Photovoltaic module defect detection method based on image processing

Abstract

The invention relates to the technical field of photovoltaic module quality detection, and provides a photovoltaic module defect detection method based on image processing, and the method comprises the steps: collecting the surface image data of a photovoltaic module through an optical sensor, transmitting the surface image data to a processing unit for analysis, and visually displaying the defect data to evaluate the quality. The detection defects comprise cracks, subfissures or hot spots of the battery piece. The analysis process comprises the following steps: acquiring a plurality of groups of image data queues of the same batch of components under different illumination conditions; decomposing into a brightness distribution item and a texture feature item, and calculating a noise interference amplitude after segmenting into an image sub-queue and a texture sub-item; determining the noise interference amplitude of each group by combining the same area data of all components under the same condition; the deviation degree is calculated according to the brightness distribution item difference, and the final noise interference amplitude is determined by combining the deviation degree difference of the adjacent illumination conditions; and screening a high-quality image data queue as a detection result based on the final noise interference amplitude.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic module quality detection, and more specifically, to a photovoltaic module defect detection method based on image processing. Background Art

[0002] Quality inspection of photovoltaic modules has always been a critical component of the rapid development of the photovoltaic industry. Existing methods for detecting defects in photovoltaic modules primarily rely on manual visual inspection or simple image processing techniques. Manual inspection is not only inefficient but also susceptible to subjective factors, leading to unstable and inconsistent test results. While traditional image processing technology can assist with detection to a certain extent, it is often only able to identify relatively obvious defects. Its ability to detect subtle defects such as microcracks, hidden cracks, or hot spots is limited. Furthermore, it is sensitive to lighting conditions and environmental changes, making it prone to misjudgment.

[0003] Existing image processing technologies typically employ single image analysis methods, such as threshold segmentation or simple edge detection algorithms. These methods struggle to accurately distinguish defects from normal areas when processing images with complex backgrounds or uneven lighting. Furthermore, traditional methods often overlook the variations in the image features of photovoltaic modules under varying lighting conditions, resulting in insufficiently accurate and reliable inspection results.

[0004] While implementing the embodiments of the present invention, the inventors discovered at least the following problems or defects in the existing technology: existing detection methods cannot effectively address the impact of illumination variations on image analysis, making it difficult to accurately identify minor defects. Furthermore, they lack precise assessment and correction mechanisms for image noise interference, resulting in insufficient stability in detection results. These issues have severely impacted the efficiency and accuracy of photovoltaic module defect detection, hindering the further development of the photovoltaic industry. Summary of the Invention

[0005] The present invention provides a photovoltaic module defect detection method based on image processing.

[0006] In a first aspect of the present invention, a photovoltaic module defect detection method based on image processing is provided, comprising: an image acquisition device acquires surface image data of photovoltaic modules through its optical sensor and transmits the data to a processing unit for analysis, and then visualizes the analyzed defect data on a display terminal to perform quality assessment of photovoltaic modules, wherein defect types include cell cracks, hidden cracks or hot spots; the analysis process comprises: S1, obtaining a plurality of groups of surface image data queues of modules of the same batch under different lighting conditions of photovoltaic modules.

[0007] S2: Decompose each group of image data queues into brightness distribution items and texture feature items; based on the time points corresponding to all extreme values ​​in the brightness distribution items, split each group of image data queues and texture feature items into multiple image sub-queues and texture sub-items; calculate the noise interference amplitude of each image sub-queue based on the number of extreme values ​​in each texture sub-item and the difference between adjacent extreme values.

[0008] S3, determining the noise interference amplitude of each group of image data queues based on the noise interference amplitude of each image sub-queue and the image data of the same area in the image data queues of all components under the same conditions.

[0009] S4, calculating the deviation of each group of image data queues based on the difference in brightness distribution items of the image data queues of all components under the same conditions; determining the final noise interference amplitude of each group of image data queues based on the difference in deviation between the noise interference amplitude of each group and the image data queues with the same sequence number under adjacent lighting conditions.

[0010] S5, based on the final noise interference amplitude of each group of image data queues, screening the corresponding high-quality image data queues of each group; and using the image data in the high-quality image data queues as defect detection results.

[0011] Furthermore, in S1 , the plurality of groups of surface image data queues are queues formed by photovoltaic module surface images arranged in order of acquisition time.

[0012] Furthermore, in S2, the method for calculating the noise interference amplitude of each image sub-queue in each group of image data queues includes: For the i-th group of image data queues of components from the same batch under arbitrary lighting conditions, the wavelet decomposition algorithm is used to decompose the queue into a brightness distribution item, a texture feature item, and a residual item; the Canny edge detection algorithm is used to extract the extreme points of the brightness distribution item; based on the time points corresponding to all extreme values ​​in the brightness distribution item, the i-th group of image data queues and texture feature items are divided into multiple image sub-queues and texture sub-items; the gradient of the j-th texture sub-item of the i-th group of image data queues is calculated to extract its extreme points; the number of extreme values ​​in the j-th texture sub-item of the i-th group of image data queues is divided by the number of pixels in the j-th image sub-item of the i-th group of image data queues, which is defined as the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queues; the absolute standard deviation of the differences between adjacent extreme values ​​in the j-th texture sub-item of the i-th group of image data queues is calculated, and the result is multiplied by the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queues to obtain the noise interference amplitude of the j-th image sub-item of the i-th group of image data queues.

[0013] Furthermore, in S2, the calculation formula for the noise interference amplitude of the j-th image sub-queue of the i-th group of image data queue is: ;in, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue, is the number of extreme values ​​of the j-th texture sub-item in the i-th group of image data queue, is the number of pixels of the jth image sub-queue of the i-th group of image data queue, is the kth extreme value in the jth texture sub-item of the i-th group image data queue, is the mean of the absolute values ​​of the differences between adjacent extreme values.

[0014] Furthermore, in S3, the method for determining the noise interference amplitude of each group of image data queues includes: among all image data queues under the mth lighting condition of the same batch of components, intercepting the reference image subqueues of other groups of image data queues at the start and end time points of the jth image subqueue of the i-th group of image data queue; defining the average of the jth image subqueue of the i-th group of image data queue and all reference image subqueues as the standard image subqueue; using the SSIM algorithm to calculate the structural similarity between the jth image subqueue of the i-th group of image data queue and the corresponding standard image subqueue; in the i-th group of image data queue under the m-th lighting condition, calculating the sum of the noise interference amplitudes of all image subqueues, and multiplying the ratio of the noise interference amplitude of the j-th image subqueue to the sum by the structural similarity to define it as the noise interference amplitude of the i-th group of image data queue under the m-th lighting condition.

[0015] Furthermore, in S3, the calculation formula for the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is: ;in, is the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue under the m-th illumination condition, is the number of image subqueues in the i-th group image data queue under the m-th illumination condition, It is the SSIM similarity between the j-th image sub-queue of the i-th group of image data queue under the m-th illumination condition and the standard image sub-queue.

[0016] Furthermore, in S4, the method for determining the final noise interference amplitude of each group includes: among all image data queues under the mth illumination condition, using Hausdorff distance to calculate the difference between the brightness distribution item of the i-th group of image data queue and the brightness distribution items of other groups, and defining the normalized mean of the difference between the i-th group of image data queue and other groups as the deviation of the i-th group of image data queue; calculating the deviation difference of the image data queues with the same serial number under adjacent illumination conditions, and taking the absolute value mean as the abnormality coefficient of each group under the m-th illumination condition; multiplying the negative correlation normalized value of the abnormality coefficient with the noise interference amplitude to obtain the final noise interference amplitude of the i-th group of image data queue under the m-th illumination condition.

[0017] Furthermore, in S4, the deviation calculation formula of the i-th group of image data queue under the m-th illumination condition is: ;in, is the deviation of the i-th group of image data queue under the m-th illumination condition, is the total number of groups, is the brightness distribution item of the i-th group of image data queue under the m-th lighting condition.

[0018] The calculation formula of the anomaly coefficient is: ;in, is the abnormal coefficient of the i-th group of image data queue under the m-th illumination condition, is the total number of lighting conditions.

[0019] The final noise interference amplitude calculation formula is: ;in, and is the mean and standard deviation of all abnormal coefficients.

[0020] Furthermore, in S5, the method for screening the high-quality image data queue includes: setting a threshold ,in and is the mean and standard deviation of all final noise interference amplitudes, is the preset coefficient.

[0021] If the final noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is , it is determined to be a high-quality image data queue.

[0022] Morphological closing operation is used to eliminate noise on high-quality image data queue and output defect detection results.

[0023] The above-described embodiments of the present invention have at least the following beneficial effects: First, by collecting and analyzing multiple image data queues and combining them with the decomposition of brightness distribution and texture features, the present invention effectively reduces the impact of varying lighting conditions on detection results. Furthermore, by accurately calculating and correcting the noise interference amplitude, the quality of image data can be improved, enabling more accurate identification of defects such as microcracks, hidden cracks, and hot spots on the surface of photovoltaic modules. This approach improves detection stability and reliability, reducing the risk of misjudgments and missed detections.

[0024] Secondly, the present invention further optimizes the visualization of defect detection results by setting a threshold to filter high-quality image data and using morphological closing operations to eliminate noise. This processing method not only improves detection efficiency but also provides more intuitive and accurate data support for PV module quality assessment, helping to promptly identify potential problems and ensure the performance and service life of PV modules. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood by reading the following detailed description with reference to the accompanying drawings, in which several embodiments of the present invention are shown by way of example and not limitation.

[0026] in: Figure 1 A schematic flow chart of a photovoltaic module defect detection method based on image processing provided in one embodiment of the present invention.

[0027] Figure 2 A schematic structural diagram of a photovoltaic module defect detection system based on image processing provided by one embodiment of the present invention.

[0028] Figure 3 The figure schematically shows the structure of an electronic device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The principles and spirit of the present invention will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are provided solely to enable those skilled in the art to better understand and implement the present invention, and are not intended to limit the scope of the present invention in any way. Rather, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0030] Those skilled in the art will appreciate that the embodiments of the present invention may be implemented as a system, apparatus, device, method, or computer program product. Therefore, the present invention may be implemented in the following forms: entirely in hardware, entirely in software (including firmware, resident software, microcode, etc.), or in a combination of hardware and software.

[0031] It should be noted that any number of elements in the drawings is for illustration only and not for limitation, and any naming is only for distinction and does not have any limiting meaning.

[0032] Reference below Figure 1 , Figure 1 Schematic diagram of a process flow of a photovoltaic module defect detection method based on image processing provided by an embodiment of the present invention. Figure 1 As shown, a photovoltaic module defect detection method based on image processing includes: The image acquisition device collects image data of the photovoltaic module surface through its optical sensor and transmits it to the processing unit for analysis. The analyzed defect data is then visualized on the display terminal to perform quality assessment of the photovoltaic module. The defect types include cell cracks, hidden cracks or hot spots.

[0033] The analysis process includes: S1, obtaining multiple groups of surface image data queues of photovoltaic components of the same batch under different lighting conditions.

[0034] S2: Decompose each group of image data queues into brightness distribution items and texture feature items; based on the time points corresponding to all extreme values ​​in the brightness distribution items, split each group of image data queues and texture feature items into multiple image sub-queues and texture sub-items; calculate the noise interference amplitude of each image sub-queue based on the number of extreme values ​​in each texture sub-item and the difference between adjacent extreme values.

[0035] S3, determining the noise interference amplitude of each group of image data queues based on the noise interference amplitude of each image sub-queue and the image data of the same area in the image data queues of all components under the same conditions.

[0036] S4, calculating the deviation of each group of image data queues based on the difference in brightness distribution items of the image data queues of all components under the same conditions; determining the final noise interference amplitude of each group of image data queues based on the difference in deviation between the noise interference amplitude of each group and the image data queues with the same sequence number under adjacent lighting conditions.

[0037] S5, based on the final noise interference amplitude of each group of image data queues, screening the corresponding high-quality image data queues of each group; and using the image data in the high-quality image data queues as defect detection results.

[0038] It should be noted that the image acquisition device uses its optical sensor to capture image data from the photovoltaic module surface and transmits it to the processing unit for analysis. The analyzed defect data is then visualized on a display terminal to assess the quality of the photovoltaic modules. The image acquisition device is a device used to capture images of the photovoltaic module surface. The optical sensor is its core component, converting optical signals from the module surface into electrical signals to generate image data. The processing unit analyzes and processes the captured image data, identifying defect features within the image and transmitting the analysis results to the display terminal for visualization. The display terminal displays the analysis results, allowing users to intuitively understand the quality status of the photovoltaic modules. Defect types include cell cracks, hidden cracks, and hot spots, which are common failure modes of photovoltaic modules. Cracks and hidden cracks can affect the mechanical strength and electrical performance of the cells, while hot spots can cause localized overheating, impacting the module's service life. This approach enables rapid and accurate assessment of photovoltaic module quality.

[0039] Specifically, the optical sensor in the image acquisition device can be a high-resolution camera or other imaging device, and the image data it collects can reflect the details of the surface of the photovoltaic module. The processing unit usually includes image processing software and hardware equipment, which can perform complex analysis on the collected image data. The analysis process involves decomposing and processing the brightness distribution, texture features, etc. of the image to identify possible defects. The brightness distribution item refers to the distribution of pixel brightness in the image, and the texture feature item refers to the detailed features of the texture in the image. Analysis of these features can help distinguish between normal areas and defective areas. For example, a crack in a cell may appear as a discontinuous line in the texture feature, while a hot spot may appear as a localized high-brightness area in the brightness distribution. The image data queue refers to a queue formed by images of the photovoltaic module surface arranged in order of acquisition time, which is used to record image changes at different time points for subsequent analysis.

[0040] Preferably, during the image acquisition process, the acquisition parameters can be adjusted according to different lighting conditions to ensure the quality of the image data. For example, when the light is weak, the exposure time can be appropriately increased or the sensitivity of the sensor can be improved. In the processing unit, the wavelet decomposition algorithm can be used to decompose the image data queue to extract the brightness distribution item and the texture feature item. The wavelet decomposition algorithm is a mathematical tool that can decompose an image into different frequency components. Through this decomposition, the detailed features in the image can be more clearly identified. During the analysis process, high-quality image data queues can also be screened by setting a threshold. The setting of the threshold can be determined based on the statistical characteristics of the noise interference amplitude. For example, the threshold can be set to the mean of the noise interference amplitude plus several times the standard deviation. The noise interference amplitude refers to the degree of interference introduced in the image due to environmental factors or equipment errors. By calculating this value, poor quality image data can be excluded, thereby improving the accuracy of defect detection.

[0041] In some embodiments, in S1 , the plurality of groups of surface image data queues are queues formed by photovoltaic assembly surface images arranged in order of acquisition time.

[0042] It should be noted that in the photovoltaic module defect detection method, the multiple surface image data queues consist of images of the photovoltaic module surface arranged in chronological order. These multiple surface image data queues refer to sequences of images acquired multiple times under varying lighting conditions. This chronological arrangement reflects changes in the surface condition of the photovoltaic module under varying lighting conditions, providing rich information for subsequent defect detection. For example, by comparing images taken at different time points, defect signatures caused by changes in lighting or the module's own condition can be more clearly identified.

[0043] Specifically, multiple surface image data queues refer to image sequences generated by capturing multiple images of a PV module under different lighting conditions. Each image data queue contains multiple chronologically ordered images that reflect the surface conditions of the PV module under specific lighting conditions. For example, reflections and shadows on the PV module surface may vary under different lighting intensities or angles. The number of images in each image data queue can be set based on actual needs. For example, one image can be captured every minute for 10 minutes, resulting in a queue of 10 images. This time-series image data queue provides rich information for subsequent analysis, helping to identify defects caused by varying lighting conditions or changes in the module's internal state. Furthermore, lighting conditions refer to the ambient lighting conditions at the time of image capture, including parameters such as intensity, angle, and uniformity. By capturing images under different lighting conditions, a more comprehensive understanding of the surface characteristics of the PV module can be achieved.

[0044] Preferably, when collecting multiple sets of surface image data queues, automated equipment can be used to control the collection time interval and lighting conditions. For example, a timer can be set to trigger image acquisition at regular intervals, such as every 30 seconds, and the intensity or angle of the light source can be adjusted simultaneously to obtain images under different lighting conditions. When processing this image data, each set of image data queues can be preprocessed, such as removing noise and correcting geometric distortion, to improve image quality. Feature extraction can then be performed on the images in each set of image data queues, such as calculating brightness distribution and texture features. These features will serve as the basis for subsequent analysis. In this way, the collected image data queues can be ensured to be of high quality and high consistency, providing strong support for accurate defect detection in photovoltaic modules.

[0045] In some embodiments, in S2, a method for calculating the noise interference amplitude of each image subqueue in each group of image data queues includes: for the i-th group of image data queues of components from the same batch under any lighting conditions, using a wavelet decomposition algorithm to decompose the queue into a brightness distribution item, a texture feature item, and a residual item; using a Canny edge detection algorithm to extract the extreme value points of the brightness distribution item; based on the time points corresponding to all extreme values ​​in the brightness distribution item, dividing the i-th group of image data queues and texture feature items into multiple image subqueues and texture subitems; and performing a calculation on the j-th texture feature item of the i-th group of image data queues. The gradient of the texture sub-item of the i-th group of image data queue is calculated, and its extreme value points are extracted; the number of extreme values ​​in the j-th texture sub-item of the i-th group of image data queue is divided by the number of pixels of the j-th image sub-item of the i-th group of image data queue, and the result is defined as the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queue; the absolute value standard deviation of the difference between adjacent extreme values ​​in the j-th texture sub-item of the i-th group of image data queue is calculated, and the result is multiplied by the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queue to obtain the noise interference amplitude of the j-th image sub-item of the i-th group of image data queue.

[0046] It should be noted that in the photovoltaic module defect detection method, the calculation method of the noise interference amplitude of each image sub-queue in each group of image data queues includes: for the i-th group of image data queues of the same batch of modules under arbitrary lighting conditions, using the wavelet decomposition algorithm to decompose the queue into brightness distribution items, texture feature items and residual items; using the Canny edge detection algorithm to extract the extreme points of the brightness distribution items; based on the time points corresponding to all extreme values ​​in the brightness distribution items, dividing the i-th group of image data queues and texture feature items into multiple image sub-queues and texture sub-items; performing gradient calculation on the j-th texture sub-item of the i-th group of image data queues to extract its extreme points; dividing the number of extreme values ​​in the j-th texture sub-item of the i-th group of image data queues by the number of pixels of the j-th image sub-item of the i-th group of image data queues, defining it as the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queues; calculating the absolute value standard deviation of the difference between adjacent extreme values ​​in the j-th texture sub-item of the i-th group of image data queues, and comparing it with ... The texture fluctuation coefficients of the image sub-queues are multiplied together to obtain the noise interference amplitude of the j-th image sub-queue in the i-th image data queue. The core of this process is to quantify the degree of noise interference in the image by decomposing the image data queue and extracting key features, thereby providing a more accurate image data foundation for subsequent defect detection.

[0047] Specifically, the wavelet decomposition algorithm is a mathematical tool that decomposes an image into different frequency components. Through this decomposition, the image data sequence can be decomposed into a brightness distribution term, a texture feature term, and a residual term. The brightness distribution term reflects the overall distribution of pixel brightness in the image; the texture feature term contains the detailed characteristics of the image texture; and the residual term typically contains noise and other irregularities. The Canny edge detection algorithm is a classic edge detection method used to extract extreme points in the brightness distribution term. These extreme points typically correspond to areas of significant image change, such as the edges of cracks or hot spots. Based on the time points corresponding to these extreme points, the image data sequence can be divided into multiple image sub-sequences, each of which corresponds to image changes over a period of time. The texture fluctuation coefficient is defined by dividing the number of extreme points in the texture sub-sequence by the number of pixels in the image sub-sequence. It reflects the degree of texture fluctuation. The absolute standard deviation of the differences between adjacent extreme points further quantifies the severity of texture changes. Multiplying these two parameters yields the noise interference amplitude, a metric used to measure the degree of noise interference within an image sub-sequence.

[0048] Preferably, during implementation, the parameters of the wavelet decomposition algorithm can be optimized to better suit the characteristics of photovoltaic module images. For example, appropriate wavelet basis functions and decomposition levels can be selected to ensure that the decomposed brightness distribution and texture feature terms clearly reflect the key information in the image. When using the Canny edge detection algorithm, the edge detection threshold parameters can be adjusted to improve the accuracy and robustness of extreme point extraction. For the calculation of texture fluctuation coefficients and noise interference amplitudes, appropriate statistical methods can be selected based on actual needs. For example, when calculating the absolute standard deviation of the difference between adjacent extreme values, a sliding window method can be used to smooth the data to reduce the impact of local fluctuations. Through these optimization measures, the noise interference amplitude can be calculated more accurately, thereby improving the accuracy of defect detection.

[0049] In some embodiments, in S2, the calculation formula for the noise interference amplitude of the j-th image sub-queue in the i-th group of image data queues is: ;in, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue, is the number of extreme values ​​of the j-th texture sub-item in the i-th group of image data queue, is the number of pixels of the jth image sub-queue of the i-th group of image data queue, is the kth extreme value in the jth texture sub-item of the i-th group image data queue, is the mean of the absolute values ​​of the differences between adjacent extreme values.

[0050] It's important to note that the formula for calculating the noise interference amplitude quantifies the degree of noise interference within an image subqueue. This formula comprehensively assesses the noise level within an image subqueue by combining the number of extreme values ​​in the texture sub-item, the number of pixels in the image subqueue, and the absolute standard deviation of the differences between adjacent extreme values. The number of extreme values ​​in the texture sub-item reflects the complexity of the texture features, the number of pixels represents the size of the image subqueue, and the absolute standard deviation of the differences between adjacent extreme values ​​measures the severity of texture fluctuations. By combining these factors, we can more accurately assess noise interference within an image, providing a more reliable basis for subsequent defect detection.

[0051] Specifically, each parameter in the formula has a clear physical meaning. The number of extreme values ​​in a texture sub-item refers to the total number of extreme points that appear in the texture sub-item after gradient calculation. These extreme points are usually associated with texture changes in the image, such as the edges of cracks or hot spots. The number of pixels in an image sub-queue refers to the total number of pixels contained in the sub-queue, which reflects the size of the sub-queue. The absolute standard deviation of the difference between adjacent extreme values ​​is used to measure the degree of fluctuation between extreme points. A larger standard deviation indicates more drastic texture changes and may contain more noise. The formula calculates the noise interference amplitude by multiplying the texture fluctuation coefficient (that is, the ratio of the number of extreme values ​​to the number of pixels) by the absolute standard deviation of the difference between adjacent extreme values. This calculation method comprehensively considers the complexity and degree of fluctuation of the texture and can more comprehensively reflect the noise level in the image.

[0052] Preferably, in practical applications, the calculation process of the noise interference amplitude can be further optimized. For example, when calculating the texture fluctuation coefficient, the number of extreme values ​​can be normalized to eliminate the influence of differences in the size of different image sub-queues. Furthermore, when calculating the absolute standard deviation of the difference between adjacent extreme values, more robust statistical methods such as median filtering or weighted averaging can be used to reduce the interference of outliers.

[0053] Furthermore, the weighting parameters in the formula can be adjusted based on the actual characteristics of PV module images to better balance the impact of texture complexity and fluctuation on the noise interference amplitude. These optimization measures can more accurately assess noise interference in images, thereby improving the accuracy and reliability of defect detection.

[0054] In some embodiments, in S3, the method for determining the noise interference amplitude of each group of image data queues includes: among all image data queues under the mth lighting condition of the same batch of components, intercepting the reference image subqueues of other groups of image data queues at the start and end time points of the jth image subqueue of the i-th group of image data queue; defining the average of the jth image subqueue of the i-th group of image data queue and all reference image subqueues as the standard image subqueue; using the SSIM algorithm to calculate the structural similarity between the jth image subqueue of the i-th group of image data queue and the corresponding standard image subqueue; in the i-th group of image data queue under the m-th lighting condition, calculating the sum of the noise interference amplitudes of all image subqueues, and multiplying the ratio of the noise interference amplitude of the j-th image subqueue to the sum by the structural similarity to define it as the noise interference amplitude of the i-th group of image data queue under the m-th lighting condition.

[0055] It should be noted that in the photovoltaic module defect detection method, the method for determining the noise interference amplitude of each image data queue includes intercepting a reference image subqueue, calculating structural similarity, and adjusting the noise interference amplitude based on the similarity. The core of this process is to evaluate the noise level of the current image subqueue by comparing it with other image data. Using structural similarity to adjust the noise interference amplitude, this method more accurately reflects the quality of the image data. This method can effectively reduce misjudgments caused by differences in lighting conditions or errors in image acquisition equipment, thereby improving the reliability of defect detection.

[0056] Specifically, the reference image subqueue refers to an image segment captured from other image data queues of the same batch of components at the same time point as the current image subqueue. These reference image subqueues are used as a reference to evaluate the noise level of the current image subqueue. The standard image subqueue is obtained by calculating the mean of the current image subqueue and all reference image subqueues. It represents the ideal state of the image at the current time point. Structural similarity (SSIM) is an algorithm that measures the structural similarity of images. By calculating the structural similarity between the current image subqueue and the standard image subqueue, the degree of noise interference in the current image subqueue can be quantified. Finally, by combining the noise interference amplitude of the current image subqueue with the structural similarity, an adjusted noise interference amplitude can be obtained. This value better reflects the quality of the image under actual lighting conditions.

[0057] Preferably, during implementation, the calculation of structural similarity can be optimized. For example, the weight parameters in the SSIM algorithm can be adjusted to better suit the characteristics of photovoltaic module images. When capturing the reference image sub-sequence, the capture range can be dynamically adjusted based on changes in lighting conditions to ensure that the reference image sub-sequence accurately reflects the image characteristics under the current lighting conditions.

[0058] Furthermore, when calculating the noise interference amplitude, normalization can be introduced to eliminate differences in image data under different lighting conditions. Through these optimization measures, the quality of image data can be more accurately assessed, thereby improving the accuracy and reliability of defect detection.

[0059] In some embodiments, in S3, the calculation formula for the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is: ;in, is the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue under the m-th illumination condition, is the number of image subqueues in the i-th group image data queue under the m-th illumination condition, It is the SSIM similarity between the j-th image sub-queue of the i-th group of image data queue under the m-th illumination condition and the standard image sub-queue.

[0060] It should be noted that the noise interference amplitude calculation method quantitatively assesses the noise level of an image data queue under specific lighting conditions. This method extracts a baseline image subqueue from another image data queue and compares it with the current image subqueue, measuring the similarity between the two using the structural similarity (SSIM) algorithm. Finally, the adjusted noise interference amplitude is calculated by combining the noise interference amplitude of the current image subqueue with the structural similarity. This method effectively reflects the differences in image quality under different lighting conditions, providing a more accurate image data foundation for subsequent defect detection.

[0061] Specifically, illumination conditions refer to the ambient lighting conditions during image acquisition, including factors such as illumination intensity, angle, and uniformity. Under the mth illumination condition, the noise interference amplitude of the jth image subqueue in the i-th image data queue is obtained by comparing it with a baseline image subqueue from another image data queue. The baseline image subqueue is an image segment captured from another image data queue at the same time as the current image subqueue and serves as a reference. The standard image subqueue is obtained by calculating the mean of the current image subqueue and all baseline image subqueues. It represents the ideal image state at the current time point. Structural similarity (SSIM) is an algorithm that measures image structural similarity. By calculating the structural similarity between the current image subqueue and the standard image subqueue, the degree of noise interference in the current image subqueue can be quantified. Finally, by combining the noise interference amplitude of the current image subqueue with the structural similarity, an adjusted noise interference amplitude is obtained, which better reflects the image quality under actual lighting conditions.

[0062] Preferably, during implementation, the calculation of structural similarity can be optimized. For example, the weight parameters in the SSIM algorithm can be adjusted to better suit the characteristics of photovoltaic module images. When capturing the reference image sub-sequence, the capture range can be dynamically adjusted based on changes in lighting conditions to ensure that the reference image sub-sequence accurately reflects the image characteristics under the current lighting conditions.

[0063] Furthermore, when calculating the noise interference amplitude, normalization can be introduced to eliminate differences in image data under different lighting conditions. Through these optimization measures, the quality of image data can be more accurately assessed, thereby improving the accuracy and reliability of defect detection.

[0064] In some embodiments, in S4, the method for determining the final noise interference amplitude of each group includes: among all image data queues under the mth lighting condition, using Hausdorff distance to calculate the difference between the brightness distribution items of the i-th group of image data queue and the brightness distribution items of other groups, and defining the normalized mean of the difference between the i-th group of image data queue and other groups as the deviation of the i-th group of image data queue; calculating the deviation difference of the image data queues with the same serial number under adjacent lighting conditions, and taking the absolute value mean as the abnormality coefficient of each group under the mth lighting condition; multiplying the negative correlation normalized value of the abnormality coefficient with the noise interference amplitude to obtain the final noise interference amplitude of the i-th group of image data queue under the mth lighting condition.

[0065] It should be noted that the final noise interference amplitude for each group is determined by quantitatively analyzing the differences in the brightness distribution of the image data queues and comprehensively evaluating the image data quality by combining the deviation and anomaly coefficient. The deviation measures the difference between the current image data queue and other image data queues, while the anomaly coefficient reflects the change in the deviation of the image data queue under adjacent lighting conditions. By combining the anomaly coefficient with the noise interference amplitude, the final noise interference amplitude can be obtained, thereby more accurately screening high-quality image data queues and providing a more reliable basis for defect detection.

[0066] Specifically, the brightness distribution item refers to the overall distribution of pixel brightness in the image, which reflects the light and dark variation characteristics of the image. The deviation is obtained by calculating the difference between the brightness distribution item of the current image data queue and the brightness distribution item of other image data queues, and is usually quantified using methods such as the Hausdorff distance. The Hausdorff distance is a method for measuring the similarity between two point sets, which can effectively reflect the structural differences between images. The anomaly coefficient is obtained by calculating the absolute mean of the deviation differences of image data queues with the same sequence number under adjacent lighting conditions. It reflects the changing trend of the deviation of image data queues under different lighting conditions. The final noise interference amplitude is obtained by multiplying the negative correlation normalized value of the anomaly coefficient by the noise interference amplitude. This method can effectively eliminate the impact of changes in lighting conditions on image quality assessment.

[0067] Preferably, during implementation, the calculation process for the deviation and anomaly coefficient can be optimized. For example, when calculating the deviation, a multi-scale Hausdorff distance can be used to account for differences in image structure at different scales. When calculating the anomaly coefficient, a time series analysis method can be introduced to more accurately capture the impact of changing lighting conditions on the image data queue.

[0068] Furthermore, when calculating the final noise interference amplitude, the weight of the negatively correlated normalized value of the anomaly coefficient can be adjusted according to actual needs to better balance the impact of deviation and noise interference amplitude on the final result. Through these optimization measures, the quality of image data can be more accurately assessed, thereby improving the accuracy and reliability of defect detection.

[0069] In some embodiments, in S4, the deviation calculation formula of the i-th group of image data queue under the m-th lighting condition is: ;in, is the deviation of the i-th group of image data queue under the m-th illumination condition, is the total number of groups, is the brightness distribution item of the i-th group of image data queue under the m-th lighting condition.

[0070] The calculation formula of the anomaly coefficient is: ;in, is the abnormal coefficient of the i-th group of image data queue under the m-th illumination condition, is the total number of lighting conditions.

[0071] The final noise interference amplitude calculation formula is: ;in, and is the mean and standard deviation of all abnormal coefficients.

[0072] It's important to note that the formulas for calculating deviation, anomaly coefficient, and final noise interference amplitude are key steps in quantifying the characteristic differences and noise levels of PV module image data queues under different lighting conditions. Deviation measures the difference in brightness distribution between the current image data queue and other image data queues. The anomaly coefficient reflects the change in deviation between image data queues under adjacent lighting conditions. The final noise interference amplitude combines the anomaly coefficient and the noise interference amplitude to assess image data quality. These formulas provide a quantitative basis for selecting high-quality image data queues, thereby ensuring the accuracy and reliability of defect detection.

[0073] Specifically, the deviation is calculated by calculating the difference between the brightness distribution of the current image data queue and the brightness distribution of other image data queues. The brightness distribution refers to the overall distribution of pixel brightness within an image, reflecting the image's brightness and shading characteristics. The Hausdorff distance is a method for measuring the similarity between two point sets and can effectively reflect structural differences between images. To calculate the deviation, the brightness distribution of all image data queues is compared pairwise using the Hausdorff distance, and the normalized mean is taken as the deviation. The anomaly coefficient is calculated by calculating the mean of the absolute value of the deviation differences for image data queues with the same sequence number under adjacent lighting conditions. It reflects the changing trend of the deviation of image data queues under different lighting conditions. The final noise interference amplitude is calculated by multiplying the negative correlation normalized value of the anomaly coefficient by the noise interference amplitude. This method effectively eliminates the impact of varying lighting conditions on image quality assessment. The negative correlation normalized value of the anomaly coefficient refers to the negative correlation value of the normalized anomaly coefficient, ensuring that the final noise interference amplitude is inversely proportional to image quality.

[0074] Preferably, during implementation, the calculation process for the deviation and anomaly coefficient can be optimized. For example, when calculating the deviation, a multi-scale Hausdorff distance can be used to account for differences in image structure at different scales. When calculating the anomaly coefficient, a time series analysis method can be introduced to more accurately capture the impact of changing lighting conditions on the image data queue.

[0075] Furthermore, when calculating the final noise interference amplitude, the weight of the negatively correlated normalized value of the anomaly coefficient can be adjusted according to actual needs to better balance the impact of deviation and noise interference amplitude on the final result. Through these optimization measures, the quality of image data can be more accurately assessed, thereby improving the accuracy and reliability of defect detection.

[0076] In some embodiments, in S5, the method for screening the high-quality image data queue includes: setting a threshold ,in and is the mean and standard deviation of all final noise interference amplitudes, is the preset coefficient; if the final noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is , it is determined to be a high-quality image data queue; the morphological closing operation is used to eliminate noise in the high-quality image data queue and output the defect detection result.

[0077] It should be noted that the method for selecting high-quality image data queues assesses image data quality based on the final noise interference amplitude and uses a threshold to distinguish high-quality image data queues from low-quality image data queues. The core of this method is to quantify the noise level of the image data and select image data with less noise interference, thereby improving the accuracy and reliability of defect detection. The final noise interference amplitude is calculated by combining the anomaly coefficient and the noise interference amplitude, providing a more comprehensive reflection of image data quality. By setting a threshold and selecting high-quality image data queues, a more accurate image foundation is provided for subsequent defect detection.

[0078] Specifically, the threshold is a key parameter used to distinguish the quality of image data, and it is usually set according to the statistical characteristics of the final noise interference amplitude. The final noise interference amplitude is obtained by multiplying the negatively correlated normalized value of the anomaly coefficient by the noise interference amplitude, reflecting the quality difference of the image data under different lighting conditions. The setting of the threshold can be based on the mean and standard deviation of the final noise interference amplitude. For example, the threshold is set to the mean plus several times the standard deviation. If the final noise interference amplitude of the image data queue is less than or equal to the threshold, it is judged to be a high-quality image data queue. In addition, morphological closing operation is an image processing method used to eliminate noise in the image. It smoothes small holes and small bumps in the image through dilation and erosion operations, thereby improving the clarity and coherence of the image. After screening out the high-quality image data queue, the morphological closing operation can be used to further optimize the image quality and provide clearer image data for defect detection.

[0079] Preferably, during the implementation process, the threshold setting can be optimized. For example, the preset coefficients in the threshold calculation formula can be adjusted according to the actual characteristics of the photovoltaic module image data to better adapt to different detection needs. The preset coefficients can be adjusted based on experience or experimental data to ensure that the screened image data queue can meet the accuracy requirements of defect detection. When performing morphological closing operations, the size and shape of the structural elements of the expansion and corrosion operations can be adjusted to adapt to different types of noise and defect characteristics. For example, for the detection of small cracks, smaller structural elements can be used to retain more detailed information; for larger hot spots, larger structural elements can be used to better eliminate noise. Through these optimization measures, high-quality image data queues can be screened more accurately, and more reliable image data support can be provided for subsequent defect detection.

[0080] The above-mentioned embodiments of the present invention have the following beneficial effects: the present invention can improve the accuracy and reliability of photovoltaic module defect detection, and can effectively overcome the limitations of single illumination condition detection through dynamic collection and analysis of image data under multiple illumination conditions. By adopting the decomposition method of brightness distribution terms and texture feature terms, effective information and noise interference in the image can be accurately identified. Combined with wavelet decomposition and Canny edge detection technology, extreme points in image features can be accurately extracted. By calculating the texture fluctuation coefficient and the difference between adjacent extreme values, the noise interference level of each image sub-queue can be quantitatively evaluated. Combined with the SSIM algorithm and Hausdorff distance calculation, the deviation of detection results under different illumination conditions can be dynamically calibrated. Finally, high-precision defect detection results can be output through morphological processing.

[0081] This method can improve the sensitivity of traditional detection methods to lighting conditions. By establishing a correlation model between noise interference amplitude and deviation, it can effectively suppress the interference of environmental factors on the detection results. Cross-validation using multiple sets of data from the same batch of components can enhance the robustness of the detection system, while normalization based on anomaly coefficients can adaptively adjust the detection threshold under different lighting conditions. Ultimately, through a high-quality image screening mechanism, the accuracy of the output results can be ensured. At the same time, this method is compatible with the detection requirements of photovoltaic modules of different specifications, improving detection efficiency while reducing false detection and missed detection rates, providing reliable technical support for the quality assessment of photovoltaic modules.

[0082] like Figure 2 As shown, some embodiments of a photovoltaic module defect detection system based on image processing include: an image acquisition device equipped with an optical sensor and a processing unit in communication with the image acquisition device, for acquiring and analyzing surface image data of photovoltaic modules, and transmitting defect detection results to a display terminal; the processing unit includes: an acquisition module 201, for obtaining multiple groups of surface image data queues of modules of the same batch of photovoltaic modules under different lighting conditions.

[0083] Decomposition module 202 is used to decompose each group of image data queues into brightness distribution items and texture feature items; based on the time points corresponding to all extreme values ​​in the brightness distribution items, each group of image data queues and texture feature items is divided into multiple image sub-queues and texture sub-items; based on the number of extreme values ​​in each texture sub-item and the difference between adjacent extreme values, the noise interference amplitude of each image sub-queue is calculated.

[0084] The calibration module 203 is configured to determine the noise interference amplitude of each group of image data queues based on the noise interference amplitude of each image sub-queue and the image data of the same area in the image data queues of all components under the same conditions.

[0085] Evaluation module 204 is used to calculate the deviation of each group of image data queues based on the difference in brightness distribution items of the image data queues of all components under the same conditions; and determine the final noise interference amplitude of each group of image data queues based on the difference in deviation between the noise interference amplitude of each group and the image data queues with the same sequence number under adjacent lighting conditions.

[0086] The output module 205 is used to screen the high-quality image data queues corresponding to each group based on the final noise interference amplitude of each group of image data queues; and use the image data in the high-quality image data queues as defect detection results.

[0087] It is understandable that the modules described in the photovoltaic module defect detection system based on image processing are similar to those in the reference Figure 1 The steps in the photovoltaic module defect detection method based on image processing correspond to each other. Therefore, the operations, features, and beneficial effects described above for the photovoltaic module defect detection method based on image processing are also applicable to the photovoltaic module defect detection system based on image processing and the modules included therein, and will not be repeated here.

[0088] Reference below Figure 3 , which shows a schematic structural diagram of an electronic device 300 suitable for implementing some embodiments of the present invention. The electronic devices in some embodiments of the present invention may include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), in-vehicle terminals (e.g., in-vehicle navigation terminals), and fixed terminals such as digital TVs and desktop computers. Figure 3 The terminal device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0089] like Figure 3 As shown, electronic device 300 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 301, which can perform various appropriate actions and processes based on programs stored in a read-only memory (ROM) 302 or programs loaded from a storage device 308 into a random access memory (RAM) 303. RAM 303 also stores various programs and data required for the operation of electronic device 300. Processing device 301, ROM 302, and RAM 303 are interconnected via a bus 304. An input / output (I / O) interface 305 is also connected to bus 304.

[0090] Typically, the following devices may be connected to the I / O interface 305: an input device 306 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 307 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 308 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 309. The communication device 309 may allow the electronic device 300 to communicate with other devices wirelessly or by wire to exchange data. Figure 3 The electronic device 300 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead. Figure 3 Each block shown in the figure may represent one device, or may represent multiple devices as needed.

[0091] Furthermore, the storage medium of the embodiment of the present application stores program instructions that can implement all the above methods, wherein the program instructions can be stored in the above storage medium in the form of a software product, including a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, or a terminal device such as a computer, server, mobile phone, or tablet.

[0092] The above descriptions merely illustrate some preferred embodiments of the present invention and the underlying technical principles. Those skilled in the art should understand that the scope of the invention encompassed by the embodiments of the present invention is not limited to technical solutions formed by specific combinations of the aforementioned technical features. It also encompasses other technical solutions formed by any combination of the aforementioned technical features or their equivalents, without departing from the aforementioned inventive concept. For example, a technical solution formed by replacing the aforementioned features with (but not limited to) technical features with similar functions disclosed in the embodiments of the present invention.

Claims

1. A photovoltaic module defect detection method based on image processing, characterized in that: include: The image acquisition device uses its optical sensor to collect image data of the photovoltaic module surface and transmits it to the processing unit for analysis. The analyzed defect data is then visualized on the display terminal to perform quality assessment of the photovoltaic module. Defect types include cell cracks, hidden cracks or hot spots. The analysis process includes: S1, obtaining multiple groups of surface image data queues of photovoltaic modules from the same batch under different lighting conditions; S2, decomposing each group of image data queues into brightness distribution items and texture feature items; based on the time points corresponding to all extreme values ​​in the brightness distribution items, dividing each group of image data queues and texture feature items into multiple image sub-queues and texture sub-items; calculating the noise interference amplitude of each image sub-queue based on the number of extreme values ​​and the difference between adjacent extreme values ​​in each texture sub-item; S3, determining the noise interference amplitude of each group of image data queues based on the noise interference amplitude of each image sub-queue and image data of the same area in image data queues of all modules under the same conditions; S4, calculating the deviation of each group of image data queues based on the difference in brightness distribution items of image data queues of all modules under the same conditions; determining the final noise interference amplitude of each group of image data queues based on the difference in the deviation of each group of noise interference amplitude and image data queues with the same sequence number under adjacent lighting conditions; S5, screening the corresponding high-quality image data queues based on the final noise interference amplitude of each group of image data queues; and using the image data in the high-quality image data queues as defect detection results.

2. The photovoltaic module defect detection method based on image processing according to claim 1, characterized in that: In S1 , the plurality of surface image data queues are queues formed by photovoltaic module surface images arranged in order of acquisition time.

3. The photovoltaic module defect detection method based on image processing according to claim 2, characterized in that: In S2, the method for calculating the noise interference amplitude of each image sub-queue in each group of image data queues includes: for the i-th group of image data queues of components from the same batch under arbitrary lighting conditions, the queue is decomposed into brightness distribution items, texture feature items and residual items using a wavelet decomposition algorithm; the extreme points of the brightness distribution items are extracted using a Canny edge detection algorithm; based on the time points corresponding to all extreme values ​​in the brightness distribution items, the i-th group of image data queues and texture feature items are divided into multiple image sub-queues and texture sub-items; the j-th texture sub-item of the i-th group of image data queues is subjected to the wavelet decomposition algorithm; the extreme points of the brightness distribution items are extracted using a Canny edge detection ... Calculate the row gradient and extract its extreme points; divide the number of extreme values ​​in the j-th texture sub-item of the i-th group of image data queue by the number of pixels of the j-th image sub-item of the i-th group of image data queue, and define it as the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queue; calculate the absolute value standard deviation of the difference between adjacent extreme values ​​in the j-th texture sub-item of the i-th group of image data queue, and multiply it by the texture fluctuation coefficient of the j-th image sub-item of the i-th group of image data queue to obtain the noise interference amplitude of the j-th image sub-item of the i-th group of image data queue.

4. The photovoltaic module defect detection method based on image processing according to claim 3, characterized in that: In S2, the calculation formula for the noise interference amplitude of the j-th image sub-queue of the i-th group of image data queue is: ;in, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue, is the number of extreme values ​​of the j-th texture sub-item in the i-th group of image data queue, is the number of pixels of the jth image sub-queue of the i-th group of image data queue, is the kth extreme value in the jth texture sub-item of the i-th group image data queue, is the mean of the absolute values ​​of the differences between adjacent extreme values.

5. The photovoltaic module defect detection method based on image processing according to claim 4, characterized in that: In S3, the method for determining the noise interference amplitude of each group of image data queues includes: among all image data queues under the mth illumination condition of the same batch of components, intercepting the reference image subqueues of other groups of image data queues at the start and end time points of the jth image subqueue of the ith group of image data queue; defining the average of the jth image subqueue of the ith group of image data queue and all reference image subqueues as the standard image subqueue; using the SSIM algorithm to calculate the structural similarity between the jth image subqueue of the ith group of image data queue and the corresponding standard image subqueue; in the ith group of image data queue under the mth illumination condition, calculating the sum of the noise interference amplitudes of all image subqueues, and multiplying the ratio of the noise interference amplitude of the jth image subqueue to the sum by the structural similarity to define the noise interference amplitude of the ith group of image data queue under the mth illumination condition.

6. The photovoltaic module defect detection method based on image processing according to claim 5, characterized in that: In S3, the calculation formula for the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is: ;in, is the noise interference amplitude of the i-th group of image data queue under the m-th illumination condition, is the noise interference amplitude of the jth image sub-queue of the i-th group of image data queue under the m-th illumination condition, is the number of image subqueues in the i-th group image data queue under the m-th illumination condition, It is the SSIM similarity between the j-th image sub-queue of the i-th group of image data queue under the m-th illumination condition and the standard image sub-queue.

7. The photovoltaic module defect detection method based on image processing according to claim 6, characterized in that: In S4, the method for determining the final noise interference amplitude of each group includes: using Hausdorff distance to calculate the difference between the brightness distribution item of the i-th group of image data queue and the brightness distribution items of other groups in all image data queues under the m-th lighting condition, and defining the normalized mean of the difference between the i-th group of image data queue and other groups as the deviation of the i-th group of image data queue; calculating the deviation difference of the image data queues with the same serial number under adjacent lighting conditions, and taking the absolute value mean as the abnormality coefficient of each group under the m-th lighting condition; multiplying the negative correlation normalized value of the abnormality coefficient by the noise interference amplitude to obtain the final noise interference amplitude of the i-th group of image data queue under the m-th lighting condition.

8. The photovoltaic module defect detection method based on image processing according to claim 7, characterized in that: In S4, the deviation calculation formula of the i-th group of image data queue under the m-th illumination condition is: ;in, is the deviation of the i-th group of image data queue under the m-th illumination condition, is the total number of groups, is the brightness distribution item of the i-th group of image data queue under the m-th illumination condition; the calculation formula of the abnormal coefficient is: ;in, is the abnormal coefficient of the i-th group of image data queue under the m-th illumination condition, is the total number of illumination conditions; the final noise interference amplitude calculation formula is: ;in, and is the mean and standard deviation of all abnormal coefficients.

9. The photovoltaic module defect detection method based on image processing according to claim 8, characterized in that: In S5, the method for screening the high-quality image data queue includes: setting a threshold ,in and is the mean and standard deviation of all final noise interference amplitudes, is the preset coefficient; if the final noise interference amplitude of the i-th group of image data queue under the m-th illumination condition is , it is determined to be a high-quality image data queue; the morphological closing operation is used to eliminate noise in the high-quality image data queue and output the defect detection result.

Images