Online quality sensing method for multi-type battery pieces in a flower basket

Abstract

The application provides an online quality sensing method for multiple types of battery pieces in a flower basket, mainly including the following three steps: S1, three-camera-based image acquisition and splicing, S2, based on Blob analysis and morphological principles, obtaining the multiple types of battery piece regions in the flower basket, and S3, matching the characteristic value parameters of the finished and semi-finished battery piece regions, for the obtained finished and semi-finished battery piece regions, the characteristic values such as the center of gravity, length and width are obtained through region decomposition and other methods, and the characteristic values are screened to remove other interference regions, to obtain accurate finished and semi-finished battery piece regions, and then the threshold range of the related characteristic values set in the algorithm is matched, to judge whether the selected finished and semi-finished battery piece regions meet the set standard, so that the online quality sensing of the multiple types of battery pieces in the flower basket is realized quickly and stably.

Description

Technical Field

[0001] This invention relates to the field of online quality sensing, and more specifically to an online quality sensing method for multiple types of battery cells within a flower basket. Background Technology

[0002] With the rise of the photovoltaic industry, solar cells have become increasingly important in the energy sector due to their high efficiency, low degradation, and high reliability. From raw material selection to finished cell acceptance, almost every stage of the solar cell production process requires baskets as temporary transfer carriers for finished and semi-finished cells. However, in actual production, defects in the preceding stages, such as microcracks, unstable automated operation, and improper human handling, can lead to missing or damaged cells within the baskets during transfer. This can cause blockages and even damage to intact cells in the automated production process. Therefore, quality control of the finished and semi-finished cells within the baskets is crucial for improving cell quality at each stage of the process and reducing debugging time and frequency.

[0003] Currently, regarding online quality perception, due to the optical characteristics of area array camera lenses and limitations imposed by camera manufacturing processes, the obtained images may suffer from distortion. Furthermore, various types of battery cells within the basket are prone to defects such as missing cells, fragments, warping, and double or multiple cells. Summary of the Invention

[0004] To address the shortcomings of existing technologies, an online quality sensing method for multiple types of battery cells within a flower basket is provided.

[0005] An online quality sensing method for multiple types of battery cells within a flower basket mainly includes the following steps:

[0006] S1. Image acquisition and stitching based on three cameras: three sets of camera lenses are used to acquire original images of various types of battery cells in the flower basket. The field of view of the three sets of visual imaging modules is a rectangular plane larger than the horizontal short side of the flower basket. The three sets of fields of view overlap with each other, and the overlapping part accounts for 20%-30% of the single field of view.

[0007] S2. Based on Blob analysis and morphological principles, obtain the multi-type battery cell regions within the flower basket: Using the grayscale value opening operation method and the top-hat transformation principle, the three sets of images are segmented into target pixels and background pixels of finished and semi-finished battery cell regions. The finished and semi-finished battery cell regions are obtained through dynamic threshold segmentation. Morphological operations are used to perform opening and closing operations on the target regions to remove the influence of noise points. Then, connectivity analysis is performed on the finished and semi-finished battery cell regions to aggregate the target images into connected bodies of target pixels or spots.

[0008] S3. Match the feature value parameters of the finished and semi-finished battery cell regions: For the obtained finished and semi-finished battery cell regions, obtain their centroid, length, width and other feature values ​​through region decomposition, and filter the feature values ​​to remove other interfering regions to obtain accurate finished and semi-finished battery cell regions. Then match them with the threshold range of relevant feature values ​​set in the algorithm to determine whether the selected finished and semi-finished battery cell regions meet the set standards.

[0009] Preferably, in S2, based on Blob analysis and morphological principles, the regions of various types of battery cells within the flower basket are obtained, and the original image is subjected to grayscale erosion operation using the structuring element b, as shown in equation (1):

[0010]

[0011] Right now:

[0012]

[0013] Then, the grayscale value of the result F in equation (1) is expanded using the structural element b, as shown in equation (3):

[0014]

[0015] Right now:

[0016]

[0017] Combining equations (1) and (3) yields equation (5), which completes the grayscale value opening operation of the original image f.

[0018]

[0019] To correct uneven illumination using a top-hat transform, the top-hat transform of the original image f is defined as f minus its opening operation:

[0020]

[0021] The image after the top-hat transformation is subjected to contrast enhancement, as shown in equation (7):

[0022] g'=T hat (f)*Mult+Add (7)

[0023] Where g' is the grayscale value after contrast enhancement, Mult is the multiplier factor, and Add is the grayscale supplement value.

[0024] Preferably, the specific steps for obtaining the finished and semi-finished battery cell regions through dynamic threshold segmentation in S2 are as follows: Set a structural element with a mask size of r*r, and calculate the mean gray value m(x,y) and standard deviation s(x,y) within the domain of this structural element, as shown in equations (8) and (9):

[0025]

[0026]

[0027] Then the threshold T(x,y) in this domain is:

[0028]

[0029] In equation (10), R is the dynamic range of the standard deviation, and K is a user-defined correction coefficient.

[0030] Preferably, in S2, morphological operations are used to open the target region, satisfying the following formula:

[0031]

[0032] In equation (11), A is the set of selected battery cell regions, B is the structuring element, and ∪ represents the union of all sets within the curly braces.

[0033] Preferably, the specific steps in S3 to obtain the characteristic values ​​such as centroid, length, and width are as follows: divide the planar region X into n finite simple figures X1, X2, X3…X n Define the centroid of a finite simple figure as C. i The area is A i Then the x-coordinate and y-coordinate of the centroid of the planar region satisfy equations (12) and (13):

[0034]

[0035]

[0036] Assuming that the number of battery cells in each image is N, the centroid position of each battery cell can be obtained from equations (12) and (13). Calculate the spacing between adjacent solar cells:

[0037]

[0038] By comparing the spacing between adjacent cells obtained by equation (14) with the set range, the defect detection of missing cells of various types of battery cells in the basket is completed.

[0039] Preferably, the specific steps for obtaining the minimum bounding rectangle of the battery cell region when matching the feature value parameters of the finished and semi-finished battery cell regions in S3 are as follows:

[0040] The minimum bounding rectangle of the battery cell region can be calculated using the principal axis method and rotation method based on vertex chain code and Discrete Green's algorithm. Assuming the number of battery cells in each image is N, the length and width of the minimum bounding rectangle for each battery cell region are (H1, W1), (H2, W2), ..., (H...). N W N Iterate through the minimum bounding rectangle of each cell region. If the length and width of the rectangle meet the actual size tolerance of the cell, then each cell is intact and without defects. Otherwise, the cell has defects such as fragmentation, stacking, or warping, as shown in equations (15) and (16).

[0041]

[0042]

[0043] In equations (15) and (16), H and W are the theoretical values ​​of the length and width of the solar cell, respectively. and These are the absolute values ​​of the differences between the theoretical and actual measured values ​​in the length and width directions of the solar cell, respectively.

[0044] Beneficial effects:

[0045] (1) This invention provides an online quality perception method for multiple types of battery cells in a flower basket. In view of the problem that the optical characteristics of the area array camera lens and the limitations of the camera manufacturing process will cause the obtained image to be distorted, this invention uses three sets of vision systems to complete the acquisition of the original images of multiple types of battery cells in the flower basket. The three sets of fields of view have overlapping parts, and the overlapping part accounts for 20%-30% of the single field of view. This is used to realize image stitching based on the data acquired by the three sets of vision imaging modules to complete the detection of the entire flower basket. At the same time, the 20%-30% setting can remove the severely distorted areas of the camera at the visual boundary, reduce the quality perception error caused by camera imaging distortion, and reduce the detection error caused by distortion.

[0046] (2) To address the defects commonly found in various types of battery cells within the basket, such as missing cells, fragments, warping, and double or multiple cells, this invention develops a detection algorithm based on blob analysis and morphological principles. The algorithm employs a grayscale opening operation to segment three sets of images into target pixels and background pixels for finished and semi-finished battery cell regions. Next, to remove the influence of noise points, morphological operations such as opening and closing operations are performed on the target region. Then, connectivity analysis is performed on the finished and semi-finished battery cell regions to aggregate the target image into connected components of target pixels or spots, obtaining their length, width, and angle features. These features are then filtered to remove other interfering regions, resulting in accurate finished and semi-finished battery cell regions. Finally, the algorithm matches these regions with threshold ranges of relevant feature values ​​to determine if they meet the set standards, achieving fast and stable online quality perception of various types of battery cells within the basket. Attached Figure Description

[0047] Figure 1 This is a mechanical structure diagram of the flower basket and the battery cells inside the flower basket in the embodiment.

[0048] Figure 2 This is a flowchart of the online quality sensing method for multiple types of battery cells in the flower basket in the embodiment.

[0049] Figure 3 This is a real-life photograph of the online quality sensing method for multiple types of battery cells inside the flower basket in the embodiment. Detailed Implementation

[0050] To enhance understanding of the present invention, the present invention will be further described in detail below with reference to embodiments and accompanying drawings. These embodiments are only used to explain the present invention and do not constitute a limitation on the scope of protection of the present invention.

[0051] This invention provides an online quality sensing method for various types of solar cells within a flower basket. The flower basket is a carrier for finished and semi-finished solar cells, such as... Figure 1As shown, the basket has a hollow rectangular structure. Depending on the size of the solar cells and the size of the production machine, the basket size ranges from 550mm * 250mm * 200mm to 700mm * 260mm * 270mm, compatible with solar cell sizes of 166mm * 166mm to 210mm * 210mm. In actual production, factors such as pre-process defects in finished and semi-finished solar cells (e.g., microcracks), unstable automated operation, and improper human operation can all lead to missing or damaged finished and semi-finished solar cells within the basket during transit. This can cause card-related pauses in the automated production process and even damage to undamaged solar cells. Based on these problems, this invention proposes an online quality sensing method for multiple types of solar cells within the basket. This method can improve the accuracy of extracting different types of solar cell regions within the basket while maintaining speed, thereby improving the accuracy of the online quality sensing method for multiple types of solar cells within the basket.

[0052] In the image acquisition section, this method uses three sets of camera lenses to acquire original images of various types of battery cells within the flower basket. The field of view of the three visual imaging modules is a rectangular plane larger than the horizontal shorter side of the flower basket. The three fields of view overlap, with the overlap accounting for 20%-30% of each single field of view. This overlap is used to stitch together images based on the data acquired by the three visual imaging modules, completing the detection of the entire flower basket. Simultaneously, the 20%-30% setting removes areas with severe camera distortion at the visual boundaries, reducing quality perception errors caused by camera imaging distortion. In this embodiment, the camera model is MV-CA050-20GM, the lens model is MVL-KF2528M-12MPE, and the light source model is OPT-FMV700-350K-W. The camera shoots vertically onto the flower basket through an open-aperture light source, with a black background board of the same size installed directly below the flower basket. Furthermore, this detection system uses a Windows 10 operating system, an Intel(R) Core(TM) i5-9500 processor with a clock speed of 3.0GHz, and 16GB of memory.

[0053] This method acquires real-time images of various types of battery cells within a flower basket using a tri-lens camera, determining the ROI detection range for each camera's images. In acquiring finished and semi-finished battery cells, the method, based on blob analysis and morphological principles, first uses grayscale opening operations to segment the three images into target pixels and background pixels for finished and semi-finished battery cell regions. Second, to remove noise, morphological operations such as opening and closing operations are performed on the target region. Next, connectivity analysis is performed on the finished and semi-finished battery cell regions to aggregate the target images into connected components of target pixels or spots. Then, features such as centroid, length, and width are acquired, and these features are filtered to remove other interfering regions, resulting in accurate finished and semi-finished battery cell regions. Finally, these regions are matched against threshold ranges set internally by the algorithm to determine if they meet the set standards, achieving fast and stable online quality perception of various types of battery cells within the flower basket.

[0054] The specific steps are as follows:

[0055] S1. Acquiring original images of multiple types of battery cells in the flower basket based on trinocular vision: Real-time acquisition of image information of multiple types of battery cells in the flower basket through a trinocular camera. The field of view of the three sets of vision imaging modules is a rectangular plane larger than the horizontal short side of the flower basket. The three sets of fields of view have overlapping parts, and the overlapping part accounts for 20%-30% of the single field of view. This is used to realize image stitching based on the data acquired by the three sets of vision imaging modules to complete the detection of the entire flower basket.

[0056] S2, see also Figure 3 Simultaneously, three cameras captured raw images of various types of battery cells within a flower basket. Battery cell regions were extracted from each image. Based on blob analysis and morphological principles, the three images were first segmented into target pixels and background pixels for finished and semi-finished battery cell regions using grayscale opening operations. Secondly, to remove noise, morphological operations such as opening and closing operations were performed on the target regions. Next, connectivity analysis was performed on the finished and semi-finished battery cell regions to aggregate the target images into connected components of target pixels or blobs. The human-machine interface of the detection system was developed using Visual Studio integrated development environment (IDE) and programmed in C / C++. Halcon (an open-source cross-platform computer vision library) was also used to program and experiment with the relevant functions of each module. The specific steps are as follows:

[0057] (1) Opening operation based on grayscale image: After grayscale opening, the image removes bright details that are smaller than the structuring element, while keeping the overall grayscale and larger bright areas unchanged. First, the original image f is eroded using a structuring element b of size 1*15, and then the result is dilated using b. When the origin of b is located at (x,y), the erosion of image f by a flat structuring element b at (x,y) is defined as the minimum value of the region of image f that coincides with b. To find the erosion of f by b, the origin of the structuring element is placed at the position of each pixel in the image. The erosion at any position is determined by the minimum value selected from all values ​​of f contained in the region that coincides with b, as shown in formula (1):

[0058]

[0059] Right now:

[0060]

[0061] Similarly, grayscale dilation is performed on the structure obtained from equation (1):

[0062]

[0063] Right now:

[0064]

[0065] Therefore, the opening operation of a grayscale image satisfies formula (5):

[0066]

[0067] (2) To correct the effects of uneven illumination, the Top-hat transform can remove objects from the bright battery area in a dark background using a structuring element through an opening operation, and then obtain an image retaining only the removed components through a difference operation. The Top-hat transform of the original image f is defined as f minus its opening operation:

[0068]

[0069] Then, the image after the top-hat transformation is subjected to appropriate contrast enhancement, as shown in formula (7):

[0070] g'=T hat (f)*Mult+Add (7)

[0071] In the formula, g' is the gray value after contrast enhancement, Mult is the multiplier factor that increases the contrast after subtracting the two images, and Add is the gray value supplement that increases the overall brightness of the image after subtraction.

[0072] (3) Dynamic thresholding: Based on the subtracted image obtained in (2), the battery cell region is extracted using a dynamic thresholding algorithm for subsequent matching of battery cell feature value parameters. A structuring element with a mask size of r*r is set, and the mean gray value m(x,y) and standard deviation s(x,y) within the neighborhood of this structuring element are calculated, as shown in equations (8) and (9):

[0073]

[0074]

[0075] Then the threshold T(x,y) in this domain is:

[0076]

[0077] In the formula, R is the dynamic range of the standard deviation, and K is a user-defined correction coefficient.

[0078] (4) The generated rectangular structuring element is used to perform an opening operation on the extracted battery cell region. The opening operation smooths the contour of the battery cell region, breaks up narrow necks, and eliminates fine protrusions, thereby removing background interference. The morphological opening operation satisfies formula (11):

[0079]

[0080] In the formula, A is the set of selected battery cell regions, B is the structuring element, and ∪{·} represents the union of all sets within the curly braces.

[0081] S3. Matching of characteristic parameters of finished and semi-finished battery cells:

[0082] (1) A planar region X can be divided into n finite simple figures X1, X2, X3…X n The centroid of these simple shapes is C. i The area is A i Then the x-coordinate and y-coordinate of the centroid of the planar region satisfy equations (12) and (13):

[0083]

[0084]

[0085] Assuming that the number of battery cells in each image is N, the centroid position of each battery cell can be obtained from equations (12) and (13). Calculate the spacing between adjacent solar cells:

[0086]

[0087] By comparing the spacing between adjacent cells obtained by equation (14) with the set range, the defect detection of missing cells of various types of battery cells in the basket is completed.

[0088] (2) Obtain the minimum bounding rectangle of the battery cell region. The minimum bounding rectangle of the battery cell region can be calculated using the principal axis method and rotation method based on vertex chain code and Discrete Green's algorithm. Assuming the number of battery cells in each image is N, the length and width of the minimum bounding rectangle of each battery cell region are H1, W1), (H2, W2)...(H... N W N Iterate through the minimum bounding rectangle of each solar cell region. If the rectangle's length and width meet the actual dimensional tolerances of the solar cell, then each solar cell is considered intact and without defects. The pool sheet has defects such as fragmentation, stacking, and warping, as shown in equations (15) and (16):

[0089]

[0090]

[0091] In the formula, H and W are the theoretical values ​​of the length and width of the solar cell, respectively. and These are the absolute values ​​of the differences between the theoretical and actual measured values ​​in the length and width directions of the solar cell, respectively.

[0092] Example:

[0093] The method described uses various types of battery cells inside a flower basket as the testing object, and its specific implementation is as follows:

[0094] S1 acquires raw images of various types of solar cells inside the flower basket. The system uses a trinocular vision system to acquire these raw images. The field of view of each of the three vision modules is a rectangular plane larger than the horizontal shorter side of the flower basket. The three fields of view overlap by 20%-30% of each individual field of view, used to stitch together images from the data acquired by the three vision modules to complete the inspection of the entire flower basket. The camera model is MV-CA050-20GM, the lens model is MVL-KF2528M-12MPE, and the light source model is OPT-FMV700-350K-W. The camera takes images vertically onto the flower basket through an open-aperture light source, with a black background of the same size installed directly below the basket. Furthermore, the inspection system uses a Windows 10 operating system, an Intel(R) Core(TM) i5-9500 processor with a clock speed of 3.0GHz, and 16GB of memory.

[0095] S2 uses the Blob analysis and morphological principles proposed in this invention to extract the battery cell region. First, a grayscale opening operation method is used to segment the three images into target pixels and background pixels for finished and semi-finished battery cell regions. Second, to remove the influence of noise points, morphological operations such as opening and closing operations are performed on the target region. Next, connectivity analysis is performed on the finished and semi-finished battery cell regions to aggregate the target image into a connected body of target pixels or spots. The algorithm flow is as follows:

[0096] (1) Opening operation based on grayscale image. After grayscale opening, the image removes bright details that are smaller than the structuring element, while keeping the overall grayscale and larger bright areas unchanged. First, the original image f is eroded using a structuring element b of size 1*15, and then the result is dilated using b. When the origin of b is located at (x,y), the erosion of image f by a flat structuring element b at (x,y) is defined as the minimum value of the region of image f that coincides with b. To find the erosion of f by b, the origin of the structuring element is placed at the position of each pixel in the image. The erosion at any position is determined by the minimum value selected from all values ​​of f contained in the region that coincides with b, as shown in formula (17):

[0097]

[0098] Right now:

[0099]

[0100] Similarly, when When the origin is located at position (x, y), the dilation of the image f by the flat structuring element b at (x, y) is defined as the dilation of the image f by the element b in the flat structuring element b at position (x, y). Maximum value of overlapping regions:

[0101]

[0102] Right now:

[0103]

[0104] Therefore, the opening operation of a grayscale image satisfies formula (5);

[0105] (2) To correct the effects of uneven illumination, the Top-hat transformation can be used to remove objects from the original image of the bright battery cell area under dark background conditions by using a structuring element through an opening operation, and then obtaining an image that retains only the deleted components through a difference operation. The Top-hat transformation of the original image f is defined by formula (6) as f minus its opening operation. Then, the image after the Top-hat transformation is subjected to appropriate contrast enhancement, as shown in formula (21):

[0106] g′=T hat (f)*1+0 (21)

[0107] In the formula, g' is the gray value after contrast enhancement, g1 is the gray value of the original image, g2 is the gray value of the image after gray opening operation, 1 is the multiplier factor to increase the contrast after subtracting the two images, and 0 is the gray supplement value to increase the overall brightness of the image after subtraction.

[0108] (3) Dynamic threshold segmentation. Based on the subtracted image obtained in [2], the battery cell region is extracted using a dynamic threshold segmentation algorithm for subsequent matching of battery cell feature value parameters. The mask size is set to a 5*5 structuring element, and the mean gray value m(x,y) and standard deviation s(x,y) in the neighborhood of this structuring element are calculated as shown in equations (22) and (23):

[0109]

[0110]

[0111] Then the threshold T(x,y) in this domain is:

[0112]

[0113] In the formula, R is the dynamic range of the standard deviation, and K is a user-defined correction coefficient.

[0114] (4) The generated rectangular structuring element is used to perform an opening operation on the extracted cell region. The opening operation smooths the contour of the cell region, breaks up narrow necks, and eliminates fine protrusions, thereby removing background interference. The morphological opening operation satisfies formula (25):

[0115]

[0116] In the formula, A represents the set of filtered battery cell regions, B is a structuring element of size 1*35, and ∪{·} represents the union of all sets within the curly braces. Battery cell region extraction effect.

[0117] 3. Matching of feature value parameters of finished and semi-finished battery cells.

[0118] (1) The extracted battery cell region X is divided into n finite simple shapes X1, X2, X3…X n The centroid of these simple shapes is C. i The area is A i Then the x-coordinate and y-coordinate of the centroid of the planar region Equations (26) and (27):

[0119]

[0120]

[0121] The number of battery cells in each image is 36. The centroid position of each battery cell can be obtained from equations (26) and (27). Then the spacing between adjacent solar cells can be calculated by iterating through them:

[0122]

[0123] By comparing the spacing between adjacent cells obtained by equation (28) with the set range, the defect detection of missing cells of various types of battery cells in the basket is completed.

[0124] (2) Obtain the minimum bounding rectangle of the battery cell region. The minimum bounding rectangle of the battery cell region can be calculated using the principal axis method and rotation method based on vertex chain code and Discrete Green's algorithm. Since there are 36 battery cells in each image, the length and width of the minimum bounding rectangle for each battery cell region are H1, W1), (H2, W2)...(H... 36 W 36 ), iterate through the minimum bounding rectangle of each cell region. If the length and width of the rectangle meet the actual size tolerance of the cell, it means that each cell is intact and without defects. Otherwise, the cell has defects such as fragmentation, stacking, and warping, as shown in equations (29) and (30):

[0125]

[0126]

[0127] In the formula, H and W are the theoretical values ​​of the length and width of the solar cell, respectively. and These are the absolute values ​​of the differences between the theoretical and actual measured values ​​in the length and width directions of the solar cell, respectively.

[0128] As a further improvement, the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. An online quality sensing method for multiple types of battery cells within a flower basket, characterized in that, Includes the following steps: S1. Image acquisition and stitching based on three cameras: three sets of camera lenses are used to acquire original images of various types of battery cells in the flower basket. The field of view of the three sets of visual imaging modules is a rectangular plane larger than the horizontal short side of the flower basket. The three sets of fields of view overlap with each other, and the overlapping part accounts for 20%-30% of the single field of view. S2. Based on Blob analysis and morphological principles, obtain the multi-type battery cell regions within the flower basket: Using the grayscale value opening operation method and the top-hat transformation principle, the three sets of images are segmented into target pixels and background pixels of finished and semi-finished battery cell regions. The finished and semi-finished battery cell regions are obtained through dynamic threshold segmentation. Morphological operations are used to perform opening and closing operations on the target regions to remove the influence of noise points. Then, connectivity analysis is performed on the finished and semi-finished battery cell regions to aggregate the target images into connected bodies of target pixels or spots. S3. Match the feature value parameters of the finished and semi-finished battery cell regions: For the obtained finished and semi-finished battery cell regions, obtain their centroid, length, and width feature values ​​through region decomposition, and filter the feature values ​​to remove other interfering regions to obtain accurate finished and semi-finished battery cell regions. Then, match them with the threshold range of relevant feature values ​​set in the algorithm to determine whether the selected finished and semi-finished battery cell regions meet the set standards. S2, based on Blob analysis and morphological principles, obtains the regions of various types of battery cells within the flower basket. The original image is then subjected to grayscale erosion using the structuring element b, as shown in equation (1). Right now: Then, the grayscale value of the result F in equation (1) is expanded using the structural element b, as shown in equation (3): Right now: F(x,y)⊕b(s,t)=max{F(sx,ty)+b(s,t)|(sx),(ty)∈D,(x,y)∈S} (4) Combining equations (1) and (3) yields equation (5), which completes the grayscale value opening operation of the original image f. To correct uneven illumination using a top-hat transform, the top-hat transform of the original image f is defined as f minus its opening operation: The image after the top-hat transformation is subjected to contrast enhancement, as shown in equation (7): g'=T hat (f)*Mult+Add (7) Where g' is the grayscale value after contrast enhancement, Mult is the multiplier factor, and Add is the grayscale supplement value.

2. The online quality sensing method for multiple types of battery cells in a flower basket according to claim 1, characterized in that, The specific steps for obtaining the finished and semi-finished battery cell regions through dynamic threshold segmentation in S2 are as follows: Set a structural element with a mask size of r*r, and calculate the gray mean m(x,y) and standard deviation s(x,y) within the domain of this structural element, as shown in equations (8) and (9): Then the threshold T(x,y) in this domain is: In equation (10), R is the dynamic range of the standard deviation, and K is a user-defined correction coefficient.

3. The online quality sensing method for multiple types of battery cells in a flower basket according to claim 1, characterized in that, In S2, morphological operations are used to open the target region, satisfying the following formula: In equation (11), A is the set of selected battery cell regions, B is the structuring element, and ∪ represents the union of all sets within the curly braces.

4. The online quality sensing method for multiple types of battery cells in a flower basket according to claim 1, characterized in that, The specific steps for obtaining the centroid, length, and width characteristic values ​​of S3 are as follows: Divide the planar region X into n finite simple figures X1, X2, X3…X n Define the centroid of a finite simple figure as C. i The area is A i Then the x-coordinate and y-coordinate of the centroid of the planar region satisfy equations (12) and (13): Assuming that the number of battery cells in each image is N, the centroid position of each battery cell can be obtained from equations (12) and (13). Calculate the spacing between adjacent solar cells: By comparing the spacing between adjacent cells obtained by equation (14) with the set range, the defect detection of missing cells of various types of battery cells in the basket is completed.

5. The online quality sensing method for multiple types of battery cells in a flower basket according to claim 1, characterized in that, The specific steps for obtaining the minimum bounding rectangle of the battery cell region when matching the feature value parameters of the finished and semi-finished battery cell regions in S3 are as follows: The minimum bounding rectangle of the battery cell region can be calculated using the principal axis method and rotation method based on vertex chain code and Discrete Green's algorithm. Assuming the number of battery cells in each image is N, the length and width of the minimum bounding rectangle for each battery cell region are (H1, W1), (H2, W2), ..., (H...). N W N ), iterate through the minimum bounding rectangle of each cell region. If the length and width of the rectangle meet the actual size tolerance of the cell, it means that each cell is intact and without defects. Otherwise, the cell has defects such as fragmentation, stacking, or warping, as shown in equations (15) and (16): In equations (15) and (16), H and W are the theoretical values ​​of the length and width of the solar cell, respectively. and These are the absolute values ​​of the differences between the theoretical and actual measured values ​​in the length and width directions of the solar cell, respectively.

Images