Photovoltaic panel defect detection method and device, electronic equipment and program product

Abstract

The invention provides a photovoltaic panel defect detection method and device, electronic equipment and a program product, and the method comprises the steps: carrying out the detection of a significant defect on a photovoltaic panel PL image through an image processing algorithm which is higher in efficiency and is based on gray analysis; and further detecting fine defects of which the defect types are not detected by the method through a deep learning algorithm with relatively high detection precision. Through preliminary detection of the first defect detection algorithm, the amount of data required to be detected by deep learning is reduced, the efficiency of the whole defect detection process is improved, further through mutual combination of different defect detection algorithms, detection of different types of defects can be realized, the comprehensiveness of defect detection is ensured, the accuracy of defect detection is improved, and the detection efficiency is improved. The efficiency of photovoltaic panel defect detection is improved, and meanwhile the detection precision is guaranteed.

Description

Technical Field

[0001] The present application relates to the field of image analysis technology, and in particular to a method, device, electronic equipment and program product for detecting defects in photovoltaic panels. Background Art

[0002] With the increasing demand for energy and the improvement of environmental awareness, solar photovoltaic panels (hereinafter referred to as photovoltaic panels) have been widely used. Photovoltaic panels are the core components of photovoltaic power generation, but defects are inevitable during their installation and operation, affecting their power generation efficiency. Therefore, defect detection of photovoltaic panels is an indispensable part of the solar photovoltaic industry. At present, defect detection is performed by combining machine vision and deep learning algorithms to detect electroluminescence (EL) images of photovoltaic panels.

[0003] However, since the deep learning-based algorithm includes multiple layers of neural networks, each layer needs to perform a large number of operations in the detection task, which results in a large amount of computation. Therefore, although the deep learning algorithm has a high defect detection accuracy, it consumes a lot of computing resources.

[0004] Therefore, how to improve the efficiency of photovoltaic panel defect detection while ensuring the accuracy of defect detection is a technical problem that needs to be solved urgently by technical personnel in this field. Summary of the invention

[0005] In view of this, the embodiments of the present application provide a method, device, electronic device and program product for detecting defects in photovoltaic panels, which realize defect detection by combining defect detection algorithms with different detection efficiency and accuracy, thereby improving the efficiency of the overall defect detection process while ensuring the accuracy of detection.

[0006] In a first aspect, an embodiment of the present application provides a photovoltaic panel defect detection method, the method comprising: acquiring photoluminescence (PL) images of multiple photovoltaic cells in a photovoltaic panel; comparing the PL images of the multiple photovoltaic cells with image templates of normal photovoltaic cells in turn, and determining the images of photovoltaic cells with a similarity lower than a first threshold as the PL images of target photovoltaic cells, the PL images of the target photovoltaic cells being the PL images of abnormal photovoltaic cells having abnormal areas in the photovoltaic cells; performing defect detection on the PL images of the target photovoltaic cells based on a grayscale image analysis algorithm to determine the defect type of the target photovoltaic cells; acquiring electroluminescence (EL) images of target photovoltaic cells of undetermined defect types among the multiple photovoltaic cells; performing secondary defect detection on the EL images of target photovoltaic cells of undetermined defect types among the multiple photovoltaic cells through a deep learning algorithm to determine the defect types of the target photovoltaic cells of undetermined defect types.

[0007] In one possible implementation, PL images of multiple photovoltaic units in a photovoltaic panel are obtained, including: irradiating the photovoltaic panel with a laser beam emitted by a laser carried by a drone, and obtaining the PL image of the photovoltaic panel through a camera carried by the drone; and pre-segmenting the PL image of the photovoltaic panel to obtain PL images of multiple photovoltaic units.

[0008] In one possible implementation, secondary defect detection is performed on the electroluminescent EL image of a target photovoltaic unit of undetermined defect type among multiple photovoltaic units through a deep learning algorithm to determine the defect type of the target photovoltaic unit of undetermined defect type, including: inputting the EL image of the target photovoltaic unit of undetermined defect type into a trained neural network model; performing feature extraction and classification on the EL image of the target photovoltaic unit of undetermined defect type through the trained neural network model to determine the defect type of the EL image of the target photovoltaic unit of undetermined defect type.

[0009] In one possible implementation, obtaining an electroluminescent EL image of a target photovoltaic cell of undetermined defect type includes: irradiating a photovoltaic cell adjacent to the target photovoltaic cell of undetermined defect type with a laser beam emitted by a laser carried by a drone to generate current, so that the target photovoltaic cell receives the current generated by the adjacent photovoltaic cell and emits electroluminescence; and obtaining an EL image of the target photovoltaic cell of undetermined defect type by a camera carried by the drone.

[0010] In one possible implementation, defect detection is performed on a PL image of a target photovoltaic unit based on a grayscale value image analysis algorithm to determine the defect type of the target photovoltaic unit, including: preprocessing the PL image of the target photovoltaic unit to obtain a grayscale image; performing grayscale deviation analysis on the grayscale image to determine a target area of ​​the image of the target photovoltaic unit; and performing defect detection on the target area using an image processing algorithm based on grayscale value analysis to determine the defect type of the target area.

[0011] In a possible implementation, an image processing algorithm based on gray value analysis is used to perform defect detection on a target area to determine the defect type of the target area, including: performing connected domain processing on the target area to obtain a target connected domain; determining a gray distribution characteristic coefficient of the pixel points in the target connected domain according to a gray deviation value of each pixel point in the target connected domain and the distribution of the pixel points in the target connected domain; when the gray distribution characteristic coefficient is greater than a preset threshold, determining the defect type of the target area to be a crack, otherwise determining the defect type of the target area to be a scratch, wherein the gray deviation value of each pixel point in the target connected domain and the distribution of the pixel points in the target connected domain are used to determine the defect type of the target area. The distribution of pixel points in a connected domain determines the grayscale distribution characteristic coefficient of pixel points in a target connected domain, including: calculating the absolute value of the sum of the row deviation value and the column deviation value of each pixel point in the target connected domain to obtain a grayscale deviation coefficient, wherein the row deviation is the deviation between the pixel point and the grayscale mean of all pixel points in its row, and the column deviation is the deviation between the pixel point and the grayscale mean of all pixel points in its column; counting the number of pixel points in each row and the number of pixel points in each column in the target connected domain and performing normalization processing to obtain an area range coefficient; calculating the product of the grayscale deviation coefficient and the area range coefficient and performing normalization processing to obtain a grayscale distribution characteristic coefficient;.

[0012] In one possible implementation, after determining the defect type of the target photovoltaic panel battery cell, it also includes: marking the defects of the target photovoltaic cell according to the defect type included in the image of the target photovoltaic cell; presenting a topological image of the photovoltaic panel array on a user interface, the photovoltaic panel array including photovoltaic panels; after receiving a user click operation on the photovoltaic panel, presenting images of multiple photovoltaic cells and corresponding detection information on the user interface, the detection information including position information, marking information and defect level information corresponding to the photovoltaic cells.

[0013] In one possible implementation, after determining the defect type of the target photovoltaic unit image, it also includes: marking the defects according to the defect type included in the target photovoltaic unit image; acquiring images of all detected photovoltaic units to form a photovoltaic panel array topology image; and displaying detection information of the photovoltaic panel array topology image and the image of the corresponding photovoltaic unit on a user interface, the detection information including the position information, marking information and defect level information corresponding to the photovoltaic unit.

[0014] In a second aspect, an embodiment of the present application provides a photovoltaic panel defect detection device, which includes: an image acquisition module for acquiring photoluminescence (PL) images of multiple photovoltaic cells in a photovoltaic panel; a target photovoltaic cell determination module for comparing the PL images of multiple photovoltaic cells with image templates of normal photovoltaic cells in turn, and determining the images of photovoltaic cells with a similarity lower than a first threshold as the PL images of target photovoltaic cells, and the PL images of target photovoltaic cells are PL images of abnormal photovoltaic cells with abnormal areas in the photovoltaic cells; a defect detection module for performing defect detection on the PL images of target photovoltaic cells based on a grayscale image analysis algorithm to determine the defect type of the target photovoltaic cells; the image acquisition module is also used to acquire electroluminescence (EL) images of target photovoltaic cells of undetermined defect types among multiple photovoltaic cells; the defect detection module is also used to perform secondary defect detection on the EL images of target photovoltaic cells of undetermined defect types among multiple photovoltaic cells through a deep learning algorithm to determine the defect types of target photovoltaic cells of undetermined defect types.

[0015] In a third aspect, an embodiment of the present application provides an electronic device comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the photovoltaic panel defect detection method described in the first aspect.

[0016] In a fourth aspect, an embodiment of the present application provides a computer program product, including a computer program / instruction, characterized in that when the computer program / instruction processor is executed, it is used to implement the photovoltaic panel defect detection method described in the first aspect.

[0017] The embodiments of the present application provide a method, device, electronic device and program product for detecting defects in photovoltaic panels. The method detects significant defects on the PL image of photovoltaic panels through an image processing algorithm based on grayscale analysis with high efficiency, and further detects subtle defects of defect types that are not detected by the above method through a deep learning algorithm with high detection accuracy. The preliminary detection of the first defect detection algorithm reduces the amount of data required for deep learning algorithm detection, and improves the efficiency of the overall defect detection process. In addition, through the combination of different algorithms, the detection of different types of defects can be achieved, ensuring the comprehensiveness of defect detection and improving the accuracy of defect detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] The accompanying drawings are used to provide a further understanding of the present disclosure and constitute a part of the specification. Together with the embodiments of the present disclosure, they are used to explain the present disclosure and do not constitute a limitation of the present disclosure. The above and other features and advantages will become more apparent to those skilled in the art by describing detailed example embodiments with reference to the accompanying drawings. In the accompanying drawings:

[0019] Figure 1 This is an application scenario diagram of a photovoltaic panel defect detection system provided in some embodiments of the present application.

[0020] Figure 2 It is an exemplary flow chart of a photovoltaic panel defect detection method provided in some embodiments of the present application.

[0021] Figure 3 This is an exemplary flow chart of a method for detecting photovoltaic panel defects using an image processing algorithm based on grayscale value analysis provided in some embodiments of the present application.

[0022] Figure 4 This is an exemplary flow chart for determining the grayscale distribution coefficient provided by some embodiments of the present application.

[0023] Figure 5 It is a schematic diagram of a user interface for presenting defect-related information provided by some embodiments of the present application.

[0024] Figure 6 It is a schematic diagram of the electroluminescence structure provided in some embodiments of the present application.

[0025] Figure 7 This is a system module diagram of a photovoltaic panel defect detection device provided in some embodiments of the present application.

[0026] Figure 8 It is a schematic diagram of the structure of an electronic device provided in some embodiments of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0028] Application Overview

[0029] In order to improve the accuracy of photovoltaic panel defect detection, in related technologies, electroluminescence (EL) images are generally obtained by electroluminescence, and then defects (especially cracks) are detected on the electroluminescence images through deep learning algorithms.

[0030] Electroluminescence images are obtained by applying voltage to a photovoltaic panel to make it emit light. Inside the photovoltaic panel, under a forward bias voltage, minority carriers are injected into the pn junction barrier region and the diffusion region, and these minority carriers continuously recombine with majority carriers to emit light. The brightness of the EL image is proportional to the minority carrier diffusion length and current density of the photovoltaic panel. When there are defects (for example, cracks) inside the photovoltaic panel, there will be obvious differences in the lifetime distribution of its minority carriers, resulting in differences in brightness in the image display. By analyzing the EL image, defects inside the photovoltaic panel can be found in a timely and clear manner. For example, a camera can be used to capture the infrared or near-infrared light emitted by the photovoltaic panel under current excitation, and the test needs to be carried out in a darkroom environment to avoid interference from external light sources.

[0031] That is to say, the EL detection process requires an external power supply (such as a special constant current power supply or an EL tester) to provide a stable current, and needs to be performed at night or in a dark field. The entire detection process is cumbersome, and is easily limited by area and site, and the cost is high.

[0032] In addition, although the accuracy of defect detection based on deep learning algorithms is relatively high, there are many photovoltaic panels in photovoltaic power stations, and each photovoltaic panel is composed of dozens of photovoltaic panel units. If deep learning algorithms are used for defect detection, it will not only consume huge computing resources, but also consume a lot of time, so defects cannot be discovered in a timely and efficient manner.

[0033] In addition, since applying reverse voltage to the photovoltaic panel will cause a certain degree of damage to the photovoltaic panel, EL detection may have an adverse effect on the power generation performance of the photovoltaic panel.

[0034] Photoluminescence (PL) is a process in which a substance re-radiates photons (or electromagnetic waves) after absorbing photons (or electromagnetic waves). The image formed on the photovoltaic panel is called a PL image. PL images can be generated by external lighting or laser irradiation, and are collected and converted into analyzable data by detectors such as photomultiplier tubes (PMTs), CCD detectors, or spectrometers. Defect detection based on photoluminescence images is a non-destructive analysis method that does not cause damage to photovoltaic panels. However, this defect detection method is generally performed during the day and is greatly affected by ambient light, especially for the detection of fine cracks, the accuracy is low.

[0035] In response to the above problems, the present application creatively proposes a method for classifying and detecting defects in photovoltaic panels, which realizes the classification detection of defects in photovoltaic panels by combining defect detection algorithms with different detection accuracies, thereby improving the efficiency of defect detection while ensuring detection accuracy.

[0036] Various non-limiting embodiments of the present application will be described in detail below with reference to the accompanying drawings.

[0037] Application scenario of exemplary photovoltaic panel defect detection method

[0038] Figure 1 This is an application scenario diagram of a photovoltaic panel defect detection system provided in some embodiments of the present application.

[0039] like Figure 1 As shown, an embodiment of the present application provides a photovoltaic panel defect detection system 100. The photovoltaic panel defect detection system 100 can be an interactive system. Among them, the photovoltaic panel defect detection system 100 may include a drone 110, a photovoltaic panel array 120, a centralized control system 130, a user 140, and a user terminal 150, wherein the photovoltaic panel array 120 may include a plurality of photovoltaic panels 121, and each photovoltaic panel 121 may include a plurality of photovoltaic units. It should be noted that the present application does not impose specific restrictions on the number of photovoltaic panels in the photovoltaic panel array and the number of photovoltaic units in the photovoltaic panel.

[0040] exist Figure 1 In the photovoltaic panel defect detection system 100 shown, the user 140 can view the defects of the current photovoltaic panel array through the user terminal 150. In the process of viewing the defects of the photovoltaic panel, the user 140 can interact with the centralized control system 130 through the user terminal 150, so that the user terminal 150 presents the current state of the photovoltaic panel array, including defective photovoltaic units and corresponding information on the defect type. Among them, based on the interaction with the user 140, the centralized control system 130 can first capture an image of the photovoltaic panel array 120 through the camera carried by the drone 110, and then determine the defect information of the photovoltaic panel array in the captured image through a defect detection algorithm, and present the defect information of the photovoltaic panel array through the user terminal 150. For more information about the centralized control system 130 determining photovoltaic panel defect information, please refer to Figure 2 and its related description.

[0041] In some embodiments, the centralized control system 130 may include a processor 131, a memory 132, and a network 133. Among them, the processor 131 can process data and / or information obtained from other devices or system components. The processor can execute program instructions based on these data, information and / or processing results to perform one or more functions described in this application. In some embodiments, the processor 131 may include one or more sub-processing devices (for example, a single-core processing device or a multi-core multi-core processing device). As an example only, the processor 131 may include a central processing unit (CPU), a graphics processing unit (GPU), a physical processing unit (PPU), a digital signal processor (DSP), etc. or any combination thereof.

[0042] The memory 132 may be used to store data and / or instructions. The memory 132 may include one or more storage components, each of which may be an independent device or a part of another device. In some embodiments, the memory 132 may include a random access memory (RAM), a read-only memory (ROM), a mass storage device, or the like, or any combination thereof. Exemplarily, the mass storage device may include a magnetic disk, an optical disk, a solid-state disk, or the like.

[0043] In some embodiments, the memory 132 may be configured to store a computer program related to the photovoltaic panel defect detection method shown in the present application. When the computer program is executed (such as called by the processor 131), the photovoltaic panel defect detection method shown in the embodiment of the present application may be implemented.

[0044] The network 133 can connect the components of the system and / or connect the system with external resources. The network 133 allows the components to communicate with each other and with other parts outside the system, and promotes the exchange of data and / or information. In some embodiments, the network 133 can be any one or more of a wired network or a wireless network. For example, the network 133 can include a cable network, an optical fiber network, a telecommunications network, the Internet, a local area network (LAN), a wide area network (WAN), a wireless local area network (WLAN), a metropolitan area network (MAN), etc. or any combination thereof. The network connection between the various parts can be in one of the above-mentioned ways, or in multiple ways.

[0045] In the present application, the user 140 may be an operator of the user terminal 150. Generally, the user 140 may be a maintenance person or a management person of a photovoltaic power station.

[0046] The user terminal 150 refers to one or more terminal devices or software used by the user. In some embodiments, the user terminal 150 can be one or any combination of other devices with input and / or output functions such as a mobile device, a tablet computer, a laptop computer, a desktop computer, etc.

[0047] Exemplary Photovoltaic Panel Cell Defect Detection Method

[0048] Figure 2 is an exemplary flow chart of a method for detecting defects in photovoltaic panels provided in some embodiments of the present application.

[0049] like Figure 2 As shown, the process may include the following steps:

[0050] S210, acquiring photoluminescence (PL) images of a plurality of photovoltaic units in a photovoltaic panel.

[0051] Specifically, a laser light source can be used to emit a laser beam to a photovoltaic panel so that the photovoltaic panel generates a PL image, and an image acquisition device such as a camera (e.g., an InGaAs camera) can be used to capture the PL image (e.g., a fluorescent image) of the photovoltaic panel. The environment of a photovoltaic power station is generally poor, and the distribution area of ​​photovoltaic panels is very wide. Therefore, a drone can be used to carry a camera to capture the PL image of each photovoltaic panel, which can greatly save labor costs.

[0052] In some embodiments, a drone carrying a camera can be used to regularly capture photoluminescence images of the entire photovoltaic panel in a photovoltaic power station, and the captured images can be sent to a centralized control system for defect detection.

[0053] In some embodiments, the photovoltaic panel may be divided into a plurality of regions according to the photovoltaic units contained therein, and the PL image may be segmented according to the regions to obtain PL images of the plurality of photovoltaic units.

[0054] Splitting the PL image of a photovoltaic panel into PL images of multiple photovoltaic units is helpful in determining which photovoltaic units have defects to facilitate subsequent defect and performance analysis. On the other hand, it can also classify and detect photovoltaic units, thereby reducing the number of photovoltaic units for defect detection using deep learning algorithms.

[0055] S220 , comparing the PL images of the plurality of photovoltaic cells with the image template of a normal photovoltaic cell in sequence, and determining the image of the photovoltaic cell with a similarity lower than a first threshold as the PL image of the target photovoltaic cell.

[0056] The PL image of the target photovoltaic cell may be an image of an abnormal photovoltaic cell in which an abnormal region may exist among the photovoltaic cells.

[0057] In some embodiments, a PL image of a normal photovoltaic cell can be taken before detecting defects, and the PL image of the normal photovoltaic cell can be used as a template. The acquired PL images of the photovoltaic cell can be compared with the template in turn, and the similarity between the image of the photovoltaic cell and the template can be calculated. The obtained similarity can be compared with a first threshold value, and when the similarity is lower than the first threshold value, it is determined that the image of the photovoltaic cell is the image of the target photovoltaic cell, that is, the target photovoltaic cell may have defects.

[0058] In some embodiments, the first threshold may be set according to an empirical value, and illustratively, may be 85%.

[0059] Generally speaking, only a few photovoltaic units on each photovoltaic panel will fail. Therefore, after screening by the above method, the number of photovoltaic units that need to be analyzed using the grayscale value analysis algorithm and the deep learning algorithm is greatly reduced. The calculation amount of the above similarity comparison method is lower than that of the grayscale value analysis algorithm and the deep learning algorithm, and the calculation speed is faster. Therefore, the use of the similarity comparison algorithm can greatly reduce the calculation amount of the grayscale value analysis algorithm and the deep learning algorithm for defect detection in the entire photovoltaic power station, and improve the efficiency of defect detection.

[0060] S230, performing defect detection on the PL image of the target photovoltaic unit based on a gray value image analysis algorithm to determine the defect type of the target photovoltaic unit.

[0061] In some embodiments, the PL image of the target photovoltaic unit may be preprocessed first, and the preprocessing operation may be graying the image of the target photovoltaic unit to obtain the gray value of the image of the target photovoltaic unit. In order to improve the contrast of the image, the image of the target photovoltaic unit may be further subjected to histogram equalization processing to provide a better basis for subsequent image analysis. In addition, in the case where the above-mentioned image is a fluorescent image, the fluorescent image may be converted into a grayscale image first.

[0062] There is a certain difference between the grayscale value of the defective part in the preprocessed photovoltaic unit image and the grayscale value of the normal photovoltaic panel area. For example, in the PL image, the brightness of the defective area is greater than the brightness of the normal area, that is, the grayscale value of the defective area is greater than the grayscale value of the normal area. Therefore, the image can be analyzed for deviation based on the grayscale value of the pixel point in the image of the target photovoltaic unit to locate the target area (abnormal area with defects) in the image.

[0063] In some embodiments, in order to reduce the influence of the texture contained in the photovoltaic panel itself on the detection process, improve the reliability of the deviation analysis, and ensure the accuracy of the analysis results, the deviation analysis can be performed on the rows and columns where the pixels are located. For example, the row deviation value and column deviation value of each pixel in the image of the target photovoltaic unit can be calculated. Specifically, the absolute value of the difference between the mean of the grayscale values ​​of all pixels in the row where the pixel to be analyzed is located and the grayscale value of the pixel to be analyzed can be used as the row deviation value of the corresponding pixel; accordingly, the absolute value of the difference between the mean of the grayscale values ​​of all pixels in the column where the pixel to be analyzed is located and the grayscale value of the pixel to be analyzed is calculated as the column deviation value, and the row deviation value and column deviation value of the pixel to be analyzed are compared with the second threshold value respectively, and the pixel whose row deviation value and column deviation value are both greater than the second threshold value is used as the target pixel, and the area composed of the target pixel is used as the target area, which can be an abnormal area where defects may exist in the image of the target photovoltaic unit. After locating the target area, the target area can be directly focused on, and subsequent defect detection can be performed directly on the target area to reduce the amount of data detected and improve the efficiency of the subsequent defect detection process.

[0064] In some embodiments, the second threshold value may be set based on an empirical value, for example, 0.8; a small number of samples may be selected for comparison to determine the second threshold value; the second threshold value may also be adjusted based on actual conditions, which is not specifically limited here.

[0065] Image processing algorithms based on grayscale value analysis can be used to detect significant defects in target photovoltaic cell images.

[0066] Taking into account that in actual use, the probability of cracks and scratches on photovoltaic panels is relatively high, and these two defects account for a relatively large proportion and are more significant in the entire image compared to other defects, for example, an image processing algorithm based on grayscale value analysis can be used to detect and identify crack and scratch defects in the image of the target photovoltaic unit.

[0067] In some embodiments, an image processing algorithm based on grayscale value analysis may be an algorithm that implements defect detection by analyzing the distribution and changes of the grayscale values ​​of pixels in the image area to be detected. Exemplarily, the grayscale values ​​of the pixels in the target area may be obtained, the grayscale deviation values ​​(column deviation values ​​and row deviation values) of the pixels may be calculated, and the distribution of the pixels in the target area may be statistically analyzed. The grayscale distribution characteristic coefficient of the target area may be calculated based on the deviation values ​​of the pixels in the target area and the distribution of the pixels in the target area. The grayscale distribution characteristic coefficient may be used to characterize the regional characteristic information of the target area (the distribution and changes of the grayscale values). The defect type may be further determined based on the obtained grayscale distribution characteristic coefficient of the target area. The specific process of calculating the grayscale distribution characteristic coefficient and the process of determining the defect type based on the grayscale distribution characteristic coefficient may be referred to. Figure 3-Figure 5 And related descriptions will not be repeated here.

[0068] Since the image processing algorithm based on gray value analysis requires fewer computing resources than the deep learning algorithm, the time consumed in the defect detection process is relatively less, that is, the time spent in the defect detection process can be reduced. At the same time, the image processing algorithm based on gray value analysis can realize the detection of most significant defects, which can greatly reduce the amount of data detected by the subsequent deep learning algorithm, reduce the computing resources required in the overall defect detection process, and improve the efficiency of the overall defect detection process.

[0069] In other embodiments, the image processing algorithm based on gray value analysis may also be a binarization method (OSTU), a local binary pattern (LBP) algorithm, an edge detection algorithm, etc., which are not specifically limited here.

[0070] S240 , acquiring an electroluminescent EL image of a target photovoltaic cell of an undetermined defect type among the plurality of photovoltaic cells.

[0071] In some embodiments, for the target photovoltaic unit whose defect type is not determined by the aforementioned S230, an EL image of the corresponding target photovoltaic unit can also be obtained, providing a good detection basis for subsequent defect detection based on a deep learning algorithm, which is conducive to improving the accuracy of detection.

[0072] Specifically, Figure 6As shown, the laser beam generated by the laser carried by the drone 110 can illuminate the photovoltaic cell 1211 adjacent to the target photovoltaic cell 1212 of undetermined defect type, causing excess carriers and radiation recombination in the irradiated area of ​​the photovoltaic cell 1212 to generate a driving force of a lateral current, so that the current flows to the photovoltaic cell 1212, and an electroluminescent effect is generated on the photovoltaic cell 1212, thereby generating an EL image. At the same time, the EL image of the photovoltaic cell 1212 is captured by the phase light carried by the drone 110.

[0073] The above-mentioned contactless electroluminescence method is not limited by contact electroluminescence, and no additional power source is required to apply voltage at both ends of the photovoltaic panel, which reduces costs and reduces the adverse effects of powering the photovoltaic panel. For example, drones can be used to carry cameras and laser equipment for illumination and shooting at night, so that electroluminescent images can be obtained without the need for manpower to operate on site or the need to set up special dark field equipment.

[0074] S250, performing secondary defect detection on the EL image of the target photovoltaic unit of undetermined defect type among the multiple photovoltaic units by using a deep learning algorithm to determine the defect type of the target photovoltaic unit of undetermined defect type.

[0075] Taking into account the diversity of defect types of photovoltaic panels, in addition to the more obvious cracks and scratches, there may also be other small and hidden defects with finer textures such as hidden cracks, broken grids, micro cracks (also called invisible cracks). In order to avoid missed detection and false detection of defects, this application further uses a deep learning algorithm with higher detection accuracy to detect hidden and tiny defects, thereby ensuring the comprehensiveness and accuracy of defect detection.

[0076] Since the accuracy of detecting small and hidden defects by detecting PL images is poor, the present application can improve the accuracy of detecting small and hidden defects by using a deep learning algorithm to detect the EL images of the remaining target photovoltaic units.

[0077] Since the deep learning algorithm performs secondary detection mainly for small and hidden defects of the defect type that were not detected in the initial detection, when designing a neural network model for secondary defect detection, a multi-scale feature fusion module can be used for feature extraction. The features are processed by multi-scale feature fusion, and the detail information in the low-level semantic information and the abstract features in the high-level semantic information are fused together, that is, the features in different levels are aggregated together, and the features of different receptive fields are fused to improve the expression ability of the model and effectively improve the accuracy of the model in identifying subtle and hidden defects; feature enhancement modules can also be used to extract features, and the target feature expression can be enhanced through the feature enhancement module to suppress redundant information to effectively capture subtle texture information related to defects. This application does not specifically limit the type of feature extraction module in the neural network model.

[0078] In some embodiments, the EL image of the target photovoltaic panel unit to be inspected may be input into a trained neural network model, and defect detection of the EL image of the target photovoltaic panel unit to be inspected may be implemented through the trained neural network model.

[0079] After determining the defect type of the target photovoltaic panel, the defect can be marked according to the defect type. For example, if the defect type of the photovoltaic panel is detected as a significant crack by the image processing method, it can be marked as 1; if the defect type is a scratch, it can be marked as 2; if the defect type is a hidden crack detected by the deep learning algorithm, it can be marked as 3, etc. Further, according to the different types of defects and the number of defects in the corresponding target photovoltaic unit image, the defect level of the target photovoltaic panel is determined, and the staff can determine whether the target photovoltaic unit needs to be maintained according to the defect level corresponding to the target photovoltaic panel. For example, the defect level can be set to three levels: high risk level, medium risk level and low risk level. Photovoltaic units with high-level defects need to be replaced, photovoltaic units with medium risk level defects need to reduce the operating time, and photovoltaic units with low risk level defects need to remind maintenance personnel to pay attention to changes in their power generation parameters. For example, the adverse effect of explicit cracks on the power generation performance of photovoltaic panels is greater than the adverse effect of scratches on the power generation performance of photovoltaic panels, and the adverse effect of scratches on the power generation performance of photovoltaic panels is greater than the adverse effect of hidden cracks on the power generation performance of photovoltaic panels. Therefore, when determining the defect level of a photovoltaic panel, different weights can be set for photovoltaic cells with different types of defects in a photovoltaic panel, and their levels can be determined according to the number and weight of photovoltaic cells with different types of defects. For example, the weight of a photovoltaic cell with an explicit crack is greater than the weight of a photovoltaic cell with a scratch, and the weight of a photovoltaic cell with a scratch is greater than the weight of a photovoltaic cell with an invisible crack. The number of photovoltaic cells with different types of defects is weighted to obtain the defect score of the photovoltaic panel. If the defect score of the photovoltaic panel is in different threshold ranges, it belongs to different levels of defects. For example, in a photovoltaic panel, the number of photovoltaic cells with explicit cracks is 4, and its weight is 0.5, the number of photovoltaic cells with scratches is 6, and its weight is 0.3, and the number of photovoltaic cells with invisible cracks is 10, and its weight is 0.2. The defect score of the photovoltaic panel is 4*0.5+6*0.3+10*0.2=5.8. If it is greater than 5 points, it is a high level, and the system will remind the maintenance personnel to replace the photovoltaic panel. Through the above method, maintenance personnel can be reminded to pay attention or take measures in time according to the risk level of the defects, which can prevent risks before they occur and carry out appropriate treatment for different defects, thereby improving the power generation efficiency and safety of the photovoltaic power station.

[0080] In other embodiments, each column or row of photovoltaic cells of a photovoltaic panel is connected in series. If a certain photovoltaic cell has an obvious crack, the photovoltaic cells in the column or row where the photovoltaic cell is located cannot work normally. In this case, all photovoltaic cells in the column or row will be marked as photovoltaic cells with obvious cracks. In addition, in this case, the defect level can also be determined according to the percentage or number of columns or rows of photovoltaic cells with obvious cracks in the photovoltaic panel. For example, if one row of photovoltaic cells in 10 rows of photovoltaic cells of a photovoltaic panel has obvious cracks, it is a low-level defect, if two rows of photovoltaic cells have obvious cracks, it is a medium-level defect, and if three rows of photovoltaic cells have obvious defects, it is a high-level defect.

[0081] In other embodiments, when marking the defect type, it can also be marked according to the degree of impact of the defect on the photovoltaic panel, for example, a crack defect has a greater impact and is marked as 3, a scratch defect is marked as 2, etc. The specific marking rules can be made by the manufacturer of the photovoltaic panel. The defect level can be determined based on the marking information and the number of defects. When the defect level reaches a certain level, the system automatically issues a corresponding alarm reminder.

[0082] In some embodiments, images of all detected photovoltaic cells can be obtained to form a photovoltaic panel array topology image, and the photovoltaic cell array topology image and detection information of the corresponding photovoltaic panel cell image can be displayed on the user interface. Exemplarily, the detection information can include geographic location information corresponding to the photovoltaic cell cell, defect type marking information and / or defect level information.

[0083] See also Figure 6 On the overall user interface of the photovoltaic power station, multiple sites are connected to the transmission bus of the photovoltaic power station and the number of defects of each site is counted. For example, site 1 currently has 2 defects, site 2 has no defects, and site 10 has 3 defects.

[0084] When the user clicks the area corresponding to site 1 on the user interface, the user interface switches to the user interface of site 1, on which multiple photovoltaic panel arrays and corresponding defect numbers are connected to the branch line of site 1. For example, the photovoltaic panel array of group 1 has 1 defect, the photovoltaic panel array of group 2 does not have 1 defect, and the photovoltaic panel array of group 3 has 1 defect.

[0085] When the user further clicks on the area corresponding to the photovoltaic panel array of group 1 on the user interface, the user interface switches to the user interface of the photovoltaic array of group 1, on which defect indications of multiple photovoltaic panels are presented, for example, a defect occurs on the photovoltaic panel in the first row and first column.

[0086] When the user further clicks on the area corresponding to the photovoltaic panel in the first row and first column on the user interface, the user interface switches to the user interface of the photovoltaic panel in the first row and first column, on which the photovoltaic cells in the 5th row and 3rd column, the 4th row and 5th column, the 3rd row and the 7th row, and the 2nd row and the 7th column have defects. In addition, the location of the defect (e.g., the latitude and longitude or identification of the photovoltaic panel where the defect is located, the type and level of the defect, and other information) can be displayed on the user interface.

[0087] The present application can collect images of photovoltaic panels through drones, and after determining the type of defects using the above-mentioned graded defect detection method, it can centrally present defect-related information of each photovoltaic panel through the user interface and remind maintenance personnel to perform corresponding maintenance, thereby greatly reducing labor costs and improving the power generation efficiency and safety of photovoltaic power stations.

[0088] Therefore, based on the above-mentioned photovoltaic panel defect detection method, significant defects on photovoltaic panels can be detected by an image processing algorithm based on grayscale value analysis with high detection efficiency, and further, subtle defects not detected by the above-mentioned method can be detected by a deep learning algorithm with high detection accuracy. The preliminary detection of the first defect detection algorithm reduces the amount of data required for deep learning to detect, and improves the efficiency of the overall defect detection process. Further, through the combination of the two defect algorithms, different types of defects can be detected, ensuring the comprehensiveness of defect detection and improving the accuracy of defect detection, thus achieving the goal of improving the efficiency of photovoltaic panel defect detection while ensuring the accuracy of detection. In addition, the use of a contactless electroluminescent method to collect electroluminescent images further improves the accuracy of defect detection and avoids adverse effects on photovoltaic panels.

[0089] Defect detection method based on gray value analysis image processing algorithm

[0090] Refer to the following Figures 3 to 5 The process of detecting defects in a target area and determining the defect type of the target area using an image processing algorithm based on gray value analysis is further described in detail.

[0091] Figure 3 This is an exemplary flow chart of a method for detecting photovoltaic panel defects using an image processing algorithm based on grayscale value analysis provided in the present application.

[0092] like Figure 3 As shown, the process may specifically include the following contents.

[0093] S310: Perform connected domain processing on the target area to obtain a target connected domain.

[0094] In some embodiments, based on the target area in the target photovoltaic unit determined in the aforementioned S220, the target area can be processed by connecting the target area to obtain the target connected area. In the subsequent defect detection process, the features of the pixels in the target connected area can be directly analyzed to reduce the amount of data to be analyzed, improve the data efficiency of the analysis, and avoid the feature information of the pixels in other areas from interfering with the defect analysis process and affecting the accuracy of the detection results.

[0095] S320, determining the grayscale distribution characteristic coefficient of the pixel points in the target connected domain according to the grayscale deviation value of each pixel point in the target connected domain and the distribution of the pixel points in the target connected domain. The deviation value can be the deviation between the grayscale value of each pixel point in the target connected domain and the grayscale mean value of the pixels in the row or column in which it is located in the photovoltaic unit, or the deviation between each pixel point and the grayscale mean value of all pixels in the entire photovoltaic unit.

[0096] The distribution of pixels in the target connected domain may be the number of pixels in each row and the number of pixels in each column in the target connected domain. The calculation process of the deviation value may refer to the relevant description of the column deviation value and the row deviation value in the above embodiment, which will not be repeated here.

[0097] The grayscale distribution characteristic coefficient can be used to characterize the similarity and distribution characteristics of the grayscale values ​​of the pixels in the target connected domain. According to the grayscale distribution characteristic coefficient, the texture characteristics of the target connected region can be analyzed to further determine the defect type of the target connected region.

[0098] In some embodiments, Figure 4 As shown, the grayscale distribution characteristic coefficient can be obtained according to the following sub-steps:

[0099] S321. Calculate the absolute value of the sum of the row deviation value and the column deviation value of each pixel in the target connected domain to obtain a grayscale deviation coefficient.

[0100] S322: Count the number of pixels in each row and the number of pixels in each column in the target connected domain and perform normalization processing to obtain an area range coefficient.

[0101] S323, calculate the product of the grayscale deviation coefficient and the area range coefficient and perform normalization processing to obtain the grayscale distribution characteristic coefficient.

[0102] S330, when the grayscale distribution characteristic coefficient is greater than the preset threshold, the defect type of the target area is determined to be a crack, otherwise the defect type of the target area is determined to be a scratch. Cracks are caused by external forces, high temperature expansion or defects in the raw materials themselves, and may extend deep into the interior of the photovoltaic cell or even penetrate the entire cell, while scratches are generally caused by external friction and generally only exist on the surface of the photovoltaic cell. In other words, the texture depth of the crack is generally deeper than the scratch, which can be represented in the grayscale map as the grayscale deviation coefficient of the crack area is generally greater than the grayscale deviation coefficient of the scratch area. Therefore, the larger the grayscale deviation coefficient, the greater the probability that the defect is a crack, and vice versa, the greater the probability that the defect is a scratch.

[0103] In addition, considering the shape characteristics of the crack and scratch regions: cracks are generally radial or scattered, that is, they have a crack center and spread outward from the center, while scratches are generally linear. It can be seen that the regional range coefficient involved in the crack is generally greater than the regional range coefficient corresponding to the scratch. Therefore, the larger the regional range coefficient, the greater the probability that the corresponding target area is a crack defect, and vice versa, the greater the probability that the corresponding target area is a scratch defect.

[0104] Therefore, the grayscale distribution characteristic coefficient obtained by combining the grayscale deviation coefficient and the area range coefficient calculated in S322 and S323 is positively correlated with the probability of crack defects. That is, the larger the grayscale distribution characteristic coefficient, the greater the probability that the corresponding target area is a crack defect, and the smaller the grayscale distribution characteristic coefficient, the smaller the probability that the corresponding target area is a scratch defect.

[0105] In some embodiments, the defect type of the target area can be determined by comparing the grayscale distribution characteristic coefficient obtained in S320 with a preset threshold. Exemplarily, the grayscale distribution characteristic coefficient is compared with a preset threshold, and when the grayscale distribution characteristic coefficient is greater than the preset threshold, the defect type of the target area is determined to be a crack, otherwise the defect type of the target area is determined to be a scratch. The above preset threshold can be obtained based on sample data analysis.

[0106] Exemplary Devices

[0107] Combination of the above Figures 1 to 6 , describes the method embodiment of the present application in detail, and the device embodiment of the present application is described in detail below. It should be understood that the description of the method embodiment corresponds to the description of the device embodiment, so the part not described in detail can refer to the previous method embodiment.

[0108] The present application also provides a photovoltaic panel defect detection device, including a module for implementing the method for layered photovoltaic panel defect detection provided by the present application. Figure 7As shown, the figure is a system module diagram of a photovoltaic panel defect detection device 700 provided in some embodiments of the present application. The photovoltaic panel defect detection device 700 provided in the present application may include an image acquisition module 710, a determination module 720, and a defect detection module 730.

[0109] The image acquisition module 710 may be used to acquire photoluminescence (PL) images of a plurality of photovoltaic units in a photovoltaic panel.

[0110] The determination module 720 can be used to compare the PL images of multiple photovoltaic cells with the image templates of normal photovoltaic cells in turn, and determine the images of photovoltaic cells with similarities lower than a first threshold as the PL images of target photovoltaic cells, and the PL images of the target photovoltaic cells are the PL images of abnormal photovoltaic cells having abnormal areas in the photovoltaic cells.

[0111] The defect detection module 730 may be used to perform defect detection on the PL image of the target photovoltaic unit based on a gray value image analysis algorithm to determine the defect type of the target photovoltaic unit.

[0112] The image acquisition module 710 may also be used to acquire an electroluminescent EL image of a target photovoltaic cell of which defect type has not been determined among the plurality of photovoltaic cells.

[0113] The defect detection module 730 can also be used to perform secondary defect detection on the EL image of the target photovoltaic cell with undetermined defect type among multiple photovoltaic cells through a deep learning algorithm to determine the defect type of the target photovoltaic cell with undetermined defect type.

[0114] For the specific definition of the photovoltaic panel defect detection device, please refer to the definition of the photovoltaic panel defect detection method mentioned above, which will not be repeated here. Each module in the above photovoltaic panel defect detection device can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0115] Exemplary electronic devices and program products

[0116] The present application also provides an electronic device, such as Figure 8 The electronic device 800 provided in the present application includes a memory 810, a processor 820, and an input / output interface 830. The memory 810, the processor 820, and the input / output interface 830 are connected through an internal connection path, the memory 810 is used to store instructions, and the processor 820 is used to execute the instructions stored in the memory 810 to control the input / output interface 830 to receive input data and information, and output data such as operation results.

[0117] It should be understood that in the embodiment of the present application, the processor 820 can adopt a general central processing unit (CPU), a microprocessor, an application specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs to implement the technical solution provided in the embodiment of the present application.

[0118] The memory 810 may include a read-only memory and a random access memory, and provides instructions and data to the processor 620. A portion of the processor 820 may also include a nonvolatile random access memory. For example, the processor 820 may also store information on the device type.

[0119] During the implementation process, each step of the above method can be completed by the hardware integrated logic circuit in the processor 820 or the instruction in the form of software. The vehicle auxiliary braking method disclosed in the embodiment of the present application can be directly embodied as a hardware processor for execution, or it can be executed by a combination of hardware and software modules in the processor. The software module can be located in a mature storage medium in the field such as a random access memory, a flash memory, a read-only memory, a programmable read-only memory or an electrically erasable programmable memory, a register, etc. The storage medium is located in the memory 810, and the processor 820 reads the information in the memory 810 and completes the steps of the above method in combination with its hardware. To avoid repetition, it is not described in detail here. The present application also provides a computer program product, including a computer program / instruction. When the computer program / instruction processor in the computer program product provided by the present application is executed, the photovoltaic panel defect detection method provided by the present application can be implemented.

[0120] All the above optional technical solutions can be arbitrarily combined to form optional embodiments of the present application, and will not be described one by one here.

[0121] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0122] It should be noted that in the apparatus, equipment and method of the present application, each module or each step can be decomposed and / or recombined. These decompositions and / or recombinations should be regarded as equivalent schemes of the present application. The above description of the disclosed aspects is provided to enable any technician in the field to make or use the present application. Various modifications to these aspects are very obvious to those skilled in the art, and the general principles defined here can be applied to other aspects without departing from the scope of the present application. Therefore, the present application is not intended to be limited to the above-mentioned aspects, but to the widest scope consistent with the principles disclosed herein and novel features.

[0123] The above description is intended to illustrate and describe the technical solution of the present application. In addition, this description is not intended to limit the embodiments of the present application to the scope of the above disclosed forms. Although multiple exemplary aspects and embodiments have been discussed in the above content, those skilled in the art can easily obtain other variations, modifications, changes, additions and sub-combinations based on the above content.

[0124] It should be noted that, in the description of this application, the terms "first", "second", "third", etc. are used for descriptive purposes only and cannot be understood as indicating or implying relative importance. In addition, in the description of this application, unless otherwise specified, the meaning of "plurality" is two or more.

[0125] The above are only preferred embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent substitutions, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A photovoltaic panel defect detection method, characterized in that: The method comprises: Acquire photoluminescence (PL) images of multiple photovoltaic cells in a photovoltaic panel; Comparing the PL images of the plurality of photovoltaic cells with the image templates of normal photovoltaic cells in sequence, determining the images of photovoltaic cells with similarity lower than a first threshold as the PL images of target photovoltaic cells, wherein the PL images of the target photovoltaic cells are the PL images of the abnormal photovoltaic cells having abnormal areas in the photovoltaic cells; Performing defect detection on the PL image of the target photovoltaic unit based on a gray value image analysis algorithm to determine the defect type of the target photovoltaic unit; Acquire an electroluminescent EL image of a target photovoltaic cell of an undetermined defect type among the plurality of photovoltaic cells; A secondary defect detection is performed on the EL image of a target photovoltaic cell of an undetermined defect type among the plurality of photovoltaic cells by using a deep learning algorithm to determine the defect type of the target photovoltaic cell of an undetermined defect type.

2. The photovoltaic panel defect detection method according to claim 1, characterized in that: The step of acquiring PL images of a plurality of photovoltaic units in a photovoltaic panel comprises: irradiating the photovoltaic panel with a laser beam emitted by a laser carried by the drone, and acquiring a PL image of the photovoltaic panel with a camera carried by the drone; The PL image of the photovoltaic panel is pre-segmented to obtain PL images of the plurality of photovoltaic units.

3. The photovoltaic panel defect detection method according to claim 1, characterized in that: The method of performing secondary defect detection on the electroluminescent EL image of the target photovoltaic unit of undetermined defect type among the plurality of photovoltaic units by using a deep learning algorithm to determine the defect type of the target photovoltaic unit of undetermined defect type comprises: Inputting the EL image of the target photovoltaic unit of the undetermined defect type into the trained neural network model; The trained neural network model is used to extract features and classify the EL image of the target photovoltaic unit with undetermined defect type, so as to determine the defect type of the EL image of the target photovoltaic unit with undetermined defect type.

4. The photovoltaic panel defect detection method according to claim 1, characterized in that: The acquiring of an EL image of a target photovoltaic cell of an undetermined defect type among the plurality of photovoltaic cells comprises: The laser beam emitted by the laser device carried by the drone irradiates the photovoltaic cell adjacent to the target photovoltaic cell of the undetermined defect type to generate current, so that the target photovoltaic cell receives the current generated by the adjacent photovoltaic cell and emits electroluminescence; An EL image of the target photovoltaic unit of the undetermined defect type is acquired by a camera carried by the drone.

5. The photovoltaic panel defect detection method according to claim 1, characterized in that: The method of performing defect detection on the PL image of the target photovoltaic unit based on the gray value image analysis algorithm to determine the defect type of the target photovoltaic unit includes: Preprocessing the PL image of the target photovoltaic unit to obtain a grayscale image; Performing grayscale deviation analysis on the grayscale image to determine a target area of ​​the image of the target photovoltaic unit; The target area is subjected to defect detection by using an image processing algorithm based on the gray value analysis to determine the defect type of the target area.

6. The photovoltaic panel defect detection method according to claim 5, characterized in that: The method of using an image processing algorithm based on gray value analysis to detect defects in a target area and determine the defect type of the target area includes: Performing connected domain processing on the target area to obtain a target connected domain; Determine the grayscale distribution characteristic coefficient of the pixel points in the target connected domain according to the grayscale deviation value of each pixel point in the target connected domain and the distribution of the pixel points in the target connected domain; When the grayscale distribution characteristic coefficient is greater than a preset threshold, the defect type of the target area is determined to be a crack; otherwise, the defect type of the target area is determined to be a scratch. Wherein, determining the grayscale distribution characteristic coefficient of the pixel points in the target connected domain according to the grayscale deviation value of each pixel point in the target connected domain and the distribution of the pixel points in the target connected domain comprises: Calculate the absolute value of the sum of the row deviation value and the column deviation value of each pixel point in the target connected domain to obtain a grayscale deviation coefficient, wherein the row deviation is the deviation of the grayscale mean of the pixel point and all the pixels in its row, and the column deviation is the deviation of the grayscale mean of the pixel point and all the pixels in its column; Counting the number of pixels in each row and the number of pixels in each column in the target connected domain and performing normalization processing to obtain a region range coefficient; The product of the grayscale deviation coefficient and the area range coefficient is calculated and normalized to obtain a grayscale distribution characteristic coefficient.

7. The photovoltaic panel defect detection method according to claim 1, characterized in that: After determining the defect type of the image of the target photovoltaic unit, the method further includes: marking defects of the target photovoltaic unit according to defect types of defects included in the image of the target photovoltaic unit; presenting a topological image of a photovoltaic panel array on a user interface, the photovoltaic panel array including the photovoltaic panel; After receiving a click operation on the photovoltaic panel from the user, images of the plurality of photovoltaic units and corresponding detection information are presented on the user interface, the detection information including position information, marking information and defect level information corresponding to the photovoltaic units.

8. A photovoltaic panel defect detection device, comprising: An image acquisition module, used for acquiring photoluminescence (PL) images of a plurality of photovoltaic units in a photovoltaic panel; A determination module, configured to compare the PL images of the plurality of photovoltaic cells with the image templates of normal photovoltaic cells in sequence, and determine the images of photovoltaic cells with similarities lower than a first threshold as the PL images of target photovoltaic cells, wherein the PL images of the target photovoltaic cells are PL images of abnormal photovoltaic cells having abnormal areas in the photovoltaic cells; A defect detection module, used to perform defect detection on the PL image of the target photovoltaic unit based on a gray value image analysis algorithm to determine the defect type of the target photovoltaic unit; The image acquisition module is also used to acquire the electroluminescent EL image of the target photovoltaic unit with undetermined defect type among the multiple photovoltaic units; the defect detection module is also used to perform secondary defect detection on the EL image of the target photovoltaic unit with undetermined defect type among the multiple photovoltaic units through a deep learning algorithm to determine the defect type of the target photovoltaic unit with undetermined defect type.

9. An electronic device, characterized in that: The electronic device comprises: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the photovoltaic panel defect detection method according to any one of claims 1 to 7.

10. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instruction processor is executed, it is used to implement the photovoltaic panel defect detection method according to any one of claims 1 to 7.

Images