A method and system for detecting defects in a photovoltaic panel

Abstract

The application discloses a kind of photovoltaic panel defect detection method and system, wherein the defect detection method includes obtaining photovoltaic panel image to be detected and pre-processing;Using the MSRCR enhancement method of fusion high-pass filter to the photovoltaic panel image is carried out image enhancement processing and obtains the grid line image of photovoltaic panel;Using the defect detection method based on region growth to the grid line image is carried out first class defect detection and obtains the first class defect detection result containing crack and missing angle;Wherein, the method disclosed in the application proposes the MSRCR enhancement method of fusion high-pass filter, and then the photovoltaic panel image is processed, can more accurately obtain the grid information in image and as far as possible reduce the influence of noise, obtain more suitable grid line image of photovoltaic panel detection, improve defect detection precision.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of photovoltaic panel defect detection, and particularly relates to a photovoltaic panel defect detection method and system. BACKGROUND

[0002] With the growth of energy demand and the improvement of environmental awareness, solar photovoltaic cells (hereinafter referred to as photovoltaic panels) have been widely applied and promoted. As an important part of solar power generation, the quality and performance of photovoltaic panels directly affect the power generation efficiency and service life of the solar photovoltaic system. Therefore, it is crucial to accurately and effectively detect the quality and performance of photovoltaic panels. At present, the commonly used photovoltaic panel detection methods mainly include manual detection and automated detection. Manual detection methods are usually visual inspection, microscopic detection and other methods, but manual detection is low in efficiency and unstable in detection results. In comparison, with the development and improvement of machine vision and image processing technology, the use of machine vision technology for automated defect detection has shown the advantages of high efficiency and precision, and has been widely applied in the market.

[0003] Since the detection method of electroluminescence (EL) is easy to cause damage to photovoltaic panels, the detection method of photoluminescence is more preferred. Although certain research results have been achieved for defect detection of photovoltaic panels, there are still many challenges and difficulties. Among them, the non-uniform illumination of photovoltaic panel images and the complexity of defect features such as grid lines and cracks are the difficulties that need to be solved in current research. Some researchers have also proposed partial solutions to these difficulties. Ma Xiaolong et al. proposed an image enhancement algorithm based on regional partial overlap to address the non-uniform illumination phenomenon in collected images, which has small computational complexity and obvious effect. The algorithm obtains the adjustment scale of the regional pixel gray value according to the brightness difference value, then expands the adjustment range of the pixel gray value from the region to the pixel point by using the image interpolation method, and finally adjusts each pixel point to eliminate the non-uniform illumination phenomenon. Hu Wanjie et al. constructed a local low-pass filter to remove the grid lines, then combined the histogram equalization method and the adaptive threshold method for image segmentation, and then used the Two-Pass algorithm to detect the connected domains in the binary segmentation image and marked them with different colors. The connected domains with an area greater than a threshold value were marked as defects. Although the above research has achieved certain results, there are problems such as few detected defect categories and excessive loss of image details.

[0004] By observing the to-be-tested region of the photovoltaic panel, the defects can be divided into two categories, the first category of which contains cracks and missing corners, and the common feature of the defects in this category is that they span multiple grid lines, the growth direction of the cracks can be an arbitrary direction of a broken line, and the missing corners must be at the four corners of the photovoltaic panel and exhibit straight lines or curves. The second category is spots, bad points and the like, and the defects in this category exhibit black spots of different sizes randomly distributed on the photovoltaic panel. In use, the first category of defects has higher destructive power, which is usually caused by internal factors such as lattice defects and material stress, so the detection and evaluation of the first category of defects should be paid more attention to in the quality evaluation of the photovoltaic panel. SUMMARY

[0005] The present application aims at the technical problem in the prior art that the first category of defects is difficult to accurately identify, and provides a defect detection method and system for a photovoltaic panel. The defect detection method provided by the present application proposes a MSRCR enhancement method fused with a high-pass filter, and then processes the photovoltaic panel image, which can more accurately obtain the grid information in the image and reduce the influence of noise as much as possible, obtain a grid line image more suitable for photovoltaic panel detection, and improve the defect detection precision.

[0006] A defect detection method for a photovoltaic panel, comprising the following steps:

[0007] S1: obtaining a photovoltaic panel image to be detected and performing pretreatment;

[0008] S2: performing image enhancement processing on the photovoltaic panel image by using a MSRCR enhancement method fused with a high-pass filter to obtain a grid line image of the photovoltaic panel;

[0009] Wherein, the MSRCR algorithm is first used to enhance the photovoltaic panel image, and then the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum, the frequency spectrum is input into the high-pass filter, and then the filtered frequency spectrum is converted to the spatial domain to obtain a grid line image;

[0010] S3: performing first category defect detection on the grid line image by using a region growing-based defect detection method to obtain a first category defect detection result;

[0011] Wherein, the first category of defects at least contains cracks and missing corners.

[0012] Photoluminescence is the process by which a light source excites the semiconductor material in a photovoltaic panel to produce fluorescence of a specific wavelength. The fluorescence is weaker at defect locations, appearing as darker areas. When photographing a photovoltaic panel with defects, the lighting in the entire image is uneven, leading to increased noise and increasing the likelihood of misjudging the location of grid line breaks during detection. To make the grid lines in the image clearer and easier for subsequent processing, this invention provides an image enhancement process combining high-pass filtering and the MSRCR algorithm to obtain the grid line image of the photovoltaic panel. This technique can improve overall lighting information, effectively enhance image details and texture information, and maintain detection accuracy while reducing computational load.

[0013] Alternatively, the high-pass filter can be represented as follows:

[0014]

[0015] In the above formula, H(u,v) represents the value of the high-pass filter at a point (u,v) on the spectrogram, where 1 indicates passage and 0 indicates rejection; D(u,v) represents the distance of a point (u,v) on the spectrogram from the image center; M and N represent the length and width of the spectrogram, where... is the coordinate of the center of the spectrogram image; d represents the radius of the circle drawn with the center of the spectrogram image, which is a constant; w represents the bandwidth of the high-pass filter, and

[0016] Further, optionally, the defect detection method further includes: performing the following transformation on the grayscale values ​​of the grid line image:

[0017] B(x,y)=f(x,y)I(x,y)

[0018]

[0019] In the formula, B(x,y) represents the gray value of the transformed pixel (x,y), f(x,y) represents the transformation coefficient, and I(x,y) represents the pixel value of the (x,y) coordinate of the pixel in the raster image. Take any point in the raster image and 5 pixels in the up and down directions, and calculate the difference between the mean of the gray values ​​of the 4 pixels other than the center point and the mean of the gray values ​​of the selected 5 pixels, denoted as S. Q and q are the specific values ​​of the transformation coefficient under the corresponding conditions, and satisfy Q > 1 and q < 1.

[0020] Alternatively, the process of using a region-growing-based defect detection method to perform first-type defect detection on the grid line image to obtain the first-type defect detection result is as follows:

[0021] First, a through-line perpendicular to the grid line is generated in the middle of the grid line image;

[0022] Then, a seed point is started at the intersection of the through connection line and the leftmost or rightmost gate line, and growth is performed in three directions of up, down, right and left.

[0023] In the growth process, if the top or bottom position of a gate line is different from the top / bottom position of most of the remaining gate lines in the height direction by more than a set threshold, the corresponding top / bottom position is included in the first type of defect position.

[0024] Further optionally, the defect detection method further comprises:

[0025] The pre-processed photovoltaic panel image is subjected to grid line removal processing.

[0026] The grid line-removed image is then subjected to enhancement processing.

[0027] The enhanced image is subjected to threshold segmentation processing, and then the stain position is determined.

[0028] The stain position and the first type of defect detection result are spliced and fused to obtain a detection result containing the first type of defect and the second type of defect.

[0029] The second type of defect is a black spot of varying size randomly distributed at various positions on the photovoltaic panel.

[0030] Further optionally, the pre-processing of the photovoltaic panel image in step 1 includes using a Hough straight line transformation algorithm to correct the tilt of the photovoltaic panel image, and then cropping the ROI region of the corrected photovoltaic panel image.

[0031] In a second aspect, the present application provides a detection system based on the defect detection method, which comprises:

[0032] An image acquisition and preprocessing module for acquiring and preprocessing a photovoltaic panel image to be detected.

[0033] A grid line image generation module for performing image enhancement processing on the photovoltaic panel image using an MSRCR enhancement method with a fused high-pass filter to obtain a grid line image of the photovoltaic panel.

[0034] In which, the MSRCR algorithm is first used to enhance the photovoltaic panel image, and then the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum, the frequency spectrum is input into the high-pass filter, and then the filtered frequency spectrum is converted to the spatial domain to obtain the grid line image.

[0035] A first type of defect detection module for performing first type of defect detection on the grid line image using a region growth-based defect detection method to obtain a first type of defect detection result; wherein the first type of defect at least includes cracks and missing corners.

[0036] Further optionally, the system further comprises a second type of defect detection module and a fusion module, the second type of defect detection module is configured to detect a second type of defect in the photovoltaic panel image, wherein the second type of defect detection module comprises a grid line removal module, an enhancement module, and a threshold segmentation module;

[0037] The grid line removal module is configured to perform grid line removal processing on the photovoltaic panel image after the preprocessing.

[0038] The enhancement module is configured to perform enhancement processing on the image after the grid line removal.

[0039] The threshold segmentation module is configured to perform threshold segmentation processing on the image after the enhancement processing, and further determine a stain position.

[0040] The fusion module is configured to splice and fuse the stain position and the first type of defect detection result to obtain a detection result containing the first type of defect and the second type of defect.

[0041] The second type of defect is a black spot with different sizes randomly distributed at various positions on the photovoltaic panel.

[0042] In a third aspect, the present application provides an electronic terminal, comprising:

[0043] One or more processors;

[0044] A memory storing one or more computer programs;

[0045] The processor invokes the computer program to perform:

[0046] The steps of a photovoltaic panel defect detection method.

[0047] In a fourth aspect, the present application provides a readable storage medium storing a computer program, the computer program is invoked by a processor to perform:

[0048] The steps of a photovoltaic panel defect detection method.

[0049] Advantages

[0050] 1. The photovoltaic panel defect detection method and system provided by the technical scheme of the present application set the MSRCR enhancement method of the fusion high-pass filter for the first type of defect, and further process the photovoltaic panel image to obtain a grid line image, that is, the grid information in the image can be more accurately obtained, the illumination information is improved, the grid line detail information is locally enhanced, and a grid line image more suitable for photovoltaic panel detection is obtained, and the defect detection precision is improved.

[0051] 2. The technical scheme of the present application is further optimized, and on the basis of the first type of defect detection, the second type of defect detection is also realized, that is, a more comprehensive detection of a type of defects containing cracks and missing corners and another type of defects containing stains and bad points.

[0052] 3. The technical scheme of the present application is further optimized, and a through connection line perpendicular to the grid lines is generated in the middle of the grid line image; and a seed point is started from the intersection of the through connection line and the leftmost or rightmost grid line, and growth is performed in the up, down, right / left three directions. Through the above technical means, the region growth is improved, the blindness of determining the seed point is solved, the calculation amount is reduced, and the detection efficiency is improved. BRIEF DESCRIPTION OF DRAWINGS

[0053] Figure 1 is the overall flowchart of the defect detection method provided by the present application;

[0054] Figure 2 is the pretreatment flowchart of the photovoltaic panel image, wherein (a) is the original image, (b) is the inclination correction schematic diagram, and (c) is the ROI cropping schematic diagram;

[0055] Figure 3 is the image enhancement effect diagram, wherein (a), (b), (c), (d), (e) and (f) correspond to the original image, the effect diagram after logarithmic transformation, the effect diagram after histogram equalization, the effect diagram after CLAHE processing, the effect diagram after SSR processing and the effect diagram after MSRCR processing, respectively;

[0056] Figure 4 is the high-pass filter schematic diagram, wherein (a) and (b) correspond to the traditional high-pass filter and the high-pass filter of the present application, respectively;

[0057] Figure 5 is the grid line extraction effect diagram based on high-pass filtering, wherein (a), (b) and (c) correspond to the photovoltaic cell spectrum diagram, the effect diagram after frequency spectrum filtering and the schematic diagram of extracting grid lines, respectively;

[0058] Figure 6 is the image enhancement result diagram fused with high-pass filtering, wherein (a), (b), (c) and (d) correspond to the original image, the schematic diagram of linear superposition processing, the schematic diagram of difference processing and the schematic diagram of the method of the present application, respectively;

[0059] Figure 7 is the grid line connection diagram, wherein (a) and (b) correspond to the schematic diagram of the through connection line and the first type of defect and the drawing schematic diagram of the through connection line, respectively;

[0060] Figure 8The first type of defect detection result image is shown, where (a), (b), (c), and (d) correspond to the schematic diagram of region growth, the image to be segmented, the region growth result, and the schematic diagram of the first type of defect detection result, respectively.

[0061] Figure 9 The final defect detection result image is shown below. (a), (b), (c), (d), (e), and (f) correspond to the original image, the schematic diagram of the grid extraction, the schematic diagram after SSR enhancement, the schematic diagram after threshold segmentation, the schematic diagram of blemish localization, and the schematic diagram of the fusion localization result, respectively.

[0062] Figure 10 The images are comparisons of image enhancement effects, where (a), (b), and (c) correspond to the original image, MSRCR, and the method described in this invention, respectively. Detailed Implementation

[0063] This invention provides a defect detection method for photovoltaic panels. Its core is an MSRCR enhancement method that integrates a high-pass filter, which processes the photovoltaic panel image to obtain a grid line image. This grid line image improves illumination information and locally enhances grid line detail information. The invention will be further described below with reference to embodiments.

[0064] Example 1:

[0065] The defect detection method for photovoltaic panels provided in this embodiment of the invention includes the following steps:

[0066] S1: Acquire the image of the photovoltaic panel to be inspected and perform preprocessing.

[0067] like Figure 2 As shown, during actual image capture of photovoltaic panels, some panels are improperly placed, resulting in a certain degree of tilt. To facilitate subsequent processing, this tilt needs to be corrected. Since the photoluminescence detection images are all captured in a dark room, all areas except the photovoltaic panels are black. Using the Hough linear transform algorithm, the edge information on all four sides can be effectively extracted from the photovoltaic panel image. By calculating the average slope of the four edges, a better image correction effect can be obtained. Finally, the ROI (region of interest) is determined based on the coordinates of the four corner points of the photovoltaic panel in the corrected image and then cropped.

[0068] Since the Hough linear transform algorithm for image correction and ROI region cropping are existing technologies, they will not be described in detail.

[0069] S2: The photovoltaic panel image is enhanced using the MSRCR enhancement method with a fused high-pass filter to obtain the grid line image of the photovoltaic panel; wherein, as shown... Figure 5As shown, the photovoltaic panel image is first enhanced using the MSRCR algorithm, then the enhanced image is converted from the spatial domain to the frequency domain to obtain a spectrum. The spectrum is then input into the high-pass filter, and finally the filtered spectrum is converted back to the spatial domain to obtain a grating image.

[0070] A high-pass filter is represented as follows:

[0071]

[0072] In the above formula, H(u,v) represents the value of the high-pass filter at a point (u,v) on the spectrogram, where 1 indicates passage and 0 indicates rejection; D(u,v) represents the distance of a point (u,v) on the spectrogram from the image center; M and N represent the size of the spectrogram, i.e., its length and width. is the coordinate of the center of the spectrogram image; d represents the radius of the circle drawn with the center of the spectrogram image, which is a constant; w represents the bandwidth of the high-pass filter, and Experiments showed that the optimal parameters for extracting high-frequency components from photovoltaic cells using this high-pass filter were w=5 and d=55, which demonstrated good extraction performance.

[0073] This invention combines a high-pass filter and MSRCR to obtain images more suitable for photovoltaic panel inspection. To prevent the grid lines from affecting existing defects during fusion, a weighted fusion method is designed based on the texture structure characteristics of the grid lines and the first type of defect. In defective areas, the grayscale value of the grid lines is significantly weaker than that of normal areas. Since the grid lines are located vertically, five pixels are taken from any point in the image, plus five pixels in the vertical direction. The difference between the mean grayscale value of the four pixels excluding the center point and the mean grayscale value of the five pixels is calculated and denoted as S. The calculation formula is as follows:

[0074]

[0075] In the above formula, I(x,y) represents the pixel value at coordinates (x,y) in the raster image. Since the grayscale value of the defect area is significantly lower than that of the raster area, when S is greater than a certain threshold λ (empirical value), it is considered that the location is a possible defect. The value of λ is determined experimentally. Due to uneven illumination, the raster extracted from the original image is a line with uneven brightness. Therefore, when S is approximately equal to 0, it represents the background area other than the raster and the defect. When S is other values, it represents the raster. Based on this characteristic, the grayscale value of the image after MSRCR enhancement is transformed using the following formula:

[0076] B(x,y)=f(x,y)I(x,y)

[0077]

[0078] In the above formula, B(x, y) represents the gray value of the pixel point (x, y) after transformation, and f(x, y) represents the transformation coefficient. For the defective position, the gray value is increased, so that the defective position looks a little brighter; for the normal part of the grid line, the gray value is reduced, so that the grid line looks darker; and for the background part, the gray value is not changed, so that the grid line of the entire image is more prominent. It should be noted that the specific value of the above transformation coefficient is the best example of the present application, but the present application is not limited thereto, and in other feasible examples, the adjustment is made based on the principle of "increasing the gray value for the defective position; reducing the gray value for the normal part of the grid line; and not changing the gray value for the background part". In addition, the standard of S being approximately equal to 0 is set according to the actual required accuracy, that is, the approximately equal range centered on 0 is set according to the accuracy requirement, and the present application does not limit the specific value thereof.

[0079] The MSRCR enhancement method fused with the high-pass filter in the technical scheme of the present application is considered as follows:

[0080] Photoluminescence is the generation of fluorescent light of a certain wavelength by exciting the semiconductor material in the photovoltaic panel by a light source. The fluorescent light is weak at the defective position, which appears as a darker area. When there is a photovoltaic panel with a shooting defect, the illumination of the entire image is not uniform, which will increase the noise in the image and easily cause misjudgment of the grid line fracture position during detection. In order to make the grid line in the image clearer and easier to process in the subsequent steps. However, histogram equalization, logarithmic change, CLAHE, SSR and MSRCR in the image enhancement method cannot achieve better results. Among them, logarithmic transformation can only process the darker parts around the image, but has little effect on the enhancement of the grid line, and instead makes the similarity between the grid line of the high-light area and the background higher. Histogram equalization, CLAHE and SSR have excellent enhancement effect on the grid line, but the processing of the dark part around the photovoltaic panel is not in place, which produces a halo and affects the normal area of the photovoltaic panel, which easily leads to misjudgment of the edge area defect. MSRCR suppresses the appearance of the halo and retains most of the details of the image, but the enhancement of the grid line in the image is still insufficient and the differentiation from the background is not high. In addition, the traditional high-pass filter can pass high frequencies and filter or attenuate low frequencies, so as to sharpen the image and highlight the boundary. However, due to the uneven illumination in the photovoltaic panel image, if the traditional high-pass filter is used, the noise information in the image will be enlarged during enhancement, which will affect the final detection result. The specific effects of various algorithms are shown in Table 1. Figure 3 Therefore, the MSRCR enhancement method fused with the high-pass filter is proposed in the present application, which can more accurately obtain the grid information in the image and reduce the influence of noise as much as possible.

[0081] It needs to be explained that, compared with some other image fusion methods, the linear superposition does not consider the characteristics of the photovoltaic panel image, and the defect area is over-enhanced. The difference fusion method emphasizes the grid line features too much. The method of the present application comprehensively considers the two points, and according to the characteristics of the photovoltaic panel, the image is fused by using the above process, so that a better image enhancement effect can be obtained.

[0082] S3: a defect detection method based on region growing is used to perform first-class defect detection on the grid line image to obtain a first-class defect detection result. First, a through connection line perpendicular to the grid line is generated in the middle of the grid line image; and then a seed point is started at the intersection of the through connection line and the leftmost or rightmost grid line, and growth is performed in the upward, downward and right / left directions. In the growth process, if the top or bottom position of the grid line is different from the top / bottom position of the remaining majority of the grid lines in the height direction by more than a threshold value, the corresponding top / bottom position is included in the first-class defect position.

[0083] The first-class defect will cause the photovoltaic panel to have a broken grid, that is, the grid bar is disconnected at a position. The edge detection combined with the straight line detection method will generate many fragmented line segments, which are difficult to distinguish, and greatly interfere with the judgment of the defect position, and the robustness is not strong. In order to accurately extract the first-class defect position, the region growing method is used to find the breakpoint along the grid line. When the grid can be extended to the top and bottom, it indicates that there is no defect at this position. If it cannot be reached, the disconnected position is the coordinate of the first-class defect. Due to different images, the grid line spacing will be slightly different, and it is quite difficult to find the seed point of each grid line.

[0084] Therefore, as shown in Figure 7 The present application generates a horizontal line perpendicular to the grid line in the center of the image. The connection line penetrates all the grid lines of the photovoltaic panel, and the gray value is equivalent to that of the grid line. The first-class defect may be distributed above, below or on the through connection line. However, regardless of which way, the region growing will be blocked by the crack defect, so it is only necessary to judge the top and bottom positions of the grid line to know the area of the first-class defect. Since the distribution of the through connection line and the grid line is known, according to this prior information, the time and space overhead of region growing can be greatly reduced.

[0085] As shown in Figure 8 The present embodiment takes the intersection of the connection line and the leftmost grid line as the starting seed point, and only needs to grow in the upward, downward and right directions. When the coordinates of a top point are too far from the coordinates of other top points, it can be judged that the point is included in the first-class defect position.

[0086] Therefore, if the top or bottom position of a grid line in the growth process is different from the top / bottom position of the majority of the remaining grid lines by more than a threshold value (an empirical value) in the height direction, the corresponding top / bottom position is included in the first type of defect position. Among them, "the majority of the remaining grid lines" refers to the majority of the grid lines whose top / bottom positions are close, and the top / bottom position of this part of the grid lines is used as a standard to determine whether the top / bottom of the grid line in the growth process is a defective part.

[0087] S4: Perform grid line removal processing on the pre-processor photovoltaic panel image; then perform enhancement processing on the grid line removed image. Among them, the SSR image enhancement processing is preferred.

[0088] S5: Perform threshold segmentation processing on the enhanced image to determine the stain position; wherein the stain position and the first type of defect detection result are spliced and fused to obtain a detection result containing the first type of defect and the second type of defect.

[0089] In the photovoltaic panel, there are also second type of defects such as bad points. Since these points are not necessarily on the grid lines, the previous method cannot obtain good detection results. In order to better detect stains, the present application uses the method of threshold segmentation after grid line removal to find the stains on the photovoltaic panel. Considering that the stains in the image also belong to high-frequency information, the image grid removal step will cause this part of the information to be weakened, so it is necessary to perform image enhancement after grid removal. Therefore, using SSR enhancement after grid removal can highlight the positions of each stain, and using threshold segmentation can separate the low gray value stain positions from the high gray value background, and find the positions of each stain. Similarly, the first type of defect will also be detected in this step, and it is fused with the result of region growing to remove the overlapping part with the first type of defect, so as to obtain all the defect positions of the photovoltaic panel and complete the photovoltaic panel defect detection. The effect diagram of the specific detection process can be seen in Figure 9 .

[0090] It should be noted that image grid removal, threshold segmentation processing, and determining the stain position according to the image after threshold segmentation processing, i.e. how to determine the stain position, are all prior art means, so they will not be specifically stated.

[0091] Verification:

[0092] The photoluminescence image data of the photovoltaic cell is collected as a sample, the main defects in the sample image are corner defects, cracks, stains and other defects, and the defects are divided into two categories in the application. Two experiments are carried out to verify the defect detection method. In the first group of experiments, different photovoltaic panel images are selected for enhancement, and the image enhancement method of the application is evaluated by subjective comparison and objective evaluation index. The second group of experiments is based on the existing data set to propose a defect detection evaluation index, and the defect detection effect of other algorithms is compared to prove the feasibility of the method proposed in the application.

[0093] In the image enhancement experiment, part of the images are randomly selected to compare the improved algorithm with the classical algorithm. MSRCR can effectively enhance the image detail information by using the multi-scale Retinex algorithm to enhance the image contrast, but in the photovoltaic panel detection, the panel is mainly white, and the algorithm is difficult to adapt to the scene change of each photovoltaic panel in the parameter, and the enhancement effect of the grid line is not good. Compared with MSRCR, the algorithm of the application strengthens the column direction texture of the photovoltaic panel, and there is no loss of detail. In order to better evaluate the enhancement effect, the image is judged by using the no-reference objective evaluation index NIQE with strong theoretical basis, which establishes a series of characteristics that can represent the image quality, and uses a multivariate Gaussian model to describe it, and the mathematical expression is:

[0094]

[0095] In the above formula, v1, v2 and ∑1, ∑2 are the mean vector and covariance matrix of the multivariate Gaussian model of the natural image and the multivariate Gaussian model of the distorted image, and the smaller the value is, the better the quality of the image is. Five groups of images are selected for comparison test, as shown in Table 1, and the NIQE value of each experiment is lower than that of MSRCR, which shows the superiority of the method in improving the image quality.

[0096] In the defect detection experiment, in order to evaluate the defect detection effect, the effect is judged by counting, and the number of the first and second defects is recorded, and compared with the defect detection method of removing the grid line. As the same as the image enhancement, five groups of images are used for comparison experiment, and the results are shown in Table 2: compared with the grid removal method, the accuracy of the first defect detection is improved by 11.76%, and the accuracy of the two defects is 84.1%. Thus, the reliability of the defect detection method is further proved.

[0097] Table 1

[0098]

[0099] Table 2

[0100]

[0101] It should be noted that in some embodiments, the first type of defects is detected, and thus the defect detection method is composed of steps S1-S3; in other embodiments, in addition to detecting the first type of defects in the manner of steps S1-S3, other existing methods or the manner of S4-S5 can also be used to detect the second type of defects, and the present application does not make specific limitations thereon. In addition, the execution of S4 and S5 can be synchronized with S2-S3 or executed before S2-S3.

[0102] Embodiment 2:

[0103] The embodiment provides a detection system based on the defect detection method, which comprises: an image acquisition and preprocessing module, a grid line image generation module, and a first type of defect detection method.

[0104] The image acquisition and preprocessing module is configured to acquire and preprocess a photovoltaic panel image to be detected. The grid line image generation module is configured to perform image enhancement processing on the photovoltaic panel image by using an MSRCR enhancement method fused with a high-pass filter to obtain a grid line image of the photovoltaic panel. Specifically, the MSRCR algorithm is first used to enhance the photovoltaic panel image, and then the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum image. The frequency spectrum image is input into the high-pass filter, and then the filtered frequency spectrum image is converted to the spatial domain to obtain the grid line image. The first type of defect detection module is configured to perform first type of defect detection on the grid line image by using a region growing-based defect detection method to obtain a first type of defect detection result. The first type of defect at least includes cracks and missing corners.

[0105] In other feasible embodiments, the detection system further comprises a second type of defect detection module and a fusion module. The second type of defect detection module is configured to detect a second type of defect in the photovoltaic panel image. The second type of defect detection module comprises a grid line removal module, an enhancement module, and a threshold segmentation module. The grid line removal module is configured to perform grid line removal processing on the photovoltaic panel image after preprocessing. The enhancement module is configured to perform enhancement processing on the grid line removed image. The threshold segmentation module is configured to perform threshold segmentation processing on the enhanced image to determine a stain position. The fusion module is configured to splice and fuse the stain position and the first type of defect detection result to obtain a detection result containing the first type of defect and the second type of defect.

[0106] The implementation process of each specific module can refer to the above method, which will not be repeated here. It should be understood that the division of the above functional modules is only a logical division, and there can be another division way in actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. At the same time, the integrated unit can be realized in the form of hardware or software function unit. For example, the hardware for obtaining the photovoltaic panel image is a camera / camera.

[0107] Embodiment 3:

[0108] The embodiment provides an electronic terminal, which includes one or more processors and a memory storing one or more computer programs; wherein the processor invokes the computer program to perform the steps of a photovoltaic panel defect detection method. The specific implementation is as follows:

[0109] S1: obtaining a photovoltaic panel image to be detected and preprocessing.

[0110] S2: performing image enhancement processing on the photovoltaic panel image by using an MSRCR enhancement method with a fusion high-pass filter to obtain a grid line image of the photovoltaic panel; wherein the MSRCR algorithm is first used to enhance the photovoltaic panel image, then the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum, the frequency spectrum is input into the high-pass filter, and finally the filtered frequency spectrum is converted to the spatial domain to obtain the grid line image.

[0111] S3: performing first-class defect detection on the grid line image by using a region growing-based defect detection method to obtain a first-class defect detection result.

[0112] In some implementations, the following is also performed:

[0113] performing grid line removal processing on the photovoltaic panel image after preprocessing; then performing enhancement processing on the image after removing the grid line; and performing threshold segmentation processing on the image after enhancement processing, thereby determining a stain position; wherein the stain position and the first-class defect detection result are spliced and fused to obtain a detection result containing the first-class defect and the second-class defect.

[0114] The specific implementation process refers to the related statements of Embodiment 1.

[0115] The memory can include a high-speed RAM memory, and can also include a non-volatile memory, such as at least one disk memory.

[0116] If the memory, the processor is independently implemented, the memory, the processor and the communication interface can be connected with each other through a bus and complete the communication between each other. The bus can be an industry standard architecture bus, an external device interconnection bus or an extended industry standard architecture bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0117] Optionally, in a specific implementation, if the memory, the processor is integrated on a chip, the memory, the processor can complete the communication between each other through an internal interface.

[0118] It should be understood that, in the embodiments of the present application, the processor can be a central processing unit (CPU), and the processor can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The memory can include read-only memory and random access memory, and provide instructions and data for the processor. A part of the memory can also include non-volatile random access memory. For example, the memory can also store device type information.

[0119] Embodiment 4:

[0120] The embodiment provides a readable storage medium which stores a computer program, the computer program is called by a processor to realize steps of a defect detection method of a photovoltaic panel. In a specific implementation:

[0121] S1: obtaining a photovoltaic panel image to be detected and performing preprocessing.

[0122] S2: adopting an MSRCR enhancement method of a fusion high-pass filter to perform image enhancement processing on the photovoltaic panel image to obtain a grid line image of the photovoltaic panel; wherein the MSRCR algorithm is first used to enhance the photovoltaic panel image, then the enhanced image is converted from a spatial domain to a frequency domain to obtain a frequency spectrum, the frequency spectrum is input into the high-pass filter, and finally the filtered frequency spectrum is converted to the spatial domain to obtain the grid line image.

[0123] S3: adopting a defect detection method based on region growing to perform first-class defect detection on the grid line image to obtain a first-class defect detection result.

[0124] In some implementations, the following are also performed:

[0125] The pre-processor photovoltaic panel image is subjected to grid line removal processing, and then the image after the grid line removal is subjected to enhancement processing. The image after the enhancement processing is subjected to threshold segmentation processing, and then the location of the stain is determined. The location of the stain is spliced and fused with the first type of defect detection result to obtain a detection result containing the first type of defect and the second type of defect.

[0126] The specific implementation process is described with reference to the related statements of Embodiment 1.

[0127] The readable storage medium is a computer readable storage medium, which can be an internal storage unit of the controller, such as a hard disk or a memory of the controller. For example, the terrain feature model constructed in the present application exists in the hard disk, and then the computer program for performing the fusion step is stored in the memory, so that the fusion process is realized by relying on the memory. The readable storage medium can also be an external storage device of the controller, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. Further, the readable storage medium can include both the internal storage unit of the controller and the external storage device. The readable storage medium is used to store the computer program and other programs and data required by the controller. The readable storage medium can also be used to temporarily store data that has been output or will be output.

[0128] Based on this understanding, the technical solutions of the present application, essentially or the part that contributes to the prior art, or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute all or part of the steps of the method described in various embodiments of the present application. The aforementioned readable storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0129] The computer program instructions can also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart block or blocks or in conjunction with the flowcharts described above. Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks Figure 1 one or more flowcharts and / or blocks

[0130] It should be emphasized that the herein described examples of the application are illustrative and not restrictive, since the application is not limited to the examples described in the detailed description. Other embodiments can be derived from the description by a person skilled in the art without departing from the scope of the application, whether modified or replaced, and they also belong to the protection scope of the application.

Claims

1. A method for defect detection of a photovoltaic panel, characterized in that: The method comprises the following steps: S1: obtaining and preprocessing a photovoltaic panel image to be detected; S2: performing image enhancement processing on the photovoltaic panel image by using an MSRCR enhancement method fused with a high-pass filter to obtain a grid line image of the photovoltaic panel; Wherein, the MSRCR algorithm is used to enhance the photovoltaic panel image first, and then the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum, the frequency spectrum is input into the high-pass filter, and finally the filtered frequency spectrum is converted to the spatial domain to obtain the grid line image; S3: performing first-type defect detection on the grid line image by using a region growing-based defect detection method to obtain a first-type defect detection result, wherein the first-type defect at least includes cracks and missing corners; The high-pass filter is expressed as follows: In the above formula, H(u, v) represents the value of the high-pass filter at a point (u, v) on the spectrum diagram, 1 represents pass, and 0 represents not pass; D(u, v) represents the distance of a point (u, v) in the spectrum diagram from the image center, M and N represent the length and width of the spectrum diagram, wherein, is the coordinate of the image center of the spectrum diagram; d represents the radius of a circle drawn with the image center of the spectrum diagram, and is a constant; w represents the bandwidth of the high-pass filter, and 2. The defect detection method of claim 1, wherein: Further comprising the following transformation on the gray value of the grid line image: B(x, y) = f(x, y)I(x, y) In the formula, B(x, y) represents the gray value of the pixel point (x, y) after transformation, f(x, y) represents a transformation coefficient, and I(x, y) represents the pixel value of the pixel point (x, y) in the grid line image; the average value of the gray values of four pixel points except the center point is calculated, and the difference between the average value of the gray values of the five selected pixel points and the average value of the gray values of the four pixel points except the center point is denoted as S; Q and q are specific values of the transformation coefficient under corresponding conditions, and Q is greater than 1 and q is less than 1; and λ is a threshold value.

3. The defect detection method of claim 1, wherein: The process of performing first-type defect detection on the grid line image by using the region growing-based defect detection method to obtain a first-type defect detection result is as follows: First, a through connection line perpendicular to the grid lines is generated in the middle of the grid line image; Then, the intersection of the through connection line and the leftmost or rightmost grid line is taken as a seed point, and growth is performed in the upward, downward, right and left directions; Wherein, if the top or bottom position of a grid line is different from the top / bottom positions of most remaining grid lines in the height direction by more than a set threshold value, the top / bottom position is included in the first-type defect position.

4. The defect detection method of claim 1, wherein: Further comprising: performing grid line removal processing on the photovoltaic panel image after preprocessing; Then, performing enhancement processing on the image after the grid line removal; performing threshold segmentation processing on the image after the enhancement processing, and then determining a stain position; splicing and fusing the stain position and the first-type defect detection result to obtain a detection result containing the first-type defect and a second-type defect; Wherein, the second-type defect is a black spot with different sizes randomly distributed on each position of the photovoltaic panel.

5. The defect detection method of claim 1, wherein: The preprocessing of the photovoltaic panel image in step 1 comprises: performing tilt correction on the photovoltaic panel image by using a Hough straight line transformation algorithm; and then performing ROI region cropping on the corrected photovoltaic panel image.

6. A system for defect detection of a photovoltaic panel, characterized by: Comprise: An image acquisition and preprocessing module for obtaining and preprocessing a photovoltaic panel image to be detected; A grid line image generation module for performing image enhancement processing on the photovoltaic panel image by using an MSRCR enhancement method fused with a high-pass filter to obtain a grid line image of the photovoltaic panel; The MSRCR algorithm is used to enhance the photovoltaic panel image, and the enhanced image is converted from the spatial domain to the frequency domain to obtain a frequency spectrum. The frequency spectrum is input into the high-pass filter, and the filtered frequency spectrum is converted to the spatial domain to obtain a grid line image. The first-type defect detection module is configured to detect the grid line image to obtain a first-type defect detection result by using a region growing-based defect detection method. The first-type defect includes at least cracks and missing corners. The high-pass filter is expressed as follows: In the above formula, H(u, v) represents the value of the high-pass filter at a point (u, v) on the spectrum diagram, 1 represents pass, and 0 represents not pass; D(u, v) represents the distance of a point (u, v) in the spectrum diagram from the image center, M and N represent the length and width of the spectrum diagram, wherein, is the coordinate of the image center of the spectrum diagram; d represents the radius of a circle drawn with the image center of the spectrum diagram, and is a constant; w represents the bandwidth of the high-pass filter, and 7. The defect detection system of claim 6, wherein: The second-type defect detection module is configured to detect a second-type defect in the photovoltaic panel image. The second-type defect detection module includes a grid line removal module, an enhancement module, and a threshold segmentation module. The grid line removal module is configured to remove grid lines from the photovoltaic panel image after preprocessing. The enhancement module is configured to enhance the image after the grid lines are removed. The threshold segmentation module is configured to perform threshold segmentation on the enhanced image to determine a stain position. The fusion module is configured to splice and fuse the stain position and the first-type defect detection result to obtain a detection result including the first-type defect and the second-type defect. The second-type defect is a black spot randomly distributed at various positions on the photovoltaic panel and having different sizes.

8. An electronic terminal, characterized by: One or more processors A memory storing one or more computer programs The processor invokes the computer program to perform The steps of the defect detection method according to any one of claims 1-5. A computer program is stored in the memory and is invoked by the processor to perform 9. A readable storage medium characterized by: The steps of the defect detection method according to any one of claims 1-5. ​

Images