A Machine Vision-Based Method and System for Measuring Laser Spot Width in Solar Cells

Abstract

This invention relates to a machine vision-based method and system for measuring the width of a laser spot on a solar cell. The method includes the following steps: image acquisition, image color space conversion, image binarization, laser spot detection, image localization and segmentation, segmented image magnification, segmented image binarization, segmented image closing operation, segmented image laser spot detection, laser spot width calculation, and result output. This machine vision-based method for measuring the width of a solar cell laser spot uses a camera to measure the laser spot width. Within a certain height range, it eliminates the need for repeated adjustments to high-magnification lenses, enabling simple operation and rapid measurement of the laser spot width. This improves measurement accuracy, offers fast measurement speed, low equipment cost, and allows for online detection with a high degree of automation.

Description

Technical Field

[0001] This invention relates to a grooving method for solar cells, and more particularly to a machine vision-based method and system for measuring the width of a laser spot on a solar cell. Background Technology

[0002] In the photovoltaic industry, solar cells are the core component of photovoltaic modules, and their production quality directly affects the performance and efficiency of the modules. Laser technology has been widely used in the production and testing of solar cells, with related processes including laser doping and laser grooving. Laser grooving forms fine grooves on the surface of the solar cell, thereby creating selective emitters or localized back fields, improving the photoelectric conversion efficiency of the cell. Laser grooving is a non-contact processing technology that optimizes electrode contact on the cell surface, featuring high precision, flexibility, fast processing speed, and low manufacturing cost, enabling high-efficiency cell technology.

[0003] The slot width is a crucial parameter affecting battery performance; excessive width or narrowness can lead to increased series resistance or carrier recombination rate, thus reducing battery efficiency. Accurate measurement of the slot width is essential for ensuring the quality of the laser grooving process. Measurement methods primarily include optical microscopy, laser scanning, confocal microscopy, and machine vision. Optical microscopy is slow, making it unsuitable for online inspection and requiring highly skilled operators. Laser scanning equipment is expensive and demands high focusing accuracy and stability of the laser beam. Confocal microscopy equipment is costly and slow, hindering large-scale production. Machine vision measurement offers high speed and is suitable for automated online inspection. Summary of the Invention

[0004] To address the above problems, this invention provides a machine vision-based method for measuring the laser spot width of battery cells. The specific technical solution is as follows:

[0005] A machine vision-based method for measuring the width of laser spot on a solar cell includes the following steps: image acquisition, acquiring a color image of the solar cell; image color space conversion to obtain an H-component image of the A-type solar cell; image binarization, removing noise using selected minimum and maximum thresholds to obtain a binary image; laser spot detection, extracting edge information of the laser spot on the solar cell from the binary image, and obtaining an image containing one or more laser spots using a contour detection algorithm; image localization and segmentation, first locating a spot region on the obtained image containing one or more laser spots, and then segmenting the H-component image of the A-type solar cell based on the localization information and the width information of the real-time acquired image; and image magnification along the horizontal axis. fx Set to 1, scaling factor along the vertical axis fyThe value is set to 10, and a bicubic interpolation method is used to enlarge the segmented image; the segmented image is binarized, and noise is removed by selecting a minimum threshold and a maximum threshold; the segmented image is closed, and based on the segmented binary image, a closing operation is performed to first connect the broken areas of the laser spot, and then refine the upper and lower boundaries of the laser spot; the segmented image is laser spot detected, and the laser spot region is extracted by a contour detection algorithm based on the closed image, and the number of this region is 1; the laser spot width is calculated, and based on the contour information of the laser spot, the fitted line width midWidth, the average value of the new width set avgWidth, and the median value of the new width set medWidth are calculated; the result is output, and the laser spot edge display image and the obtained laser spot width data are output.

[0006] Preferably, the contour detection algorithm includes the following steps: extracting the contour of the laser spot, detecting only the outermost contour, compressing horizontal, vertical, and diagonal line segments, and retaining only their endpoints; first finding all contours, storing each contour as a vector of points; and filtering out the laser spot region by area features; extracting the contour of the laser spot, detecting only the outermost contour, compressing horizontal, vertical, and diagonal line segments, and retaining only their endpoints; finding all contours, storing each contour as a vector of points; and filtering out the magnified laser spot region by area features.

[0007] Preferably, the image localization and segmentation includes the following steps: When the length of the upper boundary is assumed to be greater than a certain proportion of the width of the acquired image for the first time, this boundary is determined to be the upper boundary of the first laser spot found. The row number of the upper boundary is topPoint. The coordinates of the upper left vertex of the first spot are found to be (0, topPoint). This point is shifted upwards by 300 pixels, and the coordinates of the upper left vertex of the segmented image are (0, topPoint-300). If topPoint-300 is less than 0, then the vertex coordinates are (0, 0); Starting from the row number topPoint of the found upper boundary, the lower boundary of the laser spot is found. When the length of the lower boundary is assumed to be greater than a certain proportion of the width of the acquired image for the first time, this boundary is determined to be the lower boundary of the first laser spot found. The row number of the lower boundary is bottomPoint. The coordinates of the lower left bottom point of the first laser spot are found to be (0, bottomPoint). This point is shifted downwards by 300 pixels, and the coordinates of the lower left bottom point of the segmented image are (0, bottomPoint +300). If bottomPoint... If +300 is less than the height h of the acquired image, then the bottom point coordinates are (0, h); the width of the segmented image is calculated based on the coordinates of the upper left vertex and the lower left bottom point of the first laser spot after offset; the H component image of the type A battery cell is segmented based on the coordinates of the upper left vertex of the first laser spot after offset, the width of the segmented image, and the width of the acquired image.

[0008] Preferably, the laser spot width calculation includes the following steps: Based on the magnified laser spot region image, extract the contour information of the upper boundary of the laser spot; first, establish a dynamic array of two-dimensional points, `topPoints`; then, use a contour detection algorithm to store points that meet the conditions into `topPoints`; based on the two-dimensional point set `topPoints`, iteratively fit a straight line using the M-estimation method. The M-estimator uses Manhattan distance to robustly measure the deviation of data points from the straight line; wherein, the radial accuracy and angular accuracy are both 0.01, and the fitted upper boundary straight line of the laser spot contains a vector of four elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the straight line, (x0, vy) is a normalized vector collinear with the straight line, (vx ... y0) is a point on the straight line; based on the magnified laser spot area image, the contour information of the lower boundary of the laser spot is extracted. First, a two-dimensional point type dynamic array botPoints is established. Then, the contour detection algorithm is used to find the lower boundary point from bottom to top, and the points that meet the conditions are stored in botPoints; based on the two-dimensional point set botPoints, the straight line is iteratively fitted based on the M estimation method. The radial accuracy and angular accuracy are both 0.01. The fitted lower boundary straight line of the laser spot contains a vector of 4 elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the straight line, and (x0, y0) is a point on the straight line; the distance midWidth between the two straight lines is calculated based on the fitted upper boundary straight line parameters and lower boundary straight line parameters of the laser spot.

[0009] Furthermore, the calculation of the distance midWidth between the two lines includes the following steps: establishing a dynamic array widths, then calculating the laser spot width set based on the upper and lower boundary point sets of the laser spot and storing it in widths; deleting a certain proportion of excessively large and small width values ​​from the width set widths, and generating a new laser spot width set newWidths; calculating the average value avgWidth and median value medWidth of the new width set newWidths; and converting the pixel width into the actual physical width based on the actual calibration parameters.

[0010] A machine vision-based laser spot width measurement system for solar cells, used in the aforementioned machine vision-based method for measuring the laser spot width of solar cells, includes: a camera; a high-magnification lens mounted on the camera; a light source mounted on the high-magnification lens; a conveyor for conveying solar cells and located below the high-magnification lens; and an industrial control computer connected to the camera, the light source, and the conveyor, for acquiring images through the camera and processing the acquired images to obtain the laser spot width of the solar cells.

[0011] Compared with the prior art, the present invention has the following beneficial effects:

[0012] The present invention provides a machine vision-based method for measuring the width of a laser spot in a battery cell. The method measures the width of the laser spot using a camera. Within a certain height range, it does not require repeated adjustment of the high-magnification lens. It is simple to operate and can quickly measure the width of the laser spot, improving measurement accuracy. Furthermore, it is fast, has low equipment cost, can achieve online detection, and has a high degree of automation. Attached Figure Description

[0013] Figure 1 This is a flowchart of the measurement method of this application;

[0014] Figure 2 This is a schematic diagram of the measurement system of this application. Detailed Implementation

[0015] The present invention will now be further described with reference to the accompanying drawings.

[0016] A machine vision-based method for measuring the width of a laser spot in a battery cell measures the width of the laser spot using hardware such as a camera, a high-magnification lens, and a light source. Within a certain height range, it eliminates the need for repeated adjustments of the high-magnification lens, enabling simple operation and rapid measurement of the laser spot width, thus improving measurement accuracy.

[0017] like Figure 1 As shown, a machine vision-based method for measuring the laser spot width of a battery cell includes the following steps:

[0018] S01: Image acquisition; Coaxial light is vertically irradiated onto the solar cell to be measured. When the cell enters the field of view of the high-magnification lens, the industrial camera acquires a color image in real time through the high-magnification lens.

[0019] S02: Image color space conversion;

[0020] Comparing the edge information of the laser spot in the grayscale color space, BGR color space, and HSV color space of the Type A solar cell, the results show that the edge information of the Type A solar cell is more obvious in the HSV color space, especially in the hue H component. Therefore, the following operations are all based on the hue H component image.

[0021] S03: Image binarization;

[0022] The H component image of the type A battery cell is binarized. The H component image is an 8-bit single-channel image. Noise is removed by selecting a minimum threshold and a maximum threshold, i.e., filtering out pixels with values ​​that are too small or too large. The output image has the same size, type and number of channels as the input.

[0023] S04: Laser spot detection;

[0024] In the image obtained in S03, non-zero pixels are treated as 1 and zero pixels as 0. Based on this image, a contour detection algorithm is used to extract the contour of the laser spot. The contour retrieval mode is to detect only the outermost contour. The contour approximation method is to compress horizontal, vertical and diagonal line segments and retain only their endpoints. Based on the above method, all contours are found first, and each contour is stored as a vector of points. The laser spot area is filtered out by area features. Since the areas of the laser spots are relatively close, there are one or more laser spots in the image at this time.

[0025] S05: Image localization and segmentation;

[0026] The next step is to locate a spot region for analysis. Based on the image obtained in S04, there are other spots and other invalid information regions. Therefore, the required spot region must be located first, and then segmented according to the located region. This can include the following steps:

[0027] b01: When the size of the upper boundary is assumed to be greater than a certain proportion of the width of the original image, this boundary is determined to be the upper boundary of the first light spot found. From this, the row number at this time is topPoint. Then the coordinates of the top left vertex of the first light spot are (0, topPoint). This point is shifted upward by 300 pixels, and the coordinates of the top left vertex of the segmented image are (0, topPoint-300). If topPoint-300 is less than 0, the vertex coordinates are (0,0).

[0028] b02: Starting from the row number topPoint of the found upper boundary, find the lower boundary of the laser spot. When the size of the lower boundary is assumed to be greater than a certain proportion of the width of the original image, determine that this boundary is the lower boundary of the first spot found. From this, we can find that the row number at this time is bottomPoint. Then the coordinates of the lower left bottom point of the first spot are (0, bottomPoint). This point is shifted down by 300 pixels, and the coordinates of the lower left bottom point of the segmented image are (0, bottomPoint +300). If bottomPoint +300 is less than the height h of the acquired image, the bottom point coordinates are (0, h).

[0029] b03: Calculate the width of the segmented image based on the coordinates of the upper left vertex and lower left bottom point of the first laser spot after the offset;

[0030] b04: Segment the hue H component image based on the coordinates of the upper left vertex of the first laser spot after the offset, the width of the segmented image, and the width information of the acquired image.

[0031] S06: Segmented image magnification;

[0032] To improve the accuracy of width measurement, the scaling factor along the horizontal axis is changed. fx and the scaling factor along the vertical axis fy When magnifying and segmenting the image using two parameters, the parameter that has the greatest impact on the measurement accuracy of the laser spot width is the scaling factor along the vertical axis. fy It can be set to 10 or 20, etc., and the laser spot length can remain constant. fx Setting it to 1 employs the bicubic interpolation method, which is a bicubic interpolation method based on a 4x4 pixel neighborhood. Although this method is slightly slower when enlarging images, it is the best at enhancing image details and avoiding jagged edges. Therefore, the segmented image obtained in S05 is enlarged according to the above method.

[0033] S07: Image binarization for segmentation;

[0034] The obtained segmented image is binarized, and noise is removed by selecting the minimum and maximum thresholds, i.e., filtering out pixels with values ​​that are too small or too large. Since the image is magnified to enhance the details in the image, the two thresholds are different from the thresholds before magnification. The output is an image of the same size, type and number of channels as the input.

[0035] S08: Image segmentation closing operation;

[0036] The segmented binary image obtained from S07 contains many isolated noise points, which may affect the image quality. At the same time, there are many broken areas in the laser spot area of ​​this binary image. The degree of breakage of the spot area varies at different heights, which is not conducive to the extraction of the spot area. Therefore, morphological processing is adopted to improve robustness.

[0037] Morphological processing is widely used in machine vision. It mainly includes erosion, dilation, and the closing operation, which is a combination of the former two methods. Erosion is a basic morphological processing method, as shown in Equation (1). This method uses a specified structuring element. B To erode the source image, this structuring element determines the shape of the pixel neighborhood that takes the minimum value. B The origin moves along the boundary of the set A of white regions in the segmented binary image, and structuring elements that exceed the boundary are not included in the set. A In the middle, compare with the initial set A The image is significantly reduced, which manifests as "refinement" of the laser spot edges, noise, and disconnected areas. If only erosion is used, the disconnected areas will increase. When there are large or many defects in the spot area, it is not conducive to determining the lower boundary position.

[0038] (1);

[0039] Dilation is also a basic morphological treatment, as shown in equation (2), which uses a specified structural element. BFrom the dilated source image, this structuring element determines the shape of the neighborhood of the pixel that takes the maximum value. B The origin is along the set of white regions in the segmented binary image. A Move the boundary, structural element and set A The intersection of the two sets is not empty; compare it with the initial set. A The expansion is significant, which manifests in the image as "coarsening" of the laser spot edges, noise, and disconnected areas. If only the expansion operation is used, when there is a lot of noise smaller than the structuring element at the edge, the expansion will connect the noise and the boundary, which is not conducive to the accurate positioning of the boundary.

[0040] (2);

[0041] Closing is a composite operation that manifests in images as first dilating ("coarsening") and then eroding ("thinning") the edges of laser spots and noise. When setting the size of the structuring element, the distribution of boundary noise and the degree of connection of disconnected regions must be considered simultaneously. The shape of the structuring element is chosen to be rectangular, with the origin located at the center of the element. It should be noted that the closing operation with two iterations is, in order, dilution > dilution > erosion > erosion. The pixel extrapolation method uses constant filling. Processing the image according to the above method can achieve the effect of connecting disconnected regions while improving the accuracy of edge detection.

[0042] S09: Laser spot detection in segmented images;

[0043] Based on the image obtained in S08, the contour of the laser spot is extracted using a contour detection algorithm. The contour retrieval mode is to detect only the outermost contour. The contour approximation method is to compress horizontal, vertical, and diagonal line segments and retain only their endpoints. All contours are found first in the above manner, and each contour is stored as a vector of points. The laser spot area is filtered out by area features. It should be noted that since the image is magnified and the details in the image are enhanced, the area feature value is different from that before magnification.

[0044] S10: Calculation of laser spot width;

[0045] The width of the laser spot is calculated based on its outline information. The specific steps are as follows:

[0046] c01: Extract the contour information of the upper boundary of the laser spot. First, establish a dynamic array of two-dimensional points, topPoints, and then use a contour detection algorithm to store the points that meet the conditions into topPoints.

[0047] c02: Based on the two-dimensional point set of topPoints, the straight line is iteratively fitted using the M estimation method. Although the least squares method is the most commonly used method for fitting a straight line, this method is easily affected by noise. The noise distribution distance around the boundary of the image to be analyzed is very large. Therefore, the distance calculation formula used by the M estimator is shown in formula (3). The accuracy of the radial (distance between the origin and the straight line) and the accuracy of the angle are both 0.01. The upper boundary straight line of the laser spot fitted by the above method contains a vector of 4 elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the straight line, and (x0, y0) is a point on the straight line.

[0048] (3);

[0049] In the formula, distance D The Manhattan distance (L1 norm) is used to robustly measure the deviation of data points from a line; the coordinates of the data points ( x 1, y 1) is the input observation value, ( x 2, y 2) is the projection point on the straight line that is closest to the data point.

[0050] c03: Extract the contour information of the lower boundary of the laser spot. First, establish a dynamic array botPoints of two-dimensional point type. Since the image obtained according to S09 has basically no noise around the laser spot and there is a disconnected area in the spot area, the contour detection algorithm is used to find the lower boundary point from bottom to top, and then the points that meet the conditions are stored in botPoints.

[0051] c04: Based on the botPoints two-dimensional point set, the straight line is iteratively fitted based on the M estimation method. The distance calculation formula used by the M estimator is shown in formula (3). The accuracy of the radial (distance between the origin and the straight line) and the accuracy of the angle are both 0.01. The lower boundary straight line of the laser spot fitted by the above method contains a vector of 4 elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the straight line, and (x0, y0) is a point on the straight line.

[0052] c05: The distance between the two lines is calculated as midWidth based on the fitted lines of c02 and c04;

[0053] c06: First, create a dynamic array widths, then calculate the set of laser spot widths based on the upper and lower boundary point sets of the laser spots in c01 and c03 and store it in widths;

[0054] c07: After deleting a certain proportion of excessively large and excessively small width values ​​from the width set widths, a new laser spot width set newWidths is generated;

[0055] c08: Calculate the average avgWidth and median medWidth of the new width set newWidths;

[0056] c09: Convert the pixel width to the actual physical width based on the actual calibration parameters.

[0057] S11: Output the result;

[0058] The output laser spot width measurement results mainly include the laser spot edge display image and the laser spot width data obtained by the three calculation methods described in S10, including the fitted straight line width midWidth, the average value of the new width set avgWidth, and the median value of the new width set medWidth.

[0059] like Figure 2 As shown, a machine vision-based laser spot width measurement system for solar cells includes a camera 1, a high-magnification lens 2, a light source 3, a conveyor 5, and an industrial computer 6. The camera 1 can be an industrial camera, the high-magnification lens 2 is mounted on the camera 1 and located above the conveyor 5, which is a belt conveyor. The light source 3 can be a coaxial light source, which is mounted on the high-magnification lens 2. The industrial computer 6 is connected to the camera 1, the light source 3, and the conveyor 5 respectively, and is used to acquire images through the camera 1 and process the acquired images to obtain the laser spot width of the solar cells.

[0060] Camera 1 and high-magnification lens 2 are connected in sequence and vertically fixed by profiles. Coaxial light source 3 is placed horizontally in the observation hole on the right side of the high-magnification lens. Battery cell 4 is placed horizontally on conveyor 5. In order to facilitate positioning, a positioning device for battery cell 4 is set up to adjust the position before the battery cell 4 reaches the field of view of the high-magnification lens 2. Camera 1 maintains a communication connection with industrial control computer 6. Industrial control computer 6 integrates communication module, image processing module and storage module, etc.

[0061] When the battery cell 4 reaches the field of view of the high-magnification lens 2 via the conveyor 5, light emitted by the light source 3 is vertically irradiated onto the battery cell 4. At this time, a very small portion of the light is absorbed by the object, and most of the light is reflected on the surface of the battery cell 4. This reflected light enters the high-magnification lens 2. The industrial control computer 6 sends a control signal to the camera 1. When the camera 1 receives the control signal, it begins to acquire images of the battery cell 4 and feeds the image information back to the industrial control computer 6. The image processing module of the industrial control computer 6 calculates the width of the laser spot, and the storage module stores the relevant data. The implementation method in the image processing module is the machine vision-based battery cell laser spot width measurement method described in the above embodiment.

[0062] The technical principles of the present invention have been described above with reference to specific embodiments. These descriptions are merely for explaining the principles of the invention and should not be construed as limiting the scope of protection of the invention in any way. Based on this explanation, those skilled in the art can readily conceive of other specific embodiments of the invention without inventive effort, and these embodiments will all fall within the scope of protection of the claims of the present invention.

Claims

1. A method for measuring the laser spot width of a battery cell based on machine vision, characterized in that, Includes the following steps: Image acquisition: acquiring color images of the battery cells; Image color space conversion to obtain the H component image of the type A solar cell; The H component image is binarized, and noise is removed by selecting a minimum threshold and a maximum threshold to obtain a binary image; Laser spot detection involves extracting edge information of the laser spot on the battery cell from the binary image and obtaining an image containing one or more laser spots using a contour detection algorithm. Image localization and segmentation: First, locate a spot region on the image containing one or more laser spots, and then segment the H component image of the A-type battery cell based on the localization information and the width information of the real-time acquired image. The segmented image is magnified by setting the scaling factor fx along the horizontal axis to 1 and the scaling factor fy along the vertical axis to 10, and using bicubic interpolation to magnify the segmented image. The magnified segmented image is binarized, and noise is removed using the selected minimum and maximum thresholds; The segmented image after binarization is closed. Based on the segmented binary image, the closing operation is used to first connect the broken areas of the laser spot, and then refine the upper and lower boundaries of the laser spot. Laser spot detection in segmented image: The laser spot region is extracted from the closed image using a contour detection algorithm, and the number of this region is 1. The laser spot width is calculated by determining the fitted line width midWidth, the average value of the new width set avgWidth, and the median value of the new width set medWidth based on the laser spot contour information. The output includes the image showing the edge of the laser spot and the obtained laser spot width data.

2. The method for measuring the laser spot width of a battery cell based on machine vision according to claim 1, characterized in that, The contour detection algorithm includes the following steps: Extract the contour of the laser spot and set the search mode to detect only the outermost contour; The contour approximation method is used to compress horizontal, vertical, and diagonal line segments, retaining only their endpoints; Find all contours, and store each contour as a vector of points; Laser spot regions were selected based on area characteristics.

3. The method for measuring the laser spot width of a battery cell based on machine vision according to claim 1, characterized in that, The image localization and segmentation includes the following steps: When the length of the upper boundary is assumed to be greater than a certain proportion of the width of the acquired image for the first time, this boundary is determined to be the upper boundary of the first laser spot found. The number of rows of the upper boundary is topPoint. The coordinates of the upper left vertex of the first spot are found to be (0, topPoint). This point is shifted upward by 300 pixels, and the coordinates of the upper left vertex of the segmented image are (0, topPoint-300). If topPoint-300 is less than 0, then the vertex coordinates are (0,0). Starting from the row number topPoint of the found upper boundary, find the lower boundary of the laser spot. When the size of the lower boundary is first greater than a certain proportion of the width of the acquired image, determine that this boundary is the lower boundary of the first laser spot found, and the row number of the lower boundary is bottomPoint. Find the coordinates of the lower left bottom point of the first laser spot as (0, bottomPoint). Offset this point down by 300 pixels, and the coordinates of the lower left bottom point of the segmented image are (0, bottomPoint +300). If bottomPoint +300 is less than the height h of the acquired image, then the bottom point coordinates are (0, h). The width of the segmented image is calculated based on the coordinates of the top left vertex and bottom left apex of the first laser spot after offset. The H component image of the type A battery cell is segmented based on the coordinates of the upper left vertex of the first laser spot after the offset, the width of the segmented image, and the width information of the acquired image.

4. The method for measuring the laser spot width of a battery cell based on machine vision according to claim 1, characterized in that, The laser spot width calculation includes the following steps: Based on the laser spot region image obtained from the laser spot detection step of the segmented image, the contour information of the upper boundary of the laser spot is extracted. First, a dynamic array topPoints of two-dimensional point type is established, and then the contour detection algorithm is used to store the points that meet the conditions into topPoints. Based on the topPoints two-dimensional point set, a straight line is iteratively fitted using the M estimation method. The M estimator uses Manhattan distance to robustly measure the deviation of data points from the straight line. The accuracy of the radial direction and the accuracy of the angle are both 0.

01. The fitted upper boundary line of the laser spot contains a vector of 4 elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the line, and (x0, y0) is a point on the line. Based on the laser spot region image obtained from the laser spot detection step of the segmented image, the contour information of the lower boundary of the laser spot is extracted. First, a dynamic array of two-dimensional points, botPoints, is established. Then, the contour detection algorithm is used to find the lower boundary point from bottom to top. Finally, the points that meet the conditions are stored in botPoints. Based on the botPoints two-dimensional point set, a straight line is iteratively fitted using the M estimation method. The accuracy of the radial direction and the accuracy of the angle are both 0.

01. The fitted lower boundary line of the laser spot contains a vector of 4 elements (vx, vy, x0, y0), where (vx, vy) is a normalized vector collinear with the line, and (x0, y0) is a point on the line. The distance midWidth between the two lines is calculated based on the fitted upper and lower boundary line parameters of the laser spot.

5. The method for measuring the laser spot width of a battery cell based on machine vision according to claim 4, characterized in that, The calculation of the distance midWidth between two lines includes the following steps: Create a dynamic array widths, and then calculate the set of laser spot widths based on the set of upper and lower boundary points of the laser spot and store it in widths; After deleting a certain proportion of excessively large and excessively small width values ​​from the width set widths, a new laser spot width set newWidths is generated; Calculate the average avgWidth and median medWidth of the new width set newWidths; Based on the actual calibration parameters, the pixel width is converted into the actual physical width.

6. A machine vision-based laser spot width measurement system for solar cells, used in the machine vision-based laser spot width measurement method for solar cells as described in claim 1, characterized in that, include: camera; A high-magnification lens is mounted on the camera; The light source is located on the high-magnification lens; A conveyor for transporting battery cells, located below the high-magnification lens; as well as An industrial control computer is connected to the camera, the light source, and the conveyor, respectively, and is used to acquire images through the camera and process the acquired images to obtain the laser spot width of the battery cell.

Images