A solar panel quality inspection method

Abstract

The present invention discloses a solar panel quality inspection method, which belongs to the field of industrial quality inspection, including obtaining a complete solar panel image and grayscale, and then cutting it into small solar panel areas according to the average grayscale value of the complete image; comparing the area and size of each selected target area with the average value of the small solar panel, and counting the positions and sizes of all target areas that meet the verification conditions; breaking up the small solar panel images, traversing each position where the target area exists in sequence according to the row and column coordinates, and filling the positions; indexing and encoding according to the position and size of the small solar panel; cutting all small solar panel images into individuals, and then sending them into a trained neural network for reasoning to determine whether each small solar panel image has defects. The present invention adopts the above-mentioned solar panel quality inspection method, which improves the efficiency of quality inspection, has low cost and higher accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of industrial quality inspection, and in particular to a solar panel quality inspection method. Background Art

[0002] Solar panels are panels that use the photovoltaic effect of semiconductor materials under light conditions to directly convert solar energy into electrical energy. They are the most direct way to use solar energy. Currently, most solar panels are made of single-crystal silicon. Since electricity can be generated wherever there is sunlight, solar photovoltaic power generation is suitable for a variety of occasions, from large power stations to small portable chargers.

[0003] The existing quality inspection methods in solar panel production are mostly manual quality inspection, which has defects such as low efficiency and low accuracy. Therefore, there is an urgent need for a method with high detection efficiency and high accuracy to promptly detect whether there are small defects in the solar cells in the overall panel during the quality inspection process so that they can be replaced in time and reduce losses. Summary of the invention

[0004] The purpose of the present invention is to provide a solar panel quality inspection method to solve the problems mentioned in the above background technology.

[0005] To achieve the above object, the present invention provides a solar panel quality inspection method, comprising the following steps:

[0006] S1. In the quality inspection process of solar panel production, the complete picture of the solar panel is obtained through an industrial camera and grayscaled, and then the small solar panel area is cut out as the target area according to the average grayscale value of the complete picture;

[0007] S2, verifying and counting the target areas in S1, comparing the area and size of each selected target area with the average value of the small solar panels, and counting the positions and sizes of all target areas that meet the verification conditions;

[0008] S3, breaking up the small solar panel images, and traversing each position where the target area exists in sequence according to the row and column coordinates. If the position already has a small solar panel included in the statistics, skip it; if not, fill in the position;

[0009] S4, determining the positions and sizes of all small solar panels through S2 and S3, and indexing and encoding the small solar panels according to their positions and sizes;

[0010] S5. All small solar panel images are cut into individuals and then sent to the trained neural network for inference to determine whether each small solar panel image has defects. If the image has defects, a record is made. After all small solar panel images are identified, the positions of the defects are reversely calculated based on the codes of the recorded small solar panel images with defects, and then marked on the complete image and returned to the user.

[0011] Preferably, the complete picture of the solar panels in S1 is adaptively cropped according to the principle that the areas of all small solar panels are consistent, and the bright part thereof is the picture of the small solar panels.

[0012] Preferably, the verification condition of S2 is that the area of ​​the target region is within the allowable error range of the average value of the small solar panels.

[0013] Preferably, the specific content of S3 is: for each solar cell on the target area that meets the verification conditions, the average row coordinates and the average column coordinates are calculated respectively. When traversing to the missing position, find out whether there is a recorded row and column benchmark. If so, fill it in. If not, recursively calculate and fill it in according to the most recent benchmark value.

[0014] Preferably, the encoding method in S4 is: the number of rows of small solar panel pictures * the number of small solar panel pictures in each row + the number of columns of small solar panel pictures, to obtain the position code of each small solar panel picture in the complete solar panel picture, and then connect this code with the file name of the complete solar panel picture, and rename each small solar panel picture to obtain the specific position of each small solar panel picture.

[0015] Therefore, the present invention adopts the above-mentioned solar panel quality inspection method, and the beneficial effects are as follows:

[0016] (1) Compared with the method of directly using target detection, the segmentation and classification method proposed in the present invention has higher accuracy;

[0017] (2) Compared with target detection methods, classification methods have lower training and application costs;

[0018] (3) Compared with the target detection method, the present invention has a faster running speed and can improve the efficiency of production and quality inspection;

[0019] (4) The specific location of defective solar cells can be quickly obtained.

[0020] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1is an overall flow chart of an embodiment of the present invention;

[0022] Figure 2 It is a schematic diagram of a complete grayscale image of a solar panel according to an embodiment of the present invention;

[0023] Figure 3 A schematic diagram of a target area selected according to brightness value according to an embodiment of the present invention;

[0024] Figure 4 A schematic diagram of breaking up a small solar panel image according to an embodiment of the present invention;

[0025] Figure 5 A schematic diagram of a small solar panel with defects in an embodiment of the present invention;

[0026] Figure 6 This is a picture of a small individual solar panel that is finally cut out according to an embodiment of the present invention. DETAILED DESCRIPTION

[0027] Example

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only 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.

[0029] Reference Figure 1 The present invention discloses a solar panel quality inspection method, comprising the following steps:

[0030] S1. In the quality inspection process of solar panel production, the complete picture of the solar panel is obtained through an industrial camera and grayscaled (such as Figure 2 ), and then cut out a small solar panel area as the target area according to the average gray value of the complete image.

[0031] Calculate the average brightness of the entire image. Small solar panel images are usually bright, and edges are usually dark. Select small solar panel areas based on the brightness values, such as Figure 3 shown.

[0032] The areas and boundaries of solar panels can be easily distinguished, but if there are large differences in the grayscale values ​​of the panels, many panels cannot be distinguished normally.

[0033] S2. Verify and count the target areas in S1, compare the area and size of each selected target area with the average value of the small solar panels, and count the positions and sizes of all target areas that meet the verification conditions.

[0034] The verification condition is: the area of ​​the target area is within the error tolerance of the average value of the small solar panels. Since the area of ​​each solar panel should be basically the same, assuming the error is 10%, when the area of ​​the traversed solar panel is within the error tolerance range (90%~110%), the size and position of the solar panel are considered to be credible. Therefore, the position of the solar panel is recorded.

[0035] After recording the positions and sizes of the solar panels with reliable locations, if there are solar panels with an incorrect number due to abnormal area and location, further filling processing is required to fill in the positions where the solar panels should be located according to the positions of the reliable solar panels.

[0036] S3, break up the small solar panel images, such as Figure 4 , traverse each location where the target area exists in turn according to the row and column coordinates. If the location already has a small solar panel included in the statistics, skip it. If not, fill in the location.

[0037] The average row coordinates and column coordinates of each solar cell in a trusted position are calculated respectively. On the one hand, the credibility of each solar cell can be verified, and on the other hand, a more accurate benchmark can be provided for the missing blocks. When traversing to the missing position, first find out whether there is a recorded row and column benchmark. If it exists, it can be filled. If not, it is recursively calculated and filled according to the nearest benchmark value. For example: suppose that there is an abnormality in the image block of the solar cell in the first column and second row, and its position needs to be filled. Because its area is abnormal according to the second step, its position coordinates will not be recorded. At this time, find out whether its row and column have recorded row and column average coordinates. If so, replace its row coordinates with the average row coordinates of the second row, and replace its column coordinates with the average column coordinates of the first column. At this time, the position of the solar cell in the first column and second row can be filled.

[0038] Because the panels may have cracks, uneven brightness, etc., it is easy to have scattered small pieces of solar panel images instead of complete solar panel images, such as Figure 5 As shown, the scattered small solar panel images in the previous step need to be completed and verified.

[0039] S4. Determine the positions and sizes of all small solar panels through S2 and S3, and index and encode them according to the positions and sizes of the small solar panels.

[0040] Since the previous steps have recorded the trusted positions of all small images, that is, their row and column coordinates, we can encode them according to their trusted positions. The encoding method is: the number of rows of small images * the number of small images in each row + the number of columns of small images, and we can get the position code of each small image in a single large image. Then connect this code with the file name of the large image and rename it to each small image, and we can get the specific position of each small image, that is, which position of which large image it belongs to.

[0041] After the previous steps of breaking up, checking, filling, and encoding, all the required information of the small image is obtained. At this time, each small image is cut according to the coordinate information and the average area of ​​the small image, and a large solar panel image is divided into many small solar chip images. Figure 6 shown.

[0042] S5. Adaptively crop the complete solar panel image based on the principle that the area of ​​all small solar panels is consistent. All small solar panel images are cropped into individuals and then sent to the trained neural network for reasoning to determine whether each small solar panel image has defects. If the image has defects, a record is made. After all small solar panel images are identified, the location of the defects is reversely calculated based on the recorded codes of the small solar panel images with defects, and then marked on the large image and returned to the user.

[0043] Specifically, the file name of each small solar panel image is decoded, and the inference result is marked on the complete solar panel according to the decoded information for display, so that a complete solar panel can be fed back to the customer to show whether there are defects and where the defects exist. The result is then returned to the customer to determine whether the quality inspection has passed. This completes the automatic quality inspection process of the entire solar panel.

[0044] Therefore, the present invention adopts the above-mentioned solar panel quality inspection method to greatly improve the efficiency of quality inspection. Compared with manual work, it is not only low-cost but also more efficient. Secondly, compared with other visual methods that directly perform inspections on large panel images, each small panel is inspected separately, which has higher accuracy.

[0045] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. A solar panel quality inspection method, characterized in that: The following steps are involved: S1. In the quality inspection process of solar panel production, the complete picture of the solar panel is obtained through an industrial camera and grayscaled, and then the small solar panel area is cut out as the target area according to the average grayscale value of the complete picture; S2, verifying and counting the target areas in S1, comparing the area and size of each selected target area with the average value of the small solar panels, and counting the positions and sizes of all target areas that meet the verification conditions; S3, breaking up the small solar panel images, and traversing each position where the target area exists in sequence according to the row and column coordinates. If the position already has a small solar panel included in the statistics, skip it; if not, fill in the position; The specific content is: for each solar cell on the target area that meets the verification conditions, the average row coordinates and column coordinates are calculated respectively. When traversing to the missing position, find out whether there is a recorded row and column benchmark. If there is, fill it in. If not, recursively calculate and fill it in according to the nearest benchmark value; S4, determining the positions and sizes of all small solar panels through S2 and S3, and indexing and encoding the small solar panels according to their positions and sizes; S5. All small solar panel images are cut into individuals and then sent to the trained neural network for inference to determine whether each small solar panel image has defects. If the image has defects, a record is made. After all small solar panel images are identified, the positions of the defects are reversely calculated based on the codes of the recorded small solar panel images with defects, and then marked on the complete image and returned to the user.

2. A solar panel quality inspection method according to claim 1, characterized in that: According to the principle that the areas of all small solar panels are consistent, the complete image of the solar panels in S1 is adaptively cropped, and the bright part is the image of the small solar panels.

3. A solar panel quality inspection method according to claim 1, characterized in that: The verification condition of S2 is that the area of ​​the target region is within the allowable error range of the average value of the small solar panels.

4. A solar panel quality inspection method according to claim 1, characterized in that: The encoding method in S4 is: the number of rows of small solar panel pictures * the number of small solar panel pictures in each row + the number of columns of small solar panel pictures, to obtain the position code of each small solar panel picture in the complete solar panel picture, then connect this position code with the file name of the complete solar panel picture, and rename each small solar panel picture to obtain the specific position of each small solar panel picture.

Images