Defect detection method, device and equipment for solar module and storage medium
Through the dual-model detection method and deep learning network, the defect detection problem of solar module images is solved, and efficient and accurate defect recognition and positioning is achieved.
Patent Information
- Application Number
- CN202510133996.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-02-06
- Publication Date
- 2025-07-08
AI Technical Summary
In the prior art, the defect detection accuracy of solar energy modules is low, and it is difficult to meet the detection needs of different defect types, and the human eye observation efficiency is low.
The dual-model detection method is adopted, including a global detection model and a local detection model, combined with a corner detection model, and the defect type and location of the solar module images are identified through a deep learning network, and further re-judgment of the hidden crack defects.
It improves the accuracy and efficiency of defect detection of solar modules, can automatically identify and mark defects of different types and complex forms, and reduces missed and missed inspections.
Smart Images
Figure CN120278945A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of solar module manufacturing, and particularly to a method, device, equipment and storage medium for detecting defects in solar modules. Background Art
[0002] Solar photovoltaics, as a clean and environmentally friendly new energy source, is an important development direction for renewable energy in various countries. Solar modules are the core components of photovoltaic power generation. During the production process of solar modules, various defects that affect the performance of the modules may occur.
[0003] In related technologies, most defects in the appearance of modules are directly identified by visual inspection by humans. This method is inefficient and has poor recognition accuracy. Some module defect detections are carried out through machine vision solutions for defect recognition. However, due to the excessive number of defect types and different detection accuracy requirements for different defects in the actual production process, a single defect detection model is difficult to meet the requirements of module production.
[0004] Aiming at the problem of low defect detection accuracy of solar modules in related technologies, no effective solution has been proposed yet. Summary of the Invention
[0005] Based on this, in view of the problem of low defect detection accuracy of solar modules, it is necessary to provide a method, device, equipment and machine-readable storage medium for detecting defects in solar modules.
[0006] An embodiment of the present disclosure provides a method for detecting defects in solar modules.
[0007] The method for detecting defects in solar modules provided by the embodiment of the present disclosure includes
[0008] Obtaining an image of a solar module;
[0009] Inputting the image of the solar module into a defect detection model to determine the defect type and defect location of a target defect;
[0010] Based on the defect type, determining whether there is a hidden crack defect in the target defect. If so, cropping the image of the solar module based on the defect location to obtain a hidden crack defect image;
[0011] Inputting the hidden crack defect image into a hidden crack detection model to determine the hidden crack defect type of the hidden crack defect;
[0012] Obtaining a defect detection result of the solar module based on the defect type, the hidden crack defect type and the defect location.
[0013] In one embodiment, the defect detection model includes a global detection model and a local detection model. The steps of inputting the solar component image into the defect detection model to determine the defect type and defect location of the target defect include:
[0014] Input the solar component image into the global detection model to determine the global defect detection result of the solar component;
[0015] Segment the solar component image to obtain multiple local component images;
[0016] Input the multiple local component images into the local detection model respectively to determine the local defect detection results of the solar component;
[0017] Determine the defect type and defect location of the target defect based on the global defect detection result and the local defect detection results.
[0018] In one embodiment, the step of segmenting the solar component image to obtain multiple local component images includes:
[0019] Obtain a preset cutting number;
[0020] Based on the preset cutting number, cut the solar component image into a preset number of local component images. The local detection model magnifies each local component image, detects the local defect detection result of each local component image, and judges the distance between the same type of defects on each local component image and other local component images. If the distance is less than the preset distance, merge the two defects into one defect and merge the two local component images to determine the local defect detection result of the solar component.
[0021] In one embodiment, the steps of inputting the crack defect image into the crack detection model to determine the crack defect type of the crack defect include:
[0022] Input the crack defect image into the crack detection model,
[0023] Detect whether the crack defect includes a junction point with the edge of the solar component cell. If the junction point exists, determine the target defect as an edge crack;
[0024] If the junction point does not exist, determine the crack defect as a normal crack.
[0025] In one embodiment, the method further includes:
[0026] Input the solar component image into the corner point detection model to obtain the corner point positions of the cells in the solar component;
[0027] Determine the serial number of the solar cell in the solar module based on the corner positions;
[0028] Associate the corner positions with the defect positions to determine the serial number of the solar cell corresponding to the target defect;
[0029] Output the defect type of the target defect and the corresponding serial number of the solar cell as the final detection result.
[0030] In one embodiment, inputting the solar module image into a corner detection model; obtaining the corner positions of the solar module includes:
[0031] Input the solar module image into a corner detection model, compress the height of the solar module image to a preset height through the corner detection model, widen the solar module image to a preset width to obtain a trimmed image; identify the edges of the solar module in the trimmed image through the corner detection model, and identify the edges of each column and each row of solar cells to obtain the corner positions of each solar cell in the solar module.
[0032] In one embodiment, the determining the serial number of the solar cell in the solar module based on the corner positions includes:
[0033] Based on the identified edges of each column and each row of solar cells, obtain the length and width of each column and each row of solar cells. If the length or width of some columns or rows is much larger than that of other columns or rows, then the length or width of the part of the columns or rows is evenly divided based on the length or width of the corresponding other columns or rows, and the even division point is used as the edge of the solar cell, and the serial number of the solar cell is updated; if the length or width of some of the columns or rows is much smaller than that of other columns or rows, then delete the data of that column or row and update the serial number of the solar cell.
[0034] In a second aspect, an embodiment of the present application further provides a defect detection device for a solar module, including:
[0035] An acquisition module, configured to acquire a solar module image;
[0036] A first detection module, configured to input the solar module image into a defect detection model to determine the defect type and defect position of a target defect;
[0037] A cropping module, configured to determine whether there is a hidden crack defect in the target defect based on the defect type. If so, crop the solar module image based on the defect position to obtain a hidden crack defect image;
[0038] A second detection module, configured to input the hidden crack defect image into a hidden crack detection model to determine the type of the hidden crack defect;
[0039] A processing module, configured to obtain a defect detection result of the solar module based on the defect type, the type of the hidden crack defect, and the defect position.
[0040] In a second aspect, an embodiment of the present application further provides a defect detection device for a solar module, including a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-mentioned defect detection method for a solar module.
[0041] In a third aspect, an embodiment of the present application further provides a machine-readable storage medium. The machine-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions cause the processor to implement the above-mentioned defect detection method for a solar module.
[0042] For the above-mentioned defect detection method for a solar module, a defect detection model is used to detect defects in an image of the solar module, so as to obtain the defect position of the solar module and the corresponding defect type of the defect position; and a hidden crack detection model is used to further detect hidden crack defects. For the defect type of hidden crack, further detection is performed through the hidden crack detection model, and the defect detection of the solar module is realized based on the detection results of the dual models, which solves the specific defect types of a single and general-purpose model for rejudging hidden crack defects. Finally, the detection results of the defect detection model and the hidden crack detection model are combined to determine the defect detection result of the solar module, thereby improving the accuracy of the detection result of the defect detection model. BRIEF DESCRIPTION OF THE DRAWINGS
[0043] Figure 1 is a flowchart provided by an embodiment of the present invention;
[0044] Figure 2 is a schematic flowchart of obtaining a defect detection result of a solar module provided by an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0045] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, and are not used to limit the present application. Unless otherwise defined, the technical terms or scientific terms involved in the present application should have the general meaning understood by those of ordinary skill in the technical field to which the present application belongs. In the present application, words such as "a", "one", "a kind of", "the", "these", etc. do not indicate a limitation in quantity, and they can be singular or plural.
[0046] Please refer to Figures 1 to 2, the present invention provides a method, device, equipment, and storage medium for defect detection of solar modules. This technology can be applied to defect detection of solar modules.
[0047] For ease of understanding of this embodiment, the following describes the specific process of the embodiment of the present invention. Please refer to Figure 1 , an embodiment of a method for defect detection of a solar module in an embodiment of the present invention, includes:
[0048] A method for defect detection of a solar module, the method includes:
[0049] Step S101, obtain an image of the solar module;
[0050] The above-mentioned image of the solar module may be an EL image of the solar module, where the solar module is composed of multiple solar cell units. In application, an image acquisition device, such as an industrial camera, is used to photograph the target solar module to obtain an image of the solar module.
[0051] Step S102, input the image of the solar module into the defect detection model to determine the defect type and defect location of the target defect;
[0052] The above-mentioned defect detection model is a deep learning network architecture for defect detection. This defect detection model may be a deep learning network architecture such as the YOLO model. Before use, first use an industrial camera to collect no less than 500 EL images of solar modules, which may include images of solar modules before lamination, after lamination, and at the final inspection step, and the proportion of images containing defects is no less than 80%. The defects contained in the images need to cover all the defects that can be imaged in the EL image. These images need to ensure that the defect features on the images are clear, avoiding defect imaging with blurred features, and using a labeling tool to perform rectangular box target defect labeling on the collected component EL images. The labeled types include hidden cracks, scratches, black spots, cracks, fragments, etc. After labeling is completed, export the labels for standby training of the deep learning neural network. Then combine the images and the corresponding labels into training samples to train the defect detection model. The defect detection model learns to recognize the defect category features and locate the defect positions during training until the matching degree between the output result and the input training sample reaches a preset matching degree, and the model training is completed. At this time, the trained defect detection model can locate the defect positions in the input image of the solar module and identify the defect types at the defect positions. Among them, the defect types include hidden cracks, scratches, black spots, cracks, etc. The defect position can be represented by the position information of the bounding box for detecting the defect on the image of the solar module. If the defect in the image of the solar module is recognized, the bounding box will enclose the defect.
[0053] In this step, the acquired solar panel image is input into a pre-trained defect detection model, and the defect locations of the solar panel and the corresponding defect types at the defect locations are output. The output defect locations and the corresponding defect types at the defect locations are used as the first detection result. In this method, the target defect detection model of deep learning can automatically learn and extract defect features. After inputting the solar panel image into the trained target defect detection model, the target defect detection model detects and judges the defects in the solar panel image, automatically locates and marks the defect locations and defect types of the solar panel in the solar panel image, and can adapt to defects in different scenarios, different types, and different complex forms, realizing the automatic location and accurate marking of the defect locations and defect categories in each area of the solar panel image.
[0054] Step S103: Determine whether there is a hidden crack defect in the target defect based on the defect type. If so, crop the solar panel image based on the defect location to obtain a hidden crack defect image. A hidden crack defect is a fine crack that appears on the solar cell. Such a crack is not easily detectable by the naked eye, but will appear as a pattern with light and dark differences in the electroluminescence (EL) test. Hidden crack defects in solar panels are divided into edge hidden crack defects and ordinary hidden crack defects. Among them, an edge hidden crack defect refers to a crack that penetrates the edge of the solar cell. Such a crack not only affects the surface of the solar cell but also extends into the interior of the solar cell, seriously damaging the integrity and electrical performance of the solar cell and causing the solar cell to malfunction; an ordinary hidden crack defect refers to fine hidden cracks, cross-shaped hidden cracks, or arc-shaped cracks, etc. that appear on the solar cell. Ordinary hidden cracks have a certain impact on the integrity and performance of the solar cell. Since edge hidden crack defects have a great impact on the quality of solar panels, after the defect detection model outputs the first detection result, if there is a hidden crack defect, the image of the hidden crack defect is cropped, and then the hidden crack defect is further detected to avoid missed detection.
[0055] Step S104: Input the hidden crack defect image into a hidden crack detection model to determine the type of the hidden crack defect.
[0056] The above-mentioned hidden crack detection model is also a kind of defect detection model, which is used to detect the hidden crack defects in the image, such as identifying edge hidden cracks or ordinary hidden cracks. If it is an edge hidden crack, it has a greater impact on the quality of the solar panel. When cropping the solar panel image, the hidden crack defect and its location are the region of interest (ROI region). Cut out this region from the solar panel image, and then input the cut-out hidden crack defect image into the hidden crack detection model to re-judge the type of the hidden crack defect. By using the above-mentioned hidden crack detection model, edge hidden cracks can be directly identified, and the accuracy of defect image recognition can be improved.
[0057] Step S105: Obtain the defect detection result of the solar module based on the defect type, the type of hidden crack defect, and the defect location.
[0058] The above type of hidden crack defect is the type of hidden crack defect detected by using the hidden crack detection model. Update the type of hidden crack defect in the defect type, so as to obtain a more accurate detection result of the hidden crack defect, and improve the accuracy of defect detection. Of course, it is possible that the defect type and the type of hidden crack defect are output independently. Finally, combine the defect type and the type of hidden crack defect to comprehensively judge the defect type of the solar module corresponding to the solar module image.
[0059] The above solar defect detection method uses a defect detection model to detect defects in the solar module image of the solar module, obtains the defect location of the solar module and the corresponding defect type of the defect location; and uses a hidden crack detection model to detect hidden crack defects, which is used to rejudge whether the specific defect type of the hidden crack defect is a common hidden crack or an edge hidden crack. Finally, combine the detection results of the defect detection model and the hidden crack detection model to determine the defect detection result of the solar module, so as to improve the accuracy of the detection result of the defect detection model.
[0060] The following embodiments provide an implementation manner for determining the first detection result.
[0061] The defect detection model includes a global detection model and a local detection model. Inputting the solar module image into the defect detection model to determine the defect type and defect location of the target defect includes:
[0062] Input the solar module image into the global detection model to determine the global defect detection result of the solar module;
[0063] Segment the solar module image to obtain multiple local module images;
[0064] Input the multiple local module images into the local detection model respectively to determine the local detection result of the solar module;
[0065] Based on the global defect detection result and the local defect detection result, determine the defect type and defect location of the target defect.
[0066] Under EL imaging, some defects have problems such as small size, unclear features, and blurriness. After cutting, small-sized defects can be accurately identified. The aforementioned defect detection model includes a global detection model and a local detection model. Both of these models can be the YOLO model, or models such as EfficientDet and RetinaNet. Among them, the global detection model is used to detect the complete solar module image, focusing on identifying large-area defects. The local detection model is used to detect the local module image after segmenting the solar module image, focusing on detecting and identifying small-area defects. By combining the detection results of the global detection model and the local detection model, the detection results of the two complement each other, achieving a comprehensive detection of the solar module image, and being able to more comprehensively identify the defects in the image, avoiding false detections and missed detections.
[0067] Among them, both the global defect detection result and the local defect detection result include the defect type and the defect location. The combination of the detection results of the global detection model and the local detection model is the combination method of the global defect detection result and the local defect detection result. The detection results of the two are fused, the same content in the two detection results is fused, and the different content is combined by OR, so that in the defect type result, it contains all the detection results and there will be no phenomenon of duplicate detection results for the same defect.
[0068] Furthermore, segmenting the solar module image to obtain multiple local module images includes:
[0069] Obtaining the preset cutting quantity,
[0070] Based on the preset cutting quantity, the solar module image is cut into the preset number of local module images. The local detection model magnifies each local module image, detects the local defect detection result of each local module image, and judges the distance between the same type of defects on each local module image and other local module images. If the distance is less than the preset distance, the aforementioned two defects are merged into one defect, and the two local module images are merged to determine the local defect detection result of the solar module.
[0071] The local detection model focuses on detecting and identifying small-area defects, and the cut local module image is even smaller. Magnifying the local module image can make the defects more obvious, thereby improving the defect recognition accuracy and reducing false detections and missed detections. Among them, the method of magnifying the local module image is to magnify the local module image using the interpolation method, which is used to increase the resolution of the magnified local module image, make the defects more obvious, and facilitate the recognition of the defects. Of course, methods such as the iterative method and the finite element analysis method can also be used to magnify the local module image.
[0072] When cutting the solar module image, a fixed number of cuts are made on the solar module image. For example, the number of cuts can be 6, 12, 24, 36, etc., which can be specifically set according to the size of the solar module. Generally, the image is equally divided based on the number of cuts, or a cutting template is set. During cutting, since some defects may be at the cutting positions, the defects will be split during cutting. When detecting, it will be detected that there are defects on both of the two partial module images. If two defects are directly output based on the two partial module images, it will cause the number of output defects not to correspond to the actual number of defects, resulting in misdetection. Therefore, in this embodiment, after detecting the local defect detection results of each partial module image, the distance between the same type of defects on each partial module image is then judged. If the distance between the defects is less than the preset distance, it is judged that the two same type of defects are the same defect, and then these two defects are merged into one defect, and the partial module images containing these two same type of defects are merged, so as to ensure the accuracy of the output result of the local detection model.
[0073] The above-mentioned judgment of the distance between the same type of defects on each partial module image can be to only judge the distance between the same type of defects on adjacent partial module images, or to judge the distance between the defects at the edge of a partial module image and the same type of defects at the edge of other partial module images, or the distance between the defects at all positions on each partial module image and the same type of defects on other partial module images. To judge the distance between the same type of defects, it can be to judge the distance between the centers of the two defects, or to judge the distance between the defect coordinates. The above-mentioned preset distance can be 1mm, or 0.5mm, or other numbers, which can be set according to the resolution of the image. When merging the same type of defects, the coordinates of these two defects are merged, and the merged maximum value and merged minimum value of the two defect detection box coordinates are selected as the coordinates after merging, so as to present the defect completely in one detection result, avoiding missed detection and misjudgment of the defect.
[0074] Furthermore, input the crack defect image into the crack detection model, and determine that the types of crack defects of the crack defect include:
[0075] Input the crack defect image into the crack detection model;
[0076] Detect whether there is an intersection point between the crack defect and the edge of the solar module cell. If the intersection point exists, the crack defect is determined to be an edge crack;
[0077] If the intersection point does not exist, the crack defect is determined to be a normal crack.
[0078] The position and size of the crack defect both affect the performance of the solar cell. Moreover, when the crack defect is located at the edge of the solar cell, causing an edge crack defect, it will extend to the interior of the solar cell, severely affecting the integrity and performance of the solar cell. However, when the crack defect is located in the middle of the solar cell, the impact on the solar cell is slightly lower. Therefore, whether there is an intersection point between the crack defect and the edge of the solar cell in the solar module can directly determine whether it is an edge crack.
[0079] The following embodiments provide an implementation manner for determining the corner points.
[0080] The defect detection method for the solar module further includes:
[0081] Input the solar module image into the corner point detection model to obtain the corner point positions of the solar cells in the solar module;
[0082] Determine the serial numbers of the solar cells in the solar module based on the corner point positions;
[0083] Associate the corner point positions with the defect positions to determine the serial numbers of the solar cells corresponding to the target defects;
[0084] Output the defect type of the target defect and the corresponding serial numbers of the solar cells as the final detection result.
[0085] In this embodiment, a corner point model is used to detect the corner points of the battery, so as to obtain the positions of each solar cell in the solar module. Based on the corner points of the solar cells, the solar cells in each row and each column are numbered row by row and column by column. Thus, after associating the corner point positions with the defect positions, the final detection result shows the defect type and the row and column serial numbers of the solar cell where the defect is located, which is convenient for finding the position of the defective solar cell on the solar module.
[0086] When training the corner point detection model, the edges of the solar cell module on the solar module image and the boundaries between each column and each row of solar cells are marked. It is set that the midpoint of the width of each column marking frame and the midpoint of the height of each row marking frame are the dividing points of the small strips of solar cells and the dividing points of the small strings of solar cells, respectively. Based on this, the solar cells in each column and each row are located, and the middle position between two adjacent column marking frames is selected for small strip serial number identification, and the middle position between two adjacent row marking frames is selected for small string serial number identification. Thus, after inputting the solar module image into the corner point detection model, the corner point detection model can identify the solar cells according to the edges of the solar cell module, and then identify the solar cells in each column and each row, and mark serial numbers for the solar cells. When marking, the row number is in front and the column number is behind. For example, the solar cell located in the 24th column and the 2nd row is marked as 224, or the serial number of the solar cell is directly marked as 2 rows 24 columns, 2 strings 24 strips, or the 2nd small string and the 24th small strip, to facilitate finding the position of the solar cell.
[0087] When inputting a solar module into a defect detection model, the serial number of the corresponding solar module can be input at the same time to realize the association between the solar module image and the actual solar module. In this way, the solar module number and the corresponding defect detection result are included in the final detection result output, which facilitates the correspondence between the defect detection result and the solar module, and is more convenient for the repair personnel to perform defect repair according to the defect location, and also convenient for the staff to manage and search for solar modules.
[0088] Further, inputting the solar module image into a corner detection model, the corner positions of the solar module obtained include:
[0089] Inputting the solar module image into a corner detection model, the corner detection model compresses the height of the solar module image to a preset height and widens the solar module image to a preset width to obtain a trimmed image. The corner detection model identifies the edge of the solar module in the trimmed image and identifies the edges of each column and each row of cells, so as to obtain the corner positions of each cell in the solar module.
[0090] The purpose of compressing in the height direction is to improve the model detection speed without affecting the detection accuracy. The purpose of expanding pixels in the width direction is to avoid detection failure caused by too little black background of the component itself. With such settings, both the detection speed of the image and the detection accuracy of the image can be improved.
[0091] Determining the cell numbers of the solar module based on the corner positions includes:
[0092] Based on the identified edges of each column and each row of cells, obtain the length and width of each column and each row of cells. If the length or width of some columns or rows is much larger than the length or width of other columns or rows, then the length or width of the some columns or rows is evenly divided based on the length or width of the corresponding other columns or rows, and the evenly divided position is used as the edge of the cell, and the cell number is updated; if the length or width of some columns or rows is much smaller than the length or width of other columns or rows, then the data of the column or row is deleted, and the cell number is updated.
[0093] When the corner detection model cannot accurately identify the boundary of a certain cell, such as missing the boundary of a certain cell or misdetecting the boundary of a certain cell, then the width of the identified cell will be too wide or too narrow. According to the width of other cells, it can be judged whether the width of a certain cell is abnormal, so as to perform even division or deletion operations on the cell to achieve accurate positioning of each cell. Since the serial numbers of each column and each row of cells are marked while identifying each cell, when the number of cells is increased or deleted, the cell numbers are adjusted synchronously to ensure the accuracy of the output result.
[0094] Further, output the defect type of the target defect and the corresponding cell serial number as the final detection result. That is, if there is no defect type or the defect type is only a negligible defect, the final detection result is OK. If there are other defect types, the final detection result includes the position of the defective cell, the defect type corresponding to the cell position, and the solar module number.
[0095] On the other hand, the present invention also provides a defect detection device for a solar module, characterized in that the device includes:
[0096] An acquisition module for acquiring a solar module image;
[0097] A first detection module for inputting the solar module image into a defect detection model to determine the defect type and defect position of the target defect;
[0098] A cropping module for determining whether there is a hidden crack defect in the target defect based on the defect type. If so, crop the solar module image based on the defect position to obtain a hidden crack defect image;
[0099] A second detection module for inputting the hidden crack defect image into a hidden crack detection model to determine the hidden crack defect type of the hidden crack defect;
[0100] A processing module for obtaining the defect detection result of the solar module based on the defect type, the hidden crack defect type, and the defect position.
[0101] In this method, the first detection module performs defect detection on the solar module image containing the solar module through a trained defect detection model to obtain the defect type and defect position of the target defect on the solar module. The cropping module is used to determine whether there is a hidden crack defect in the target defect based on the defect type. If so, crop the solar module image based on the defect position to obtain a hidden crack defect image. The second detection module performs a re-judgment on the hidden crack detected by the defect detection model through a trained hidden crack detection module to determine whether the hidden crack is a normal hidden crack or an edge hidden crack, and updates the defect type detected by the defect detection model. The processing module is used to obtain the defect detection result of the solar module based on the defect type, the hidden crack defect type, and the defect position. This method comprehensively considers the detection results of the target defect detection model and the hidden crack detection model, avoids the problem of low defect accuracy caused by single deep learning model detection, and improves the accuracy of defect detection.
[0102] The above first detection module includes a global detection module and a local detection module. The global detection module is used to input a complete image into a global detection model to detect the complete solar panel image and output a global defect detection result. The local detection module is used to input the local component image after cutting the solar panel image into a local detection model to detect the local component image after cutting the solar panel image and output a local defect detection result. The local defect detection result and the global defect detection result are not exactly the same. Based on the above global defect detection result and local defect detection result, the defect detection result of the solar panel is obtained.
[0103] Further, segmenting the solar panel image to obtain multiple local component images includes:
[0104] Obtain a preset cutting quantity.
[0105] Based on the preset cutting quantity, cut the solar panel image into a preset number of local component images. The local detection model magnifies each local component image, detects the local defect detection result of each local component image, and judges the distance between the same type of defect on each local component image and other local component images. If the distance is less than the preset distance, then merge the aforementioned two defects into one defect and merge the two local component images to determine the local defect detection result of the solar panel.
[0106] Further, input the crack defect image into a crack detection model to determine the types of crack defects of the crack defect, including:
[0107] Input the crack defect image into the crack detection model;
[0108] Detect whether there is an intersection point between the crack defect and the edge of the solar panel cell. If the intersection point exists, then judge the crack defect as an edge crack;
[0109] If the intersection point does not exist, then obtain the crack size and the number of cracks of the crack defect;
[0110] If the crack size or the number of cracks is greater than or equal to the preset threshold, then judge the crack defect as a common crack;
[0111] If both the crack size and the number of cracks are less than the preset threshold, then judge the crack defect as a negligible defect.
[0112] Further, for the corner detection module of the defect detection device, the corner detection module determines the corner positions of each cell in the solar module through a trained corner detection model, determines the cell numbers of the solar module based on the corner positions of the cells, associates the corner positions with the defect positions, can determine the corresponding cell numbers of the target defects, and outputs the defect types of the target defects and the corresponding cell numbers as the final detection results.
[0113] Further, inputting the solar module image into the corner detection model to obtain the corner positions of the solar module includes:
[0114] Inputting the solar module image into the corner detection model, compressing the height of the solar module image to a preset height and widening the solar module image to a preset width by the corner detection model to obtain a trimmed image, identifying the edges of the solar module in the trimmed image by the corner detection model, and identifying the edges of each column and each row of cells to obtain the corner positions of each cell in the solar module.
[0115] Determining the cell numbers of the solar module based on the corner positions includes:
[0116] Based on the identified edges of each column and each row of cells, obtaining the lengths and widths of each column and each row of cells. If the length or width of some columns or rows is much larger than that of other columns or rows, the length or width of the part of the columns or rows is evenly divided based on the corresponding lengths or widths of other columns or rows, and the evenly divided position is used as the edge of the cell, and the cell numbers are updated; if the length or width of some columns or rows is much smaller than that of other columns or rows, the data of the columns or rows are deleted, and the cell numbers are updated.
[0117] Further, outputting the defect types of the target defects and the corresponding cell numbers as the final detection results includes that if there is no defect type or the defect type is only a negligible defect, the final detection result is OK; if there are other defect types, the final detection results include the positions of the defective cells, the corresponding defect types of the cell positions, and the solar module numbers.
[0118] On the other hand, the present invention also proposes a defect detection device for a solar module, including a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-mentioned defect detection method for the solar module. The defect detection device can be a server or a terminal device.
[0119] The defect detection device includes a processor and a memory. The memory has machine-executable instructions that can be executed by the processor, and the processor executes the machine-executable instructions to implement the above-mentioned defect detection method for the solar module.
[0120] The memory may include high-speed random access memory (RAM), and may also include non-volatile memory, such as at least one disk memory. The communication connection between the system network element and at least one other network element is implemented through at least one communication interface (which can be wired or wireless), and the Internet, wide area network, local area network, metropolitan area network, etc. can be used.
[0121] The processor may be an integrated circuit chip with signal processing capabilities. In the implementation process, each step of the above method can be completed by the integrated logic circuit in the hardware of the processor or the instructions in the form of software. The above-mentioned processor may be a general-purpose processor, including a central processing unit (CPU for short), a network processor (NP for short), etc.; it may also be a digital signal processor (DSP for short), an application specific integrated circuit (ASIC for short), a field programmable gate array (FPGA for short), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components. It can implement or execute the various methods, steps and logic block diagrams disclosed in the embodiments of the present invention. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the embodiments of the present invention can be directly embodied as being executed by the hardware decoding processor, or executed by the combination of the hardware and software modules in the decoding processor. The software module may be located in a mature storage medium in the art such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, register, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the method in the foregoing embodiments.
[0122] In another aspect, the present invention also proposes a machine-readable storage medium, characterized in that the machine-readable storage medium stores machine-executable instructions, and when the machine-executable instructions are called and executed by the processor, the machine-executable instructions cause the processor to implement the above-mentioned defect detection method for solar components.
[0123] A computer program product for a method, apparatus, device, and storage medium for defect detection of a solar module provided by an embodiment of the present invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the method in the foregoing method embodiments. For specific implementation, reference can be made to the method embodiments and will not be elaborated herein. Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described system and apparatus can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.
[0124] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, database, or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memories. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetoresistive random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc. The databases involved in the embodiments provided in the present application can include at least one of relational databases and non-relational databases. Non-relational databases can include distributed databases based on blockchain, etc., and are not limited thereto. The processors involved in the embodiments provided in the present application can be general-purpose processors, central processing units, graphics processors, digital signal processors, programmable logic devices, data processing logics based on quantum computing, etc., and are not limited thereto.
[0125] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope recorded in this specification.
[0126] The above-described embodiments merely represent several implementation manners of the present application. The description is relatively specific and detailed, but it should not be construed as a limitation on the patent scope of the present application. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present application, several modifications and improvements can still be made, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the appended claims.
Claims
1. A method for defect detection of a solar module, characterized in that, The method includes: Obtaining an image of a solar component; Inputting the image of the solar component into a defect detection model to determine the defect type and defect location of a target defect; Based on the defect type, determining whether there is a hidden crack defect in the target defect. If so, cropping the image of the solar component based on the defect location to obtain a hidden crack defect image; Inputting the hidden crack defect image into a hidden crack detection model to determine the type of the hidden crack defect; obtaining a defect detection result of the solar component based on the defect type, the type of the hidden crack defect, and the defect location.
2. The defect detection method of the solar module according to claim 1, wherein, The defect detection model includes a global detection model and a local detection model. The step of inputting the image of the solar component into the defect detection model to determine the defect type and defect location of the target defect includes: Inputting the image of the solar component into the global detection model to determine a global defect detection result of the solar component; Segmenting the image of the solar component to obtain multiple local component images; Inputting the multiple local component images into the local detection model respectively to determine local defect detection results of the solar component; Based on the global defect detection result and the local defect detection results, determining the defect type and defect location of the target defect.
3. The defect detection method for a solar module according to claim 2, wherein, The step of segmenting the image of the solar component to obtain multiple local component images includes: Obtaining a preset cutting number; Based on the preset cutting number, cutting the image of the solar component into a preset number of local component images. The local detection model magnifies each local component image, detects the local defect detection result of each local component image, and judges the distance between the same type of defects on each local component image and other local component images. If the distance is less than a preset distance, combining the two defects into one defect and combining the two local component images to determine the local defect detection result of the solar component.
4. The defect detection method of the solar module according to claim 1, characterized in that, The step of inputting the hidden crack defect image into the hidden crack detection model to determine the type of the hidden crack defect includes: Inputting the hidden crack defect image into the hidden crack detection model; Detecting whether there is an intersection point between the hidden crack defect and the edge of the solar component cell. If the intersection point exists, determining that the hidden crack defect is an edge hidden crack; If the intersection point does not exist, determining that the hidden crack defect is a common hidden crack.
5. The defect detection method of the solar module according to claim 1, characterized in that, The method further includes: Inputting the image of the solar component into a corner point detection model to obtain the corner point positions of the cells in the solar component; Based on the corner point positions, determining the cell numbers of the solar component; Associating the corner point positions with the defect locations to determine the corresponding cell numbers of the target defect; outputting the defect type of the target defect and the corresponding cell numbers as a final detection result.
6. The defect detection method for a solar module as claimed in claim 5, wherein Inputting the image of the solar component into the corner point detection model; Obtaining the corner point positions of the solar component includes: Input the solar panel image into the corner detection model. Compress the height of the solar panel image to a preset height and widen the solar panel image to a preset width through the corner detection model to obtain a trimmed image. Identify the edges of the solar panel in the trimmed image through the corner detection model, and identify the edges of each column and each row of solar cells to obtain the corner positions of each solar cell in the solar panel.
7. The defect detection method of the solar module according to claim 6, wherein, Determining the serial numbers of the solar cells of the solar panel based on the corner positions includes: Based on the identified edges of each column and each row of solar cells, obtain the lengths and widths of each column and each row of solar cells. If the length or width of some columns or rows is much larger than that of other columns or rows, the length or width of the part of the columns or rows is evenly divided based on the length or width of the corresponding other columns or rows, and the evenly divided position is used as the edge of the solar cell, and the serial number of the solar cell is updated. If the length or width of some of the columns or rows is much smaller than that of other columns or rows, delete the data of that column or row, and update the serial number of the solar cell.
8. A defect detection device for a solar module, characterized in that, The device includes: An acquisition module, configured to acquire a solar panel image; A first detection module, configured to input the solar panel image into a defect detection model to determine the defect type and defect position of a target defect; A cropping module, configured to determine whether there is a hidden crack defect in the target defect based on the defect type. If so, crop the solar panel image based on the defect position to obtain a hidden crack defect image; A second detection module, configured to input the hidden crack defect image into a hidden crack detection model to determine the type of the hidden crack defect; A processing module, configured to obtain a defect detection result of the solar panel based on the defect type, the type of the hidden crack defect, and the defect position.
9. A defect detection device for a solar module, characterized in that, It includes a processor and a memory. The memory stores machine-executable instructions that can be executed by the processor. The processor executes the machine-executable instructions to implement the defect detection method of the solar panel according to any one of claims 1-7.
10. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores machine-executable instructions. When the machine-executable instructions are called and executed by a processor, the machine-executable instructions cause the processor to implement the defect detection method of the solar panel according to any one of claims 1-7.