CN-122003811-A - Deep learning-based solar panel defect detection device and method

Abstract

The invention relates to a method and a device for detecting defects of a solar panel by utilizing images of the solar panel, which are characterized in that the defects which cannot be detected in an EL image are detected by utilizing a pre-artificial intelligence part, the defects are classified according to types, and then the sizes of the defects are measured to judge whether the defects are final defects or not, so that the method and the device for detecting the defects of the solar panel have the advantages that whether the defects are final defects or not can be judged more quickly and accurately in the production process of the solar panel, the process time can be shortened, the defect detection rate can be improved, and the cost loss is reduced.

Inventors

• LI XIAOCHENG

Assignees

• 韩华思路信株式会社

Dates

Publication Date

20260508

Application Date

20241016

Priority Date

20231017

Claims (12)

1. 1. A method of detecting defects in a solar panel, the method being performed by a control section including one or more processors and a memory, the method comprising: a step of receiving an image of the solar panel, A step of measuring the size of the constituent elements based on the solar panel image to determine whether the size-related defects are present, A step of detecting defects from the solar panel image using an artificial intelligence section learned in advance and classifying the detected defects by type, and And measuring the sizes of the defects of the classified types, and determining whether defects exist according to the defect types.
2. 2. The method for detecting defects of a solar panel according to claim 1, wherein, The solar panel image is an optically captured image, not an electroluminescent image of the solar panel.
3. 3. The method for detecting defects of a solar panel according to claim 1, wherein, In the step of measuring the size to determine whether the size-related defect is present, And measuring the dimension between the battery pieces and the frame or between the battery pieces forming the solar panel so as to determine whether the solar panel is bad or not.
4. 4. The method for detecting defects of a solar panel according to claim 1, wherein, The defect type is a defect type which cannot be confirmed in an electroluminescent image of a solar panel.
5. 5. The method for detecting defects of a solar panel according to claim 4, wherein, The defect type comprises more than one of surface scratch, crack, wire shape, bubble and foreign matter.
6. 6. The method for detecting defects of a solar panel according to claim 1, wherein, In the step of measuring the sizes of the defects of the classified types, determining whether there is a defect by defect type, And determining that the defect is bad when the size of the defect is larger than a threshold value predetermined according to the defect type.
7. 7. A solar panel defect detection apparatus, comprising: an image input section for receiving a solar panel image, An artificial intelligence section, which learns in advance to classify defects of the solar panel image, and A control unit including one or more processors and a memory; The control section is configured to: Based on the solar panel image inputted through the image input part, the size of the constituent elements is measured to determine whether the size-related defects exist, Detecting defects from the solar panel image using the artificial intelligence section, and classifying the detected defects by type, And measuring the sizes of the defects of the classified types, and determining whether defects exist according to the defect types.
8. 8. The apparatus for detecting defects of a solar panel according to claim 7, The solar panel image is an optically captured image, not an electroluminescent image of the solar panel.
9. 9. The apparatus for detecting defects of a solar panel according to claim 7, The control section is configured to: And measuring the dimension between the battery pieces and the frame or between the battery pieces constituting the solar panel to determine whether the dimension-related defect exists.
10. 10. The apparatus for detecting defects of a solar panel according to claim 7, The defect type is a defect type which cannot be confirmed in an electroluminescent image of a solar panel.
11. 11. The device for detecting defects of a solar panel according to claim 10, wherein, The defect type comprises more than one of surface scratch, crack, wire shape, bubble and foreign matter.
12. 12. The apparatus for detecting defects of a solar panel according to claim 7, In the step of measuring the sizes of the defects of the classified types, determining whether there is a defect by defect type, And determining that the defect is bad when the size of the defect is larger than a threshold value predetermined according to the defect type.

Description

Deep learning-based solar panel defect detection device and method Technical Field The present invention relates to a technology for detecting defects of a solar panel, and more particularly, to a technology for detecting defects of a solar panel by deep learning. Background Although electric power is an indispensable element in daily life, chemical fuel or nuclear energy used for producing electric power is gradually exposed in such a way that it causes problems such as pollution of the earth, initiation of climate change, and the like. Accordingly, attention to so-called renewable energy sources such as tidal energy, wind power, and solar energy is also increasing as an alternative thereto, and attempts are being made to replace power generation using fossil fuel or nuclear power generation. Among them, solar power generation has been attracting attention as the most potential alternative energy source, and solar power generation uses a solar panel (photovoltaic panel) to convert solar energy into electric energy. Since the solar panel is a means for directly converting solar energy into electric energy, if the solar panel itself has a defect, the amount of generated electricity is reduced, and if the electrical defect is serious, the risk of fire due to arc or the like is increased. Therefore, it becomes important to detect defects at the stage of manufacturing the solar panel. In the past, for inspection of solar panels, the most common is the electroluminescent (Electro Luminescence, EL) measurement method. In this method, when power is supplied to the solar panel, the characteristic that the panel emits light of a natural wavelength band is utilized, and the defect of the solar panel can be detected by capturing and utilizing the obtained EL image. However, although EL images can detect various types such as cracks (cracks) from various defects of solar panels, only defective types that can be distinguished by light released by power supply can be detected, and there is a problem in that fine micro cracks cannot be detected due to poor quality of EL images, or defects are classified as excessive detection of defects, etc., and defect detection performance is degraded. Disclosure of Invention The purpose of the present invention is to detect a type of defect that cannot be detected in an EL image by using an Optical (Optical) image of a solar panel. In addition, the present invention is directed to a device and a method for detecting defects of a solar panel, which can reduce a process time and reduce a defective rate of a finally produced solar panel by detecting defects in advance in a solar panel production process. The method for detecting the defects of the solar panel is characterized by comprising the steps of receiving the solar panel image, measuring the sizes of the components based on the solar panel image to detect the defects, classifying the solar panel image according to the types of defects by utilizing an artificial intelligence part learned in advance, and measuring the sizes of the classified types of defects to determine whether the defects are final defects or not. The solar panel defect detection device is characterized by comprising an image input part, an artificial intelligence part and a control part, wherein the image input part is used for receiving a solar panel image, the artificial intelligence part is used for pre-learning to classify defects of the solar panel image, the control part comprises more than one processor and a memory, the control part is configured to measure sizes of components based on the solar panel image input by the image input part to detect the defects, the artificial intelligence part is used for classifying defects included in the solar panel image according to types, and the sizes of the classified types of defects are measured to determine the types of defects and final defects of the solar panel. According to an embodiment of the invention, the optical image of the solar panel is classified into the types of defects by using artificial intelligence, and whether the final defects are detected or not is determined according to the sizes of the types of defects, so that defects which cannot be detected in the EL image of the solar panel can be detected, and thus, the method has the advantages that the defect rate of the finally produced solar panel is reduced, and the time required for detection is shortened. Drawings Fig. 1 is a schematic flow chart of a method for detecting defects of a solar panel according to a preferred embodiment of the present invention. Fig. 2 is an example of an optical image of a solar panel. Fig. 3 is an example of a portion for measuring the size of a solar panel optical image. Fig. 4 is an example of a portion for measuring the size of a solar panel optical image. Fig. 5 is an example of the result of preprocessing of the solar panel optical image. Fig. 6 is an example of the type of defect detected in the optical ima