A method and system for detecting laser welding points of positive and negative connecting pieces of an electric core by machine vision

Abstract

This invention discloses a machine vision-based method and system for detecting laser solder joints on positive and negative electrode connectors of a battery cell. The method includes building a detection platform and acquiring three-dimensional point cloud data and two-dimensional image data of the laser solder joints on the positive and negative electrode connectors of the battery cell; preprocessing the acquired three-dimensional point cloud data and two-dimensional image data; extracting the annular contour and calculating the width parameter of the laser solder joint, and determining the passability of the solder joint annular width by comparing it with a preset standard width range; extracting the depth features of the laser solder joint area and fusing the grayscale features of the two-dimensional image data to comprehensively identify whether there is a weld penetration defect; integrating the annular width detection and weld penetration defect identification results to generate an inspection report. This invention, considering the structural characteristics of the positive and negative electrode connectors of the battery cell and combining the annular distribution characteristics of the laser solder joints, specifically designs a solder joint annular width detection process. Through two-dimensional image contour extraction, circle fitting, and center deviation verification, it achieves accurate measurement of the annular width.

Description

Technical Field

[0001] This invention relates to the field of power battery technology for new energy vehicles, and more specifically, to a machine vision-based method and system for detecting laser weld points on positive and negative connectors of battery cells. Background Technology

[0002] New energy vehicle power batteries consist of several cells connected in series or parallel via positive and negative electrode connecting plates. These connecting plates are typically rectangular aluminum sheets, fixedly connected to the positive and negative electrode tabs of the cells via laser welding. Figure 1 and Figure 2 As shown, the quality of laser welds directly affects the conductivity, structural stability, and safety of power batteries. The core indicators for evaluating weld quality are whether the weld ring width meets standards and whether there are any weld-through defects.

[0003] Currently, the inspection of laser weld joints on the positive and negative electrode connectors of battery cells mainly relies on manual visual inspection, supplemented by simple tools. This method suffers from low inspection efficiency, poor accuracy, poor consistency, and high labor intensity, making it difficult to meet the needs of large-scale production. Existing machine vision-based welding inspection methods are mostly applied to parts such as the top cover or electrode sheets of power batteries, with few specific inspection solutions for the annular weld joints on the positive and negative electrode connectors of battery cells. Furthermore, these methods are not optimized for the structural characteristics of rectangular aluminum sheets and the annular shape of the weld joints, making it difficult to accurately extract the outline of the weld joint annulus, resulting in large errors in the measurement of the annulus width. At the same time, the identification of weld penetration defects often relies on single features, which are easily affected by factors such as aluminum sheet surface reflection and changes in lighting, resulting in insufficient identification accuracy and failing to simultaneously meet the dual requirements of annulus width measurement and weld penetration defect identification.

[0004] Therefore, there is an urgent need to develop a machine vision inspection method and system that can accurately and efficiently detect the width of the laser solder joint ring and the weld penetration defect of the positive and negative electrode connecting piece of the battery cell, and is suitable for large-scale production scenarios. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of the prior art and provide a machine vision-based laser solder joint detection method and system for positive and negative connectors of battery cells, so as to achieve accurate measurement of the solder joint ring width and efficient identification of weld penetration defects, improve detection efficiency and accuracy, and ensure the production quality and safety of power batteries.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] A machine vision-based method for detecting laser solder joints on positive and negative connectors of a battery cell includes the following steps:

[0008] Step S1: Build a testing platform to acquire three-dimensional point cloud data and two-dimensional image data of the laser welding points of the positive and negative electrode connection pieces of the battery cell;

[0009] Step S2: Preprocess the acquired 3D point cloud data and 2D image data;

[0010] Step S3: Based on the preprocessed two-dimensional image data, extract the circular contour and calculate the width parameter of the laser welding point. By comparing it with the preset standard width range, determine the qualification of the welding point circular width.

[0011] Step S4: Based on the preprocessed 3D point cloud data, extract the depth features of the laser welding point area and fuse the grayscale features of the 2D image data to comprehensively identify whether the welding point has a weld penetration defect.

[0012] Step S5: Integrate the results of the ring width detection and weld penetration defect identification, generate an inspection report, mark unqualified weld points and trigger an early warning.

[0013] Furthermore, in step S1, the detection platform includes a 3D vision module for acquiring three-dimensional point cloud data of the solder joints, a 2D industrial camera for acquiring two-dimensional image data of the solder joints, a light source module for providing uniform illumination, and a stage for fixing the battery cell and positive and negative electrode connecting pieces.

[0014] Furthermore, in step S2, the preprocessing includes: applying Gaussian filtering, adaptive histogram equalization, threshold segmentation, and morphological opening and closing operations to the two-dimensional image data.

[0015] Furthermore, in step S2, the preprocessing includes: applying radius filtering, iterative nearest point registration, and pass-through filtering to the 3D point cloud data.

[0016] Furthermore, in step S3, the detection of the weld joint ring width includes:

[0017] Step S31: The Canny edge detection algorithm is used to extract the inner and outer edge contours of the laser welding point.

[0018] Step S32: Perform circle fitting on the extracted inner edge contour and outer edge contour respectively to obtain the center coordinates and corresponding radius of the inner edge circle, and the center coordinates and corresponding radius of the outer edge circle.

[0019] Step S33: Calculate the center deviation between the inner edge circle and the outer edge circle. If the center deviation is greater than the preset center deviation threshold, the weld point ring outline is determined to be irregular and marked as unqualified in width detection.

[0020] Step S34: If the center deviation is less than or equal to the preset center deviation threshold, the ring width is calculated based on the radii of the inner and outer edge circles, and the ring width is compared with the preset standard width range. If the ring width is within the preset standard width range, the weld ring width is deemed qualified; otherwise, it is deemed unqualified.

[0021] Furthermore, in step S4, the identification of weld burn-through defects includes:

[0022] Step S41: Extract the depth values ​​of all points within the solder joint area, and calculate the average depth and standard deviation of the depth within the solder joint area;

[0023] Step S42: Set a depth threshold. If there is a point in the solder joint area with a depth value less than the depth threshold, and the absolute value of the difference between the depth value of the point and the average depth is greater than 3 times the depth standard deviation, then the point is determined to be a suspected solder burn-through point.

[0024] Step S43: Extract the gray value of the image area corresponding to the suspected weld penetration point. If the difference between the gray value of the area and the average gray value of the surrounding weld point area is greater than the preset gray value threshold, then the point is confirmed as a weld penetration defect.

[0025] Step S44: Count the number and size of weld penetration defects. If the number of weld penetration defects is greater than a preset number threshold, or the size of a single weld penetration defect is greater than a preset size threshold, then it is determined that the weld joint has a weld penetration defect.

[0026] Furthermore, it also includes step S6, periodic adaptive calibration, which uses standard solder joint samples for periodic calibration, generates calibration coefficients, and corrects subsequent test results.

[0027] The present invention also provides a machine vision-based laser solder joint detection system for positive and negative connectors of battery cells, the system comprising:

[0028] The testing platform module provides a stable testing environment, including a stage for fixing the battery cells and positive and negative electrode connecting pieces, and a light source module for providing illumination.

[0029] The data acquisition module is used to acquire 3D point cloud data and 2D image data of the weld joints;

[0030] The data preprocessing module is used to preprocess the acquired 3D point cloud data and 2D image data;

[0031] The ring width detection module is used to extract the outline of the weld point ring based on the preprocessed two-dimensional image data, calculate the ring width, and compare and judge its passability.

[0032] The weld penetration defect identification module is used to extract depth features based on preprocessed 3D point cloud data and combine them with grayscale features of 2D image data to identify weld penetration defects.

[0033] The result integration and early warning module is used to integrate the results of ring width detection and weld penetration defect identification, generate a detection report and trigger an early warning.

[0034] The calibration module is used for periodic calibration to ensure stable detection accuracy.

[0035] Furthermore, the ring width detection module includes a contour extraction unit, a circle fitting unit, and a width comparison unit. The contour extraction unit uses the Canny edge detection algorithm to extract the inner and outer edges of the solder joint ring. The circle fitting unit uses the least squares method to fit the inner and outer edges to obtain the center coordinates and radius. The width comparison unit calculates the center deviation and the ring width, compares them with a preset center deviation threshold and a standard width range, and outputs the detection result.

[0036] Furthermore, the weld penetration defect identification module includes a depth feature extraction unit, a suspected point screening unit, a grayscale verification unit, and a defect determination unit. The depth feature extraction unit extracts the depth value of the weld point area and calculates the average depth and depth standard deviation. The suspected point screening unit filters out suspected weld penetration points based on the depth threshold and depth standard deviation. The grayscale verification unit extracts the image grayscale value of the suspected weld penetration point and compares it with the surrounding area to verify the weld penetration defect. The defect determination unit counts the number and size of weld penetration defects and outputs the defect identification result.

[0037] The beneficial effects of this invention are:

[0038] 1. This invention addresses the structural characteristics of the positive and negative electrode connecting pieces of the battery cell and combines them with the annular distribution characteristics of laser solder joints. It specifically designs a solder joint annular width detection process, which achieves accurate measurement of the annular width through two-dimensional image contour extraction, circle fitting, and center deviation verification. The measurement error can be controlled within 0.04mm, solving the problem of insufficient accuracy in annular width measurement in existing methods and ensuring that the solder joint connection strength meets the requirements.

[0039] 2. This invention uses a combination of three-dimensional point cloud depth features and two-dimensional image grayscale features to identify weld penetration defects. It not only uses three-dimensional depth information to accurately capture depth anomalies in the weld penetration area, but also uses two-dimensional grayscale information to verify the authenticity of the defect. This effectively avoids false detection and missed detection caused by aluminum sheet surface reflection and light interference. The weld penetration defect recognition rate can reach over 99%, improving the reliability of defect recognition.

[0040] 3. This invention adopts visual automated inspection throughout the entire process, which improves the inspection efficiency by more than 5 times compared with manual inspection. It can adapt to the pace requirements of large-scale production of power batteries, while eliminating the influence of subjective human factors, ensuring the consistency and stability of inspection results, and reducing the intensity of manual labor and inspection costs.

[0041] 4. This invention incorporates a periodic adaptive calibration module, which periodically calibrates using standard solder joint samples to correct detection errors in a timely manner, ensuring the stability of long-term detection accuracy. Simultaneously, it integrates the detection results to generate detailed reports, marks unqualified solder joints, and triggers early warnings, facilitating timely handling by staff and improving the efficiency of production quality control.

[0042] 5. The detection method and system of the present invention are highly versatile and can be adapted to the laser welding point detection of positive and negative connecting pieces of battery cells of different specifications and sizes. It does not require extensive adjustment of equipment parameters, thus reducing equipment debugging costs and has broad industrial application value. Attached Figure Description

[0043] Figure 1 This is a schematic diagram of a structure for the positive and negative electrode connection piece of a battery cell in the prior art.

[0044] Figure 2 This is a schematic diagram of a typical morphology of a laser weld joint in the prior art.

[0045] Figure 3 This is a flowchart of a machine vision-based laser solder joint detection method for positive and negative connectors of battery cells in this embodiment.

[0046] Figure 4 This is a framework diagram of a machine vision-based laser solder joint detection system for positive and negative connectors of battery cells in this embodiment.

[0047] Figure 5 This is a topographic image of the laser weld joint before image preprocessing in this embodiment;

[0048] Figure 6 This is a morphological image of the laser weld joint after image preprocessing in this embodiment;

[0049] Figure 7 This is a rendering of the circular ring feature extraction in this embodiment;

[0050] Figure 8 This is a rendering of a burn-through defect in this embodiment;

[0051] Figure 9 This is a rendering illustrating a scenario where the solder joints of the positive and negative electrode connecting pieces of the battery cell are defective in this embodiment.

[0052] Figure 10 This is a rendering showing a qualified solder joint of the positive and negative electrode connecting pieces of the battery cell in this embodiment.

[0053] Figure reference numerals: Detection platform module 1, data acquisition module 2, data preprocessing module 3, annular width detection module 4, contour extraction unit 41, circle fitting unit 42, width comparison unit 43, weld penetration defect identification module 5, depth feature extraction unit 51, suspected point screening unit 52, grayscale verification unit 53, defect judgment unit 54, result integration and early warning module 6, calibration module 7. Detailed Implementation

[0054] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0055] Example: A machine vision-based method for detecting laser solder joints on positive and negative connectors of battery cells, such as... Figure 3 As shown, it includes the following steps:

[0056] Step S1: Build a testing platform to obtain three-dimensional point cloud data and two-dimensional image data of the laser welding points of the positive and negative electrode connecting pieces of the battery cell; wherein, the positive and negative electrode connecting pieces of the battery cell are rectangular aluminum sheets, which are connected to the positive and negative electrode tabs of the battery cell by laser welding, and the laser welding points are distributed in a ring at the connection between the connecting pieces and the tabs.

[0057] Furthermore, the inspection platform includes a stage, a light source module, a 3D vision module, and a 2D industrial camera. The stage is a high-precision motorized stage with a positioning accuracy of ±0.01mm, used to fix the battery cells and positive and negative electrode connecting pieces, ensuring stable workpiece position during inspection. The light source module uses a ring-shaped supplementary light source with adjustable brightness from 0-1000 lux and rotation angle from 0-360°. For example, to address the surface reflectivity of a rectangular aluminum sheet, the light source angle can be adjusted to 45° and the brightness to 500 lux to reduce reflective interference. The 3D vision module uses a structured light 3D camera with a resolution of 0.006mm and a frame rate of 15fps, used to acquire 3D point cloud data of the weld joints and obtain their depth information. The 2D industrial camera is a 5-megapixel camera with a frame rate of 30fps and a lens focal length of 16mm, used to acquire 2D image data of the weld joints and obtain their contour information.

[0058] During testing, the battery cell and positive and negative electrode connecting pieces are fixed on the stage. The positions of the 3D vision module and 2D industrial camera are adjusted so that the lens is aimed at the laser welding point area. At the same time, the angle and brightness of the light source are adjusted to reduce reflection interference. The acquisition equipment is started to simultaneously acquire the three-dimensional point cloud data and two-dimensional image data of the welding point.

[0059] Step S2 involves preprocessing the acquired 3D point cloud data and 2D image data. The results before and after preprocessing are shown in the figure. Figure 5 and Figure 6 As shown.

[0060] Two-dimensional image data preprocessing: A 5×5 Gaussian filter kernel is used to filter the two-dimensional image to eliminate image noise; an adaptive histogram equalization algorithm is used to enhance image contrast and improve the distinction between the solder joint contour and the background area; the Otsu threshold segmentation algorithm is used to automatically determine the segmentation threshold and separate the solder joint area from the background area; then, morphological opening operation (kernel size of 3×3) is used to remove small impurities in the image, and morphological closing operation is used to fill the small holes in the solder joint contour to obtain a clear two-dimensional contour image of the solder joint.

[0061] 3D point cloud data preprocessing: A radius filtering algorithm is used, with a filtering radius of 0.01 mm and a minimum neighbor number of 5, to remove outliers from the point cloud; the iterative nearest point (ICP) algorithm is used to register the point cloud with the reference point on the surface of the connector as a reference, correcting the point cloud position deviation, and the registration error is controlled within 0.005 mm; a through filtering algorithm is used, with a Z-axis depth range set, to extract the 3D point cloud of the solder joint area, and remove redundant point cloud data in the non-solder joint area of ​​the connector to reduce the amount of data for subsequent processing.

[0062] Step S3: Based on the preprocessed 2D image data, extract the annular contour and calculate the width parameter of the laser weld point. By comparing it with the preset standard width range, determine the qualification of the weld point annular width. The annular feature extraction effect is as follows: Figure 7 As shown.

[0063] Specifically, this step includes the following steps:

[0064] Step S31: Using the Canny edge detection algorithm, with a low threshold of 50 and a high threshold of 150, extract the circular edge of the laser welding point in the preprocessed two-dimensional contour image to obtain the inner and outer edge contours of the circular ring.

[0065] Step S32: Using the least squares method, circle fitting is performed on the extracted inner and outer edge contours to obtain the coordinates of the center of the inner edge circle. and radius The coordinates of the center of the outer edge circle and radius ;

[0066] Step S33: Calculate the center deviation between the inner edge circle and the outer edge circle. The calculation formula is as follows:

[0067]

[0068] Set a preset center deviation threshold (e.g., 0.02mm). If the center deviation... If the deviation exceeds the preset center deviation threshold, the weld point ring outline is determined to be irregular and marked as unqualified in width detection.

[0069] Step S34, if the center of the circle deviates If the deviation is less than or equal to the preset center deviation threshold, calculate the ring width. The calculation formula is as follows:

[0070]

[0071] Set the preset standard width range It can be adjusted according to the welding point requirements of different specifications of connecting pieces, such as... ;like If the width of the weld ring is deemed acceptable; ,or If so, the width of the weld ring is deemed unqualified, and the width deviation value is recorded (e.g., ...). =0.7mm, deviation value is -0.1mm).

[0072] Step S4: Based on the preprocessed 3D point cloud data, extract the depth features of the laser welding point area and fuse them with the grayscale features of the 2D image data to comprehensively identify whether there is a weld penetration defect. The effect of weld penetration defect is as follows: Figure 8 As shown.

[0073] Specifically, this step includes the following steps:

[0074] Step S41: Based on the preprocessed 3D point cloud data, extract the depth values ​​of all points within the solder joint area and calculate the average depth of the solder joint area. and depth standard deviation For example, the average depth of a certain solder joint area =2.5mm, depth standard deviation =0.03mm;

[0075] Step S42, set the depth threshold The settings are based on the thickness of the connecting piece and the welding process, such as =2.3mm, if the depth value in the solder joint area is less than the depth threshold The point, and the depth value of that point. With average depth The absolute value of the difference is greater than (like If the point is suspected to be a weld burn-through point, then the point is determined to be a suspected weld burn-through point.

[0076] Step S43: Based on the preprocessed two-dimensional image data, extract the grayscale value of the image region corresponding to the suspected weld penetration point, and set a preset grayscale threshold (e.g., the preset grayscale threshold is 50, and the average grayscale value of the surrounding weld point region is 30). If the difference between the grayscale value of this region and the average grayscale value of the surrounding weld point region is greater than the preset grayscale threshold (e.g., the grayscale value of this region is greater than 80, and the grayscale difference with the surrounding region is greater than 50), then the point is confirmed as a weld penetration defect (the grayscale value of the weld penetration region is significantly higher than that of the surrounding region due to increased light transmittance).

[0077] Step S44: Set the preset quantity threshold to 1 and the preset size threshold to 0.1mm. 2 The number and size of weld penetration defects are counted. If the number of weld penetration defects exceeds a preset threshold, or the size of a single weld penetration defect exceeds a preset size threshold, the weld joint is determined to have weld penetration defects and is marked as unqualified; otherwise, it is determined to have no weld penetration defects and is marked as qualified.

[0078] Step S5: Integrate the results of the ring width detection and weld penetration defect identification, generate an inspection report, mark unqualified weld points and trigger an early warning.

[0079] The inspection report includes the cell number, connector number, solder joint location, ring width measurement, ring width qualification status, weld penetration defect identification results (whether there is weld penetration, the number and size of weld penetration), reasons for non-compliance (such as ring width being too narrow, presence of weld penetration defects), and inspection time; non-conforming solder joints are marked with coordinates, and an audible and visual alarm (flashing red light, buzzer prompt) is triggered, and a signal warning is sent to the production control system to notify staff to handle non-conforming solder joints in a timely manner.

[0080] Step S6, periodic adaptive calibration, involves performing periodic calibration using standard solder joint samples to generate calibration coefficients, which are then used to correct subsequent test results. Specifically:

[0081] The preset calibration cycle is set to 24 hours. When the calibration cycle is reached, a standard solder joint sample (e.g., the standard value of the ring width is known to be 1.0 mm, with no weld penetration defects) is used for testing. The deviation between the measured ring width and the standard value is calculated to generate a calibration coefficient. For example, if the measured width of the standard solder joint is 1.02 mm, the deviation is +0.02 mm, and the calibration coefficient is -0.02 mm. In subsequent tests, the measured ring width is subtracted from the calibration coefficient to obtain the corrected actual width, ensuring the stability of the test accuracy.

[0082] This embodiment also provides a machine vision-based laser solder joint detection system for positive and negative connectors of battery cells, such as... Figure 4 As shown, the system includes a detection platform module 1, a data acquisition module 2, a data preprocessing module 3, a ring width detection module 4, a weld penetration defect identification module 5, a result integration and early warning module 6, and a calibration module 7.

[0083] The detection platform module 1 is responsible for providing a stable detection environment, including a stage and a light source module. The stage is a high-precision electric stage with a positioning accuracy of ±0.01mm, used to fix the battery cell and positive and negative electrode connecting pieces (rectangular aluminum sheets) to ensure the stability of the workpiece position during the detection process. The light source module is a ring-shaped supplementary light source with brightness adjustable in the range of 0-1000 lux and angle adjustable in the range of 0-360°, adapted to the reflective characteristics of the rectangular aluminum sheet surface to reduce reflective interference.

[0084] The data acquisition module 2 includes a 3D vision module and a 2D industrial camera. The 3D vision module is a structured light 3D camera with a resolution of 0.006mm and a frame rate of 15fps, used to acquire 3D point cloud data of the laser welding points of the positive and negative electrode connecting pieces of the battery cell and obtain the depth information of the welding points. The 2D industrial camera is a 5-megapixel camera with a frame rate of 30fps and a lens focal length of 16mm, used to acquire 2D image data of the welding points and obtain the contour information of the welding points. The data acquisition module 2 transmits the acquired 3D point cloud data and 2D image data to the data preprocessing module 3.

[0085] The data preprocessing module 3 is responsible for preprocessing the 3D point cloud data and 2D image data acquired by the data acquisition module 2, eliminating noise, lighting interference and redundant data, and outputting the preprocessed 3D point cloud data and 2D image data.

[0086] The data preprocessing module 3 includes a two-dimensional image preprocessing unit and a three-dimensional point cloud preprocessing unit. The two-dimensional image preprocessing unit uses Gaussian filtering algorithm, adaptive histogram equalization algorithm, Otsu threshold segmentation algorithm and morphological opening and closing operation to achieve image denoising, contrast enhancement and contour extraction. The three-dimensional point cloud preprocessing unit uses radius filtering algorithm, ICP registration algorithm and pass-through filtering algorithm to achieve point cloud denoising, registration and redundant data removal.

[0087] The ring width detection module 4 is responsible for extracting the outline of the solder joint ring based on the preprocessed two-dimensional image data, fitting the inner edge circle and the outer edge circle, calculating the ring width, comparing it with the preset standard, and outputting the ring width detection result.

[0088] The ring width detection module 4 includes a contour extraction unit 41, a circle fitting unit 42, and a width comparison unit 43. The contour extraction unit 41 uses the Canny edge detection algorithm to extract the inner and outer edges of the weld ring. The circle fitting unit 42 uses the least squares method to fit the inner and outer edges to obtain the center coordinates and radius. The width comparison unit 43 calculates the center deviation and the ring width, compares them with the preset threshold and standard range, and outputs the detection result (pass / fail, deviation value).

[0089] The weld penetration defect identification module 5 is responsible for extracting the depth features of the weld point based on the preprocessed 3D point cloud data, combining the grayscale features of the 2D image data, identifying whether there is a weld penetration defect, and outputting the weld penetration defect identification result.

[0090] The weld penetration defect identification module 5 includes a depth feature extraction unit 51, a suspected point screening unit 52, a grayscale verification unit 53, and a defect judgment unit 54. The depth feature extraction unit 51 extracts the depth value of the weld point area and calculates the average depth and depth standard deviation. The suspected point screening unit 52 filters out suspected weld penetration points based on the depth threshold and standard deviation. The grayscale verification unit 53 extracts the image grayscale value of the suspected weld penetration point and compares it with the surrounding area to verify the weld penetration defect. The defect judgment unit 54 counts the number and size of weld penetration defects and outputs the defect identification result (whether there is weld penetration, the number and size of weld penetration, and whether it is qualified or unqualified).

[0091] The result integration and early warning module 6 is responsible for integrating the output results of the ring width detection module 4 and the weld penetration defect identification module 5, generating a test report, marking the coordinates of unqualified weld points, and triggering audible and visual warnings and signal warnings. The test report is stored in the system database and can be queried at any time. The audible and visual warnings are implemented through red lights and buzzers, and the signal warnings are sent to the production control system to facilitate timely handling of unqualified weld points by staff.

[0092] Calibration module 7 is responsible for setting a preset calibration cycle (e.g., 24 hours), performing periodic calibration using standard solder joint samples, calculating calibration coefficients, and calibrating subsequent test results to ensure the stability of test accuracy. Calibration module 7 stores calibration records, allowing for traceability of the calibration process. If the calibration error exceeds a preset range, it triggers an equipment maintenance warning.

[0093] To verify the effectiveness of the method proposed in this embodiment in identifying the weld ring width and weld penetration defects, 1000 laser weld joints of the positive and negative electrode connecting pieces of the battery cell were inspected. Among them, 800 weld joints were found to be qualified (ring width 0.8-1.2mm, no weld penetration defects). Figure 10 As shown, there are 200 defective welds (100 of which have unacceptable ring width, 50 have burn-through defects, and 50 have both unacceptable width and burn-through defects). Figure 9 As shown, the defective weld point is a burn-through defect.

[0094] Experimental results show that the average measurement error of the weld ring width of the detection system in this embodiment is 0.038 mm, and the measurement accuracy reaches 99.2%; the identification rate of weld penetration defects is 99.5%, the missed detection rate is 0.3%, and the false detection rate is 0.2%; the detection efficiency is 120 pieces / hour, which is 5 times higher than manual detection (20 pieces / hour); the detection consistency is good, and the deviation of different batches of detection results is less than 0.01 mm, which fully meets the detection requirements of laser weld joints of positive and negative electrode connecting pieces of battery cells in the production process of power batteries.

[0095] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A machine vision-based method for detecting laser solder joints on positive and negative connectors of a battery cell, characterized in that, Includes the following steps: Step S1: Build a testing platform to obtain three-dimensional point cloud data and two-dimensional image data of the laser welding points of the positive and negative electrode connection pieces of the battery cell; Step S2: Preprocess the acquired 3D point cloud data and 2D image data; Step S3: Based on the preprocessed two-dimensional image data, extract the circular contour and calculate the width parameter of the laser welding point. By comparing it with the preset standard width range, determine the qualification of the welding point circular width. Step S4: Based on the preprocessed 3D point cloud data, extract the depth features of the laser welding point area and fuse the grayscale features of the 2D image data to comprehensively identify whether the welding point has a weld penetration defect. Step S5: Integrate the results of the ring width detection and weld penetration defect identification, generate an inspection report, mark unqualified weld points and trigger an early warning.

2. The method for detecting laser solder joints of positive and negative connectors in a battery cell using machine vision according to claim 1, characterized in that, In step S1, the detection platform includes a 3D vision module for acquiring three-dimensional point cloud data of the solder joints, a 2D industrial camera for acquiring two-dimensional image data of the solder joints, a light source module for providing uniform illumination, and a stage for fixing the battery cells and positive and negative electrode connecting pieces.

3. The method for detecting laser solder joints of positive and negative connectors in a battery cell using machine vision according to claim 1, characterized in that, In step S2, the preprocessing includes: applying Gaussian filtering, adaptive histogram equalization, threshold segmentation, and morphological opening and closing operations to the two-dimensional image data.

4. The machine vision method for detecting laser solder joints of positive and negative connectors in a battery cell according to claim 1, characterized in that, In step S2, the preprocessing includes: applying radius filtering, iterative nearest point registration, and pass-through filtering to the 3D point cloud data.

5. The machine vision method for detecting laser solder joints of positive and negative connectors in a battery cell according to claim 1, characterized in that, In step S3, the detection of the weld joint ring width includes: Step S31: The Canny edge detection algorithm is used to extract the inner and outer edge contours of the laser welding point. Step S32: Perform circle fitting on the extracted inner edge contour and outer edge contour respectively to obtain the center coordinates and corresponding radius of the inner edge circle, and the center coordinates and corresponding radius of the outer edge circle. Step S33: Calculate the center deviation between the inner edge circle and the outer edge circle. If the center deviation is greater than the preset center deviation threshold, the weld point ring outline is determined to be irregular and marked as unqualified in width detection. Step S34: If the center deviation is less than or equal to the preset center deviation threshold, the ring width is calculated based on the radii of the inner and outer edge circles, and the ring width is compared with the preset standard width range. If the ring width is within the preset standard width range, the weld ring width is deemed qualified; otherwise, it is deemed unqualified.

6. The method for detecting laser solder joints of positive and negative connectors in a battery cell using machine vision according to claim 1, characterized in that, In step S4, the identification of weld burn-through defects includes: Step S41: Extract the depth values ​​of all points within the solder joint area, and calculate the average depth and standard deviation of the depth within the solder joint area; Step S42: Set a depth threshold. If there is a point in the solder joint area with a depth value less than the depth threshold, and the absolute value of the difference between the depth value of the point and the average depth is greater than 3 times the depth standard deviation, then the point is determined to be a suspected solder burn-through point. Step S43: Extract the gray value of the image area corresponding to the suspected weld penetration point. If the difference between the gray value of the area and the average gray value of the surrounding weld point area is greater than the preset gray value threshold, then the point is confirmed as a weld penetration defect. Step S44: Count the number and size of weld penetration defects. If the number of weld penetration defects is greater than a preset number threshold, or the size of a single weld penetration defect is greater than a preset size threshold, then it is determined that the weld joint has a weld penetration defect.

7. The method for detecting laser solder joints of positive and negative connectors in a battery cell using machine vision according to claim 1, characterized in that, It also includes step S6, periodic adaptive calibration, which uses standard solder joint samples for periodic calibration, generates calibration coefficients, and corrects subsequent test results.

8. A machine vision-based laser solder joint detection system for positive and negative connectors of a battery cell, used to implement the method of claim 1, characterized in that, The system includes: The testing platform module (1) is used to provide a stable testing environment, including a stage for fixing the battery cell and positive and negative electrode connecting pieces, and a light source module for providing illumination. The data acquisition module (2) is used to acquire three-dimensional point cloud data and two-dimensional image data of the weld joint; The data preprocessing module (3) is used to preprocess the collected 3D point cloud data and 2D image data; The ring width detection module (4) is used to extract the outline of the weld ring based on the preprocessed two-dimensional image data, calculate the ring width and compare and judge the qualification. The weld penetration defect identification module (5) is used to extract depth features based on the preprocessed three-dimensional point cloud data and identify weld penetration defects by combining the grayscale features of the two-dimensional image data. The result integration and early warning module (6) is used to integrate the results of ring width detection and weld penetration defect identification, generate a detection report and trigger an early warning; The calibration module (7) is used to perform periodic calibration to ensure stable detection accuracy.

9. The machine vision-based laser solder joint detection system for positive and negative connectors of a battery cell according to claim 8, characterized in that, The ring width detection module (4) includes a contour extraction unit (41), a circle fitting unit (42), and a width comparison unit (43). The contour extraction unit (41) uses the Canny edge detection algorithm to extract the inner and outer edges of the weld ring. The circle fitting unit (42) uses the least squares method to perform circle fitting on the inner and outer edges to obtain the center coordinates and radius. The width comparison unit (43) calculates the center deviation and the ring width, compares them with the preset center deviation threshold and the standard width range, and outputs the detection result.

10. The machine vision-based laser solder joint detection system for positive and negative connectors of a battery cell according to claim 8, characterized in that, The weld penetration defect identification module (5) includes a depth feature extraction unit (51), a suspected point screening unit (52), a grayscale verification unit (53), and a defect judgment unit (54). The depth feature extraction unit (51) extracts the depth value of the weld point area and calculates the average depth and depth standard deviation. The suspected point screening unit (52) filters out suspected weld penetration points based on depth threshold and depth standard deviation; the grayscale verification unit (53) extracts the image grayscale value of the suspected weld penetration point and compares it with the surrounding area to verify the weld penetration defect. The defect determination unit (54) counts the number and size of weld penetration defects and outputs the defect identification results.

Images