

[0058] Image compensation system based on electroluminescence detection10

[0059] Electroluminescent camera module 11

[0060] Light intensity acquisition module 12

[0061] Light intensity removal module 13

[0062] Filter building blocks 14

[0063] Enhanced image generation module 15

[0064] Electronic devices 20

[0065] Processor 21

[0066] Memory 22

[0067] Method steps S100-S500 DETAILED DESCRIPTION

[0068] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all of the embodiments.

[0069] It should be noted that, in the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art in the present application. The terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application.

[0070] It should be noted that, in the embodiments of the present application, words such as "first" and "second" are only used for the purpose of distinguishing descriptions, and cannot be understood as indicating or implying relative importance, nor can they be understood as indicating or implying order. Features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of the embodiments of the present application, words such as "exemplary" or "for example" are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" or "for example" in the embodiments of the present application should not be interpreted as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of words such as "exemplary" or "for example" is intended to present related concepts in a concrete way.

[0071] Based on the implementations in this application, all other implementations obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0072] Electroluminescence (EL) detection technology plays an important role in the quality control of photovoltaic panels. It can effectively identify material defects such as cracks and hidden cracks. However, traditional EL detection needs to be carried out in a dark room, which limits its scope of application. Although daylight EL detection can be used in outdoor environments, it is affected by ambient light, and the image quality and defect recognition accuracy are low. Ordinary EL detection cannot be directly applied to installed photovoltaic systems due to its strict requirements on environmental conditions.

[0073] At present, due to the strong background noise in sunlight, ambient light will interfere with EL signals, resulting in a decrease in image quality. Although the influence of ambient light can be weakened to a certain extent through signal processing, filtering and synchronization technology, there is still a large gap in the clarity, contrast and defect recognition accuracy of sunlight EL images compared with ordinary EL inspection.

[0074] In view of this, the present application provides an image compensation method, system and storage medium for electroluminescence detection, which can significantly improve the accuracy and efficiency of photovoltaic panel defect detection. By removing background light interference and noise, the corrected image obtained significantly improves the image clarity and contrast, making the identification of tiny cracks and hidden defects more accurate. The method can be applied under different environmental conditions, especially in sunny outdoor environments. Through daylight EL detection technology, on-site detection of photovoltaic panels is more convenient and flexible, reducing dependence on laboratory conditions. Combined with generative adversarial network technology, it can automatically optimize the image processing process, reduce manual intervention, improve detection efficiency, shorten detection time, and provide timely feedback for the maintenance of photovoltaic power stations.

[0075] Figure 1FIG. 1 is a flow chart of an image compensation method based on electroluminescence detection provided by an embodiment of the present application. Figure 1 The image compensation method based on electroluminescence detection shown includes at least the following steps: S100: applying a reverse voltage to a photovoltaic panel, and using an electroluminescence camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescence image and at least two background images; S200: acquiring the intensity of the background light based on at least two background images; S300: removing the intensity of the background light from the electroluminescence image to obtain a corrected image without background light interference; S400: calculating the local variance of pixels in the corrected image to obtain the noise level in the corrected image, and constructing a filter based on the noise level; S500: applying a generative adversarial network to process the corrected image to generate an enhanced image.

[0076] S100: applying a reverse voltage to the photovoltaic panel, and using an electroluminescent camera to capture a surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images.

[0077] In an embodiment of the present application, the image compensation method based on electroluminescence detection includes in step S100 applying a reverse voltage to the photovoltaic panel, and using an electroluminescence camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescence image and at least two background images.

[0078] It is understandable that by applying a reverse voltage, the defective area inside the photovoltaic panel is stimulated to emit light, thereby obtaining an electroluminescent image with diagnostic value, providing basic data for subsequent defect identification. The electroluminescent camera can capture these signals and provide accurate image data for subsequent processing, ensuring the effectiveness and accuracy of the processing in subsequent steps. At the same time, by obtaining at least two background images, the subsequent extraction of background light intensity can be achieved.

[0079] S200: Acquire the intensity of background light according to at least two background images.

[0080] In the embodiment of the present application, the image compensation method based on electroluminescence detection includes acquiring the intensity of background light according to at least two background images in step S100.

[0081] In the embodiment of the present application, the calculation formula of light intensity is:

[0082]

[0083] Where N is the number of background images, I i (x,y) is the light intensity of pixel (x,y) in the i-th image.

[0084] It is understandable that by averaging multiple background images, the ambient light intensity that may interfere with the signal during the electroluminescence detection process can be more accurately estimated, thereby providing effective parameter support for the subsequent background light removal step. This processing process ensures that the acquired background light intensity value is stable and reliable, so that it is more targeted when removing interference.

[0085] S300: removing the intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference.

[0086] In the embodiment of the present application, the image compensation method based on electroluminescence detection includes, in step S300, removing the intensity of background light from the electroluminescence image to obtain a corrected image without background light interference.

[0087] It is understandable that this process is designed to eliminate interference caused by ambient light, making the remaining image information closer to the true luminous characteristics of the photovoltaic panel, thereby improving the accuracy of subsequent defect identification. Through this processing, the corrected image obtained will more clearly reflect the defects on the surface of the photovoltaic panel, making subsequent analysis more efficient.

[0088] S 400: Calculate the local variance of pixels in the corrected image to obtain the noise level in the corrected image, and construct a filter according to the noise level.

[0089] In the embodiment of the present application, the image compensation method based on electroluminescence detection includes calculating the local variance of pixels in the corrected image in step S400 to obtain the noise level in the corrected image, and constructing a filter according to the noise level, and the filter is used to remove the noise in the corrected image.

[0090] In the embodiment of the present application, the calculation formula of the local variance is:

[0091]

[0092] in, To correct the local variance of the pixel point (x, y) in the image, M is the size of the local window, m is the radius of the window, and i and j are the relative position offsets of the pixel points.

[0093] It can be understood that the calculation of local variance can effectively evaluate the noise distribution in the image and provide a basis for the subsequent design of adaptive filters, thereby effectively improving image quality and reducing the impact of noise interference on defect recognition. In this way, the filter can be dynamically adjusted to adapt to different noise conditions to ensure the clarity of the final image.

[0094] S 500: Applying a generative adversarial network to process the corrected image to generate an enhanced image.

[0095] In the embodiment of the present application, the image compensation method based on electroluminescence detection includes applying a generative adversarial network to process the corrected image in step S500 to generate an enhanced image, and the enhanced image is used to identify defects of the photovoltaic panel.

[0096] It is understandable that the generative adversarial network can enhance the details of the image while retaining effective information through adversarial training, so that the final output enhanced image can more clearly show the defect characteristics in the photovoltaic panel, thereby improving the accuracy and reliability of defect identification. Through the training of the adversarial network, the model can learn rich feature representations, thereby achieving more accurate defect detection and reducing the possibility of misjudgment and missed judgment.

[0097] The image compensation method based on electroluminescence detection implemented in this application has many beneficial effects and can significantly improve the accuracy and efficiency of photovoltaic panel defect detection. First, by applying a reverse voltage to the photovoltaic panel, the luminous signal inside the photovoltaic material is stimulated, making the potential defect area visible. This process provides a rich information basis for subsequent image analysis, ensuring that defects that may affect the performance of photovoltaic panels can be identified at an early stage.

[0098] Secondly, the method accurately obtains the intensity of the background light by collecting at least two background images, and eliminates the interference of ambient light on the detection results by calculating the average value. This design significantly improves the reliability of image processing, so that the obtained electroluminescent image can more realistically reflect the state of the photovoltaic panel. After removing the influence of background light, the subsequent corrected image will be clearer, effectively reducing noise and interference, thereby enhancing the accuracy of the detection process.

[0099] Furthermore, the local variance of the corrected image is calculated to evaluate the noise level, which provides a basis for the construction of an adaptive filter. This filter can be dynamically adjusted according to the noise distribution in the image to ensure the effectiveness of the noise removal process. This process not only improves the quality of the corrected image, but also provides clearer image data for subsequent defect identification, significantly improving the performance of the defect detection algorithm.

[0100] In addition, the design of using a generative adversarial network for image processing makes the generation process of enhanced images more intelligent and automated. The generative adversarial network can retain effective detail information during the processing process, while enhancing the contrast and clarity of the image, making the defects of the photovoltaic panel more obvious. This technological innovation greatly improves the accuracy of defect detection, can effectively reduce the incidence of missed detection and false detection, and provides technical support for the reliability of photovoltaic panels.

[0101] Finally, this method is not only suitable for defect detection of photovoltaic panels, but also has good scalability and adaptability, and can be adjusted and optimized under different environmental conditions. This makes the application prospects of this technology in the photovoltaic industry broad, which can not only improve production efficiency but also reduce maintenance costs, thus achieving a good balance between economic benefits and environmental protection.

[0102] In an embodiment of the present application, the adversarial network includes a generator and a discriminator. The generator is used to reduce the adversarial network of the enhanced image. The loss function of the adversarial network is:

[0103]

[0104] Among them, G(z) represents the enhanced image output by the generator, D(G(z)) is the judgment result of the discriminator on the generator output image, G is the generator, D is the discriminator, and z is the random noise input. Represents the probability distribution p z (z) is the expected value of the random noise z extracted from

[0105] It can be understood that the Generative Adversarial Network (GAN) consists of a generator and a discriminator. The main task of the generator is to generate enhanced images to more effectively identify defects in photovoltaic panels. The loss function L G The design of is to promote the generator to generate more realistic images by minimizing the difference between the generated image and the real image. Here G(z) represents the enhanced image output by the generator, and D(G(z)) is the judgment result of the discriminator on this image. The purpose of the discriminator is to evaluate whether the image generated by the generator is realistic, and to provide feedback to the generator for improvement. The random noise z is obtained from a specific probability distribution p z (z) is extracted to ensure that the generator can generate diverse images. This design not only improves the diversity of generated images, but also enhances the learning ability of the generator, enabling it to gradually generate higher quality images. This method uses a continuous competition mechanism to enable the generator and the discriminator to promote each other, thereby achieving high-quality enhanced image generation, thereby effectively supporting the detection of photovoltaic panel defects.

[0106] In an embodiment of the present application, a generative adversarial network is applied to process a corrected image to generate an enhanced image, including: applying a perceptual loss function to optimize the enhanced image to improve the quality of enhanced image compensation. The formula of the perceptual loss function is:

[0107]

[0108] Among them, φ i is the pre-trained convolutional neural network feature extraction layer, I real Represents the light intensity value of the corresponding pixel point (x, y) in the real image.

[0109] It is understandable that when applying the generative adversarial network to process the corrected image, a perceptual loss function is introduced to optimize the quality of the enhanced image. In the formula of the perceptual loss function, φ i represents the pre-trained convolutional neural network feature extraction layer, and I real is the actual image. The introduction of the perceptual loss function allows the generator to consider not only the similarity between the generated image and the real image at the pixel level, but also the similarity in the feature space. This mechanism significantly improves the quality of the generated images because the network focuses on capturing high-level features of the image, such as shape, texture, and color, rather than just simple pixel matching. In this way, the generator is able to generate more realistic and usable images, thereby improving the accuracy and reliability of photovoltaic panel defect detection. This approach ensures that the enhanced images are not only more visually appealing, but also provide more useful information in the process of feature extraction and defect identification.

[0110] In the embodiment of the present application, the generative adversarial network is used to process the corrected image to generate an enhanced image, and also includes: applying gradient calculation to introduce structural information into the enhanced image, and the structural information is used to improve the details of the enhanced image to improve the accuracy of identifying defects in the photovoltaic panel. The formula for gradient calculation is:

[0111]

[0112] Among them, G(x,y) represents the gradient amplitude of image I at point (x,y), and are the partial derivatives of the image in the x and y directions respectively.

[0113] It is understandable that the application of gradient calculation to introduce structural information in the enhanced image is also an important step to improve the detection effect. By calculating the gradient of the image, the edge and detail information in the image can be effectively extracted, which is crucial for the defect identification of photovoltaic panels. Edge information can usually indicate the outline and changes of the object, and in photovoltaic panel detection, defects often appear as slight changes in the surface structure. After introducing structural information, the details and contrast of the enhanced image will be improved, thereby improving the accuracy of photovoltaic panel defect identification. This method effectively combines image processing and deep learning technology, so that the defect detection method can maintain high performance in different situations, ensuring the safety and reliability of photovoltaic panels.

[0114] In the embodiment of the present application, the intensity of the background light is removed from the electroluminescent image to obtain a corrected image without background light interference, including: removing the intensity of the background light by a linear correction algorithm. To obtain the corrected image, the formula of the linear correction algorithm is:

[0115] I correct(x,y)=I EL (x,y)-I bg (x,y)

[0116] Among them, I correct (x, y) is the corrected image, I EL (x,y) is the electroluminescence image.

[0117] It can be understood that by removing the intensity of the background light from the electroluminescent image, a corrected image without background light interference can be effectively obtained. In this way, the image quality problems caused by background light interference can be minimized, thereby improving the accuracy of subsequent image processing and analysis. This linear correction method is simple and efficient, and can quickly eliminate the influence of background light, making the obtained corrected image more realistic and convenient for subsequent defect detection.

[0118] In an embodiment of the present application, the image compensation method based on electroluminescence detection further includes obtaining the material type of the photovoltaic panel, and obtaining the linear adjustment parameter of the photovoltaic panel according to the material type of the photovoltaic panel. The corrected image is corrected according to the linear adjustment parameter to form an adapted image. The calculation formula of the adapted image is:

[0119] I adapted (x,y)=k m I correct (x,y)+b m

[0120] Among them, k m and b m is a linear adjustment parameter.

[0121] It is understandable that the material type of the photovoltaic panel is also taken into consideration. According to the material type of the photovoltaic panel, a specific linear adjustment parameter can be obtained to further correct the corrected image to form an adapted image. m and b m The linear adjustment parameters related to the type of photovoltaic panel material are introduced. The introduction of this step enables the image adaptation process to be personalized according to the optoelectronic properties of different materials, ensuring that the generated adapted images have higher accuracy and reliability, thereby providing more effective information in subsequent defect detection.

[0122] In the embodiment of the present application, the filter is an adaptive filter, and the output of the adaptive filter is:

[0123]

[0124] Where W is the size of the adaptive filter.

[0125] It can be understood that the adaptive filter is used to further process the corrected image to reduce noise and improve image quality. The adaptive filter dynamically adjusts its response based on the local characteristics of the image, which can effectively suppress noise while retaining important details in the image. The application of this technology further enhances the clarity of the image, making the defects of the photovoltaic panel easier to identify in subsequent processing and analysis.

[0126] Specifically, the double summation symbol in the formula indicates the accumulation of all pixel values ​​within the window. The range here is [-W / 2, W / 2], so the formed window is actually a square area centered at (x, y). It is the averaging factor of the pixel values ​​within the window, ensuring that the output F(x,y) is the mean of all pixel values ​​in the area and maintaining the range of the output value.

[0127] It is understandable that through this adaptive filter, the main content of the formula calculation is the average value of the current pixel (x, y) and its surrounding pixel values, with the purpose of reducing random noise in the image through smoothing. The adaptive filter can dynamically adjust its filtering effect according to the local characteristics of the image, so that different areas are processed differently, which helps to preserve edges and details in different image areas. The output of the adaptive filter achieves noise reduction and image quality by calculating the local pixel mean. This process not only helps to improve accuracy in subsequent defect detection, but also plays a key role in the overall image compensation method.

[0128] In the embodiment of the present application, the image compensation method based on electroluminescence detection further includes obtaining shape features, texture features and color features in the enhanced image, and inputting the shape features, texture features and color features into a classifier to identify defects of the photovoltaic panel.

[0129] It is understandable that the image compensation method based on electroluminescence detection also includes extracting shape features, texture features and color features in the enhanced image. These feature information provides an important basis for defect identification. Inputting shape features, texture features and color features into the classifier can effectively identify defects in photovoltaic panels and enhance the intelligence level of the defect detection system. This process not only improves the accuracy of defect identification, but also makes the entire detection process more efficient and automated, thereby reducing the need for manual intervention and saving detection costs and time.

[0130] In the embodiment of the present application, the shape feature extraction method uses the Canny edge detection algorithm to extract the image edge. The edge detection formula is:

[0131]

[0132] Among them, E(x,y) is the edge strength, and is the gradient of the image.

[0133] It can be understood that by calculating the gradient of the image, the Canny algorithm can effectively identify areas with large intensity changes in the image, thereby extracting clear edges. This edge information is crucial for subsequent shape feature analysis and helps identify defects in photovoltaic panels. The main purpose of edge detection is to find the contours and boundaries of objects in the image and provide basic data for subsequent feature analysis. This process usually includes smoothing the image, calculating the gradient, applying non-maximum suppression and threshold processing, and the final edge image can intuitively reflect the shape information of the object in the image.

[0134] In the embodiment of the present application, texture features are used to describe the texture information of the surface in the image. The gray level co-occurrence matrix (GLCM) is used here. The gray level co-occurrence matrix is ​​used to extract texture information by calculating the symbiotic relationship between pixel gray values. The formula of the gray level co-occurrence matrix is:

[0135]

[0136] Among them, δ is just a function, and I(x,y) is the grayscale value of the image.

[0137] It can be understood that by constructing GLCM, the relationship between different gray levels in the image can be obtained, thereby extracting useful texture features. This method effectively captures the subtle differences in the image surface, making texture analysis more accurate. GL CM not only provides texture distribution information, but also provides a rich data source for subsequent feature extraction. By analyzing GLCM, many important texture features can be extracted, thereby achieving accurate identification of surface defects of photovoltaic panels.

[0138] In the embodiment of the present application, information such as contrast, uniformity and entropy is then extracted from the gray level co-occurrence matrix. The calculation formula of contrast is:

[0139]

[0140] It is understandable that contrast reflects the magnitude of grayscale value changes in an image and is an important indicator for determining the complexity of image texture. A high contrast value usually indicates that there are obvious texture changes in the image, while a low contrast indicates that the image is relatively flat. The calculation of contrast helps to identify small defects and uneven areas in the image, providing an important basis for subsequent image analysis and defect detection.

[0141] In the embodiment of the present application, the distribution of different colors in the image is calculated by a color histogram, and the calculation formula of the color histogram is:

[0142]

[0143] Where H(c) is the frequency of color c in the color histogram, and R is the image area.

[0144] It is understandable that the color histogram can effectively display the distribution of different colors in the image, which is crucial for identifying color differences and other color-related defects on photovoltaic panels. The extraction of color features makes the image analysis more comprehensive, not only considering shape and texture, but also introducing color information, which helps to improve the accuracy of defect identification.

[0145] In the embodiment of the present application, the extracted shape features, texture features and color features are fused to form a feature vector F. The expression of F is:

[0146] F=[f1,f2,…,f n ]

[0147] Among them, f1 is the i-th feature, and then the feature vector is input into the classifier.

[0148] It can be understood that the construction of feature vectors integrates multiple feature information into a unified representation, allowing the classifier to understand the image features more comprehensively, thereby effectively improving the recognition rate of photovoltaic panel defects. Through this feature fusion method, the information in the image can be maximized to improve the accuracy and efficiency of the analysis.

[0149] Figure 2 FIG. 1 is a schematic diagram of an image compensation system module based on electroluminescence detection provided by an embodiment of the present application. Figure 2 As shown, the image compensation system 10 based on electroluminescence detection includes at least the following parts: an electroluminescence camera module 11 , a light intensity acquisition module 12 , a light intensity removal module 13 , a filter construction module 14 , and an enhanced image generation module 15 .

[0150] In the embodiment of the present application, the electroluminescent camera module 11 is used to apply a reverse voltage to the photovoltaic panel, and the electroluminescent camera is used to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images. For details, please refer to Figure 1 And the corresponding description will not be repeated here.

[0151] In the embodiment of the present application, the light intensity acquisition module 12 is used to acquire the light intensity of the background light according to at least two background images. The calculation formula of the light intensity is:

[0152]

[0153] Where N is the number of background images, Ii (x,y) is the light intensity of the pixel (x,y) in the ith image. Figure 1 And the corresponding description will not be repeated here.

[0154] In the embodiment of the present application, the light intensity removal module 13 is used to remove the light intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference. For details, please refer to Figure 1 And the corresponding description will not be repeated here.

[0155] In the embodiment of the present application, the filter construction module 14 is used to calculate the local variance of the pixels in the corrected image to obtain the noise level in the corrected image, and to construct a filter according to the noise level. The filter is used to remove the noise in the corrected image. The calculation formula of the local variance is:

[0156]

[0157] in, To correct the local variance of the pixel point (x, y) in the image, M is the size of the local window, m is the radius of the window, i and j are the relative position offsets of the pixel points, please refer to Figure 1 And the corresponding description will not be repeated here.

[0158] In the embodiment of the present application, the enhanced image generation module 15 applies a generative adversarial network to process the modified image to generate an enhanced image, which is used to identify defects of the photovoltaic panel. For details, please refer to Figure 1 And the corresponding description will not be repeated here.

[0159] Figure 3 This is an electronic device 20 provided in an embodiment of the present application. Figure 3 As shown, the electronic device 20 includes at least the following parts: a processor 21 and a memory 22 .

[0160] In the embodiment of the present application, the memory 22 is used to store instructions executable by the processor 21, and the processor 21 is configured to implement the following when executing the instructions: Figure 1 The image compensation method based on electroluminescence detection is shown.

[0161] In an embodiment of the present application, a computer-readable storage medium includes instructions, and the instructions instruct a device to execute the method of the first aspect. For example, the instructions instruct the device to execute Figure 1 The image compensation method based on electroluminescence detection shown in steps S100 to S700.

[0162] The program operating in the electronic device 20 involved in one embodiment of the present application may be a program (a program that enables a computer to function) that controls a central processing unit (CPU) and the like to realize the functions of the above-mentioned embodiment involved in one scheme of the present invention. Then, the information processed by these devices is temporarily stored in a random access memory (RAM) during its processing, and then stored in various ROMs such as a read-only memory (Flash ROM) and a hard disk drive (HDD), and is read, modified, and written by the CPU as needed.

[0163] It should be noted that part of the electronic device 20 of the above embodiment may be realized by a computer. In this case, a program for realizing the control function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer and executed.

[0164] It should be noted that the "computer" mentioned here refers to a computer built into the electronic device 20, and uses a computer including hardware such as an OS and peripheral devices. In addition, "computer-readable recording medium" refers to removable media such as floppy disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into the computer.

[0165] Furthermore, "computer-readable recording media" may include: media that dynamically store programs for a short period of time, such as communication lines when sending programs via networks such as the Internet or communication lines such as telephone lines; and media that store programs for a fixed period of time, such as volatile memories inside computers that serve as servers or clients in this case. In addition, the above-mentioned program may be a program for realizing a part of the above-mentioned functions, or a program that can realize the above-mentioned functions by combining with a program already recorded in a computer.

[0166] In addition, the electronic device 20 in the above-mentioned embodiment can also be implemented as a collection (device group) composed of multiple devices. Each device constituting the device group can have a part or all of the functions or functional blocks of the electronic device 20 in the above-mentioned embodiment. As a device group, it is sufficient to have all the functions or functional blocks of the electronic device 20.

[0167] It is understandable that the image compensation method based on electroluminescence detection, the image compensation system 10 based on electroluminescence detection, the electronic device 20 and the storage medium provided in the embodiment of the present application, through the combination of multiple advanced technologies, significantly improve the accuracy and efficiency of photovoltaic panel defect detection, and have important application value and broad promotion prospects. This method not only improves the quality control level of photovoltaic panels, but also makes positive contributions to the sustainable development of the photovoltaic industry.

[0168] Those skilled in the art should recognize that the above embodiments are only used to illustrate the present application and are not intended to be limiting of the present application. As long as they are within the spirit and scope of the present application, appropriate changes and modifications to the above embodiments are within the scope of protection claimed in the present application.

Claims

1. An image compensation method based on electroluminescence detection, applied to detecting defects of photovoltaic panels, characterized in that: The method comprises: Applying a reverse voltage to the photovoltaic panel, and using an electroluminescent camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images; The intensity of the background light is obtained according to at least two of the background images, and the calculation formula of the intensity is: Where N is the number of background images, I i (x,y) is the light intensity of the pixel (x,y) in the i-th image; removing the intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference; The local variance of the pixels in the corrected image is calculated to obtain the noise level in the corrected image, and a filter is constructed according to the noise level, the filter is used to remove the noise in the corrected image, and the calculation formula of the local variance is: in, To correct the local variance of the pixel (x, y) in the image, M is the size of the local window, m is the radius of the window, and i and j are the relative position offsets of the pixel points; A generative adversarial network is applied to process the corrected image to generate an enhanced image, wherein the enhanced image is used to identify defects of the photovoltaic panel.

2. The image compensation method based on electroluminescence detection according to claim 1, characterized in that: The adversarial network includes a generator and a discriminator. The generator is used to reduce the adversarial network of the enhanced image. The loss function of the adversarial network is: Among them, G(z) represents the enhanced image output by the generator, D(G(z)) is the judgment result of the discriminator on the generator output image, G is the generator, D is the discriminator, and z is the random noise input. Represents the probability distribution p z (z) is the expected value of the random noise z extracted from 3. The image compensation method based on electroluminescence detection according to claim 2, characterized in that: The applying a generative adversarial network to process the corrected image to generate an enhanced image includes: The enhanced image is optimized by applying a perceptual loss function to improve the quality of the enhanced image compensation. The formula of the perceptual loss function is: Among them, φ i is the pre-trained convolutional neural network feature extraction layer, I real Represents the light intensity value of the corresponding pixel point (x, y) in the real image.

4. The image compensation method based on electroluminescence detection according to claim 2, characterized in that: The applying a generative adversarial network to process the corrected image to generate an enhanced image also includes: Gradient calculation is applied to introduce structural information into the enhanced image, and the structural information is used to improve the details of the enhanced image to improve the accuracy of identifying defects of the photovoltaic panel. The formula for the gradient calculation is: Among them, G(x,y) represents the gradient amplitude of image I at point (x,y), and are the partial derivatives of the image in the x and y directions respectively.

5. The image compensation method based on electroluminescence detection according to claim 1, characterized in that: The step of removing the intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference comprises: The intensity of the background light is removed by a linear correction algorithm to obtain the corrected image. The formula of the linear correction algorithm is: I correct (x,y)=I EL (x,y)-I bg (x,y) Among them, I correct (x, y) is the corrected image, I EL (x,y) is the electroluminescence image.

6. The image compensation method based on electroluminescence detection according to claim 5, characterized in that: The method further comprises: Obtaining the material type of the photovoltaic panel; Acquiring a linear adjustment parameter of the photovoltaic panel according to a material type of the photovoltaic panel; The corrected image is corrected according to the linear adjustment parameter to form an adapted image. The calculation formula of the adapted image is: I adapted (x,y)=k m ·I correct (x,y)+b m Among them, k m and b m is a linear adjustment parameter.

7. The image compensation method based on electroluminescence detection according to claim 1, characterized in that: The filter is an adaptive filter, and the output of the adaptive filter is: Where W is the size of the adaptive filter.

8. The image compensation method based on electroluminescence detection according to claim 1, characterized in that: The method further comprises: Acquiring shape features, texture features and color features in the enhanced image; The shape feature, the texture feature and the color feature are input into a classifier to identify defects of the photovoltaic panel.

9. An image compensation system based on electroluminescence detection, used to execute the image compensation method based on electroluminescence detection according to any one of claims 1 to 8, characterized in that: The system comprises: An electroluminescent camera module, used for applying a reverse voltage to the photovoltaic panel, and using an electroluminescent camera to capture the surface luminescence signal of the photovoltaic panel to form an electroluminescent image and at least two background images; The light intensity acquisition module is used to acquire the light intensity of the background light according to at least two background images. The calculation formula of the light intensity is: Where N is the number of background images, I i (x,y) is the light intensity of the pixel (x,y) in the i-th image; A light intensity removal module, used for removing the light intensity of the background light from the electroluminescent image to obtain a corrected image without background light interference; A filter construction module is used to calculate the local variance of pixels in the modified image to obtain the noise level in the modified image, and to construct a filter according to the noise level. The filter is used to remove the noise in the modified image. The calculation formula of the local variance is: in, To correct the local variance of the pixel (x, y) in the image, M is the size of the local window, m is the radius of the window, and i and j are the relative position offsets of the pixel points; The enhanced image generation module applies a generative adversarial network to process the corrected image to generate an enhanced image, and the enhanced image is used to identify defects of the photovoltaic panel.

10. A computer-readable storage medium, characterized in that: The device comprises instructions for instructing the device to execute the image compensation method based on electroluminescence detection as claimed in any one of claims 1 to 8.

Images