Method and device for detecting defects of photovoltaic module

Abstract

The invention provides a photovoltaic module defect detection method and device, and the method comprises the steps: responding to the triggering of an event for carrying out the defect detection of a target photovoltaic module, and collecting superimage data corresponding to the target photovoltaic module through superpixel camera equipment; preprocessing the superimage data, and inputting the preprocessed superimage data into the physical interpretable segmentation model to obtain physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physical interpretable segmentation model; fusing the physical feature data and the semantic feature data to obtain defect mark data corresponding to the target photovoltaic module; and respectively performing connected domain analysis and area filtering processing on the defect mark data, and determining a defect detection result corresponding to the target photovoltaic module. By means of the method, the accuracy, efficiency, safety and adaptability of defect detection on the photovoltaic module are improved.

Description

Technical Field

[0001] This application relates to the field of photovoltaic product image processing technology, and in particular to a method and apparatus for detecting defects in photovoltaic modules. Background Technology

[0002] Before leaving the factory, photovoltaic modules must undergo complete online inspection for defects including microcracks, poor solder joints, broken grids, foreign objects, and potential-induced degradation. Since the traditional method of electroluminescence combined with manual visual inspection is too inefficient for production line inspection, a dual-camera inspection method combining an electroluminescent camera and an RGB visible light camera is generally used to detect defects in photovoltaic modules.

[0003] In the dual-camera inspection method that combines an electroluminescent camera with an RGB visible light camera, the electroluminescent camera must be connected to a high current and can only be inspected in a dark room environment, which reduces the safety and adaptability of defect inspection of photovoltaic modules. In addition, the detected results have a low dimension and lack the ability to detect and compare high-level defects. Furthermore, the above method only uses affine transformation camera parallax, and parameters need to be adjusted separately for different models of photovoltaic modules, which reduces the accuracy and efficiency of defect inspection of photovoltaic modules. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide a method and apparatus for detecting defects in photovoltaic modules. The method involves acquiring multi-channel super-image data of the target photovoltaic module through a single exposure using a super-pixel camera in a bright-field environment. The pre-processed super-image data is then input into a physically interpretable segmentation model to obtain physical feature data and semantic feature data. These physical and semantic feature data are then fused, connected component analysis is performed, and area filtering is applied to determine the defect detection result of the target photovoltaic module. This method enables multi-dimensional defect detection of photovoltaic modules under both unpowered and bright-field environments, improving the accuracy and efficiency of defect detection, and consequently enhancing the safety and adaptability of photovoltaic module defect detection.

[0005] This application provides a method for detecting defects in photovoltaic modules, the method comprising: In response to an event that triggers defect detection of a target photovoltaic module, a super-pixel camera device is used to acquire super-image data corresponding to the target photovoltaic module; The hyperimage data is preprocessed and then input into a preset physically interpretable segmentation model to obtain physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model. The physical feature data and the semantic feature data are fused to obtain the defect marking data corresponding to the target photovoltaic module; The defect marker data is subjected to connected component analysis and area filtering to determine the defect detection results corresponding to the target photovoltaic module.

[0006] Furthermore, the acquisition of hyperpixel data corresponding to the target photovoltaic module using a hyperpixel camera device includes: The super-pixel camera device is calibrated and standardized to determine that the super-pixel camera device is in a calibrated state. In response to the fact that the super-pixel camera device is in a calibrated state, the super-pixel camera device is configured to determine that the super-pixel camera device is in a working state; In response to the super-pixel camera device being in working state, the lens of the super-pixel camera device is controlled to be perpendicular to the target photovoltaic module, and the super-pixel camera device is controlled to acquire the super-image data to be processed corresponding to the target photovoltaic module through a single exposure; Thermal drift compensation and polarization information correction are performed on the super-image data to be processed to obtain the super-image data corresponding to the target photovoltaic module; wherein, the super-image data includes spectral channel data, polarization channel data and phase channel data.

[0007] Furthermore, the super-pixel camera device includes a spectral imaging unit, a polarization imaging unit, and a phase imaging unit; the calibration and standardization of the super-pixel camera device to determine that the super-pixel camera device is in a calibrated state includes: The polarization imaging unit is calibrated using a preset calibration polarizer and polarization light source; In response to the polarization imaging unit being calibrated, the mapping relationship between each superpixel and physical coordinates in the superpixel camera device is calibrated based on a preset calibration pixel accuracy value; In response to the calibration of the mapping relationship, the phase plate in the phase imaging unit is surface calibrated; In response to the surface of the phase plate being calibrated, the superpixel camera device is determined to be in a calibrated state.

[0008] Furthermore, configuring the super-pixel camera device to determine that the super-pixel camera device is in a working state includes: The band settings for the spectral imaging unit are configured based on the preset band length and number of bands. The polarization channel of the polarization imaging unit is set based on a preset polarization angle; The phase of the phase imaging unit is set based on the preset phase information; In response to the fact that the spectral imaging unit, the polarization imaging unit, and the phase imaging unit have all been set, it is determined that the superpixel camera device is in a working state.

[0009] Furthermore, the physically interpretable segmentation model includes a physically interpretable branch model and a semantic segmentation branch model; the preprocessing of the hyperimage data and the input of the preprocessed hyperimage data into the preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model includes: The hyperimage data is compressed to obtain compressed hyperimage data, and the compressed hyperimage data is preprocessed to obtain target hyperimage data. The target super-image data is input into the physically interpretable branch model and the semantic segmentation branch model respectively to obtain the physical feature data corresponding to the target photovoltaic module output by the physically interpretable branch model and the semantic feature data corresponding to the target photovoltaic module output by the semantic segmentation branch model; wherein, both the physically interpretable branch model and the semantic segmentation branch model are equipped with a grid line period consistency supervision mechanism.

[0010] Furthermore, the defect labeling data includes defect boundary features, defect type features, confidence level, background environment features, and spatial location features corresponding to each defect; the process of performing connected component analysis and area filtering on the defect labeling data to determine the defect detection result corresponding to the target photovoltaic module includes: Based on the defect boundary features, the defect marking data is binarized to obtain the binary map data corresponding to the defect marking data; Perform connected component analysis on the binary graph data to determine at least one connected region corresponding to the defect marker data; Calculate the pixel area corresponding to each of the connected regions, and perform area filtering on the defect marking data based on the pixel area to obtain the target defect marking data; Based on the target defect marking data, the defect detection results corresponding to the target photovoltaic module are determined; wherein, the defect detection results include the defect type result, defect location result, and confidence level result for each defect.

[0011] Furthermore, the detection method also includes: The defect detection results are uploaded to the manufacturing execution system, and in response to receiving the component identification information returned by the manufacturing execution system, the target photovoltaic module is marked according to the component identification information.

[0012] This application embodiment also provides a photovoltaic module defect detection device, the detection device comprising: The camera control module is used to acquire hyperimage data corresponding to the target photovoltaic module using a hyperpixel camera device in response to an event that triggers defect detection of the target photovoltaic module. The industrial control computer processing module is used to preprocess the hyperimage data and input the preprocessed hyperimage data into a preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model. The data fusion module is used to fuse the physical feature data and the semantic feature data to obtain the defect marking data corresponding to the target photovoltaic module; The data processing module is used to perform connected component analysis and area filtering on the defect marking data to determine the defect detection results corresponding to the target photovoltaic module.

[0013] Furthermore, when the camera control module is used to acquire super-image data corresponding to the target photovoltaic module using the super-pixel camera device, the camera control module is used to: The super-pixel camera device is calibrated and standardized to determine that the super-pixel camera device is in a calibrated state. In response to the fact that the super-pixel camera device is in a calibrated state, the super-pixel camera device is configured to determine that the super-pixel camera device is in a working state; In response to the super-pixel camera device being in working state, the lens of the super-pixel camera device is controlled to be perpendicular to the target photovoltaic module, and the super-pixel camera device is controlled to acquire the super-image data to be processed corresponding to the target photovoltaic module through a single exposure; Thermal drift compensation and polarization information correction are performed on the super-image data to be processed to obtain the super-image data corresponding to the target photovoltaic module; wherein, the super-image data includes spectral channel data, polarization channel data and phase channel data.

[0014] Furthermore, the super-pixel camera device includes a spectral imaging unit, a polarization imaging unit, and a phase imaging unit; when the camera control module is used to calibrate and standardize the super-pixel camera device to determine that the super-pixel camera device is in a calibrated state, the camera control module is used to: The polarization imaging unit is calibrated using a preset calibration polarizer and polarization light source; In response to the polarization imaging unit being calibrated, the mapping relationship between each superpixel and physical coordinates in the superpixel camera device is calibrated based on a preset calibration pixel accuracy value; In response to the calibration of the mapping relationship, the phase plate in the phase imaging unit is surface calibrated; In response to the surface of the phase plate being calibrated, the superpixel camera device is determined to be in a calibrated state.

[0015] Furthermore, when configuring the super-pixel camera device to determine if the super-pixel camera device is in a working state, the camera control module is used to: The band settings for the spectral imaging unit are configured based on the preset band length and number of bands. The polarization channel of the polarization imaging unit is set based on a preset polarization angle; The phase of the phase imaging unit is set based on the preset phase information; In response to the fact that the spectral imaging unit, the polarization imaging unit, and the phase imaging unit have all been set, it is determined that the superpixel camera device is in a working state.

[0016] Furthermore, the physically interpretable segmentation model includes a physically interpretable branch model and a semantic segmentation branch model; when the industrial control computer processing module preprocesses the hyperimage data and inputs the preprocessed hyperimage data into the preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model, the industrial control computer processing module is used to: The hyperimage data is compressed to obtain compressed hyperimage data, and the compressed hyperimage data is preprocessed to obtain target hyperimage data. The target super-image data is input into the physically interpretable branch model and the semantic segmentation branch model respectively to obtain the physical feature data corresponding to the target photovoltaic module output by the physically interpretable branch model and the semantic feature data corresponding to the target photovoltaic module output by the semantic segmentation branch model; wherein, both the physically interpretable branch model and the semantic segmentation branch model are equipped with a grid line period consistency supervision mechanism.

[0017] Furthermore, the defect marking data includes defect boundary features, defect type features, confidence level, background environment features, and spatial location features corresponding to each defect; when the data processing module performs connected component analysis and area filtering on the defect marking data to determine the defect detection result corresponding to the target photovoltaic module, the data processing module is used to: Based on the defect boundary features, the defect marking data is binarized to obtain the binary map data corresponding to the defect marking data; Perform connected component analysis on the binary graph data to determine at least one connected region corresponding to the defect marker data; Calculate the pixel area corresponding to each of the connected regions, and perform area filtering on the defect marking data based on the pixel area to obtain the target defect marking data; Based on the target defect marking data, the defect detection results corresponding to the target photovoltaic module are determined; wherein, the defect detection results include the defect type result, defect location result, and confidence level result for each defect.

[0018] Furthermore, the detection device also includes a result uploading module, which is used for: The defect detection results are uploaded to the manufacturing execution system, and in response to receiving the component identification information returned by the manufacturing execution system, the target photovoltaic module is marked according to the component identification information.

[0019] This application embodiment also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, the steps of the photovoltaic module defect detection method described above are performed.

[0020] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, performs the steps of the photovoltaic module defect detection method described above.

[0021] The photovoltaic module defect detection method and apparatus provided in this application include: in response to an event triggering defect detection of a target photovoltaic module, acquiring super-image data corresponding to the target photovoltaic module using a super-pixel camera device; preprocessing the super-image data and inputting the preprocessed super-image data into a preset physically interpretable segmentation model to obtain physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model; fusing the physical feature data and the semantic feature data to obtain defect labeling data corresponding to the target photovoltaic module; and performing connected component analysis and area filtering on the defect labeling data to determine the defect detection result corresponding to the target photovoltaic module.

[0022] Compared with existing technologies such as traditional electroluminescence combined with manual visual inspection, and dual-camera detection methods using electroluminescence cameras combined with RGB visible light cameras, this method acquires multi-channel hyperimage data of the target photovoltaic module through a single exposure in a bright field environment using a hyperpixel camera. The preprocessed hyperimage data is then input into a physically interpretable segmentation model to obtain physical feature data and semantic feature data. These physical and semantic feature data are then fused, connected component analysis is performed, and area filtering is applied to determine the defect detection results of the target photovoltaic module. This method enables multi-dimensional defect detection of photovoltaic modules in both unpowered and bright field environments, improving the accuracy and efficiency of defect detection, and consequently enhancing the safety and adaptability of photovoltaic module defect detection.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is one of the flowcharts for a method of detecting defects in a photovoltaic module provided in an embodiment of this application; Figure 2 A second flowchart illustrating a method for detecting defects in photovoltaic modules provided in this application embodiment; Figure 3 This is one of the structural schematic diagrams of a photovoltaic module defect detection device provided in an embodiment of this application; Figure 4 This is a second schematic diagram of a photovoltaic module defect detection device provided in an embodiment of this application; Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0027] Research has found that photovoltaic modules must undergo complete online inspection before leaving the factory to detect defects such as microcracks, poor solder joints, broken grids, foreign objects, and potential-induced degradation. Since the traditional method of electroluminescence combined with manual visual inspection is too inefficient for production line inspection, a dual-camera inspection method combining an electroluminescent camera and an RGB visible light camera is generally used to detect defects in photovoltaic modules.

[0028] For example, in a dual-camera inspection method that combines an electroluminescent camera with an RGB visible light camera, the photovoltaic module is powered on online, and a near-infrared image with a size of "3200×1800" is captured by the electroluminescent camera. At the same time, a 25-megapixel RGB visible light camera is triggered to capture a color image of the front of the photovoltaic module. The two images are matched using a template under a given pulse from the PLC, and the grayscale threshold is manually set to complete the defect segmentation.

[0029] Currently, in the dual-camera inspection method that combines an electroluminescent camera with an RGB visible light camera, the electroluminescent camera must connect the photovoltaic module to a high current, for example, 25 A, 60 V DC power. Each shift requires 4000 connection cycles, and the inspection must be carried out in a dark room environment. This poses a high risk of arcing or burning, reducing the safety and adaptability of defect inspection of photovoltaic modules.

[0030] Furthermore, the detection results only include electroluminescence grayscale and RGB three channels, resulting in low dimensionality and a lack of comparison for detecting high-level defects. In addition, the above method only uses affine transformation camera parallax, which leads to a high false positive rate in edge areas. Parameters need to be adjusted individually for different types of photovoltaic modules, reducing the accuracy and efficiency of defect detection for photovoltaic modules.

[0031] Based on this, this application provides a method for detecting defects in photovoltaic modules. By using a super-pixel camera device for a single exposure in a bright-field environment, multi-channel super-image data of the target photovoltaic module is acquired. The preprocessed super-image data is input into a physically interpretable segmentation model to obtain physical feature data and semantic feature data. The physical feature data and semantic feature data are then fused, connected component analysis is performed, and area filtering is applied to determine the defect detection result of the target photovoltaic module. This method achieves multi-dimensional defect detection of photovoltaic modules under both unpowered and bright-field environments, improving the accuracy and efficiency of defect detection, and consequently enhancing the safety and adaptability of photovoltaic module defect detection.

[0032] Please see Figure 1 , Figure 1 This is one of the flowcharts for a method of detecting defects in a photovoltaic module provided in an embodiment of this application. Figure 1 As shown in the embodiments of this application, the method for detecting defects in photovoltaic modules includes: S101. In response to an event that triggers defect detection of the target photovoltaic module, the super-pixel camera device is used to collect super-image data corresponding to the target photovoltaic module.

[0033] It should be noted that the photovoltaic module defect detection method provided in this application embodiment can be applied to a photovoltaic module defect detection system. The photovoltaic module defect detection system may include a super-pixel camera device, a liquid crystal tunable filter snapshot module, a bright field light source, an FPGA compression unit, an industrial control computer for deploying model inference, etc.

[0034] In this step, the target photovoltaic module is first placed at the designated position in the inspection station, ensuring that it is perpendicular to the lens of the super-pixel camera. Then, a precision clamp is used to fix the target photovoltaic module, and a laser alignment device is used to confirm that the target photovoltaic module is perpendicular to the lens of the super-pixel camera to ensure uniform illumination. Finally, the target photovoltaic module triggers the photoelectric sensor, triggering an event to detect defects in the target photovoltaic module, so as to further utilize the super-pixel camera to collect the corresponding super-image data of the target photovoltaic module.

[0035] Here, the target photovoltaic module refers to a photovoltaic module for which defect detection is desired using the method described in the embodiments of this application.

[0036] Photovoltaic modules, also known as solar panels, are the core power generation units that directly convert sunlight into electrical energy. They are the most basic and important hardware components of a photovoltaic power generation system. A standard crystalline silicon photovoltaic module is usually composed of the following five layers: a top cover, an encapsulating film, a cell array, a backsheet, a frame, and a junction box.

[0037] Here, a super-pixel camera device may include a "spectro-polarization-phase" super-pixel camera (SP). 3 C), the super-pixel camera is a deep fusion of spectral imaging, polarization imaging and phase imaging technologies.

[0038] The hyperpixel camera device includes a spectral imaging unit, a polarization imaging unit, and a phase imaging unit. The spectral imaging unit employs hyperspectral imaging technology to achieve sub-nanometer spectral resolution and acquires spectral information across multiple continuous bands through interferometric spectral splitting or metasurface spectral chips to form a unified three-dimensional data cube. The polarization imaging module integrates a micro-polarization filter array to capture polarized light signals at four polarization angles: 0°, 45°, 90°, and 135°. It then synthesizes images of polarization degree and polarization angle using algorithms, transforming microscopic differences into visual comparisons. The phase imaging module employs interferometry to obtain depth information by measuring the phase difference of light waves and also supports indirect ranging, enabling precise measurement of the time difference of reflected photons.

[0039] For example, the front end of the spectral imaging unit uses a tunable liquid crystal filter, set to 25 nm steps, 400 to 1000 nm, a total of 24 bands, and the switching time is set to 0.2 milliseconds to achieve a "one-shot, full-spectrum" snapshot; the focal plane of the polarization imaging unit uses a 2×2 superpixel period, with 4 sub-pixels per period, and aluminum nanowire grid polarizers with 0°, 45°, 90°, and 135° respectively attached; the phase imaging unit has an 8-order silicon dioxide phase plate deposited on its surface, and the phase is obtained by unwrapping through Fourier transform.

[0040] In this embodiment, the super-imaging data includes spectral channel data, polarization channel data, and phase channel data. Specifically, the super-imaging data may include a 24-channel super-imaging data, which is a fusion of spectral, polarization, and phase information. The spectral channels (e.g., 12 to 16 channels) can provide continuous spectral information covering the visible to near-infrared bands, resolving the submolecular characteristics of matter; the polarization channels (e.g., 4 channels) are polarized light signals in four directions: 0°, 45°, 90°, and 135°; and the phase channels (e.g., 8 to 10 channels) can obtain depth information through different phase modulations.

[0041] For example, the representation of super-image data may include a wavelength of 1000 nm, with each band consisting of 25 nm bands; polarization in four directions: 0°, 45°, 90°, and 135°; and a phase of 0-2π.

[0042] In one possible implementation of this application, in specific implementation, the step S101 of acquiring the hyper-image data corresponding to the target photovoltaic module using a hyper-pixel camera device may include: S1011. The super-pixel camera device is calibrated and standardized to determine that the super-pixel camera device is in a calibrated state.

[0043] In one possible implementation of this application, step S1011 may include: S10111. The polarization imaging unit is calibrated using a preset calibration polarizer and polarization light source.

[0044] Here, the accuracy of polarization information is ensured by calibrating the polarization imaging unit.

[0045] S10112. In response to the polarization imaging unit being calibrated, the mapping relationship between each superpixel and physical coordinates in the superpixel camera device is calibrated based on a preset calibration pixel accuracy value.

[0046] For example, during the calibration of the "super-pixel-physical coordinate" lookup table, the mapping relationship between each super-pixel and its physical coordinates is calibrated using the factory-calibrated 0.07 pixel precision value.

[0047] S10113. In response to the calibration of the mapping relationship, the phase plate in the phase imaging unit is surface calibrated.

[0048] Here, we ensure that the surface of the 8th-order silicon dioxide phase plate is free of scratches to ensure the accuracy of phase information acquisition.

[0049] S10114. In response to the surface of the phase plate being calibrated, the super-pixel camera device is determined to be in a calibrated state.

[0050] For example, the system automatically runs the calibration program, using a standard polarized light source (polarization degree of 100%) for calibration. After calibration is completed, the system displays "Polarization calibration completed, error less than 0.07 pixels". At the same time, the system confirms that the "superpixel-physical coordinate" lookup table has been loaded and the error is within the calibration range.

[0051] S1012. In response to the super-pixel camera device being in a calibrated state, the super-pixel camera device is configured to determine that the super-pixel camera device is in a working state.

[0052] In one possible implementation of this application, in specific implementation, the step of configuring the super-pixel camera device in step S1012 to determine that the super-pixel camera device is in a working state may include: S10121. Based on the preset band length and number of bands, the band settings are performed on the spectral imaging unit.

[0053] For example, configure the bands of the tunable liquid crystal filter, set the band length to 400-1000nm, in 25nm increments, for a total of 24 bands.

[0054] S10122. Based on a preset polarization angle, the polarization channel of the polarization imaging unit is set.

[0055] Here, it is confirmed that the polarization imaging unit has been configured with four polarization channels: 0°, 45°, 90°, and 135°.

[0056] S10123. Based on preset phase information, the phase imaging unit is phase-set.

[0057] Here, we confirm that the phase plate is correctly configured and can acquire phase information from 0 to 2π.

[0058] S10124. In response to the fact that the spectral imaging unit, the polarization imaging unit and the phase imaging unit have all been set, it is determined that the super-pixel camera device is in a working state.

[0059] In this step, in response to the fact that the spectral imaging unit, polarization imaging unit, and phase imaging unit have all been set, the exposure time is set according to the component reflectivity and ambient light (e.g., 5-10 ms to ensure sufficient light signal is captured), and the super-pixel camera device is determined to be in working condition.

[0060] S1013. In response to the super-pixel camera device being in working state, control the lens of the super-pixel camera device to be perpendicular to the target photovoltaic module, and control the super-pixel camera device to acquire the super-image data to be processed corresponding to the target photovoltaic module through a single exposure.

[0061] Here, the super-pixel camera device is controlled to simultaneously acquire spectral channel data of 24 bands, polarization channel data of 4 bands and phase channel data in a single exposure, and it is confirmed that the 24 channels of super-image data to be processed have been completely acquired without any missing or abnormal data.

[0062] S1014. Perform thermal drift compensation and polarization information correction on the super-image data to be processed to obtain the super-image data corresponding to the target photovoltaic module.

[0063] In this step, a phase correlation and quadratic surface fitting algorithm is used to compensate for the pixel drift of the super-image data to be processed caused by thermal expansion, and a "super-pixel-physical coordinate" lookup table is used to correct the polarization information in the super-image data to be processed, so as to obtain the super-image data (e.g., an array of [height, width, number of channels]) corresponding to the target photovoltaic module.

[0064] For example, thermal drift is first detected, and a drift of 0.25 pixels is found; then, phase correlation and quadratic surface fitting algorithms are used for compensation, and the drift is less than 0.01 pixels after compensation; next, the polarization information is corrected using a "superpixel-physical coordinate" lookup table; finally, 24 channels of superpixel data are generated, represented as an array of [2048, 2048, 24].

[0065] S102. The hyperimage data is preprocessed, and the preprocessed hyperimage data is input into a preset physically interpretable segmentation model to obtain physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model.

[0066] In the embodiments of this application, the physically interpretable segmentation model may include a dual-branch semantic segmentation network model (Physics-Interpretable SegmentationNet, PIS-Net) that combines physical interpretability with deep learning semantic understanding. PIS-Net breaks through the limitations of traditional semantic segmentation networks. Through the dual-branch architecture of "physical interpretability and semantic segmentation", it integrates physical characteristics and semantic information, and solves the dual problems of the cause of defects and the specific content of defects in photovoltaic module detection.

[0067] Here, the physically interpretable segmentation model includes a physically interpretable branch model and a semantic segmentation branch model.

[0068] In this embodiment of the application, the physical feature data includes, but is not limited to, polarization features, phase features, and spectral features; the semantic feature data includes, but is not limited to, defect boundaries, defect types, confidence levels, background environment, and spatial location.

[0069] In one possible implementation of this application, step S102 may include: S1021. The super-image data is compressed to obtain compressed super-image data, and the compressed super-image data is preprocessed to obtain target super-image data.

[0070] In this step, the data is compressed using FFGA (Field Programmable Gate Array) in a 2x2 lossless manner (e.g., 2×2 compression) to obtain compressed super-image data, thereby reducing the amount of super-image data.

[0071] Furthermore, the compressed hyperimage data is preprocessed, including grayscale conversion and standardization, to obtain the target hyperimage data.

[0072] S1022. Input the target super-image data into the physically interpretable branch model and the semantic segmentation branch model respectively to obtain the physical feature data corresponding to the target photovoltaic module output by the physically interpretable branch model and the semantic feature data corresponding to the target photovoltaic module output by the semantic segmentation branch model.

[0073] In the embodiments of this application, the physically interpretable branch model verifies the absorption coefficient to analyze the spectral reflectance of different bands and identify changes in the optical properties of the component material; the birefringence is calculated through polarization information to identify crystal structure anomalies; and physical feature data such as physical feature maps are generated to transform physical properties into visual features.

[0074] The semantic segmentation branch model uses a lightweight Swin-Transformer to encode the input data and extract high-level semantic features; it enhances multi-scale feature representation through the ASPP (Spatial Pyramid Pooling) module; and it focuses on key regions through an attention mechanism to avoid background noise interference.

[0075] Both the physically interpretable branch model and the semantic segmentation branch model are equipped with a grid period consistency supervision mechanism. Using "grid period consistency" as the supervision signal for the physically interpretable branch model and the semantic segmentation branch model, and based on the grid-consistency loss function, the physically interpretable segmentation model can automatically align the feature space when detecting switching of photovoltaic modules of any pattern.

[0076] Here, when training the physically interpretable segmentation model, style randomization data augmentation is introduced, such as color jitter and polarization perturbation, which forces the physically interpretable branch model to retain only physically relevant features without the need for manual relabeling of the training data.

[0077] For example, the physically interpretable branching model uses the silicon absorption coefficient and bifold ratio as priors; the semantic branches are encoded using the Swing-Transformer.

[0078] S103. The physical feature data and the semantic feature data are fused to obtain the defect marking data corresponding to the target photovoltaic module.

[0079] In this step, the physical feature data is first projected onto the channel space corresponding to the semantic feature data through convolution; then, the projected physical feature data and semantic feature data are concatenated along the channel dimension to obtain the concatenated data; finally, the concatenated data is compressed into the channel space corresponding to the semantic feature data through convolution to obtain the defect marking data corresponding to the target photovoltaic module.

[0080] In this embodiment of the application, the defect marking data includes defect boundary features, defect type features, confidence level, background environment features, and spatial location features corresponding to each defect.

[0081] For example, an example of defect labeling data is shown in the table below.

[0082]

[0083] Here, defect types include, but are not limited to, microcracks, poor solder joints, broken gates, foreign matter, and potential-induced decay.

[0084] S104. Perform connected component analysis and area filtering on the defect marking data to determine the defect detection result corresponding to the target photovoltaic module.

[0085] In this embodiment of the application, the defect detection results include the defect type result, defect location result, and confidence result for each defect.

[0086] In one possible implementation of this application, step S104 may include: S1041. Based on the defect boundary features, the defect marking data is binarized to obtain the binary map data corresponding to the defect marking data.

[0087] In this step, based on a preset binarization threshold, the continuous values ​​in the defect marking data are binarized to obtain the binary map data corresponding to the defect marking data.

[0088] S1042. Perform connected component analysis on the binary graph data to determine at least one connected region corresponding to the defect marker data.

[0089] In this step, adjacent 1-pixel points in the binary image data are grouped into a connected component to perform 8-connected region detection on the defect marking data (e.g., diagonal direction, to avoid missing slender defects such as microcracks), and to determine at least one connected region corresponding to the defect marking data.

[0090] S1043. Calculate the pixel area corresponding to each of the connected regions, and perform area filtering on the defect marking data based on the pixel area to obtain the target defect marking data.

[0091] Here, the area threshold range is pre-set based on the physical size range of defects in the photovoltaic module.

[0092] In this step, based on the pixel area, regions with pixel areas within the area threshold range are retained, and regions with areas outside the area threshold range are removed to obtain target defect marking data.

[0093] S1044. Based on the target defect marking data, determine the defect detection result corresponding to the target photovoltaic module.

[0094] In this step, for each defect in the defect detection results, the defect type result, defect location result, and confidence result are obtained. The defect type result is extracted from the target defect type features in the target defect labeling data, that is, the maximum value in the target defect type features is taken. The defect location result is obtained from the target spatial location features in the target defect labeling data to obtain the center coordinates. The confidence result is obtained from the target confidence of the target defect labeling data to take the average value.

[0095] Optional, please refer to Figure 2 , Figure 2 This is a second flowchart illustrating a method for detecting defects in photovoltaic modules provided in an embodiment of this application. Figure 2 As shown in the figure, the photovoltaic module defect detection method provided in this application embodiment includes step S105 in addition to steps S101 to S104. Specifically, step S105 is used to describe the method of cutting the target photovoltaic module based on the defect detection result corresponding to the target photovoltaic module.

[0096] S105. Upload the defect detection results to the manufacturing execution system, and in response to receiving the component identification information returned by the manufacturing execution system, mark the target photovoltaic module according to the component identification information.

[0097] Here, the Manufacturing Execution System (MES) is the core digital hub connecting the upper-level planning management (such as ERP) and the lower-level industrial control (such as PLC and equipment) of an enterprise, and plays a key role in the intelligent manufacturing system of "connecting the upper and lower levels and connecting the whole".

[0098] In this step, the defect detection results are uploaded to the manufacturing execution system, and the manufacturing execution system determines the component identification information of the target photovoltaic module based on the defect detection results; then, in response to receiving the component identification information returned by the manufacturing execution system, the target photovoltaic module is marked according to the component identification information.

[0099] The component identification information is used to identify whether the target photovoltaic module is a qualified product. For example, unqualified products are marked and unloaded to form a closed production loop and ensure that the process is controllable and traceable.

[0100] The photovoltaic module defect detection method provided in this application acquires multi-channel super-image data of the target photovoltaic module through a single exposure of a super-pixel camera device in a bright field environment. The preprocessed super-image data is input into a physically interpretable segmentation model to obtain physical feature data and semantic feature data. The physical feature data and semantic feature data are then fused, connected component analysis is performed, and area filtering is applied to determine the defect detection result of the target photovoltaic module. This method realizes multi-dimensional defect detection of photovoltaic modules in both unpowered and bright field environments, improving the accuracy and efficiency of defect detection of photovoltaic modules, and thus enhancing the safety and adaptability of defect detection of photovoltaic modules.

[0101] Please see Figure 3 , Figure 4 , Figure 3 This is one of the structural schematic diagrams of a photovoltaic module defect detection device provided in an embodiment of this application. Figure 4 This is a second schematic diagram of a photovoltaic module defect detection device provided in an embodiment of this application. Figure 3 As shown, the detection device 300 includes: The camera control module 310 is used to collect hyperpixel data corresponding to the target photovoltaic module in response to an event that triggers the defect detection of the target photovoltaic module. The industrial control computer processing module 320 is used to preprocess the hyperimage data and input the preprocessed hyperimage data into a preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model. The data fusion module 330 is used to fuse the physical feature data and the semantic feature data to obtain the defect marking data corresponding to the target photovoltaic module; The data processing module 340 is used to perform connected component analysis and area filtering on the defect marking data to determine the defect detection result corresponding to the target photovoltaic module.

[0102] Furthermore, when the camera control module 310 is used to acquire super-image data corresponding to the target photovoltaic module using the super-pixel camera device, the camera control module 310 is used to: The super-pixel camera device is calibrated and standardized to determine that the super-pixel camera device is in a calibrated state. In response to the fact that the super-pixel camera device is in a calibrated state, the super-pixel camera device is configured to determine that the super-pixel camera device is in a working state; In response to the super-pixel camera device being in working state, the lens of the super-pixel camera device is controlled to be perpendicular to the target photovoltaic module, and the super-pixel camera device is controlled to acquire the super-image data to be processed corresponding to the target photovoltaic module through a single exposure; Thermal drift compensation and polarization information correction are performed on the super-image data to be processed to obtain the super-image data corresponding to the target photovoltaic module; wherein, the super-image data includes spectral channel data, polarization channel data and phase channel data.

[0103] Furthermore, the super-pixel camera device includes a spectral imaging unit, a polarization imaging unit, and a phase imaging unit; when the camera control module 310 is used to calibrate and standardize the super-pixel camera device to determine that the super-pixel camera device is in a calibrated state, the camera control module 310 is used to: The polarization imaging unit is calibrated using a preset calibration polarizer and polarization light source; In response to the polarization imaging unit being calibrated, the mapping relationship between each superpixel and physical coordinates in the superpixel camera device is calibrated based on a preset calibration pixel accuracy value; In response to the calibration of the mapping relationship, the phase plate in the phase imaging unit is surface calibrated; In response to the surface of the phase plate being calibrated, the superpixel camera device is determined to be in a calibrated state.

[0104] Furthermore, when configuring the super-pixel camera device to determine if the super-pixel camera device is in a working state, the camera control module 310 is used to: The band settings for the spectral imaging unit are configured based on the preset band length and number of bands. The polarization channel of the polarization imaging unit is set based on a preset polarization angle; The phase of the phase imaging unit is set based on the preset phase information; In response to the fact that the spectral imaging unit, the polarization imaging unit, and the phase imaging unit have all been set, it is determined that the superpixel camera device is in a working state.

[0105] Furthermore, the physically interpretable segmentation model includes a physically interpretable branch model and a semantic segmentation branch model; when the industrial control computer processing module 320 preprocesses the hyperimage data and inputs the preprocessed hyperimage data into the preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model, the industrial control computer processing module 320 is used to: The hyperimage data is compressed to obtain compressed hyperimage data, and the compressed hyperimage data is preprocessed to obtain target hyperimage data. The target super-image data is input into the physically interpretable branch model and the semantic segmentation branch model respectively to obtain the physical feature data corresponding to the target photovoltaic module output by the physically interpretable branch model and the semantic feature data corresponding to the target photovoltaic module output by the semantic segmentation branch model; wherein, both the physically interpretable branch model and the semantic segmentation branch model are equipped with a grid line period consistency supervision mechanism.

[0106] Furthermore, the defect marking data includes defect boundary features, defect type features, confidence level, background environment features, and spatial location features corresponding to each defect; when the data processing module 340 performs connected component analysis and area filtering on the defect marking data to determine the defect detection result corresponding to the target photovoltaic module, the data processing module 340 is used to: Based on the defect boundary features, the defect marking data is binarized to obtain the binary map data corresponding to the defect marking data; Perform connected component analysis on the binary graph data to determine at least one connected region corresponding to the defect marker data; Calculate the pixel area corresponding to each of the connected regions, and perform area filtering on the defect marking data based on the pixel area to obtain the target defect marking data; Based on the target defect marking data, the defect detection results corresponding to the target photovoltaic module are determined; wherein, the defect detection results include the defect type result, defect location result, and confidence level result for each defect.

[0107] Furthermore, such as Figure 4 As shown, the detection device 300 further includes a result uploading module 350, which is used for: The defect detection results are uploaded to the manufacturing execution system, and in response to receiving the component identification information returned by the manufacturing execution system, the target photovoltaic module is marked according to the component identification information.

[0108] The photovoltaic module defect detection device provided in this application acquires multi-channel super-image data of the target photovoltaic module through a single exposure of a super-pixel camera in a bright field environment. The preprocessed super-image data is input into a physically interpretable segmentation model to obtain physical feature data and semantic feature data. The physical feature data and semantic feature data are then fused, connected component analysis is performed, and area filtering is applied to determine the defect detection result of the target photovoltaic module. This device enables multi-dimensional defect detection of photovoltaic modules in both unpowered and bright field environments, improving the accuracy and efficiency of defect detection, and thus enhancing the safety and adaptability of defect detection for photovoltaic modules.

[0109] Please see Figure 5 , Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 5 As shown, the electronic device 500 includes a processor 510, a memory 520, and a bus 530.

[0110] The memory 520 stores machine-readable instructions executable by the processor 510. When the electronic device 500 is running, the processor 510 and the memory 520 communicate via the bus 530. When the machine-readable instructions are executed by the processor 510, they can perform the operations described above. Figure 1 as well as Figure 2 The steps of the photovoltaic module defect detection method in the illustrated method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0111] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 as well as Figure 2 The steps of the photovoltaic module defect detection method in the method embodiment shown are described in detail in the method embodiment, and will not be repeated here.

[0112] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0113] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0114] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0115] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0116] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0117] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for detecting defects in photovoltaic modules, characterized in that, The detection method includes: In response to an event that triggers defect detection of a target photovoltaic module, a super-pixel camera device is used to acquire super-image data corresponding to the target photovoltaic module; The hyperimage data is preprocessed and then input into a preset physically interpretable segmentation model to obtain physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model. The physical feature data and the semantic feature data are fused to obtain the defect marking data corresponding to the target photovoltaic module; The defect marker data is subjected to connected component analysis and area filtering to determine the defect detection results corresponding to the target photovoltaic module.

2. The method according to claim 1, characterized in that, The acquisition of hyperpixel data corresponding to the target photovoltaic module using a hyperpixel camera device includes: The super-pixel camera device is calibrated and standardized to determine that the super-pixel camera device is in a calibrated state. In response to the fact that the super-pixel camera device is in a calibrated state, the super-pixel camera device is configured to determine that the super-pixel camera device is in a working state; In response to the super-pixel camera device being in working state, the lens of the super-pixel camera device is controlled to be perpendicular to the target photovoltaic module, and the super-pixel camera device is controlled to acquire the super-image data to be processed corresponding to the target photovoltaic module through a single exposure; Thermal drift compensation and polarization information correction are performed on the super-image data to be processed to obtain the super-image data corresponding to the target photovoltaic module; wherein, the super-image data includes spectral channel data, polarization channel data and phase channel data.

3. The method according to claim 2, characterized in that, The super-pixel camera device includes a spectral imaging unit, a polarization imaging unit, and a phase imaging unit; the calibration and standardization of the super-pixel camera device to determine that the super-pixel camera device is in a calibrated state includes: The polarization imaging unit is calibrated using a preset calibration polarizer and polarization light source; In response to the polarization imaging unit being calibrated, the mapping relationship between each superpixel and physical coordinates in the superpixel camera device is calibrated based on a preset calibration pixel accuracy value; In response to the calibration of the mapping relationship, the phase plate in the phase imaging unit is surface calibrated; In response to the surface of the phase plate being calibrated, the superpixel camera device is determined to be in a calibrated state.

4. The method according to claim 3, characterized in that, The configuration settings for the super-pixel camera device to determine that the super-pixel camera device is in a working state include: The band settings for the spectral imaging unit are configured based on the preset band length and number of bands. The polarization channel of the polarization imaging unit is set based on a preset polarization angle; The phase of the phase imaging unit is set based on the preset phase information; In response to the fact that the spectral imaging unit, the polarization imaging unit, and the phase imaging unit have all been set, it is determined that the superpixel camera device is in a working state.

5. The method according to claim 1, characterized in that, The physically interpretable segmentation model includes a physically interpretable branch model and a semantic segmentation branch model; the preprocessing of the hyperimage data and the input of the preprocessed hyperimage data into the preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model include: The hyperimage data is compressed to obtain compressed hyperimage data, and the compressed hyperimage data is preprocessed to obtain target hyperimage data; The target super-image data is input into the physically interpretable branch model and the semantic segmentation branch model respectively to obtain the physical feature data corresponding to the target photovoltaic module output by the physically interpretable branch model and the semantic feature data corresponding to the target photovoltaic module output by the semantic segmentation branch model; wherein, both the physically interpretable branch model and the semantic segmentation branch model are equipped with a grid line period consistency supervision mechanism.

6. The method according to claim 1, characterized in that, The defect labeling data includes defect boundary features, defect type features, confidence level, background environment features, and spatial location features corresponding to each defect; The process of performing connected component analysis and area filtering on the defect-marked data to determine the defect detection result corresponding to the target photovoltaic module includes: Based on the defect boundary features, the defect marking data is binarized to obtain the binary map data corresponding to the defect marking data; Perform connected component analysis on the binary graph data to determine at least one connected region corresponding to the defect marker data; Calculate the pixel area corresponding to each of the connected regions, and perform area filtering on the defect marking data based on the pixel area to obtain the target defect marking data; Based on the target defect marking data, the defect detection results corresponding to the target photovoltaic module are determined; wherein, the defect detection results include the defect type result, defect location result, and confidence level result for each defect.

7. The method according to claim 1, characterized in that, The detection method further includes: The defect detection results are uploaded to the manufacturing execution system, and in response to receiving the component identification information returned by the manufacturing execution system, the target photovoltaic module is marked according to the component identification information.

8. A device for detecting defects in photovoltaic modules, characterized in that, The detection device includes: A camera control module is used to acquire hyper-image data corresponding to the target photovoltaic module using a hyper-pixel camera device in response to an event that triggers defect detection of the target photovoltaic module. The industrial control computer processing module is used to preprocess the hyperimage data and input the preprocessed hyperimage data into a preset physically interpretable segmentation model to obtain the physical feature data and semantic feature data corresponding to the target photovoltaic module output by the physically interpretable segmentation model. The data fusion module is used to fuse the physical feature data and the semantic feature data to obtain the defect marking data corresponding to the target photovoltaic module; The data processing module is used to perform connected component analysis and area filtering on the defect marking data to determine the defect detection results corresponding to the target photovoltaic module.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. The machine-readable instructions are executed by the processor to perform the steps of the photovoltaic module defect detection method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the photovoltaic module defect detection method as described in any one of claims 1 to 7.

Images