Photovoltaic cell crack detection method and system based on image processing

Abstract

The present invention relates to the field of image processing technology, and in particular to a method and system for detecting cracks in a photovoltaic cell based on image processing; the method comprises: collecting a photoluminescence image of a photovoltaic cell; obtaining a plurality of super-pixel blocks of the image; calculating a saliency value of each super-pixel block; marking the crack defect level of a super-pixel block when the saliency value is greater than a first set value as high; marking the crack defect level of a super-pixel block when the saliency value is less than or equal to the first set value and greater than a second set value as medium; that is, the solution of the present invention can accurately detect cracks in a photovoltaic cell.

Description

Technical Field

[0001] The present invention relates to the field of image processing technology, and more specifically, to a method and system for detecting cracks in a photovoltaic cell based on image processing. Background Art

[0002] With the rapid development of the photovoltaic industry, the output of photovoltaic cells has also been greatly increased.

[0003] During the production process of photovoltaic panels, problems such as hidden cracks, uneven diffusion, and virtual printing may occur. The existence of these defects may affect the photoelectric conversion efficiency of photovoltaic cells, reduce the battery life, and affect the stability of the photovoltaic system. Therefore, it is necessary to detect defects in photovoltaic cells.

[0004] Taking cracks in photovoltaic cells as an example, in the existing technology, PL (Photoluminescence) detection technology is usually used to detect and analyze various process links such as silicon wafers, diffusion, etching, electrode printing, and cells, so as to quickly and effectively locate problems in the production process and provide reliable guarantee for product quality.

[0005] Figure 1 The image of the photovoltaic cell is obtained by using the PL photoluminescence principle. Since defects will cause the minority carrier concentration in the area to decrease and thus weaken its fluorescence effect, it will appear as dark dots, lines or certain areas after imaging. Therefore, by collecting the photoluminescence image of the photovoltaic cell, although it can show tiny cracks, micro defects or other non-uniform areas on the photovoltaic cell that may be overlooked, the cracks are too small to be easily noticed and therefore easily missed during detection.

[0006] At the same time, although different crack levels may not directly affect the appearance of the battery, they may cause different changes in optical and electrical properties. Therefore, it is particularly important to detect the degree of crack defects.

[0007] In the related technology, such as the patent application document with application publication number CN109447978A and titled "A method for classifying electroluminescent image defects of photovoltaic solar cells", a classifier is used to classify defects of photovoltaic cells. However, it is highly dependent on the data set. If the data set is not comprehensive or has deviations, the accuracy of the classification of defects of photovoltaic cells will be affected. Summary of the invention

[0008] The purpose of the present invention is to propose a photovoltaic cell crack detection method and system based on image processing to solve the problem that the defects of photovoltaic cells may not be accurately classified in the prior art; to this end, the present invention provides solutions in the following two aspects.

[0009] In a first aspect, the present invention provides a photovoltaic cell crack detection method based on image processing, comprising:

[0010] Collecting images of photoluminescence from photovoltaic cells;

[0011] Acquire multiple superpixel blocks of the image;

[0012] Calculate the saliency value of each superpixel block; mark the crack defect level of the superpixel block whose saliency value is greater than a first set value as high; mark the crack defect level of the superpixel block whose saliency value is less than or equal to the first set value and greater than a second set value as medium;

[0013] The significant values ​​are: ;in, , , are the morphological coefficient and dissimilarity value of the i-th superpixel block and the j-th superpixel block, respectively. , are the distribution characteristic values ​​of the ith superpixel block and the jth superpixel block, respectively. K is the number of superpixel blocks within the target range of the ith superpixel block. is the normalized value of the color Euclidean distance between the i-th superpixel block and the j-th superpixel block in the LAB space, exp( ) is an exponential function; the distribution characteristic value is inversely correlated with the similarity between the corresponding superpixel block and the image, and the ratio of the number of pixels in the corresponding superpixel block to the maximum number of pixels in all superpixel blocks; the morphological coefficient represents the shape similarity of the two superpixel blocks; the target range is the set of the first K other superpixel blocks whose dissimilarity values ​​between the selected i-th superpixel block and all other superpixel blocks are sorted from small to large, K≥2.

[0014] In the above scheme, the image of the photovoltaic cell is segmented by thresholding, and the image of the photovoltaic cell is segmented into superpixels according to the segmentation result. The distribution characteristic value of each superpixel block and the morphological coefficient between each superpixel block and its minimum circumscribed rectangle are calculated according to the pixel points in each superpixel block in the image to measure the characteristic differences between the superpixel blocks. That is, the present invention can make the characteristics of the image block containing cracks more prominent based on the appearance characteristics of the photovoltaic cell and the distribution characteristics in each superpixel block, which is beneficial to the detection of cracks on the cell.

[0015] Optionally, the distribution characteristic value for: ; is the similarity between the i-th superpixel block and the image, is the number of pixels in the ith superpixel block, N is the maximum number of pixels in all superpixel blocks; the similarity is the cosine similarity between the grayscale feature vector and the grayscale reference vector, the grayscale feature vector is a sequence composed of the number of different grayscale values ​​of pixels in the superpixel block, and the grayscale reference vector is a sequence composed of the number of different grayscale values ​​of pixels in the image.

[0016] In the above scheme, by obtaining the similarity between each super pixel block and the image, the distribution characteristic value of the attribute of each super pixel block itself can be characterized.

[0017] Optionally, the morphology coefficient is a first morphology coefficient , specifically:

[0018] ;

[0019] in, is the distribution characteristic value of the i-th superpixel block, is the distribution characteristic value of the jth superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

[0020] In the above scheme, the similarity between two super pixel blocks can be accurately determined by the height, width, distribution characteristic value and the ratio of the area where the pixel block is located to the area of ​​its minimum circumscribed rectangle.

[0021] Optionally, the morphology coefficient is a second morphology coefficient , specifically:

[0022] ;

[0023] in, , is the grayscale mean of the i-th superpixel block and the j-th superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

[0024] The above scheme can accurately determine the similarity between two super pixel blocks through the height, width, grayscale mean and the ratio of the area where the pixel block is located to the area of ​​its minimum circumscribed rectangle.

[0025] Optionally, the specific process of obtaining the multiple super pixel blocks of the image is:

[0026] Segmenting the image to obtain a segmentation result;

[0027] Each connected region in the segmentation result is obtained through a connected region marking algorithm; the center point of each connected region is used as the seed point for superpixel segmentation, and the image of the photovoltaic cell is superpixel segmented to obtain the corresponding superpixel block, wherein the coordinates of the center point are the mean of the horizontal coordinates and the mean of the vertical coordinates of all the pixel points in the connected region.

[0028] The above scheme first segments the image, and can quickly obtain suitable seed points to perform superpixel segmentation of the image.

[0029] Optionally, the multiple super pixel blocks of the image are directly obtained by using a simple linear iterative clustering method.

[0030] Optionally, the segmentation result is obtained by adopting Otsu algorithm.

[0031] In a second aspect, a photovoltaic cell crack detection system based on image processing comprises:

[0032] processor;

[0033] A memory stores computer instructions for photovoltaic cell crack detection based on image processing. When the computer instructions are executed by the processor, the system executes the above-mentioned photovoltaic cell crack detection method based on image processing.

[0034] The beneficial effects of the present invention are:

[0035] The solution of the present invention determines the distribution characteristic value and morphological coefficient of each superpixel block based on the appearance characteristics of the photovoltaic cell to calculate the significance value of each superpixel block, which can make the characteristics of the image block corresponding to the superpixel block containing cracks more prominent, which is beneficial to the detection of cracks on the photovoltaic cell. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] By reading the following detailed description with reference to the accompanying drawings, the above and other objects, features and advantages of the exemplary embodiments of the present invention will become readily understood. In the accompanying drawings, several embodiments of the present invention are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0037] Figure 1 An image of the photoluminescence of a photovoltaic cell is schematically shown;

[0038] Figure 2 The flowchart schematically shows the steps of the photovoltaic cell crack detection method based on image processing in this embodiment;

[0039] Figure 3 A schematic diagram schematically shows the connected areas in this embodiment;

[0040] Figure 4 A schematic diagram of super-pixel blocks after super-pixel segmentation of an image in this embodiment is shown;

[0041] Figure 5 The structural block diagram of the photovoltaic cell crack detection system based on image processing in this embodiment is schematically shown. DETAILED DESCRIPTION

[0042] 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 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 those skilled in the art without creative work are within the scope of protection of the present invention.

[0043] The scenario targeted by the present invention is photovoltaic cells without obvious defects. Photovoltaic cells without obvious defects can be suspected qualified cells obtained by manually screening out relatively obvious defects (such as cracks, pollution, scratches, black spots, etc.) in advance.

[0044] like Figure 2 As shown, the photovoltaic cell crack detection method based on image processing in this embodiment includes the following steps:

[0045] Step S1, collecting a photoluminescence image of a photovoltaic cell.

[0046] In this embodiment, a PL detector (including a laser emitting device and an image acquisition device) is arranged at a suitable angle, and its working principle is as follows: the laser emitting device is turned on, the light source emits near-infrared light to the photovoltaic cell, and the image acquisition device performs image acquisition to capture the light emitted by the photovoltaic cell. The image acquisition device is a camera.

[0047] Step S2, obtaining multiple super pixel blocks of the image.

[0048] In one embodiment, the specific process of obtaining multiple super pixel blocks is:

[0049] First, the image is segmented to obtain a segmentation result. Specifically, the image of the photovoltaic cell is segmented using the Otsu algorithm to obtain a segmentation result.

[0050] Secondly, the connected region marking algorithm is used to obtain each connected region in the segmentation result. Since the pixels in the segmentation result are actually only black and white, it is necessary to use the connected region marking algorithm to obtain each connected region in the segmentation result (the region where the white points are located), such as Figure 3 shown.

[0051] Among them, the connected component labeling algorithm is an existing technology and will not be described in detail here.

[0052] Then, the center point of each connected area is used as the seed point for superpixel segmentation, and the image of the photovoltaic cell is segmented into superpixels to obtain each superpixel block, such as Figure 4 The coordinates of the center point are the average of the horizontal coordinates and the average of the vertical coordinates of all the pixels in the connected area.

[0053] The above-mentioned superpixel segmentation can adopt the SLIC (simple linear iterative clustering) method; since the superpixel segmentation process is an existing technology, it will not be introduced in detail here.

[0054] Since the grid lines are evenly distributed on the surface of the photovoltaic cell, the grid lines in the image evenly divide the photovoltaic cell into different areas (usually the parts of the image with the same structure of the cell should have similar structures). Therefore, this feature can be used to perform threshold segmentation on the image of the photovoltaic cell for subsequent saliency detection.

[0055] In another embodiment, the image does not need to be segmented, and the SLIC method can be directly used to perform superpixel segmentation on the image; specifically, the image is evenly divided into multiple initial cluster center grids, and each pixel is assigned to the nearest cluster center by combining the distance metric of spatial and color information, and then the cluster center is updated until the cluster center changes insignificantly, thereby obtaining multiple superpixel blocks.

[0056] It should be noted that the difference between the above two embodiments lies in the method of obtaining the initial seed point (ie, the cluster center).

[0057] Step S3, using the CA algorithm to calculate the saliency value of each superpixel block.

[0058] Among them, the CA (Context-Aware) algorithm is a saliency detection algorithm based on local features and global features. When calculating the saliency value, it first divides the image at a certain scale into K image blocks, then converts the image from RGB space to Lab space, and calculates the color Euclidean distance and spatial Euclidean distance between any two regions, and finally obtains the difference between the two regions (proportional to the color Euclidean distance and inversely proportional to the spatial Euclidean distance). The greater the difference between the regions, the greater the saliency value of the pixels in the region. Among them, the formula of the saliency value in the existing CA algorithm is the existing technology and will not be repeated here.

[0059] In this embodiment, the problem that the color distances between superpixel blocks in the image of the photovoltaic cell are very close is taken into consideration. Based on the existing color Euclidean distance, the distribution characteristic value of each superpixel block is introduced to characterize the characteristics of each superpixel block.

[0060] Specifically, the process of obtaining the distribution feature value of each superpixel block is as follows:

[0061] First, a grayscale histogram of the image is obtained, and a grayscale reference vector of the image is formed by the column height of each grayscale value in the grayscale histogram; illustratively, the number of pixels with a grayscale value of 0, the number of pixels with a grayscale value of 1, ..., the number of pixels with a grayscale value of 255 is formed into a sequence containing 256 pixel numbers, and the sequence is a grayscale reference vector.

[0062] Secondly, the grayscale histogram of each superpixel block is obtained, and the grayscale feature vector of the corresponding superpixel block is composed of the column height of each grayscale value in each superpixel block.

[0063] Then, the similarity between the grayscale reference vector and the grayscale feature vector is calculated, and based on the similarity, the distribution feature value of each superpixel block is determined.

[0064] The distribution eigenvalue is calculated as follows: ;in, is the distribution characteristic value of the i-th superpixel block, is the number of pixels in the ith superpixel block, and N is the maximum number of pixels in all superpixel blocks.

[0065] The above similarity can be cosine similarity. Specifically, the similarity for: ; is the grayscale feature vector corresponding to the i-th superpixel block, is the grayscale reference vector of the photovoltaic cell image, is the cosine similarity function.

[0066] Similarity The value range of is [-1,1], The value range of is [0,1], It can represent the difference in overall color distribution between the ith superpixel block and the image. The larger the value, the smaller the difference in color distribution between the ith superpixel block and the image. The smaller the value, the greater the difference in color distribution between the ith superpixel block and the image.

[0067] The above similarity calculation is due to the fact that when the pixels in the image have different distribution conditions, the grayscale histogram corresponding to the image will also present different forms, such as single peak or multi-peak, smooth change or drastic change. The superpixel block containing battery cell cracks and the superpixel block without cracks will have certain differences in the form of the grayscale histogram. Therefore, the distribution characteristics of the pixels in the superpixel block can be reflected by using the grayscale reference vector and the grayscale feature vector of the superpixel block.

[0068] For ease of calculation, add 1 and divide by 2. Normalize.

[0069] Since the accuracy of superpixel segmentation has certain limitations, superpixel blocks with similar color distribution may be segmented into different sizes, resulting in the grayscale histograms corresponding to the superpixel blocks having similar shapes but different overall sizes. This also makes the grayscale feature vectors corresponding to the superpixel blocks in these two cases have a large gap with the grayscale reference vector (such as vector (1, 2, 3) and vector (11, 12, 13). The difference between the values ​​in the vectors is very small and they have similar shapes, but when the two vectors are calculated with other vectors for cosine similarity, there will be a large gap between the two cosine similarities obtained), which will affect the measurement of the feature differences between superpixel blocks, and thus affect the identification of cracks on the battery cell. Therefore, by right Correction is made to reduce the impact of the number of pixels in a superpixel block on the relative difference in the distribution eigenvalues ​​between superpixel blocks. is the relative size of the number of pixels contained in the ith superpixel block. The larger the value, the larger the area of ​​the ith superpixel block. right The greater the downward correction effect of , the closer the distribution characteristic values ​​between super-pixel blocks with similar pixel distribution characteristics are. The smaller the value, the smaller the area of ​​the i-th super-pixel block. The closer it is to 1, the right The smaller the downward correction effect is, the greater the gap between the distribution feature values ​​of super-pixel blocks with dissimilar pixel distribution features can be maintained.

[0070] Furthermore, in this embodiment, the morphological coefficients between super-pixel blocks are calculated according to the distribution characteristic value of each super-pixel block in the image and the bounding rectangle of each super-pixel block.

[0071] In one embodiment, the morphology coefficient may be a first morphology coefficient , specifically:

[0072] ;

[0073] in, is the first morphological coefficient of the i-th superpixel block and the j-th superpixel block in the image, is the distribution characteristic value of the i-th superpixel block, is the distribution characteristic value of the jth superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

[0074] The area of ​​the super pixel block mentioned above is the number of pixels in the super pixel block.

[0075] is the difference in aspect ratio between the ith superpixel block and the jth superpixel block. Since superpixel segmentation segments the image according to the characteristics of pixel points, there is a certain difference in morphology between the superpixel block containing cracks and the image block without cracks. The smaller the value, the more similar the morphologies of the two superpixel blocks are. The more likely the two superpixel blocks are to have similar morphologies (for example, both are flat or long strips), the larger the morphological coefficients of the two superpixel blocks. The larger the value, the greater the possible difference in morphology between the two superpixel blocks. That is, the more likely the two superpixel blocks are to have different morphologies, the smaller the morphological coefficient between the two superpixel blocks.

[0076] is the difference in distribution characteristics between two superpixel blocks, It reflects the difference in color distribution characteristics between the two superpixel blocks. The larger the value, the greater the difference in color distribution characteristics between the two superpixel blocks, and the smaller the morphological coefficient between the two superpixel blocks; the smaller the value, the smaller the difference in color characteristics between the two superpixel blocks, and the larger the morphological coefficient between the two superpixel blocks. and They respectively reflect the shapes of the two superpixel blocks. It reflects the difference in shape between the two superpixel blocks. The larger the value, the greater the difference in shape between the two superpixel blocks, that is, the more likely the two superpixel blocks are to be dissimilar, and the smaller the morphological coefficient is at this time; the smaller the value, the more likely the shapes of the two superpixel blocks are to be similar, and the larger the morphological coefficient is at this time. The larger the value of is, the smaller the morphological coefficient between two superpixel blocks is. The smaller the value of , the larger the morphological coefficient between two superpixel blocks.

[0077] As another embodiment, the shape coefficient can also be a second shape coefficient , specifically:

[0078] ;

[0079] in, , is the grayscale mean of the i-th superpixel block and the j-th superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

[0080] The reason for calculating the morphological coefficient above is that the photovoltaic cell is evenly divided into several areas by the grid lines, and there is a high similarity between different areas. This also means that there may be a high similarity between two areas with a large distance in the photovoltaic cell image. Then, the spatial Euclidean distance cannot accurately determine the gap between areas in the photovoltaic cell (the photovoltaic cell contains a large number of repetitive areas). Therefore, the morphological coefficient is introduced to characterize the properties of the superpixel block itself.

[0081] In this embodiment, the process of obtaining the saliency value of each superpixel block is as follows:

[0082] First, obtain the Euclidean color distance. Convert the photovoltaic cell image from RGB color space to LAB color space, calculate the average pixel value of all pixels in each superpixel block on the L channel, the average pixel value on the A channel, and the average pixel value on the B channel, calculate the Euclidean color distance between each superpixel block and other superpixel blocks, and normalize the color distance.

[0083] Secondly, according to the distribution characteristic values ​​of the two superpixel blocks, the color distance between the two superpixel blocks and the morphological coefficients between the two superpixel blocks, the dissimilarity value between each superpixel block and other superpixel blocks is calculated, and then the saliency value of each superpixel block is calculated.

[0084] Specifically, the dissimilarity values ​​are: ; is the dissimilarity value between the ith superpixel block and the jth superpixel block, is the distribution characteristic value of the i-th superpixel block, is the distribution characteristic value of the jth superpixel block, is the morphological coefficient of the i-th superpixel block and the j-th superpixel block in the image.

[0085] It represents the value of the color and distribution difference between two super-pixel blocks. This value reflects whether the two super-pixel blocks are similar in color and color distribution. The larger the value, the more dissimilar the two super-pixel blocks are, and the larger the dissimilarity value between the two super-pixel blocks is. The smaller the value, the more similar the two super-pixel blocks are, and the smaller the dissimilarity value between the two super-pixel blocks is. It reflects whether the two superpixel blocks are similar in shape features. The larger the value, the more similar the two superpixel blocks are in morphological features, and the smaller the dissimilarity value between the two superpixel blocks. The smaller the value, the greater the difference in morphological features between the two superpixel blocks, and the greater the dissimilarity value between the two superpixel blocks.

[0086] Among them, the significant value is ; is the saliency value of the i-th superpixel block in the image, is the dissimilarity value between the ith superpixel block and the jth superpixel block, exp( ) is an exponential function with the natural constant e as the base, and K is the number of superpixel blocks within the target range of the ith superpixel block.

[0087] is the comprehensive non-significant metric value of the ith superpixel block, which reflects the feature difference between the ith superpixel block and the superpixel blocks in the target range. The larger the value, the more prominent the feature of the ith superpixel block, the more significant the superpixel block, and the larger the significance value of the superpixel block. The smaller the value, the less prominent the feature of the ith superpixel block, the less significant the superpixel block, and the smaller the significance value of the superpixel block.

[0088] The value of K is 3. Specifically, the dissimilarity values ​​are sorted from small to large, and other superpixel blocks corresponding to the first three dissimilarity values ​​are selected. At this time, the target range is the set of the first three smaller dissimilarity values. Of course, it can also be determined according to actual conditions.

[0089] The above method of obtaining the K super pixel blocks closest to the target pixel block and calculating the saliency value can accurately evaluate the crack defect.

[0090] The above scheme is based on the particularity of the surface structure of photovoltaic cells, and improves the existing CA algorithm to better identify cracks on photovoltaic cells in images.

[0091] Step S4, marking the crack defect level of the super pixel block when the significance value is greater than the first set value as high; marking the crack defect level of the super pixel block when the significance value is less than or equal to the first set value and greater than or equal to the second set value as medium.

[0092] In this embodiment, when the significance value is less than the second set value, there is no crack on the photovoltaic cell or the defect degree of the crack is very small.

[0093] In this embodiment, the first setting value is 0.6, and the second setting value is 0.2. Of course, the setting values ​​can also be determined according to actual conditions.

[0094] The above-mentioned high and medium grading can be determined based on experience. For example, when the crack defect is more obvious, it is marked as a high level of defect; when the crack defect is not too obvious, it can be marked as medium; when the crack defect is very small or non-existent, it can be marked as a low level, and the low level can be considered as qualified for the photovoltaic cell.

[0095] The solution of the present invention performs threshold segmentation on the image of the photovoltaic cell, and performs super-pixel segmentation on the image of the photovoltaic cell according to the segmentation result to obtain multiple super-pixel blocks of the image, and then collects the distribution characteristics and shape of the super-pixel blocks, calculates the saliency value of each super-pixel block, and judges whether there is a crack in the image and the degree of the crack defect according to the saliency value of the super-pixel block. Obtaining the saliency area in the image by this method can better identify the cracks on the photovoltaic cell panel, and to a certain extent, can reduce the occurrence of missed detection of cell defects.

[0096] The present invention also provides a photovoltaic cell crack detection system based on image processing. Figure 5 As shown, the system includes a processor and a memory, wherein the memory stores computer program instructions, and when the computer program instructions are executed by the processor, the photovoltaic cell crack detection method based on image processing according to the present invention is implemented.

[0097] The detection system also includes other components familiar to those skilled in the art, such as a communication bus and a communication interface, whose configuration and functions are known in the art and thus will not be described in detail here.

[0098] In the present invention, the aforementioned memory may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, apparatus or device. For example, a computer-readable storage medium may be any appropriate magnetic storage medium or magneto-optical storage medium, such as a resistive random access memory RRAM (Resistive Random Access Memory), a dynamic random access memory DRAM (Dynamic Random Access Memory), a static random access memory SRAM (Static Random-Access Memory), an enhanced dynamic random access memory EDRAM (Enhanced Dynamic Random Access Memory), a high-bandwidth memory HBM (High-Bandwidth Memory), a hybrid memory cube HMC (Hybrid Memory Cube), etc., or any other medium that can be used to store the required information and can be accessed by an application, a module, or both. Any such computer storage medium may be part of a device or accessible or connectable to a device. Any application or module described in the present invention may be implemented using computer-readable / executable instructions that may be stored or otherwise maintained by such a computer-readable medium.

[0099] In the description of this specification, “plurality” means at least two, such as two, three or more, etc., unless otherwise clearly and specifically defined.

[0100] Although this specification has shown and described a number of embodiments of the present invention, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Those skilled in the art will conceive of many modifications, changes and alternatives without departing from the ideas and spirit of the present invention. It should be understood that in the practice of the present invention, various alternatives to the embodiments of the present invention described herein may be employed.

Claims

1. A photovoltaic cell crack detection method based on image processing, characterized in that: include: Collecting images of photoluminescence from photovoltaic cells; Acquire multiple superpixel blocks of the image; Calculating the saliency value of each super pixel block; marking the crack defect degree of the super pixel block whose saliency value is greater than a first set value as high; Marking the crack defect level of the super pixel block whose significance value is less than or equal to the first set value and greater than the second set value as medium; The significant values ​​are: ;in, , , are the morphological coefficient and dissimilarity value of the i-th superpixel block and the j-th superpixel block, respectively. , are the distribution characteristic values ​​of the ith superpixel block and the jth superpixel block, respectively. K is the number of superpixel blocks within the target range of the ith superpixel block. is the normalized value of the color Euclidean distance between the i-th superpixel block and the j-th superpixel block in the LAB space, and exp() is an exponential function; the distribution characteristic value is inversely correlated with the similarity between the corresponding superpixel block and the image, and the ratio of the number of pixels in the corresponding superpixel block to the maximum number of pixels in all superpixel blocks; The morphological coefficient represents the shape similarity of two superpixel blocks; the target range is the set of the first K other superpixel blocks whose dissimilarity values ​​of the selected i-th superpixel block and all other superpixel blocks are sorted from small to large, K≥2.

2. The photovoltaic cell crack detection method based on image processing according to claim 1, characterized in that: The distribution characteristic value for: ; is the similarity between the ith superpixel block and the image, is the number of pixels in the ith superpixel block, N is the maximum number of pixels in all superpixel blocks; the similarity is the cosine similarity between the grayscale feature vector and the grayscale reference vector, the grayscale feature vector is a sequence composed of the number of different grayscale values ​​of pixels in the superpixel block, and the grayscale reference vector is a sequence composed of the number of different grayscale values ​​of pixels in the image.

3. The photovoltaic cell crack detection method based on image processing according to claim 1 or 2, characterized in that: The shape coefficient is the first shape coefficient , specifically: ； in, is the distribution characteristic value of the i-th superpixel block, is the distribution characteristic value of the jth superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

4. The photovoltaic cell crack detection method based on image processing according to claim 1 or 2, characterized in that: The shape coefficient is the second shape coefficient , specifically: ； in, , is the grayscale mean of the i-th superpixel block and the j-th superpixel block, and are the height and width of the bounding rectangle of the i-th superpixel block, and are the height and width of the bounding rectangle of the jth superpixel block, is the ratio of the area of ​​the i-th superpixel block to the minimum bounding rectangle of the superpixel block, is the ratio of the area of ​​the jth superpixel block to the minimum circumscribed rectangle of the superpixel block, exp( ) is an exponential function with the natural constant e as the base, and norm( ) is a linear normalization function.

5. The photovoltaic cell crack detection method based on image processing according to claim 1, characterized in that: The specific process of obtaining multiple super pixel blocks of the image is as follows: Segmenting the image to obtain a segmentation result; Each connected region in the segmentation result is obtained through a connected region marking algorithm; the center point of each connected region is used as the seed point for superpixel segmentation, and the image of the photovoltaic cell is superpixel segmented to obtain the corresponding superpixel block, wherein the coordinates of the center point are the mean of the horizontal coordinates and the mean of the vertical coordinates of all the pixel points in the connected region.

6. The photovoltaic cell crack detection method based on image processing according to claim 1, characterized in that: The multiple super pixel blocks of the image are obtained by adopting a simple linear iterative clustering method.

7. The photovoltaic cell crack detection method based on image processing according to claim 5, characterized in that: The segmentation result is obtained by using the Otsu algorithm.

8. Photovoltaic cell crack detection system based on image processing, characterized in that: include: processor; A memory storing computer instructions for photovoltaic cell crack detection based on image processing, wherein when the computer instructions are executed by the processor, the system executes the photovoltaic cell crack detection method based on image processing according to any one of claims 1 to 7.

Images