Photovoltaic module image abnormality detection method, device and equipment

Abstract

The embodiment of the present invention provides a method, device and equipment for detecting anomalies in photovoltaic component images. The method includes: acquiring a photovoltaic component image; identifying image feature points of the photovoltaic component image; identifying grid lines and dividing lines in the photovoltaic component image based on the image feature points; and performing anomaly detection on the photovoltaic component image according to the image feature points, grid lines and dividing lines. The technical solution provided by this embodiment effectively realizes the use of a non-neural network algorithm to perform anomaly detection operations on photovoltaic component images. When the type of photovoltaic component is updated, there is no need to retrain the convolutional neural network. Only fine-tuning the algorithm is required to adapt to the updated photovoltaic component type. This makes the anomaly detection method more flexible. In addition, the method can detect extremely subtle image anomalies with high detection accuracy, which is conducive to avoiding abnormal photovoltaic component images from affecting subsequent production links, further improving the practicality of the method.

Description

Technical Field

[0001] The present invention relates to the field of image technology, and in particular to a method, device and equipment for detecting abnormality of a photovoltaic component image. Background Art

[0002] Photovoltaic modules are cell arrays composed of many cells. Common photovoltaic modules have cell row and column specifications of 6*20 or 6*24. During the production process of photovoltaic modules, they need to be photographed to check the quality. Specifically, when collecting photovoltaic module images, the cells are first divided into several groups and photographed using a line scan camera. After completion, the images of each cell group are spliced ​​into a component image. The spliced ​​component image is used for subsequent quality inspection and production links.

[0003] Errors may occur during the image capture and splicing process, resulting in abnormal photovoltaic module images, affecting the normal development of subsequent production links. Photovoltaic module images with shooting or splicing errors are called abnormal shooting images. There are many forms of abnormal shooting images, and there is no specific form of expression. Common shooting abnormalities include overexposure, grid line misalignment, cell misalignment, etc. Therefore, there is an urgent need for a method for abnormal detection of photovoltaic module images to prevent abnormal photovoltaic module images from affecting subsequent product production links. Summary of the invention

[0004] The embodiments of the present invention provide a method, device and equipment for detecting abnormalities in photovoltaic module images, which can accurately identify abnormal shooting problems existing in photovoltaic module images, thereby effectively preventing abnormal photovoltaic module images from affecting subsequent product production links.

[0005] In a first aspect, an embodiment of the present invention provides a method for detecting anomalies in a photovoltaic module image, comprising:

[0006] Acquire images of photovoltaic modules;

[0007] Identifying image feature points of the photovoltaic assembly image;

[0008] Identify grid lines and dividing lines in the photovoltaic assembly image based on the image feature points;

[0009] Anomalies are detected on the photovoltaic component image according to the image feature points, grid lines and segmentation lines.

[0010] In a second aspect, an embodiment of the present invention provides a photovoltaic module image abnormality detection device, comprising:

[0011] A first acquisition module is used to acquire a photovoltaic module image;

[0012] A first recognition module, used to recognize image feature points of the photovoltaic component image;

[0013] A first processing module, used for identifying grid lines and dividing lines in the photovoltaic component image based on the image feature points;

[0014] The first detection module is used to perform abnormality detection on the photovoltaic module image according to the image feature points, grid lines and dividing lines.

[0015] In a third aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, and a processor; wherein the memory is used to store one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the method for detecting anomalies in photovoltaic component images in the first aspect mentioned above is implemented.

[0016] In a fourth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the photovoltaic component image abnormality detection method in the first aspect when executed.

[0017] In a fifth aspect, an embodiment of the present invention provides a method for extracting feature points of a photovoltaic module image, comprising:

[0018] Acquire images of photovoltaic modules;

[0019] Segmenting the photovoltaic module image based on rows and columns to obtain a plurality of sub-images;

[0020] Determining a local peak point of the sub-image;

[0021] Based on the local peak point, a feature point of the sub-image is determined, and the feature point of the photovoltaic assembly image includes feature points of multiple sub-images.

[0022] In a sixth aspect, an embodiment of the present invention provides a feature point extraction device for a photovoltaic module image, comprising:

[0023] A second acquisition module is used to acquire the photovoltaic module image;

[0024] A second segmentation module is used to segment the photovoltaic module image based on rows and columns to obtain multiple sub-images;

[0025] A second determination module, used to determine a local peak point of the sub-image;

[0026] The second processing module is used to determine the feature points of the sub-image based on the local peak point, and the feature points of the photovoltaic component image include feature points of multiple sub-images.

[0027] In the seventh aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the feature point extraction method of the photovoltaic component image in the above-mentioned fifth aspect.

[0028] In an eighth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the feature point extraction method of a photovoltaic component image in the fifth aspect when executed.

[0029] In a ninth aspect, an embodiment of the present invention provides a method for detecting anomalies in a photovoltaic module image, comprising:

[0030] In response to a request for invoking an anomaly detection service for a photovoltaic assembly image, determining a processing resource corresponding to the anomaly detection service for the photovoltaic assembly image;

[0031] The processing resources are used to perform the following steps: acquiring a photovoltaic module image; identifying image feature points of the photovoltaic module image; identifying grid lines and dividing lines in the photovoltaic module image based on the image feature points; and performing anomaly detection on the photovoltaic module image based on the image feature points, grid lines and dividing lines.

[0032] In a tenth aspect, an embodiment of the present invention provides a photovoltaic module image abnormality detection device, comprising:

[0033] A third determination module is used to determine the processing resources corresponding to the photovoltaic component image abnormality detection service in response to the request for invoking the photovoltaic component image abnormality detection service;

[0034] The third processing module is used to use the processing resources to perform the following steps: acquiring a photovoltaic component image; identifying image feature points of the photovoltaic component image; identifying grid lines and dividing lines in the photovoltaic component image based on the image feature points; and performing anomaly detection on the photovoltaic component image based on the image feature points, grid lines and dividing lines.

[0035] In the eleventh aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, a processor; wherein the memory is used to store one or more computer instructions, wherein the one or more computer instructions, when executed by the processor, implement the method for detecting abnormalities in photovoltaic component images in the ninth aspect above.

[0036] In a twelfth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the method for detecting abnormalities in a photovoltaic component image in the ninth aspect when executed.

[0037] In a thirteenth aspect, an embodiment of the present invention provides a method for extracting feature points of a photovoltaic module image, comprising:

[0038] In response to a request for invoking a feature point extraction service for a photovoltaic assembly image, determining a processing resource corresponding to the feature point extraction service for the photovoltaic assembly image;

[0039] The processing resources are used to perform the following steps: acquiring a photovoltaic component image; segmenting the photovoltaic component image based on rows and columns to obtain multiple sub-images; determining local peak points of the sub-images; and determining feature points of the sub-images based on the local peak points, wherein the feature points of the photovoltaic component image include feature points of multiple sub-images.

[0040] In a fourteenth aspect, an embodiment of the present invention provides a feature point extraction device for a photovoltaic module image, comprising:

[0041] A fourth determination module is used to determine, in response to a request for invoking a feature point extraction service for a photovoltaic assembly image, a processing resource corresponding to the feature point extraction service for the photovoltaic assembly image;

[0042] The fourth processing module is used to use the processing resources to perform the following steps: acquiring a photovoltaic component image; segmenting the photovoltaic component image based on rows and columns to obtain multiple sub-images; determining local peak points of the sub-images; and determining feature points of the sub-images based on the local peak points, wherein the feature points of the photovoltaic component image include feature points of multiple sub-images.

[0043] In the fifteenth aspect, an embodiment of the present invention provides an electronic device, comprising: a memory, and a processor; wherein the memory is used to store one or more computer instructions, wherein when the one or more computer instructions are executed by the processor, the feature point extraction method of the photovoltaic component image in the above-mentioned thirteenth aspect is implemented.

[0044] In a sixteenth aspect, an embodiment of the present invention provides a computer storage medium for storing a computer program, wherein the computer program enables a computer to implement the feature point extraction method of a photovoltaic component image in the thirteenth aspect when executed.

[0045] The technical solution provided in this embodiment effectively realizes the use of non-neural network algorithms to perform anomaly detection operations on photovoltaic module images. When the type of photovoltaic module is updated, there is no need to retrain the convolutional neural network, and only the algorithm needs to be fine-tuned to adapt to the updated photovoltaic module. This makes the anomaly detection method more flexible, and the method can detect extremely subtle image anomalies with high detection accuracy, which is conducive to avoiding abnormal photovoltaic module images from affecting subsequent production links, further improving the practicality of the method. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0047] Figure 1 A schematic diagram of misaligned grid lines provided in the related art;

[0048] Figure 2 A schematic diagram of battery cell misalignment provided in the related art;

[0049] Figure 3 A schematic diagram of overexposure in related art;

[0050] Figure 4 A schematic diagram of a process flow of a photovoltaic module image abnormality detection method provided by an embodiment of the present invention;

[0051] Figure 5 A schematic diagram of a photovoltaic module image anomaly detection method provided by an embodiment of the present invention;

[0052] Figure 6 A schematic diagram of segmenting the photovoltaic assembly image provided by an embodiment of the present invention;

[0053] Figure 7 A schematic diagram of a process for determining a local peak point of a sub-image provided by an embodiment of the present invention;

[0054] Figure 8 A schematic diagram of a mean vector corresponding to a sub-image provided by an embodiment of the present invention;

[0055] Fig. 9 A schematic diagram of a high frequency component provided by an embodiment of the present invention;

[0056] Fig.10 A schematic diagram of a local peak point of a sub-image provided by an embodiment of the present invention;

[0057] Fig.11 A schematic diagram of determining the jump position included in the initial peak point based on the second-order difference vector provided in an embodiment of the present invention;

[0058] Fig.12 A schematic diagram of a local peak point of a sub-image provided by an embodiment of the present invention;

[0059] Fig.13A schematic diagram of a process for determining the image feature points of the sub-image based on the local peak points provided in an embodiment of the present invention;

[0060] Fig.14 Schematic diagram of the process of obtaining the confidence of the local peak point provided by the embodiment of the present invention Figure 1 ;

[0061] Fig.15 Schematic diagram of the process of obtaining the confidence of the local peak point provided by the embodiment of the present invention Figure 2 ;

[0062] Fig.16 A schematic diagram of a process for determining segmentation line feature points from the image feature points provided by an embodiment of the present invention;

[0063] Fig.17 A schematic diagram of a process for obtaining a photovoltaic module image provided by an embodiment of the present invention;

[0064] Fig.18 A schematic diagram of a process for removing the invalid area in the original image to obtain the photovoltaic module image provided by an embodiment of the present invention;

[0065] Fig.19 A schematic flow chart of another method for detecting abnormalities in photovoltaic module images provided by an embodiment of the present invention;

[0066] Fig. 20 A schematic diagram of a flow chart of a method for extracting feature points of a photovoltaic module image provided by an embodiment of the present invention;

[0067] Fig.21 A schematic diagram of the principle of a photovoltaic module image shooting anomaly detection method based on feature point matching provided by an application embodiment of the present invention;

[0068] Fig. 22 A schematic diagram of sampling a plurality of image blocks provided by an application embodiment of the present invention;

[0069] Fig.23 A schematic diagram of a sub-image of a longitudinal slice provided by an application embodiment of the present invention;

[0070] Fig.24 A schematic diagram of a mean vector provided for an application embodiment of the present invention;

[0071] Fig.25 A schematic diagram of a high frequency component provided by an application embodiment of the present invention;

[0072] Fig.26 A schematic diagram of image feature points provided by an application embodiment of the present invention;

[0073] Fig. 27A schematic diagram of calculating a mean vector of a photovoltaic assembly image along a horizontal direction provided by an application embodiment of the present invention;

[0074] Fig.28 A schematic diagram of a high frequency part provided for an application embodiment of the present invention;

[0075] Fig.29 A schematic diagram of peak confidence provided for an application embodiment of the present invention;

[0076] Fig.30 A schematic diagram of determining a cell separation line according to a threshold value provided in an application embodiment of the present invention;

[0077] Fig.31 A schematic diagram of determining a cell separation line according to a threshold value provided in an application embodiment of the present invention;

[0078] Fig.32 A schematic structural diagram of a photovoltaic module image anomaly detection device provided by an embodiment of the present invention;

[0079] Fig.33 For Fig.32 A schematic diagram of the structure of an electronic device corresponding to the photovoltaic module image abnormality detection device provided in the illustrated embodiment;

[0080] Fig.34 A schematic diagram of the structure of a feature point extraction device for a photovoltaic module image provided by an embodiment of the present invention;

[0081] Fig.35 For Fig.34 A schematic diagram of the structure of an electronic device corresponding to the device for extracting feature points of a photovoltaic assembly image provided by the illustrated embodiment;

[0082] Fig.36 A schematic flow chart of another method for detecting abnormalities in photovoltaic module images provided by an embodiment of the present invention;

[0083] Fig.37 A schematic structural diagram of a photovoltaic module image anomaly detection device provided by an embodiment of the present invention;

[0084] Fig.38 For Fig.37 A schematic diagram of the structure of an electronic device corresponding to the photovoltaic module image abnormality detection device provided in the illustrated embodiment;

[0085] Fig.39 A schematic flow chart of another method for extracting feature points from a photovoltaic module image provided by an embodiment of the present invention;

[0086] Fig.40 A schematic diagram of the structure of a feature point extraction device for a photovoltaic module image provided by an embodiment of the present invention;

[0087] Fig.41 For Fig.40 A schematic structural diagram of an electronic device corresponding to the device for extracting feature points of a photovoltaic assembly image provided by the illustrated embodiment. DETAILED DESCRIPTION

[0088] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0089] The terms used in the embodiments of the present invention are only for the purpose of describing specific embodiments, and are not intended to limit the present invention. The singular forms "a", "said" and "the" used in the embodiments of the present invention and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings, and "multiple" generally includes at least two, but does not exclude the inclusion of at least one.

[0090] It should be understood that the term "and / or" used in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article generally indicates that the associated objects before and after are in an "or" relationship.

[0091] As used herein, the words "if" and "if" may be interpreted as "at the time of" or "when" or "in response to determining" or "in response to detecting", depending on the context. Similarly, the phrases "if it is determined" or "if (stated condition or event) is detected" may be interpreted as "when it is determined" or "in response to determining" or "when detecting (stated condition or event)" or "in response to detecting (stated condition or event)", depending on the context.

[0092] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a product or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such a product or system. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the product or system including the elements.

[0093] In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0094] Explanation of terms:

[0095] Photovoltaic module: an array composed of photovoltaic cells.

[0096] Grid line: A horizontal line segment in a photovoltaic cell.

[0097] Filtering: A technique that allows or blocks signals in a certain frequency band from passing through.

[0098] Non-Maximum Suppression (NMS) is a technique that searches for local maxima and filters adjacent non-maxima.

[0099] Photovoltaic modules are cell arrays composed of many cells. Common photovoltaic modules have cell row and column specifications of 6*20 or 6*24. During the production process of photovoltaic modules, they need to be photographed to check the quality. Specifically, when collecting photovoltaic module images, the cells are first divided into several groups and photographed using a line scan camera. After completion, the images of each cell group are stitched together into a component image. The stitched component image is used for subsequent quality inspection and production links. For example, solar panels can be produced from photovoltaic module images.

[0100] Errors may occur during the image capture and splicing process, resulting in abnormal images of photovoltaic modules, affecting the normal development of subsequent production links. Photovoltaic module images with capture or splicing errors are called abnormal images. There are many forms of abnormal capture, and there is no specific form. Common abnormal captures include overexposure, grid line misalignment, cell misalignment, etc. Figure 1 A schematic diagram of grid line misalignment is provided. Figure 2 A schematic diagram of battery cell misalignment is provided. Figure 3 A schematic diagram of overexposed photography is provided.

[0101] In order to ensure the normal development of subsequent production links, the relevant technology provides a method for detecting shooting anomalies based on deep convolutional neural networks. However, this method has the following defects: (1) It has a slow adaptation speed for photovoltaic module products and requires human resources to annotate photovoltaic module images. When photovoltaic module products are updated, it is necessary to train and update the model, and the entire update cycle is long. (2) Based on deep convolutional neural networks, only obvious photovoltaic module anomalies can be detected, such as: large-area overexposure and other abnormalities, and it is not sensitive to more subtle image anomalies, such as: deep convolutional neural networks are difficult to detect abnormalities such as grid line misalignment.

[0102] In order to solve the above technical problems, this embodiment proposes a photovoltaic module image abnormality detection method, device and equipment. The following is a detailed description of some embodiments of the present invention in conjunction with the accompanying drawings. In the case of no conflict between the embodiments, the following embodiments and features in the embodiments can be combined with each other.

[0103] Figure 4 A schematic diagram of a process flow of a photovoltaic module image abnormality detection method provided by an embodiment of the present invention; Figure 5 A schematic diagram of a photovoltaic module image abnormality detection method provided by an embodiment of the present invention; Figure 4 -Attached Figure 5 As shown, this embodiment provides a method for detecting anomalies in photovoltaic component images. The executor of the method may be a device for detecting anomalies in photovoltaic component images. It can be understood that the device for detecting anomalies in photovoltaic component images may be implemented as software, or a combination of software and hardware. Specifically, the device for detecting anomalies in photovoltaic component images may be implemented as a client, which may be any computing device with certain image processing capabilities.

[0104] The basic structure of the client may include: at least one processor. The number of processors depends on the configuration and type of the client. The client may also include a memory, which may be volatile, such as RAM, or non-volatile, such as read-only memory (ROM), flash memory, etc., or may include both types. The memory usually stores an operating system (OS), one or more application programs, and may also store program data, etc. In addition to the processing unit and the memory, the client also includes some basic configurations, such as a network card chip, an IO bus, a display component, and some peripheral devices, etc. Optionally, some peripheral devices may include, for example, a keyboard, a mouse, an input pen, a printer, etc. Other peripheral devices are well known in the art and will not be described in detail here. Optionally, the client may be a PC (personal computer) terminal, a handheld terminal (e.g., a smart phone, a tablet computer), etc.

[0105] Alternatively, the photovoltaic module image anomaly detection device can be implemented as a server, which refers to a device that can provide photovoltaic module image anomaly detection services in a network virtual environment. In physical implementation, the server can be any device that can provide computing services, respond to anomaly detection requests, and perform processing, for example: it can be a cluster server, a conventional server, a cloud server, a cloud host, a virtual center, etc., which can mainly include a processor, a hard disk, a memory, a system bus, etc., which is similar to a general computer architecture.

[0106] In the above embodiment, the abnormality detection device of the photovoltaic module image can be connected to the client (for obtaining the photovoltaic module image) through a network, and the network connection can be a wireless or wired network connection. If the client and the server are connected in communication, the network standard of the mobile network can be any one of 2G (GSM), 2.5G (GPRS), 3G (WCDMA, TD-SCDMA, CDMA2000, UTMS), 4G (LTE), 4G+ (LTE+), WiMax, 5G, etc.

[0107] Specifically, the photovoltaic module image abnormality detection method may include:

[0108] S401: Acquire a photovoltaic module image.

[0109] The photovoltaic module image refers to an image that needs to be detected for abnormality. In the process of producing photovoltaic modules and products corresponding to photovoltaic modules, in order to ensure the quality and efficiency of the production of photovoltaic modules and products corresponding to photovoltaic modules, the photovoltaic module image corresponding to the photovoltaic module can be obtained, and then the photovoltaic module image can be detected for abnormality. When there is an abnormality in the photovoltaic module image, it means that there may be an abnormality in the photovoltaic module, for example: the grid lines of the cells constituting the photovoltaic module are not aligned, the cells constituting the photovoltaic module are misaligned, etc.; when there is no abnormality in the photovoltaic module image, it means that there is no abnormality in the photovoltaic module.

[0110] In addition, the present embodiment does not limit the specific implementation method for obtaining the photovoltaic component image. In some examples, the abnormality detection device may be provided with a shooting device, and the photovoltaic component image can be obtained by shooting the photovoltaic component to be detected through the shooting device. In other examples, the abnormality detection device is communicatively connected to the shooting device, and the photovoltaic component image can be obtained by shooting the photovoltaic component to be detected through the shooting device, and then the obtained photovoltaic component image can be sent to the abnormality detection device, so that the abnormality detection device can stably obtain the photovoltaic component image.

[0111] Of course, the specific implementation method of obtaining the photovoltaic component image is not limited to the implementation method described above. Those skilled in the art may also adopt other methods to obtain the photovoltaic component image. For example, the photovoltaic component image may be stored in a preset area, and the photovoltaic component image may be obtained by accessing the preset area. As long as the accuracy and reliability of the photovoltaic component image can be guaranteed, it will not be repeated here.

[0112] S402: Identify image feature points of a photovoltaic module image.

[0113] After the photovoltaic module image is acquired, the photovoltaic module image may be analyzed and processed to identify the image feature points of the photovoltaic module image. In some instances, the photovoltaic module image may be analyzed and processed using a feature point extraction algorithm to identify the image feature points of the photovoltaic module image. In other instances, since the photovoltaic module is an array composed of photovoltaic cells spliced ​​together, the photovoltaic module image obtained at this time may include image blocks corresponding to the photovoltaic cells. Therefore, the identification of the image feature points of the photovoltaic module image in this embodiment may include: segmenting the photovoltaic module image based on rows and columns to obtain multiple sub-images; determining the local peak points of the sub-images; and determining the image feature points of the sub-image based on the local peak points, and the image feature points of the photovoltaic module image include the image feature points of multiple sub-images.

[0114] In order to accurately obtain the image feature points of the photovoltaic module image, after obtaining the photovoltaic module image, the photovoltaic module image can be segmented based on rows and columns, so that multiple sub-images corresponding to the photovoltaic module image can be obtained, such as Figure 6 In other examples, the photovoltaic component image may not be segmented based on rows and columns, but may be segmented according to a set size, thereby obtaining a plurality of sub-images. It is understood that the set size may be the same as or different from the row size / column size included in the photovoltaic component image. This embodiment does not limit the specific value of the set size, and those skilled in the art may adjust or configure the set size according to specific application scenarios and design requirements.

[0115] For each sub-image, each sub-image is composed of several pixels, and the feature points are only part of the feature points included in the sub-image. In order to ensure the quality and efficiency of determining the image feature points of the sub-image, after obtaining multiple sub-images, each sub-image can be analyzed and processed to determine the local peak point of the sub-image, wherein one sub-image can correspond to multiple local peak points, and the above-mentioned local peak point refers to a feature point in a preset area with an amplitude greater than that of other feature points. After obtaining the local peak point, the image feature point of the sub-image can be determined based on the local peak point. In some instances, the local peak point can be determined as the image feature point of the sub-image. For the photovoltaic module image and the sub-image, the sub-image is a part of the photovoltaic module image. Therefore, after determining the image feature points of all sub-images, the image feature points of the photovoltaic module image can be accurately identified. The image feature point is a pixel point with an amplitude greater than or equal to the set threshold.

[0116] S403: Identify grid lines and dividing lines in the photovoltaic module image based on image feature points.

[0117] The photovoltaic module image includes grid lines and dividing lines. The grid lines refer to the corresponding horizontal line segments in the photovoltaic cell sheets, and the dividing lines refer to the line segments between different photovoltaic cells. For photovoltaic modules, the characteristic values ​​(for example, gray values) corresponding to the pixel points of the grid lines and dividing lines corresponding to the photovoltaic modules in the photovoltaic module image will be greater than the characteristic values ​​corresponding to the pixel points at other positions. Therefore, after obtaining the image feature points, the photovoltaic module image can be analyzed and processed based on the image feature points to identify the grid lines and dividing lines in the photovoltaic module image.

[0118] In some instances, identifying grid lines and dividing lines in a photovoltaic module image based on image feature points may include: determining dividing line feature points among the image feature points, wherein, in a local area, the amplitude of the dividing line feature points is greater than the amplitude of other image feature points; determining grid lines in the photovoltaic module image based on the image feature points; and determining dividing lines in the photovoltaic module image based on the dividing line feature points.

[0119] Among them, the image feature points include feature points for forming a dividing line and feature points for forming a grid line. In order to accurately identify the dividing line, after acquiring the image feature points, the image feature points can be analyzed and processed to determine the dividing line feature points included in the image feature points. Specifically, the amplitude of the dividing line feature point is greater than the amplitude of other image feature points in the local area, that is, the dividing line feature point is the peak point in the local area; it can be understood that the dividing line feature point is part of the image feature points.

[0120] After acquiring the image feature points and the dividing line feature points, the grid lines in the photovoltaic module image can be determined based on the image feature points, and the dividing lines in the photovoltaic module image can be determined based on the dividing line feature points, thereby effectively ensuring the accuracy and reliability of determining the dividing lines and grid lines. It should be noted that when the determined grid lines coincide with the determined dividing lines, the coincident line segments can be determined as the dividing lines.

[0121] S404: Perform anomaly detection on the photovoltaic module image according to image feature points, grid lines and segmentation lines.

[0122] After acquiring the image feature points, grid lines and dividing lines, an abnormality detection operation can be performed on the photovoltaic component image based on the image feature points, grid lines and dividing lines. In some examples, performing an abnormality detection on the photovoltaic component image based on the image feature points, grid lines and dividing lines can include: identifying whether the image feature points match the grid lines, whether the image feature points match the dividing lines; when the image feature points match the grid lines, and the image feature points match the dividing lines, it is determined that there is no abnormality in the photovoltaic component image; when the image feature points do not match the grid lines, or the image feature points do not match the dividing lines, it is determined that there is an abnormality in the photovoltaic component image.

[0123] Specifically, after acquiring the image feature points, grid lines and dividing lines, it is possible to detect whether the image feature points match the grid lines, and whether the image feature points match the dividing lines, that is, to detect whether the image feature points coincide with the grid lines, and whether the image feature points coincide with the dividing lines; when the image feature points coincide with the grid lines, it means that the image feature points match the grid lines; when the image feature points do not coincide with the grid lines, it means that the image feature points do not match the grid lines; when the image feature points coincide with the dividing lines, it means that the image feature points match the dividing lines; when the image feature points do not coincide with the dividing lines, it means that the image feature points do not match the dividing lines.

[0124] When the recognition result is that the image feature points match the grid lines, and the image feature points match the dividing lines, it means that the image feature points in the photovoltaic module image coincide with both the identified grid lines and the dividing lines, and thus it can be determined that there is no abnormality in the photovoltaic module image. When the recognition result is that the image feature points do not match the grid lines, or the image feature points do not match the dividing lines, it means that the image feature points in the photovoltaic module image do not coincide with at least one of the identified grid lines and the dividing lines, and thus it can be determined that there is an abnormality in the photovoltaic module image, and the abnormality may include: misalignment of the cell, misalignment of the grid lines, too dark light or overexposure, etc.

[0125] In some instances, in order to improve the practicality of the method, the method in this embodiment may also include: when it is determined that there is an abnormality in the photovoltaic component image, prohibiting production operations based on the photovoltaic component image; when it is determined that there is no abnormality in the photovoltaic component image, allowing production operations based on the photovoltaic component image.

[0126] Specifically, after abnormality detection is performed on the photovoltaic module image, if it is determined that the photovoltaic module image has an abnormality, in order to ensure the quality and efficiency of the production of photovoltaic modules and related products, production operations based on the photovoltaic module image can be prohibited. When it is determined that there is no abnormality in the photovoltaic module image, it means that stable production operations can be performed based on the photovoltaic module image at this time, and then production operations based on the photovoltaic module image are allowed, which effectively ensures the quality and efficiency of the production of photovoltaic modules and related products.

[0127] The anomaly detection method provided in the present embodiment acquires a photovoltaic module image, identifies image feature points of the photovoltaic module image, identifies grid lines and dividing lines in the photovoltaic module image based on the image feature points, and performs anomaly detection on the photovoltaic module image according to the image feature points, grid lines and dividing lines. This effectively realizes the use of a non-neural network algorithm to perform anomaly detection on the photovoltaic module image. When the type of photovoltaic module is updated, there is no need to retrain the convolutional neural network, and only the algorithm needs to be fine-tuned to adapt to the updated photovoltaic module. This makes the anomaly detection method more flexible, and the method can detect extremely subtle image anomalies with high detection accuracy, which is beneficial to avoid abnormal photovoltaic module images affecting subsequent production links, further improving the practicality of the method.

[0128] Figure 7 A schematic diagram of a flow chart of determining a local peak point of a sub-image according to an embodiment of the present invention; based on the above embodiment, referring to the attached Figure 7 As shown, after the photovoltaic module image is segmented to obtain multiple sub-images, this embodiment provides an implementation method for determining the local peak point of the sub-image. Specifically, the determination of the local peak point of the sub-image in this embodiment may include:

[0129] Step S701: Obtain a mean vector corresponding to each sub-image along a direction perpendicular to the edge of the image.

[0130] Step S702: Perform low-frequency filtering on the mean vector to obtain high-frequency components.

[0131] Step S703: Determine the local peak point of the sub-image based on the high-frequency component.

[0132] After obtaining multiple sub-images, each sub-image can be analyzed and processed to obtain a mean vector corresponding to each sub-image. Specifically, for each sub-image, the sub-image can be first compressed in a direction perpendicular to the image edge (including horizontal edges and vertical edges) to obtain a line segment corresponding to the sub-image, and then all pixel points located on the line segment are averaged to obtain a mean vector corresponding to the sub-image. It can be understood that one sub-image can correspond to one mean vector.

[0133] Since the photovoltaic module image includes multiple sub-images, each sub-image corresponds to a mean vector, and therefore, a photovoltaic module image corresponds to multiple mean vectors. After obtaining the mean vector, in order to accurately identify the local peak point, the mean vector can be low-frequency filtered to obtain the high-frequency component, which can avoid the local peak point identification error caused by the close pixel values ​​of the image feature points. It can be understood that when the mean vector is low-frequency filtered, the low-frequency threshold corresponding to the low-frequency filter can be configured or adjusted by those skilled in the art according to the specific application scenario and application requirements, which will not be repeated here.

[0134] After obtaining the high-frequency components, the high-frequency components can be analyzed and processed to determine the local peak points of the sub-image; in some instances, determining the local peak points of the sub-image based on the high-frequency components may include: determining a first-order difference vector based on the high-frequency components; determining an initial peak point of the sub-image based on the first-order difference vector, the amplitude of the initial peak point being greater than the amplitude of adjacent feature points; and determining the local peak point of the sub-image among the initial peak points.

[0135] After obtaining the high-frequency components, the high-frequency components can be analyzed and processed to determine the first-order differential vector, which can include a first-order forward differential vector and a first-order backward differential vector; then the initial peak point of the sub-image can be determined based on the first-order differential vector, and the amplitude of the initial peak point is greater than the amplitude of the adjacent feature points. After obtaining the initial peak point, the local peak point of the sub-image can be determined from the initial peak point, and the local peak point can be any one or more of the initial peak points.

[0136] Specifically, among the initial peak points, determining the local peak points of the sub-image may include: sorting the initial peak points; determining the second-order difference vectors corresponding to the sorted initial peak points; based on the second-order difference vectors, determining the jump positions included in the initial peak points; determining the initial peak points corresponding to the jump positions as reference feature points; and determining the initial peak points whose amplitudes are greater than the reference feature points as the local peak points of the sub-image.

[0137] For example, after acquiring the sub-images, the sub-images can be analyzed and processed along the direction perpendicular to the image edge, so as to obtain the mean vector corresponding to each sub-image, such as Figure 8 As shown in FIG. 1 , a schematic diagram of a mean vector corresponding to a sub-image is provided. After the mean vector is obtained, a low-frequency filtering process can be performed on the mean vector to obtain a high-frequency component, such as Fig. 9 After obtaining the high-frequency component, the first-order difference vector corresponding to the high-frequency component can be determined, and the initial peak point of the sub-image can be determined based on the first-order difference vector. The number of initial peak points can be multiple, such as Fig.10As shown, the amplitude of the initial peak point is greater than the amplitude of the adjacent feature points.

[0138] After the initial peak points are obtained, the obtained initial peak points can be sorted, and then the second-order difference vectors corresponding to the sorted initial peak points are determined, and the jump positions included in the initial peak points are determined based on the second-order difference vectors, such as Fig.11 shown.

[0139] After the jump position is obtained, the initial peak point corresponding to the jump position can be determined as a reference feature point, and the initial peak point with a magnitude greater than the reference feature point can be determined as a local peak point of the sub-image, such as Fig.12 As shown, the number of determined local peak points is multiple, thereby effectively ensuring the accuracy and reliability of determining the local peak points of the sub-image.

[0140] In this embodiment, after acquiring the sub-image, the mean vector corresponding to each sub-image is obtained along a direction perpendicular to the edge of the image, and then the mean vector is low-frequency filtered to obtain a high-frequency component, and the local peak point of the sub-image is determined based on the high-frequency component, thereby effectively ensuring the accuracy of determining the local peak point and further improving the accuracy and reliability of abnormality detection of photovoltaic module images based on local peak points.

[0141] Fig.13 A schematic diagram of a process for determining image feature points of a sub-image based on local peak points provided by an embodiment of the present invention; Fig.13 As shown, this embodiment provides an implementation method for determining image feature points of a sub-image based on local peak points. Specifically, determining image feature points of a sub-image based on local peak points may include:

[0142] Step S1301: Identify redundant feature points included in local peak points.

[0143] Among them, for the photovoltaic module image, the image feature points of the photovoltaic module image are mainly used to identify the grid lines and the dividing lines, and because the grid lines and the dividing lines are straight lines, when the grid lines or the dividing lines are identified and confirmed, one or several image feature points can be used to accurately identify them. At this time, if the image feature points of the identified photovoltaic module image include redundant feature points, it will not only increase the amount of data calculation and occupy computing resources, but also reduce the quality and efficiency of identifying the grid lines and the dividing lines. Therefore, when identifying the image feature points of the photovoltaic module image, it is possible to detect whether there are redundant feature points in the image feature points. If so, the redundant feature points can be deleted.

[0144] Specifically, after obtaining the local peak points of the sub-image, redundancy detection may be performed on the local peak points to identify redundant feature points included in the local peak points, wherein the redundant feature points may include feature points with relatively close distances, feature points with low recognition accuracy, etc. In some examples, identifying the redundant feature points included in the local peak points may include: obtaining the confidence of the local peak points; and identifying the redundant feature points included in the local peak points based on the confidence.

[0145] After obtaining the local peak point, the local peak point can be analyzed and processed to obtain the confidence of the local peak point. The confidence of the local peak point is used to identify the recognition accuracy of the local peak point. The higher the confidence, the higher the recognition accuracy of the local peak point. After obtaining the confidence, the confidence can be analyzed and processed to identify the redundant feature points included in the local peak point. Specifically, based on the confidence, identifying the redundant feature points included in the local peak point may include: when the confidence is less than a preset threshold, it means that the recognition accuracy corresponding to the local peak point at this time is low, and then the local peak point corresponding to the confidence can be determined as a redundant feature point; when the confidence is greater than or equal to the preset threshold, it means that the recognition accuracy corresponding to the local peak point at this time is high, and then the local peak point corresponding to the confidence can be determined as a non-redundant feature point.

[0146] Step S1302: Delete the redundant feature points in the local peak points to obtain the image feature points of the sub-image.

[0147] After identifying that the local peak points include redundant feature points, in order to ensure the quality and efficiency of abnormality detection of the photovoltaic module image, the redundant feature points in the local peak points can be deleted, so that the image feature points of the sub-image that does not contain the redundant feature points can be obtained. When the image feature points of all sub-images in the photovoltaic module image are obtained, the image feature points of the photovoltaic module image can be obtained.

[0148] In this embodiment, the redundant feature points included in the local peak points are identified and then deleted to obtain the image feature points of the sub-image. At this time, the image feature points of the photovoltaic module image determined by the image feature points of the sub-image do not include redundant feature points. Therefore, when the abnormality detection operation is performed on the photovoltaic module image based on the image feature points of the photovoltaic module image, the quality and efficiency of the abnormality detection on the photovoltaic module image can be effectively improved, and the stability and reliability of the method can be further improved.

[0149] Fig.14 Schematic diagram of the process of obtaining the confidence of a local peak point provided by an embodiment of the present invention Figure 1 ; Reference Attachment Fig.14As shown, this embodiment provides an implementation method for obtaining the confidence of a local peak point. Specifically, obtaining the confidence of a local peak point in this embodiment may include:

[0150] Step S1401: Obtain the adjacent distances corresponding to all adjacent local peak points.

[0151] Step S1402: Determine the average distance based on all adjacent distances.

[0152] Step S1403: Determine the confidence of the local peak point based on the average distance and the adjacent distance.

[0153] Among them, after obtaining the local peak point, the distance calculation can be performed on all adjacent local peak points, so as to obtain the adjacent distances corresponding to all adjacent local peak points. After obtaining all the adjacent distances, the average distance is determined based on all the adjacent distances. Then, the average distance and the adjacent distance can be analyzed and processed to determine the confidence of the local peak point.

[0154] Specifically, based on the average distance and the adjacent distance, determining the confidence of the local peak point may include: obtaining the distance difference between the average distance and the adjacent distance; based on the distance difference, determining the confidence of the local peak point, and the confidence is negatively correlated with the size of the distance difference; in simple terms, the average distance can be regarded as the amplitude vibration center point of all local peak points. When the amplitude of the local peak point is farther away from the vibration center point, the confidence of the local peak point is lower; when the amplitude of the local peak point is closer to the vibration center point, the confidence of the local peak point is higher.

[0155] For example, the local peak points include peak point a, peak point b, peak point c, peak point d, peak point e and peak point f, and then the adjacent distances corresponding to the adjacent local peak points can be obtained, that is, the adjacent distance d1 between peak point a and peak point b, the adjacent distance d2 between peak point b and peak point c, the adjacent distance d3 between peak point c and peak point d, the adjacent distance d4 between peak point d and peak point e, and the adjacent distance d5 between peak point e and peak point f.

[0156] After obtaining the adjacent distances of the adjacent local peak points mentioned above, the average distance can be determined based on all the adjacent distances, that is, D = (d1+d2+d3+d4+d5) / 5. After that, the distance difference between the average distance and the adjacent distance can be obtained, that is, the distance difference can be obtained: |D-d1|, |D-d2|, |D-d3|, |D-d4|, |D-d5|. After obtaining the distance difference, the confidence of the local peak point can be determined based on the mapping relationship between the distance difference and the confidence of the local peak point. Specifically, the confidence is negatively correlated with the size of the distance difference, that is, the smaller the distance difference, the higher the confidence of the local peak point, and the larger the distance difference, the smaller the confidence of the local peak point.

[0157] In this embodiment, by obtaining the adjacent distances of all adjacent local peak points, the average distance is determined based on all the adjacent distances, and then the confidence of the local peak point is determined based on the average distance and the adjacent distance, thereby effectively ensuring the accuracy and reliability of determining the confidence of the local peak point.

[0158] Fig.15 Schematic diagram of the process of obtaining the confidence of a local peak point provided by an embodiment of the present invention Figure 2 ; Reference Attachment Fig.15 As shown, this embodiment provides another implementation method for obtaining the confidence of a local peak point. Specifically, obtaining the confidence of a local peak point in this embodiment may include:

[0159] Step S1501: Determine the neighborhood corresponding to the local peak point.

[0160] Step S1502: Count the number of local peak points in the neighborhood.

[0161] Step S1503: Based on the quantity, determine the confidence of the local peak point, wherein the confidence is positively correlated with the quantity.

[0162] Among them, after obtaining the local peak point, the neighborhood corresponding to the local peak point can be determined. Specifically, the area with a set distance from the local peak point can be determined as the neighborhood corresponding to the local peak point with the local peak point as the center; after obtaining the neighborhood corresponding to the local peak point, the number of local peak points in the neighborhood can be counted, and then the confidence of the local peak point can be determined based on the mapping relationship between the number and the confidence of the local peak point. Specifically, the confidence is positively correlated with the number, that is, the larger the number of local peak points included in the neighborhood, the greater the confidence of the local peak point; the smaller the number of local peak points included in the neighborhood, the smaller the confidence of the local peak point.

[0163] In this embodiment, the neighborhood corresponding to the local peak point is determined, and then the number of local peak points in the neighborhood is counted, and the confidence of the local peak point is determined based on the number. This not only provides another implementation method for determining the confidence of the local peak point, but also ensures the flexibility and reliability of determining the confidence.

[0164] Of course, those skilled in the art may also adopt other methods to obtain the confidence of the local peak point. In some examples, in order to ensure the accuracy and reliability of the confidence determination, obtaining the confidence of the local peak point may also include: obtaining the adjacent distances of all adjacent local peak points and the neighborhood corresponding to the local peak point, determining the average distance based on all adjacent distances, counting the number of local peak points located in the neighborhood, and then determining the confidence of the local peak point based on the average distance, the adjacent distance and the number of local peak points located in the neighborhood.

[0165] After obtaining the average distance and the adjacent distance, the distance difference between the average distance and the adjacent distance can be obtained, and then the confidence of the local peak point can be determined based on the distance difference and the number of local peak points located in the neighborhood. At this time, the confidence of the local peak point is negatively correlated with the distance difference, and positively correlated with the number of local peak points located in the neighborhood, that is, the confidence of the local peak point obtained by this method is determined by comprehensively considering the degree of deviation between the local peak points and the number of distributions between the local peak points, thereby effectively improving the accuracy of determining the confidence of the local peak point.

[0166] Fig.16 A schematic diagram of a flow chart of determining segmentation line feature points in image feature points provided by an embodiment of the present invention; Fig.16 As shown, this embodiment provides an implementation method for determining the segmentation line feature points. Specifically, in this embodiment, among the image feature points, determining the segmentation line feature points may include:

[0167] Step S1601: Determine peak feature points among image feature points.

[0168] Step S1602: Obtain the confidence vector corresponding to the peak feature point.

[0169] Step S1603: Perform low-frequency filtering on the confidence vector to obtain a processed vector.

[0170] Step S1604: Determine segmentation line feature points based on the processed vector.

[0171] After acquiring the image feature points, all the image feature points can be analyzed and compared to obtain the peak feature points, and then the peak feature points can be analyzed and processed to obtain the confidence vector corresponding to the peak feature points. In some examples, obtaining the confidence vector corresponding to the peak feature point can include: obtaining all intermediate feature points between the peak feature point and adjacent peak feature points; determining the maximum difference between the peak feature point and all the intermediate feature points; and determining the confidence vector corresponding to the peak feature point based on the maximum difference.

[0172] Specifically, after obtaining the peak feature point, the adjacent peak feature points (previous adjacent peak feature points and / or subsequent adjacent peak feature points) corresponding to the peak feature point can be determined, and then all the intermediate feature points between the peak feature point and the adjacent peak feature points can be obtained, and then the maximum difference corresponding to the peak feature point and all the intermediate feature points is determined, that is, the intermediate feature point with the smallest amplitude value is determined among all the intermediate feature points, and then the difference between the smallest intermediate feature point and the peak feature point is determined as the maximum difference. After obtaining the maximum difference, the maximum difference can be analyzed and processed to determine the confidence vector corresponding to the peak feature point.

[0173] After obtaining the confidence vector, in order to avoid the influence of low-frequency data on the determination of the segmentation line feature points, the confidence vector can be subjected to low-frequency filtering, so as to obtain a processed vector. After obtaining the processed vector, the processed vector can be analyzed to determine the segmentation line feature points. In some examples, based on the processed vector, determining the segmentation line feature points may include: obtaining a threshold value corresponding to the processed vector; and determining all peak feature points greater than the threshold value as segmentation line feature points.

[0174] Among them, a threshold for analyzing and processing the processed vector is pre-configured. After the processed vector is obtained, the threshold corresponding to the processed vector can be obtained, and then all the peak feature points can be analyzed and compared with the threshold, and all the peak feature points greater than the threshold are determined as dividing line feature points, that is, the amplitude of the peak feature point used to determine the dividing line is greater than the amplitude of the peak feature point used to determine the gate line, thereby effectively ensuring the accuracy and reliability of the determination of the dividing line feature points.

[0175] In this embodiment, by determining the peak feature point among the image feature points, and then obtaining the confidence vector corresponding to the peak feature point, and performing low-frequency filtering on the confidence vector to obtain the processed vector, finally the segmentation line feature point is determined based on the processed vector. This not only ensures the accurate reliability of determining the segmentation line feature point, but also improves the accuracy of determining the segmentation line based on the segmentation line feature point, further ensuring the quality and efficiency of abnormality detection of photovoltaic module images.

[0176] Fig.17 A schematic diagram of a process for obtaining a photovoltaic module image according to an embodiment of the present invention; Fig.17 As shown, this embodiment provides an implementation method for acquiring a photovoltaic component image. Specifically, acquiring a photovoltaic component image in this embodiment may include:

[0177] Step S1701: Acquire an original image corresponding to a photovoltaic module.

[0178] Step S1702: Determine the invalid area in the original image.

[0179] Step S1703: remove the invalid area in the original image to obtain a photovoltaic module image.

[0180] When it is necessary to produce photovoltaic modules and products corresponding to the photovoltaic modules, in order to ensure the quality and efficiency of the production of photovoltaic modules and products corresponding to the photovoltaic modules, the original image corresponding to the photovoltaic modules can be obtained, and the original image can include the module image area corresponding to the photovoltaic module and other image areas located at the edge of the photovoltaic module. After the original image corresponding to the photovoltaic module is obtained, the original image can be analyzed and processed to determine the invalid area included in the original image, and the invalid area refers to the area that is not related to the photovoltaic module.

[0181] In some examples, determining the invalid area in the original image may include: identifying edge information included in the original image; determining the area formed by the edge information as the valid area in the original image; and determining the area outside the valid area as the invalid area.

[0182] Among them, after acquiring the original image, the original image can be analyzed and identified to identify the edge information included in the original image. Specifically, identifying the edge information included in the original image can include: sampling the original image to obtain multiple image blocks; processing the image blocks along the edge direction to obtain the mean vector corresponding to the image blocks; determining the differential vector based on the mean vector; based on the differential vector, determining the edge information included in the original image.

[0183] In order to accurately identify the edge information included in the original image, after the original image is acquired, the original image can be sampled and processed, so as to obtain multiple image blocks corresponding to the original image. The multiple image blocks can be part of the original image. After the image block is acquired, the image block can be processed along the edge direction to obtain the mean vector corresponding to the image block. Specifically, the specific implementation method of obtaining the mean vector corresponding to the image block in this embodiment is similar to the specific implementation method of obtaining the mean vector corresponding to each sub-image in the above embodiment. For details, please refer to the above statement, which will not be repeated here.

[0184] After obtaining the mean vector, the mean vector can be analyzed and processed to determine the difference vector, which can characterize the degree of change of the pixel points in the image block. After obtaining the difference vector, the edge information included in the original image can be determined based on the difference vector, which effectively ensures the accuracy and reliability of determining the edge information included in the original image.

[0185] After identifying the edge information included in the original image, the area formed by the edge information can be determined as the valid area in the original image, and the valid area refers to the area corresponding to the photovoltaic module; after obtaining the valid area, the area outside the valid area can be determined as the invalid area, which effectively ensures the accuracy and reliability of the determination of the invalid area.

[0186] like Fig.18 As shown, after determining the invalid area, in order to improve the quality and efficiency of abnormality detection on the photovoltaic module image, the invalid area in the original image can be removed, so that the photovoltaic module image can be obtained. The photovoltaic module image at this time refers to the image after the invalid area is removed. In this way, when analyzing and processing the photovoltaic module image, the quality and efficiency of abnormality detection are effectively improved.

[0187] In this embodiment, the original image corresponding to the photovoltaic module is acquired, and then the invalid area in the original image is determined, and then the invalid area in the original image is removed to obtain the photovoltaic module image. At this time, the photovoltaic module image obtained does not include areas unrelated to the photovoltaic module. Therefore, when the photovoltaic module image is analyzed and processed, the quality and efficiency of abnormality detection on the photovoltaic module image can be effectively guaranteed.

[0188] Fig.19 A flowchart of another abnormality detection method for photovoltaic module images provided by an embodiment of the present invention; based on any of the above embodiments, refer to the attached Fig.19 As shown, after acquiring the photovoltaic component image, the method in this embodiment may further include:

[0189] Step S1901: Identify whether there is an abnormal light area in the photovoltaic component image.

[0190] Step S1902: when there is an abnormal light area in the photovoltaic component image, it is prohibited to identify the image feature points of the photovoltaic component image.

[0191] Step S1903: when there is no abnormal light area in the photovoltaic component image, it is allowed to identify image feature points of the photovoltaic component image.

[0192] After the photovoltaic module image is acquired, the photovoltaic module image may be detected and processed to detect whether there is an abnormal light area in the photovoltaic module image that can affect the abnormal detection operation of the photovoltaic module image, and the above-mentioned abnormal light area may include: a bright light area, a dim light area, etc. In some examples, identifying whether there is an abnormal light area in the photovoltaic module image may include: acquiring a grayscale image corresponding to the photovoltaic module image; extracting pixel values ​​of pixels in the grayscale image; and detecting whether there is an abnormal light area in the photovoltaic module image that can affect the abnormal detection operation of the photovoltaic module image based on the pixel values ​​of the pixels in the grayscale image.

[0193] Specifically, after acquiring the photovoltaic module image, the photovoltaic module image can be converted from the RGB space to the grayscale space, so as to obtain a grayscale image corresponding to the photovoltaic module image, and then the pixel values ​​of the pixels in the grayscale image are extracted. After the pixel values ​​of the pixels in the grayscale image are extracted, the pixel values ​​of the pixels in the grayscale image can be analyzed and processed to detect whether there is an abnormal light area in the photovoltaic module image that can affect the abnormal detection operation of the photovoltaic module image.

[0194] In some instances, based on the pixel values ​​of the pixels in the grayscale image, detecting whether there is an abnormal light area in the photovoltaic component image that can affect the abnormal detection operation on the photovoltaic component image may include: when the pixel value is less than a first pixel threshold, or the first pixel value is greater than or equal to a second pixel threshold, determining the area corresponding to the pixel value as the abnormal light area, wherein the second pixel threshold is greater than the first pixel threshold; when the pixel value is greater than or equal to the first pixel threshold, and the first pixel value is less than the second pixel threshold, determining the area corresponding to the pixel value as the normal light area.

[0195] Specifically, after obtaining the pixel value of the pixel point in the grayscale image, the pixel value can be analyzed and compared with the preset first pixel threshold and the second pixel threshold respectively. When the pixel value is less than the first pixel threshold, it means that the brightness of the pixel point corresponding to the pixel value in the grayscale image is darker, that is, the photovoltaic component image at this time includes a dark light area corresponding to the above pixel point, and then the area corresponding to the pixel value can be determined as an abnormal light area. Similarly, when the pixel value is greater than or equal to the second pixel threshold, it means that the brightness of the pixel point corresponding to the pixel value in the grayscale image is brighter, that is, the photovoltaic component image at this time includes a high light area corresponding to the above pixel point, and then the area corresponding to the pixel value can be determined as an abnormal light area.

[0196] When the pixel value is greater than or equal to the first pixel threshold and the first pixel value is less than the second pixel threshold, it means that the brightness of the pixel point corresponding to the pixel value in the grayscale image is within the normal range, and the area corresponding to the pixel value can be determined as the normal light area. This effectively realizes the ability to stably and effectively detect whether there is an abnormal light area in the photovoltaic component image that can affect the abnormal detection operation of the photovoltaic component image.

[0197] Specifically, when there is an abnormal light area in the photovoltaic module image, it means that the photovoltaic module image at this time is an abnormal image, and there is no need to perform an abnormality detection operation on the photovoltaic module image again, so the recognition of the image feature points of the photovoltaic module image can be prohibited. When there is no abnormal light area in the photovoltaic module image, at this time, in order to accurately perform an abnormality detection operation on the photovoltaic module image, it is allowed to recognize the image feature points of the photovoltaic module image, so as to perform an abnormality detection operation on the photovoltaic module image based on the image feature points, further ensuring the accuracy and reliability of the analysis and recognition of the photovoltaic module image.

[0198] Fig. 20 A schematic diagram of a method for extracting feature points from a photovoltaic module image according to an embodiment of the present invention; Fig. 20 As shown, this embodiment provides a method for extracting feature points of a photovoltaic component image. The execution subject of the method can be a device for extracting feature points of a photovoltaic component image. It can be understood that the device for extracting feature points of a photovoltaic component image can be implemented as software, or a combination of software and hardware. Specifically, the device for extracting feature points of a photovoltaic component image can be implemented as a client or a server. The client or the server can be any computing device with a certain image processing capability. Specifically, the method for extracting feature points can include:

[0199] Step S2001: Acquire a photovoltaic module image.

[0200] Among them, the photovoltaic component image refers to an image that needs to perform feature point extraction operation. Specifically, the specific implementation method and implementation effect of obtaining the photovoltaic component image in this embodiment are similar to the specific implementation method and implementation effect of step S401 in the above embodiment. For details, please refer to the above statements and will not be repeated here.

[0201] Step S2002: Segment the photovoltaic module image based on rows and columns to obtain multiple sub-images.

[0202] Step S2003: Determine the local peak point of the sub-image.

[0203] The specific implementation method and implementation effect of steps S2002 to S2003 in this embodiment are the same as those described above. Figure 4 The specific implementation methods and implementation effects of the corresponding steps in the embodiment are similar. Please refer to the above statements for details, and will not be repeated here.

[0204] Step S2004: determining feature points of the sub-image based on the local peak points, wherein the feature points of the photovoltaic assembly image include feature points of multiple sub-images.

[0205] Specifically, in order to accurately obtain the image feature points of the photovoltaic module image, after obtaining the photovoltaic module image, the photovoltaic module image can be segmented based on rows and columns, so that a plurality of sub-images corresponding to the photovoltaic module image can be obtained, such as Figure 6 In other examples, the photovoltaic component image may not be segmented based on rows and columns, but may be segmented according to a set size, thereby obtaining a plurality of sub-images. It is understood that the set size may be the same as or different from the row size or column size included in the photovoltaic component image. This embodiment does not limit the specific value of the set size, and those skilled in the art may adjust or configure the set size according to specific application scenarios and design requirements.

[0206] For each sub-image, each sub-image is composed of several pixel points, and the feature point is only a part of the feature points included in the sub-image. In order to ensure the quality and efficiency of determining the image feature points of the sub-image, after obtaining multiple sub-images, each sub-image can be analyzed and processed to determine the local peak point of the sub-image, wherein the local peak point refers to a feature point whose amplitude is greater than the amplitude of other feature points in the preset area in the preset area, and one sub-image can correspond to multiple local peak points. After obtaining the local peak point, the image feature point of the sub-image can be determined. In some examples, the local peak point can be determined as the image feature point of the sub-image; since the image feature point of the photovoltaic module image is composed of the image feature points of multiple sub-images, after determining the image feature points of all sub-images, the image feature point of the photovoltaic module image can be accurately identified, and the image feature point is a pixel point whose amplitude is greater than or equal to the set threshold.

[0207] The feature point extraction method provided in the present embodiment obtains a photovoltaic module image, segments the photovoltaic module image based on rows and columns to obtain multiple sub-images, and then determines the local peak points of the sub-images, and determines the feature points of the sub-images based on the local peak points. Since the feature points of the photovoltaic module image include the feature points of multiple sub-images, the feature point extraction operation on the photovoltaic module image is effectively guaranteed, which facilitates the abnormality detection operation on the photovoltaic module image based on the extracted image feature points, and further improves the accuracy and reliability of the feature point extraction method.

[0208] In specific applications, this application embodiment provides a method for detecting abnormal shooting of photovoltaic module images based on feature point matching to solve the abnormal shooting problem in the process of splicing the battery array. This method can complete the operation of filtering out abnormal module images and ensure the normal operation of subsequent processes of the production line. Fig.21 As shown, the shooting anomaly detection method may include: removing invalid areas, overall component determination, feature point detection, feature point fusion, cell separation line detection, point-by-point matching, and post-processing. The following is a detailed description of each of the above steps:

[0209] Step 1: Remove invalid areas.

[0210] Generally, the central area of ​​the photovoltaic module image is the area corresponding to the cell array, while the edge area of ​​the photovoltaic module image is the black edge and some text information retained during image acquisition. In order to improve the quality and efficiency of the shooting anomaly detection method, the outer frame of the cell array can be detected, and the invalid external area can be removed to retain only the valid cell area. The specific steps include the following:

[0211] Step 1.1: Sample several image blocks at the edge of the PV module image.

[0212] Reference Fig. 22 As shown, the edge of the photovoltaic module image may include a horizontal edge and a vertical edge, and a plurality of image blocks are sampled on the horizontal edge and the vertical edge, respectively. Specifically, a plurality of first image blocks may be sampled on the horizontal edge, and a plurality of second image blocks may be sampled on the vertical edge. In some examples, the density of the image blocks sampled on the horizontal edge may be the same as or different from the density of the image blocks sampled on the vertical edge, such as Fig. 22 As shown, the density of image blocks collected on the horizontal edge is relatively large, and the spacing between the first image blocks collected is relatively small; the density of image blocks collected on the vertical edge is relatively small, and the spacing between the second image blocks collected is relatively large.

[0213] Step 1.2: For each sampled image block, calculate the mean vector along the edge direction.

[0214] After obtaining a plurality of first image blocks and a plurality of second image blocks, each first image block and each second image block can be compressed, so as to obtain a vertical line corresponding to the first image block and a horizontal line corresponding to the second image block; after obtaining the vertical line and the horizontal line, the mean vectors corresponding to the vertical line and the horizontal line are calculated respectively.

[0215] Step 1.3: Calculate the difference vector for the mean vector.

[0216] In order to determine the degree of change between adjacent image blocks, after the mean vector is obtained, a difference calculation may be performed on the mean vector, thereby obtaining a difference vector.

[0217] Step 1.4: Determine the edge position based on the transition point in the differential vector.

[0218] After obtaining the differential vector, the jump point can be found based on the differential vector. The jump point is the pixel point corresponding to the process of changing from black to light or from light to black. Generally, when the differential vector is greater than the set threshold, the second point corresponding to the differential vector can be determined as the jump point.

[0219] After the transition points in the differential vector are obtained, the edge positions may be determined based on the transition points. In some examples, the median of the transition points may be obtained and determined as the edge positions.

[0220] Step 1.5: All detected edge positions are merged to determine the outer frame of the battery array, and then the outer area is removed.

[0221] Step 2: Overall component determination.

[0222] Among them, the overall judgment of the component is to detect whether there are two abnormal conditions, too dark and overexposed, in the photovoltaic component image. Specifically, after obtaining the effective area of ​​the photovoltaic component image, the grayscale value in the photovoltaic component image can be obtained, and the number of pixels with a grayscale value greater than a first threshold, or the number of pixels with a grayscale value less than a second threshold is counted, wherein the first threshold is greater than the second threshold. If there are a large number of too dark pixels or overexposed pixels in the photovoltaic component image, it means that there is an abnormal condition of too dark or overexposed in the photovoltaic component image.

[0223] When it is determined that there is an abnormal situation of being too dark or overexposed in the photovoltaic component image, there is no need to perform abnormal detection operation on the photovoltaic component image; when it is determined that there is no abnormal situation of being too dark or overexposed in the photovoltaic component image, the abnormal detection operation can continue to be performed on the photovoltaic component image.

[0224] Step 3: Feature point detection.

[0225] For the photovoltaic module image after the overall determination of the module, various linear feature points included in the photovoltaic module image can be extracted, and then the feature points can be fused to obtain the effective line segments in the photovoltaic module image, and then the dividing lines between the cells and the grid lines corresponding to the cells can be determined through the effective line segments in the photovoltaic module image. Specifically, this step can include the following sub-steps:

[0226] Step 3.1: The photovoltaic module image is cut into rows and columns to be divided into several sub-images.

[0227] After acquiring the photovoltaic module image, the photovoltaic module image can be cut according to the column, so as to obtain several sub-images, such as Fig.23 As shown, a schematic diagram of a sub-image obtained by longitudinal slicing is provided.

[0228] Step 3.2: Calculate the mean vector of the sub-image along the direction perpendicular to the image edge.

[0229] Specifically, after obtaining the sub-images, the mean of each sub-image can be calculated along the direction perpendicular to the edge of the image, so that the mean vector corresponding to the sub-image can be obtained, such as Fig.24 shown.

[0230] Step 3.3: Use a median filter to process the mean vector to obtain the low-frequency component.

[0231] Step 3.4: Subtract the low-frequency component from the mean vector and retain its high-frequency component.

[0232] Specifically, after determining the low-frequency component, the low-frequency component can be subtracted from the mean vector to obtain the high-frequency component, such as Fig.25As shown, this can effectively remove the interference of the brightness difference of the solar cells in the photovoltaic module image, ensuring the accuracy and reliability of abnormality detection of the photovoltaic module image.

[0233] Step 3.5: Use the peak detection module and the fast NMS 1D (one-dimensional) module to extract image feature points.

[0234] The peak detection module is used to detect the downward peak points in the input vector. Specifically, the peak point acquisition step includes the following steps: Step 3.511: Setting the jump point threshold. Step 3.512: Analyze and calculate the high-frequency component to obtain the first-order forward difference vector and the first-order backward difference vector. According to the positive and negative properties of the two vectors, all local minimum points corresponding to the high-frequency component are obtained, such as Fig.10 As shown. Step 3.513: Sort all local minima. Step 3.514: Calculate the second-order difference vector for the sorted local minima, as Fig.11 Step 3.515: Determine the jump position based on the second-order difference vector, and trace back the jump position in the difference vector to find the local minimum point. The local minimum point found corresponds to the local minimum point with the smallest peak value in the entire mean vector. Then, use this point as the reference threshold to obtain all local minimum points with peak values ​​greater than this point, and determine the above local minimum points as peak points, thereby completing the peak detection operation, as shown in FIG. Fig.12 As shown, the peak points determined above are the initial image feature points.

[0235] In addition, the fast NMS 1D module is used to determine the confidence based on a vector (which can be a mean vector, a first-order difference vector, a second-order difference vector, etc.) and each element (that is, the initial image feature point mentioned above), specifically based on the Euclidean distance, determine the redundant elements based on the confidence, and filter out the redundant elements. Specifically, filtering out the redundant elements may include the following steps: Step 3.521: Set the Euclidean distance threshold. Step 3.522: Sort the confidence of each element (initial image feature point) from high to low. Step 3.523: Set an identification information for each element, and the identification information is used to record whether the element is filtered out. Step 3.524: According to the sorting order, take out each element x in turn, and search for other elements y on both sides, which are divided into the following cases: a) When the identification information of element x has been set to filter out, skip x and take the next element according to the confidence level; b) When the Euclidean distance between x and y is within the threshold, set the identification information of y to filter out and search for the next y, that is, when the distance between two elements is close, one of the elements can be determined as a redundant element; c) When the Euclidean distance between x and y is outside the threshold, take the next x. Step 3.525: Filter out redundant elements based on the identification information of all the elements at the end, so that the image feature points corresponding to the photovoltaic module image can be obtained, such as Fig.26 shown.

[0236] Step 4: Feature point fusion

[0237] Step 4.1: After obtaining the image feature points, the image feature points can be grouped according to rows and columns.

[0238] Step 4.2: Calculate the confidence of the grouped image feature points.

[0239] After obtaining the image feature points, the average Euclidean distance between the image feature points and other points in the neighborhood can be obtained, and the number of image feature points in the neighborhood can be counted, and the confidence can be calculated based on the average Euclidean distance and the number of image feature points in the neighborhood.

[0240] Step 4.3: Retain image feature points whose number of feature points in the neighborhood exceeds the threshold.

[0241] Step 4.4: Use the fast NMS 1D module to filter out redundant feature points to complete the feature point fusion operation.

[0242] Step 5: After acquiring the image feature points, the grid lines and dividing lines in the photovoltaic module image can be determined based on the image feature points. In the process of determining the dividing lines, the cell dividing line detection module can be used to analyze the relationship between the grid lines and dividing lines of the cells in the photovoltaic module image to detect the horizontal cell dividing lines and filter out the interference of the grid lines. Specifically, the dividing line recognition steps include:

[0243] Step 5.1: Calculate the PV module image along the horizontal direction to obtain the mean vector, such as Fig. 27 shown.

[0244] Step 5.2: Use median filtering to calculate the low-frequency part of the mean vector.

[0245] Step 5.3: Subtract the low-frequency part from the mean vector and keep the high-frequency part, such as Fig.28 shown.

[0246] Step 5.4: Use the spike detection module and the fast NMS 1D module to extract the spike feature points, such as Fig.28 shown.

[0247] Step 5.5: Calculate the confidence for each peak feature point. The confidence is the maximum difference between it and all points between two adjacent peaks. Specifically, the distribution of confidence is as follows: Fig.29 shown.

[0248] Step 5.6: Keep the high-frequency part of the confidence vector, such as Fig.30 shown.

[0249] Step 5.7: Determine the position of the cell separation line according to the set threshold, such as Fig.30 shown.

[0250] Step 6: Match point by point.

[0251] After obtaining the image feature points, grid lines and dividing lines, the image feature points can be matched with the grid lines and dividing lines point by point to detect whether there are serious or subtle abnormalities in the photovoltaic module image. Specifically, point by point matching is to match the feature points with the grid lines, dividing lines and other features in the component image. If the matching fails, it means that there is an abnormality in the photovoltaic module image, such as Fig.31 As shown, two common abnormal situations in photovoltaic module images are provided, that is, the three image feature points located at the top are used to identify the misalignment of the battery cells in the photovoltaic module image, and the dense image feature points located at the bottom are used to identify the misalignment of the grid lines in the battery cells.

[0252] Step 7: Post-processing.

[0253] For a photovoltaic module image without abnormalities, the bounding box coordinates of each cell in the photovoltaic module image as well as the separation lines, grid line coordinates, cell row and column numbers and other information can be output.

[0254] The technical solution provided by the present application embodiment uses a peak detection module and a fast NMS 1D module to perform feature extraction operations on the photovoltaic module image, thereby obtaining rich image feature points, which can characterize relatively complete features in the photovoltaic module image, and then identify the grid lines and dividing lines in the photovoltaic module image based on the extracted image feature points, and perform point-by-point matching operations on the image feature points, grid lines and dividing lines, so that not only more serious shooting anomalies can be detected, but also extremely slight anomalies can be detected, thereby ensuring the accuracy of anomaly detection for photovoltaic module images, thereby overcoming anomalies that are easily missed due to manual observation, such as grid line misalignment and other anomalies that are also easily missed by manual observation, etc.; in addition, the method can quickly adapt to new photovoltaic modules by simply adjusting the algorithm parameters, and the update cycle is basically kept within 10 minutes, with a fast response speed, further improving the flexibility and reliability of the method.

[0255] Fig.32 A schematic diagram of a photovoltaic module image abnormality detection device provided by an embodiment of the present invention; Fig.32 As shown, this embodiment provides a photovoltaic module image abnormality detection device, which can perform the above Figure 4 The abnormality detection method of the corresponding photovoltaic component image. Specifically, the abnormality detection device may include a first acquisition module 11, a first recognition module 12, a first processing module 13 and a first detection module 14:

[0256] The first acquisition module 11 is used to acquire the photovoltaic component image.

[0257] The first recognition module 12 is used to recognize image feature points of the photovoltaic component image.

[0258] The first processing module 13 is used to identify grid lines and dividing lines in the photovoltaic module image based on image feature points.

[0259] The first detection module 14 is used to perform abnormality detection on the photovoltaic module image based on image feature points, grid lines and dividing lines.

[0260] In some instances, when the first recognition module 12 recognizes image feature points of a photovoltaic component image, the first recognition module 12 is used to perform: segmenting the photovoltaic component image based on rows and columns to obtain multiple sub-images; determining local peak points of the sub-images; and determining image feature points of the sub-image based on the local peak points, wherein the image feature points of the photovoltaic component image include image feature points of multiple sub-images.

[0261] In some instances, when the first recognition module 12 determines the local peak point of the sub-image, the first recognition module 12 is used to perform: obtaining a mean vector corresponding to each sub-image along a direction perpendicular to the edge of the image; performing a low-frequency filtering on the mean vector to obtain a high-frequency component; and determining the local peak point of the sub-image based on the high-frequency component.

[0262] In some instances, when the first recognition module 12 determines the local peak point of the sub-image based on the high-frequency component, the first recognition module 12 is used to perform: determining a first-order difference vector based on the high-frequency component; determining an initial peak point of the sub-image based on the first-order difference vector, the amplitude of the initial peak point being greater than the amplitude of the adjacent feature points; and determining the local peak point of the sub-image among the initial peak points.

[0263] In some instances, when the first recognition module 12 determines the local peak points of the sub-image among the initial peak points, the first recognition module 12 is used to perform the following operations: sorting the initial peak points; determining the second-order difference vectors corresponding to the sorted initial peak points; determining the jump positions included in the initial peak points based on the second-order difference vectors; determining the initial peak points corresponding to the jump positions as reference feature points; and determining the initial peak points having amplitudes greater than the reference feature points as the local peak points of the sub-image.

[0264] In some examples, when the first recognition module 12 determines the image feature points of the sub-image based on the local peak points, the first recognition module 12 is used to perform: identifying redundant feature points included in the local peak points; deleting the redundant feature points in the local peak points to obtain the image feature points of the sub-image.

[0265] In some examples, when the first recognition module 12 recognizes redundant feature points included in the local peak points, the first recognition module 12 is used to perform: obtaining the confidence of the local peak points; and based on the confidence, identifying the redundant feature points included in the local peak points.

[0266] In some examples, when the first recognition module 12 obtains the confidence of a local peak point, the first recognition module 12 is used to perform: obtaining adjacent distances corresponding to all adjacent local peak points; determining an average distance based on all adjacent distances; and determining the confidence of the local peak point based on the average distance and the adjacent distance.

[0267] In some instances, when the first recognition module 12 determines the confidence of the local peak point based on the average distance and the adjacent distance, the first recognition module 12 is used to perform: obtaining the distance difference between the average distance and the adjacent distance; and determining the confidence of the local peak point based on the distance difference, and the confidence is negatively correlated with the size of the distance difference.

[0268] In some instances, when the first recognition module 12 obtains the confidence of a local peak point, the first recognition module 12 is used to perform: determining a neighborhood corresponding to the local peak point; counting the number of local peak points in the neighborhood; and determining the confidence of the local peak point based on the number, wherein the confidence is positively correlated with the number.

[0269] In some instances, when the first recognition module 12 identifies redundant feature points included in local peak points based on confidence, the first recognition module 12 is used to perform: when the confidence is less than a preset threshold, the local peak point corresponding to the confidence is determined as a redundant feature point; when the confidence is greater than or equal to the preset threshold, the local peak point corresponding to the confidence is determined as a non-redundant feature point.

[0270] In some instances, when the first processing module 13 identifies grid lines and dividing lines in a photovoltaic component image based on image feature points, the first processing module 13 is used to perform: determining dividing line feature points among the image feature points, wherein, in a local area, the amplitude of the dividing line feature points is greater than the amplitudes of other image feature points; determining grid lines in the photovoltaic component image based on the image feature points; and determining dividing lines in the photovoltaic component image based on the dividing line feature points.

[0271] In some instances, when the first processing module 13 determines the segmentation line feature points among the image feature points, the first processing module 13 is used to execute: determining the peak feature points among the image feature points; obtaining the confidence vector corresponding to the peak feature point; performing low-frequency filtering on the confidence vector to obtain a processed vector; and determining the segmentation line feature points based on the processed vector.

[0272] In some instances, when the first processing module 13 obtains a confidence vector corresponding to a peak feature point, the first processing module 13 is used to execute: obtaining all intermediate feature points between the peak feature point and adjacent peak feature points; determining the maximum difference between the peak feature point and all intermediate feature points; and determining the confidence vector corresponding to the peak feature point based on the maximum difference.

[0273] In some examples, when the first processing module 13 determines the segmentation line feature points based on the processed vector, the first processing module 13 is used to execute: obtaining a threshold value corresponding to the processed vector; and determining all peak feature points greater than the threshold value as segmentation line feature points.

[0274] In some instances, when the first detection module 14 performs abnormality detection on the photovoltaic component image based on the image feature points, grid lines and dividing lines, the first detection module 14 is used to perform the following: identifying whether the image feature points match the grid lines, and whether the image feature points match the dividing lines; when the image feature points match the grid lines, and the image feature points match the dividing lines, it is determined that there is no abnormality in the photovoltaic component image; when the image feature points do not match the grid lines, or the image feature points do not match the dividing lines, it is determined that there is an abnormality in the photovoltaic component image.

[0275] In some instances, the first processing module 13 in this embodiment is also used to execute: when it is determined that there is an abnormality in the photovoltaic component image, prohibiting production operations based on the photovoltaic component image; when it is determined that there is no abnormality in the photovoltaic component image, allowing production operations based on the photovoltaic component image.

[0276] In some examples, when the first acquisition module 11 acquires a photovoltaic component image, the first acquisition module 11 is used to execute: acquiring an original image corresponding to the photovoltaic component; determining an invalid area in the original image; and removing the invalid area in the original image to obtain the photovoltaic component image.

[0277] In some examples, when the first acquisition module 11 determines an invalid area in the original image, the first acquisition module 11 is used to perform: identifying edge information included in the original image; determining the area formed by the edge information as a valid area in the original image; and determining the area outside the valid area as an invalid area.

[0278] In some instances, when the first acquisition module 11 identifies edge information included in the original image, the first acquisition module 11 is used to perform: sampling the original image to obtain multiple image blocks; processing the image blocks along the edge direction to obtain mean vectors corresponding to the image blocks; determining differential vectors based on the mean vectors; and determining edge information included in the original image based on the differential vectors.

[0279] In some examples, after acquiring the photovoltaic assembly image, the first recognition module 12 and the first processing module 13 in this embodiment are used to perform the following steps:

[0280] The first recognition module 12 is used to recognize whether there is an abnormal light area in the photovoltaic module image;

[0281] The first processing module 13 is used to prohibit the recognition of image feature points of the photovoltaic component image when there is an abnormal light area in the photovoltaic component image; and to allow the recognition of image feature points of the photovoltaic component image when there is no abnormal light area in the photovoltaic component image.

[0282] Fig.32 The device shown can perform Figure 4-Figure 19, Figure 21-Figure 31 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Figure 4-Figure 19 , Figure 21-Figure 31 The implementation process and technical effects of this technical solution refer to Figure 4-Figure 19 , Figure 21-Figure 31 The description in the illustrated embodiment will not be repeated here.

[0283] In one possible design, Fig.32 The structure of the photovoltaic module image abnormality detection device shown can be implemented as an electronic device, which can be a mobile phone, a tablet computer, a server, or other devices. Fig.33 As shown, the electronic device may include: a first processor 21 and a first memory 22. The first memory 22 is used to store the corresponding electronic device to execute the above Figure 4-Figure 19 , Figure 21-Figure 31 In the illustrated embodiment, the first processor 21 is configured to execute the program stored in the first memory 22 .

[0284] The program includes one or more computer instructions, wherein the one or more computer instructions, when executed by the first processor 21, can implement the following steps: acquiring a photovoltaic module image; identifying image feature points of the photovoltaic module image; identifying grid lines and dividing lines in the photovoltaic module image based on the image feature points; and performing anomaly detection on the photovoltaic module image based on the image feature points, grid lines and dividing lines.

[0285] Furthermore, the first processor 21 is also used to execute the aforementioned Figure 4-Figure 19 , Figure 21-Figure 31 All or part of the steps in the illustrated embodiments.

[0286] The structure of the electronic device may further include a first communication interface 23 for the electronic device to communicate with other devices or a communication network.

[0287] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by electronic devices, which includes instructions for executing the above Figure 4-Figure 19 , Figure 21-Figure 31 The procedures involved in the method for detecting abnormalities in photovoltaic component images in the method embodiment are shown.

[0288] Fig.34 A schematic diagram of a feature point extraction device for a photovoltaic module image provided by an embodiment of the present invention; Fig.34 As shown, this embodiment provides a feature point extraction device for a photovoltaic module image, and the feature point extraction device can perform the above Fig. 20The feature point extraction method corresponding to the photovoltaic assembly image. Specifically, the feature point extraction device may include a second acquisition module 31, a second segmentation module 32, a second determination module 33 and a second processing module 34:

[0289] A second acquisition module 31 is used to acquire a photovoltaic module image;

[0290] A second segmentation module 32 is used to segment the photovoltaic module image based on rows and columns to obtain multiple sub-images;

[0291] A second determination module 33, used to determine a local peak point of a sub-image;

[0292] The second processing module 34 is used to determine the feature points of the sub-image based on the local peak point, and the feature points of the photovoltaic assembly image include the feature points of multiple sub-images.

[0293] Fig.34 The device shown can perform Fig. 20 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Fig. 20 The implementation process and technical effects of this technical solution refer to Fig. 20 The description in the illustrated embodiment will not be repeated here.

[0294] In one possible design, Fig.34 The structure of the feature point extraction device of the photovoltaic module image shown can be implemented as an electronic device, which can be a mobile phone, a tablet computer, a server, etc. Fig.35 As shown, the electronic device may include: a second processor 41 and a second memory 42. The second memory 42 is used to store the corresponding electronic device executing the above Fig. 20 In the program of the feature point extraction method of the photovoltaic component image provided in the illustrated embodiment, the second processor 41 is configured to execute the program stored in the second memory 42. The program includes one or more computer instructions, wherein the one or more computer instructions can implement the following steps when executed by the second processor 41: acquiring a photovoltaic component image; segmenting the photovoltaic component image based on rows and columns to obtain multiple sub-images; determining local peak points of the sub-images; determining feature points of the sub-image based on the local peak points, wherein the feature points of the photovoltaic component image include feature points of multiple sub-images.

[0295] Furthermore, the second processor 41 is also used to execute the aforementioned Fig. 20 All or part of the steps in the illustrated embodiments.

[0296] The structure of the electronic device may further include a second communication interface 43 for the electronic device to communicate with other devices or a communication network.

[0297] The embodiment of the present invention provides a computer storage medium for storing computer software instructions used by electronic devices, which includes instructions for executing the above Fig. 20 The procedures involved in the feature point extraction method of the photovoltaic component image in the method embodiment are shown.

[0298] Fig.36 A schematic diagram of a flow chart of another abnormality detection method for photovoltaic module images provided by an embodiment of the present invention; Fig.36 As shown, this embodiment provides another abnormality detection method for photovoltaic component images. The execution subject of the method may be an abnormality detection device for photovoltaic component images. It can be understood that the abnormality detection device for photovoltaic component images can be implemented as software, or a combination of software and hardware. Specifically, the abnormality detection method for photovoltaic component images may include:

[0299] Step S3601: In response to a request for invoking a photovoltaic assembly image anomaly detection service, a processing resource corresponding to the photovoltaic assembly image anomaly detection service is determined.

[0300] Step S3602: Utilize processing resources to perform the following steps: obtain a photovoltaic module image; identify image feature points of the photovoltaic module image; identify grid lines and dividing lines in the photovoltaic module image based on the image feature points; and perform anomaly detection on the photovoltaic module image based on the image feature points, grid lines and dividing lines.

[0301] Specifically, the photovoltaic module image anomaly detection method provided by the present invention can be executed in the cloud, and a number of computing nodes can be deployed in the cloud, each of which has computing, storage and other processing resources. In the cloud, multiple computing nodes can be organized to provide a certain service, and of course, a computing node can also provide one or more services.

[0302] With respect to the solution provided by the present invention, the cloud can provide a service for completing the anomaly detection method of photovoltaic component images, which is called the anomaly detection service of photovoltaic component images. When a user needs to use the anomaly detection service of the photovoltaic component image, the anomaly detection service of the photovoltaic component image is called to trigger a request to the cloud to call the anomaly detection service of the photovoltaic component image, and the request can carry the photovoltaic component image to be detected. The cloud determines the computing node that responds to the request, and uses the processing resources in the computing node to perform the following steps: acquiring the photovoltaic component image; identifying the image feature points of the photovoltaic component image; identifying the grid lines and dividing lines in the photovoltaic component image based on the image feature points; and performing anomaly detection on the photovoltaic component image according to the image feature points, grid lines and dividing lines.

[0303] Specifically, the implementation process, implementation principle and implementation effect of the above method steps in this embodiment are the same as those in the above Figure 4-Figure 19 , Figure 21-Figure 31 The implementation process, implementation principle and implementation effect of the method steps in the embodiment shown are similar. For the parts not described in detail in this embodiment, please refer to Figure 4-Figure 19 , Figure 21-Figure 31 Description of the illustrated embodiment.

[0304] Fig.37 A schematic diagram of a photovoltaic module image abnormality detection device provided by an embodiment of the present invention; Fig.37 As shown, this embodiment provides a photovoltaic component image abnormality detection device, which can perform the above-mentioned Fig.36 The abnormality detection method of the photovoltaic component image shown in the figure, specifically, the abnormality detection device may include: a third determination module 51 and a third processing module 52; specifically,

[0305] A third determination module 51 is used to determine the processing resources corresponding to the photovoltaic component image anomaly detection service in response to a request for invoking the photovoltaic component image anomaly detection service;

[0306] The third processing module 52 is used to use processing resources to perform the following steps: obtain a photovoltaic module image; identify image feature points of the photovoltaic module image; identify grid lines and dividing lines in the photovoltaic module image based on the image feature points; and perform anomaly detection on the photovoltaic module image based on the image feature points, grid lines and dividing lines.

[0307] Fig.37 The device shown can perform Fig.36 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Fig.36 The implementation process and technical effects of this technical solution refer to Fig.36 The description in the illustrated embodiment will not be repeated here.

[0308] In one possible design, Fig.37 The structure of the photovoltaic module image abnormality detection device shown can be implemented as an electronic device, which can be a mobile phone, a tablet computer, a server, or other devices. Fig.38 As shown, the electronic device may include: a third processor 61 and a third memory 62. The third memory 62 is used to store the corresponding electronic device executing the above Fig.36 In the illustrated embodiment, the third processor 61 is configured to execute the program stored in the third memory 62 .

[0309] The program includes one or more computer instructions, wherein when the one or more computer instructions are executed by the third processor 61, the following steps can be implemented:

[0310] In response to a request for invoking an abnormality detection service for a photovoltaic assembly image, a processing resource corresponding to the abnormality detection service for a photovoltaic assembly image is determined.

[0311] The processing resources are used to perform the following steps: acquiring a photovoltaic module image; identifying image feature points of the photovoltaic module image; identifying grid lines and segmentation lines in the photovoltaic module image based on the image feature points; and performing anomaly detection on the photovoltaic module image based on the image feature points, grid lines and segmentation lines.

[0312] Furthermore, the third processor 61 is also used to execute the aforementioned Fig.36 All or part of the steps in the illustrated embodiments.

[0313] The structure of the electronic device may further include a third communication interface 63 for the electronic device to communicate with other devices or a communication network.

[0314] In addition, an embodiment of the present invention provides a computer storage medium for storing computer software instructions used by an electronic device, which includes instructions for executing Fig.36 The procedures involved in the method for detecting abnormalities in photovoltaic component images in the method embodiment are shown.

[0315] Fig.39 A schematic diagram of a flow chart of another method for extracting feature points of a photovoltaic module image provided by an embodiment of the present invention; Fig.39 As shown, this embodiment provides another method for extracting feature points of a photovoltaic component image. The execution subject of the method may be a device for extracting feature points of a photovoltaic component image. It can be understood that the device for extracting feature points of a photovoltaic component image may be implemented as software, or a combination of software and hardware. Specifically, the method for extracting feature points of a photovoltaic component image may include:

[0316] Step S3901: In response to a request for invoking a feature point extraction service for a photovoltaic assembly image, determining processing resources corresponding to the feature point extraction service for the photovoltaic assembly image.

[0317] Step S3902: Utilize processing resources to perform the following steps: obtain a photovoltaic module image; segment the photovoltaic module image based on rows and columns to obtain multiple sub-images; determine local peak points of the sub-images; and determine feature points of the sub-images based on the local peak points, wherein the feature points of the photovoltaic module image include feature points of multiple sub-images.

[0318] Specifically, the feature point extraction method of the photovoltaic module image provided by the present invention can be executed in the cloud, and a number of computing nodes can be deployed in the cloud, each of which has processing resources such as computing and storage. In the cloud, multiple computing nodes can be organized to provide a certain service, and of course, a computing node can also provide one or more services.

[0319] With respect to the solution provided by the present invention, the cloud can provide a service for completing the feature point extraction method of the photovoltaic component image, which is called the feature point extraction service of the photovoltaic component image. When the user needs to use the feature point extraction service of the photovoltaic component image, the feature point extraction service of the photovoltaic component image is called to trigger a request to the cloud to call the feature point extraction service of the photovoltaic component image, and the request can carry the photovoltaic component image to be detected. The cloud determines the computing node that responds to the request, and uses the processing resources in the computing node to perform the following steps: obtaining the photovoltaic component image; segmenting the photovoltaic component image based on rows and columns to obtain multiple sub-images; determining the local peak points of the sub-images; and determining the feature points of the sub-image based on the local peak points, wherein the feature points of the photovoltaic component image include the feature points of multiple sub-images.

[0320] Specifically, the implementation process, implementation principle and implementation effect of the above method steps in this embodiment are the same as those in the above Figure 20-Figure 31 The implementation process, implementation principle and implementation effect of the method steps in the embodiment shown are similar. For the parts not described in detail in this embodiment, please refer to Figure 20-Figure 31 Description of the illustrated embodiment.

[0321] Fig.40 A schematic diagram of a feature point extraction device for a photovoltaic module image provided by an embodiment of the present invention; Fig.40 As shown, this embodiment provides a feature point extraction device for a photovoltaic component image, and the feature point extraction device for a photovoltaic component image can perform the above Fig.39 The feature point extraction method of the photovoltaic component image shown in the figure, specifically, the feature point extraction device may include: a fourth determination module 71 and a fourth processing module 72; specifically,

[0322] A fourth determination module 71 is used to determine the processing resources corresponding to the photovoltaic assembly image feature point extraction service in response to a request for invoking the photovoltaic assembly image feature point extraction service;

[0323] The fourth processing module 72 is used to use processing resources to perform the following steps: obtain a photovoltaic module image; segment the photovoltaic module image based on rows and columns to obtain multiple sub-images; determine the local peak points of the sub-images; based on the local peak points, determine the feature points of the sub-image, and the feature points of the photovoltaic module image include the feature points of multiple sub-images.

[0324] Fig.40 The device shown can perform Fig.39 For the method of the embodiment shown in the figure, the part not described in detail in this embodiment can be referred to Fig.39 The implementation process and technical effects of this technical solution refer to Fig.39The description in the illustrated embodiment will not be repeated here.

[0325] In one possible design, Fig.40 The structure of the feature point extraction device of the photovoltaic module image shown can be implemented as an electronic device, which can be a mobile phone, a tablet computer, a server, etc. Fig.41 As shown, the electronic device may include: a fourth processor 81 and a fourth memory 82. The fourth memory 82 is used to store the corresponding electronic device executing the above Fig.39 The fourth processor 81 is configured to execute the program stored in the fourth memory 82 for the method for extracting feature points from a photovoltaic assembly image provided in the illustrated embodiment.

[0326] The program includes one or more computer instructions, wherein when the one or more computer instructions are executed by the fourth processor 81, the following steps can be implemented:

[0327] In response to a request for invoking a feature point extraction service for a photovoltaic assembly image, determining a processing resource corresponding to the feature point extraction service for the photovoltaic assembly image;

[0328] The processing resources are used to perform the following steps: acquiring a photovoltaic module image; segmenting the photovoltaic module image based on rows and columns to obtain multiple sub-images; determining local peak points of the sub-images; and determining feature points of the sub-images based on the local peak points, wherein the feature points of the photovoltaic module image include feature points of multiple sub-images.

[0329] Furthermore, the fourth processor 81 is also used to execute the aforementioned Fig.39 All or part of the steps in the illustrated embodiments.

[0330] The structure of the electronic device may further include a fourth communication interface 83 for the electronic device to communicate with other devices or a communication network.

[0331] The embodiment of the present invention provides a computer storage medium for storing computer software instructions used by electronic devices, which includes instructions for executing the above Fig.39 The procedures involved in the feature point extraction method of the photovoltaic component image in the method embodiment are shown.

[0332] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0333] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by adding a necessary general hardware platform, and of course can also be implemented by combining hardware and software. Based on such an understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a computer product, and the present invention can be in the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0334] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable device to produce a machine, so that the instructions executed by the processor of the computer or other programmable device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0335] These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable device to work in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0336] These computer program instructions may also be loaded onto a computer or other programmable device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for executing on the computer or other programmable device to implement the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0337] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0338] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0339] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0340] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting anomalies in photovoltaic module images, characterized in that: include: Acquire images of photovoltaic modules; Identifying image feature points of the photovoltaic assembly image; Identify grid lines and dividing lines in the photovoltaic assembly image based on the image feature points; Performing abnormality detection on the photovoltaic module image according to the image feature points, grid lines and segmentation lines; Wherein, identifying the image feature points of the photovoltaic assembly image includes: Segmenting the photovoltaic module image based on rows and columns to obtain a plurality of sub-images; Determine a local peak point of the sub-image, wherein the local peak point is a feature point whose amplitude is greater than that of other feature points in a preset area; Based on the local peak point, the image feature points of the sub-image are determined, and the image feature points of the photovoltaic assembly image include image feature points of multiple sub-images.

2. The method according to claim 1, characterized in that Determining a local peak point of the sub-image includes: Obtain the mean vector corresponding to each sub-image along the direction perpendicular to the edge of the image; Performing low-frequency filtering on the mean vector to obtain a high-frequency component; Based on the high frequency component, a local peak point of the sub-image is determined.

3. The method according to claim 2, characterized in that Determining a local peak point of the sub-image based on the high-frequency component includes: Based on the high frequency component, determining a first order difference vector; Based on the first-order difference vector, determining an initial peak point of the sub-image, wherein the amplitude of the initial peak point is greater than the amplitude of an adjacent feature point; Among the initial peak points, local peak points of the sub-image are determined.

4. The method according to claim 1, characterized in that: Determining the image feature points of the sub-image based on the local peak point includes: Identifying redundant feature points included in the local peak points; The redundant feature points in the local peak points are deleted to obtain the image feature points of the sub-image.

5. The method according to claim 4, characterized in that Identifying redundant feature points included in the local peak points includes: Obtaining the confidence of the local peak point; Based on the confidence level, redundant feature points included in the local peak points are identified.

6. The method according to claim 5, characterized in that Obtaining the confidence of the local peak point includes: Obtain the adjacent distances corresponding to all adjacent local peak points; Based on all neighboring distances, determine the average distance; The confidence of the local peak point is determined based on the average distance and the adjacent distances.

7. The method according to claim 5, characterized in that Obtaining the confidence of the local peak point includes: Determine a neighborhood corresponding to the local peak point; Counting the number of local peak points within the neighborhood; Based on the quantity, the confidence of the local peak point is determined, wherein the confidence is positively correlated with the quantity.

8. The method according to claim 1, characterized in that Identifying grid lines and dividing lines in the photovoltaic assembly image based on the image feature points includes: Determine a segmentation line feature point among the image feature points, wherein, in a local area, the amplitude of the segmentation line feature point is greater than the amplitude of other image feature points; Based on the image feature points, determining grid lines in the photovoltaic assembly image; Based on the segmentation line feature points, a segmentation line in the photovoltaic assembly image is determined.

9. The method according to claim 8, characterized in that Determining the segmentation line feature points among the image feature points includes: Determine a peak feature point among the image feature points, wherein the peak feature point is a feature point having a peak characteristic determined after analyzing and comparing the image feature points; Obtaining a confidence vector corresponding to the peak feature point; Performing low-frequency filtering on the confidence vector to obtain a processed vector; Based on the processed vector, the segmentation line feature point is determined.

10. The method according to claim 9, characterized in that Obtaining a confidence vector corresponding to the peak feature point, including: Acquire all intermediate feature points between the peak feature point and adjacent peak feature points; Determine the maximum difference between the peak feature point and all the intermediate feature points; Based on the maximum difference, a confidence vector corresponding to the peak feature point is determined.

11. The method according to claim 9, characterized in that Determining the segmentation line feature points based on the processed vectors includes: Obtaining a threshold value corresponding to the processed vector; All peak feature points greater than the threshold are determined as the segmentation line feature points.

12. The method according to claim 1, characterized in that The photovoltaic module image is subjected to abnormality detection according to the image feature points, grid lines and segmentation lines, including: Identify whether the image feature points match the grid lines, and whether the image feature points match the segmentation lines; When the image feature point matches the grid line, and the image feature point matches the dividing line, it is determined that there is no abnormality in the photovoltaic assembly image; When the image feature point does not match the grid line, or the image feature point does not match the dividing line, it is determined that an abnormality exists in the photovoltaic assembly image.

13. The method according to any one of claims 1 to 12, characterized in that: After acquiring the photovoltaic component image, the method further includes: Identify whether there is an abnormal light area in the photovoltaic module image; When there is an abnormal light area in the photovoltaic component image, it is prohibited to identify the image feature points of the photovoltaic component image; When there is no abnormal light area in the photovoltaic assembly image, it is allowed to identify the image feature points of the photovoltaic assembly image.

14. A method for detecting abnormality in a photovoltaic module image, characterized in that: include: In response to a request for invoking an anomaly detection service for a photovoltaic assembly image, determining a processing resource corresponding to the anomaly detection service for the photovoltaic assembly image; The processing resources are used to perform the following steps: acquiring a photovoltaic module image; identifying image feature points of the photovoltaic module image; identifying grid lines and dividing lines in the photovoltaic module image based on the image feature points; performing abnormality detection on the photovoltaic module image according to the image feature points, grid lines and dividing lines; wherein identifying the image feature points of the photovoltaic module image comprises: segmenting the photovoltaic module image based on rows and columns to obtain a plurality of sub-images; determining a local peak point of the sub-image, wherein the local peak point is a feature point having an amplitude greater than that of other feature points in a preset area; and determining the image feature points of the sub-image based on the local peak point, wherein the image feature points of the photovoltaic module image include image feature points of a plurality of sub-images.

Images