Battery pad post-welding measurement and detection method and device based on deep learning technology

Abstract

The embodiment of the invention provides a battery pad post-welding measurement and detection method and device based on a deep learning technology, and belongs to the technical field of battery measurement and detection. The method comprises the following steps: acquiring a post-welding image of a battery pad; according to the post-welding image of the battery bonding pad, obtaining a minimum external rectangle of the welding lines; obtaining a measurement value according to the minimum bounding rectangle of the welding lines; the measured value comprises the offset of the bonding pad relative to the battery and the offset of each welding line relative to the bonding pad; obtaining welding line defect segmentation data according to the welding line minimum enclosing rectangle; obtaining defect analysis data according to the welding line defect segmentation data; and according to the measurement value and the defect analysis data, obtaining a post-welding measurement detection result of the battery bonding pad. According to the method, the situation that background interference of a welding line image is serious can be effectively solved, the welding lines are accurately positioned, then accurate measurement is carried out, and defects on the welding lines are accurately detected.

Description

Technical Field

[0001] The present application relates to the field of battery measurement and detection technology, and in particular to a battery pad post-welding measurement and detection method and device based on deep learning technology. Background Art

[0002] With the rapid development of new energy vehicles and other industries, higher requirements are placed on lithium battery equipment to produce higher quality battery products. There are many welding processes in the production process of battery products, and the post-welding appearance inspection is an important link to ensure the welding quality, and it is also one of the key links to ensure the quality of the final battery product.

[0003] The existing methods for detecting weld marks after battery welding have the disadvantages of poor detection accuracy, being greatly affected by the imaging background, and having poor detection capabilities for dot weld marks and slight weld marks. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide a battery pad post-welding measurement and detection method and device based on deep learning technology.

[0005] The technical solution adopted by the present invention is:

[0006] On the one hand, an embodiment of the present invention provides a battery pad post-welding measurement and detection method based on deep learning technology, the method comprising the following steps:

[0007] Get the image of the battery pad after welding;

[0008] According to the battery pad after welding image, obtaining the minimum circumscribed rectangle of the welding mark;

[0009] Obtaining a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes an offset of the welding pad relative to the battery and an offset of each welding mark relative to the welding pad;

[0010] Obtaining weld mark defect segmentation data according to the weld mark minimum circumscribed rectangle;

[0011] Obtaining defect analysis data according to the weld defect segmentation data;

[0012] According to the measurement values ​​and defect analysis data, the post-welding measurement and detection results of the battery pad are obtained.

[0013] Further, the obtaining of the battery pad post-welding image comprises the following steps:

[0014] Obtain several pad soldering products;

[0015] Adjusting imaging parameters; the imaging parameters include light source adjustment data, lens adjustment data, focus and exposure adjustment data;

[0016] According to the post-welding product of the welding pad and the imaging parameters, a post-welding image of the battery welding pad is obtained.

[0017] Further, obtaining the minimum circumscribed rectangle of the weld mark according to the image of the battery pad after welding includes the following steps:

[0018] Using a free rectangle target detection AI algorithm, the weld lines are located from the image of the battery pad after welding, and the minimum circumscribed rectangle of each weld line is obtained;

[0019] The minimum circumscribed rectangle of the weld pattern includes the coordinates of the center point of the rectangle and its length and width.

[0020] Further, obtaining the measurement value according to the minimum circumscribed rectangle of the weld mark comprises the following steps:

[0021] According to the minimum circumscribed rectangle of the weld mark, the center coordinates and radius of the battery are extracted using a traditional algorithm;

[0022] Extracting the position coordinates and angle of the welding pad according to the minimum circumscribed rectangle of the welding pattern;

[0023] According to the center coordinates and radius of the battery, the position coordinates and angle of the welding pad, and the center and length and width of the welding mark, the offset of the welding pad relative to the battery and the offset of each welding mark relative to the welding pad are calculated;

[0024] The offset of the pad relative to the battery and the offset of each weld line relative to the pad are taken as measured values.

[0025] Furthermore, obtaining the weld mark defect segmentation data according to the weld mark minimum circumscribed rectangle comprises the following steps:

[0026] Extracting a local part of the weld pattern based on the minimum circumscribed rectangle of the weld pattern;

[0027] According to the local weld line, a segmentation module is used to extract weld line defects;

[0028] According to the weld defect, a defect category, a bounding rectangle, a defect mask and a confidence score are obtained;

[0029] The defect category, the circumscribed rectangle, the defect mask and the confidence score are used as weld defect segmentation data.

[0030] Further, obtaining defect analysis data according to the weld defect segmentation data comprises the following steps:

[0031] Extracting defect features based on the weld defect segmentation data and in combination with the defect mask to obtain defect feature data;

[0032] The defect characteristic data is analyzed and tested to obtain defect analysis data.

[0033] Further, obtaining the post-welding measurement and detection result of the battery pad according to the measurement value and the defect analysis data comprises the following steps:

[0034] According to the defect analysis data, a plurality of defect characteristic values ​​are obtained;

[0035] According to the defect characteristic value and the measurement value, combined with the defect category and defect detection requirements, the post-welding measurement and detection result of the battery pad is obtained.

[0036] On the other hand, an embodiment of the present invention further provides a battery pad post-welding measurement and detection device based on deep learning technology, the device comprising:

[0037] The first module is used to obtain the image of the battery pad after welding;

[0038] The second module is used to obtain the minimum circumscribed rectangle of the welding mark according to the image of the battery pad after welding;

[0039] The third module is used to obtain a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes an offset of the pad relative to the battery and an offset of each weld mark relative to the pad;

[0040] The fourth module is used to obtain weld mark defect segmentation data according to the minimum circumscribed rectangle of the weld mark;

[0041] A fifth module is used to obtain defect analysis data according to the weld defect segmentation data;

[0042] The sixth module is used to obtain the post-welding measurement and detection results of the battery pad according to the measurement values ​​and defect analysis data.

[0043] Furthermore, the battery pad post-welding measurement and detection device based on deep learning technology also includes:

[0044] Free rectangle target detection and positioning module, used to locate the weld mark from the background;

[0045] A measurement module, used to extract the center coordinates and radius of the battery through an algorithm; extract the position coordinates and angle of the pad through an algorithm; calculate the measurement value of the pad relative to the battery and the measurement value of each weld line relative to the pad according to the center coordinates and radius of the battery, the position coordinates and angle of the pad, and the center and length and width of the weld line;

[0046] The measurement judgment module is used to comprehensively judge whether the measured part meets the standards according to the respective size specification requirements of the battery, pad and weld mark and the specification requirements of the relative position;

[0047] An instance segmentation module, used for extracting a part of the weld mark according to the minimum circumscribed rectangle of the weld mark;

[0048] A feature extraction and analysis module, used to extract defect features from the image based on the information obtained by the instance segmentation module and the defect mask, and select the required feature type for each defect through analysis and testing;

[0049] The defect detection and judgment module is used to design detection specification requirements, post-processing rules and judgment logic for each defect, and adjust the detection standard for each defect.

[0050] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method described above.

[0051] The embodiments of the present application include at least the following beneficial effects: The present application provides a method and device for measuring and detecting a battery pad after welding based on deep learning technology. The present invention obtains a battery pad image after welding; obtains a minimum circumscribed rectangle of a weld mark according to the battery pad image after welding; obtains a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes the offset of the pad relative to the battery and the offset of each weld mark relative to the pad; obtains weld mark defect segmentation data according to the minimum circumscribed rectangle of the weld mark; obtains defect analysis data according to the weld mark defect segmentation data; obtains a battery pad post-weld measurement and detection result according to the measurement value and defect analysis data. The present invention can effectively solve the problem of severe background interference in the weld mark image, accurately locate the weld mark, and then accurately measure it, and accurately detect defects on the weld mark. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 It is a flow chart of a battery pad post-welding measurement and detection method based on deep learning technology provided by an embodiment of the present invention;

[0053] Figure 2 is a partial image of a complex weld pattern provided by an embodiment of the present invention;

[0054] Figure 3 is a schematic diagram of an imaging hardware parameter adjustment process provided by an embodiment of the present invention;

[0055] Figure 4 It is a schematic diagram of the execution flow of the algorithm processing part provided by an embodiment of the present invention;

[0056] Figure 5 is a judgment flow chart of the measurement judgment module provided in an embodiment of the present invention;

[0057] Figure 6 It is a judgment flow chart of the defect detection judgment module provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0058] In order to make the purpose, technical solutions and advantages of the present application clearer, the present application is further described in detail below in conjunction with the accompanying drawings and examples. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the attached claims.

[0059] It is understood that the terms "first", "second", etc. used in this application can be used to describe various concepts in this article, but unless otherwise specified, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another concept. For example, without departing from the scope of the embodiment of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the words "if" and "if" as used herein can be interpreted as "at the time of" or "when" or "in response to determination".

[0060] The terms "at least one", "multiple", "each", "any", etc. used in this application, at least one includes one, two or more, multiple includes two or more, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0061] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0062] The present invention relates to the technical field of battery measurement and detection, and in particular to a method and device for measuring and detecting battery pads after welding based on deep learning technology.

[0063] The present invention takes into account that the weld marks after pad welding present different forms such as continuous lines, lines composed of dense dot matrix, etc., and it is impossible or difficult to accurately achieve the measurement and detection of weld marks by using traditional visual algorithms or simply applying semantic segmentation algorithms.

[0064] The method and device of the present invention combine multiple deep learning algorithms and classical traditional algorithms to accurately measure and detect weld marks. By measuring and detecting the weld marks of each battery, unqualified products can be eliminated in time to ensure that the appearance quality of the weld marks after welding meets the requirements.

[0065] The method and device involved in the present invention adopt a variety of deep learning algorithms and combine them with classic traditional algorithms. They can effectively solve the problem of serious background interference in the weld mark image, accurately locate the weld mark, and then accurately measure and detect defects on the weld mark. Through learning, training and parameter adjustment, they can well adapt to continuous weld marks, spot weld marks and slight weld marks.

[0066] The embodiments of the present invention are further described below in conjunction with the accompanying drawings.

[0067] On the one hand, an embodiment of the present invention provides a battery pad post-welding measurement and detection method based on deep learning technology, referring to Figure 1 , the method comprises the following steps:

[0068] S100, obtaining an image of the battery pad after welding;

[0069] S200, obtaining a minimum circumscribed rectangle of a welding pattern according to an image of a battery pad after welding;

[0070] S300, obtaining a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes an offset of the welding pad relative to the battery and an offset of each welding mark relative to the welding pad;

[0071] S400, obtaining weld defect segmentation data according to the minimum circumscribed rectangle of the weld;

[0072] S500, obtaining defect analysis data according to the weld defect segmentation data;

[0073] S600: Obtain the post-welding measurement and inspection result of the battery pad according to the measurement value and the defect analysis data.

[0074] As an optional implementation, the step of obtaining the battery pad post-welding image in S100 of the present invention includes:

[0075] 1. Composition of the image acquisition system: The system consists of an industrial array camera, lens, and light source. The camera is responsible for capturing images, the lens is responsible for image quality, and the light source ensures image brightness and clarity in different environments.

[0076] 2. Environmental control: During the imaging process, ensure that the working environment is stable, avoid light changes and vibration interference, and ensure the consistency of the collected images.

[0077] 3. Parameter setting: Adjust camera parameters such as shutter speed, aperture size and exposure time to optimize imaging effects and ensure that weld details can be captured clearly.

[0078] 4. Image acquisition: Select multiple pad samples and take multiple photos to obtain image data containing different weld states (such as qualified and unqualified) to form a data set for subsequent analysis and training.

[0079] The step S200 of the present invention of obtaining the minimum circumscribed rectangle of the weld mark according to the image of the battery pad after welding comprises:

[0080] 1. Preprocess the image: De-noise the acquired image and use a filtering algorithm (such as Gaussian filtering) to reduce background noise and enhance the visibility of the weld mark.

[0081] 2. Weld mark detection: Utilize the free rectangle target detection algorithm and analyze the image through the deep learning model to identify the edge of the weld mark and locate its position.

[0082] 3. Calculation of bounding box: Calculate the minimum bounding box on the detected weld area, and obtain its center point coordinates, length and width data to provide a basis for subsequent measurements.

[0083] The present invention S300 obtains the measurement value according to the minimum circumscribed rectangle of the weld mark, including:

[0084] 1. Pad position and angle extraction: Use traditional image processing algorithms (such as Hough transform) to extract the center coordinates and radius of the battery and determine the relative position and angle of the pad.

[0085] 2. Calculation of weld center and length and width: Based on the minimum circumscribed rectangle of the weld, calculate the coordinates of the weld center point and its length and width to form a corresponding data structure.

[0086] 3. Offset calculation: Based on the relative position of the pad and the battery, calculate the offset of the pad relative to the battery, and the offset of each weld line relative to the pad. Store these values ​​as measurement results for subsequent qualified judgment.

[0087] In the embodiment S400 of the present invention, the weld mark defect segmentation data is obtained according to the minimum circumscribed rectangle of the weld mark, including:

[0088] 1. Application of instance segmentation model: Use the AI ​​algorithm of instance segmentation to extract the weld mark from the complex background and obtain the local image of the weld mark.

[0089] 2. Defect extraction: Use a deep learning model that has been trained with a large amount of annotation to identify and extract weld defects, obtain defect categories (such as pores, cracks, and missing welds), circumscribed rectangles, defect masks, and their confidence scores.

[0090] 3. Result storage: Store defect segmentation data and related information in the database for subsequent analysis and processing.

[0091] In the embodiment S500 of the present invention, defect analysis data is obtained based on the weld defect segmentation data, including:

[0092] 1. Feature extraction: By analyzing the weld defect mask and image, the defect features such as defect area, shape, edge features, etc. are extracted.

[0093] 2. Defect classification: Classify the defect data according to the defect characteristics and analyze the severity and impact of each defect.

[0094] 3. Data integration: Combine defect analysis data with segmentation results to form a complete defect information record to facilitate subsequent evaluation and decision-making.

[0095] In the embodiment S600 of the present invention, a post-welding measurement and detection result of a battery pad is obtained according to the measurement value and the defect analysis data, including:

[0096] 1. Comprehensive judgment criteria: Based on the size specification requirements of the battery, pad and weld pattern, combined with the measured values ​​and defect analysis data, the qualified status of the pad is comprehensively evaluated.

[0097] 2. Judgment logic execution: Based on the measurement value and defect situation, the qualified or unqualified judgment is executed according to the preset judgment logic. If the measurement is qualified and the defect detection is qualified, it is judged as a qualified product; otherwise, it is judged as a non-conforming product.

[0098] 3. Result output: The test results are recorded in the database and corresponding reports are generated for subsequent quality control and traceability.

[0099] The embodiment of the present invention discloses step S100 of acquiring a battery pad image after welding, including the following steps:

[0100] S110, obtaining a plurality of soldering pad products;

[0101] S120, adjusting imaging parameters; the imaging parameters include light source adjustment data, lens adjustment data, focus and exposure adjustment data;

[0102] S130, obtaining a battery pad image after welding according to the pad product after welding and imaging parameters.

[0103] The embodiment of the present invention discloses step S200 of obtaining the minimum circumscribed rectangle of the weld mark according to the image of the battery pad after welding, including the following steps:

[0104] S210, using a free rectangle target detection AI algorithm, locate the weld lines from the battery pad after welding image, and obtain the minimum circumscribed rectangle of each weld line;

[0105] S220. The minimum circumscribed rectangle of the weld pattern includes the coordinates of the center point of the rectangle and its length and width.

[0106] The embodiment of the present invention discloses step S300 of obtaining a measurement value according to a minimum circumscribed rectangle of a weld mark, comprising the following steps:

[0107] S310, extracting the center coordinates and radius of the battery using a traditional algorithm according to the minimum circumscribed rectangle of the weld mark;

[0108] S320, extracting the position coordinates and angle of the pad according to the minimum circumscribed rectangle of the weld pattern;

[0109] S330, calculating the offset of the pad relative to the battery and the offset of each weld line relative to the pad according to the center coordinates and radius of the battery, the position coordinates and angle of the pad, and the center and length and width of the weld line;

[0110] S340: Take the offset of the pad relative to the battery and the offset of each solder line relative to the pad as measurement values.

[0111] The embodiment of the present invention discloses step S400 to obtain weld mark defect segmentation data according to the minimum circumscribed rectangle of the weld mark, including the following steps:

[0112] S410, extracting a local part of the weld pattern based on the minimum circumscribed rectangle of the weld pattern;

[0113] S420, extracting weld defects using a segmentation module according to a local weld mark;

[0114] S430, according to the weld defect, obtaining a defect category, a circumscribed rectangle, a defect mask and a confidence score;

[0115] S440, using defect category, bounding rectangle, defect mask and confidence score as weld defect segmentation data.

[0116] The embodiment of the present invention discloses step S500 of obtaining defect analysis data according to weld defect segmentation data, including the following steps:

[0117] S510, extracting defect features according to the weld defect segmentation data and in combination with the defect mask to obtain defect feature data;

[0118] S520: Analyze and test the defect characteristic data to obtain defect analysis data.

[0119] The embodiment of the present invention discloses that step S600 obtains the post-welding measurement and detection result of the battery pad according to the measurement value and the defect analysis data, including the following steps:

[0120] According to the defect analysis data, multiple defect characteristic values ​​are obtained;

[0121] According to the defect characteristic value and measurement value, combined with the defect category and defect detection requirements, the post-welding measurement and detection results of the battery pad are obtained.

[0122] On the other hand, an embodiment of the present invention further provides a battery pad post-welding measurement and detection device based on deep learning technology, the device comprising:

[0123] The first module is used to obtain the image of the battery pad after welding;

[0124] The second module is used to obtain the minimum circumscribed rectangle of the welding pattern according to the image of the battery pad after welding;

[0125] The third module is used to obtain the measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes the offset of the pad relative to the battery and the offset of each weld mark relative to the pad;

[0126] The fourth module is used to obtain the weld defect segmentation data according to the minimum circumscribed rectangle of the weld;

[0127] The fifth module is used to segment the data according to the weld defect to obtain defect analysis data;

[0128] The sixth module is used to obtain the post-welding measurement and inspection results of the battery pad based on the measurement values ​​and defect analysis data.

[0129] The battery pad post-welding measurement and detection method and device of the present invention ensures effective evaluation of pad welding quality through the coordinated work of the imaging part and the algorithm processing part. The following are the execution steps of the method:

[0130] 1. Adjustment of the imaging part

[0131] 1.1 Sample selection

[0132] Select N products after soldering of pads, including typical qualified (OK) products and unqualified (NG) products, to form a sample set for inspection and measurement.

[0133] 1.2 Adjusting imaging parameters

[0134] Light source adjustment: Adjust the light source controller to make the light source brightness as high as possible to obtain a clearer image.

[0135] Lens adjustment: Adjust the lens aperture to the minimum to increase the depth of field and ensure image clarity.

[0136] Focus and exposure adjustment: Adjust the focus ring and exposure settings to ensure that the image is clear and the brightness is moderate, and make preliminary exposure adjustments.

[0137] 1.3 Refine exposure

[0138] For the selected N products, the exposure settings are further refined to ensure that the image of each product is clear, the brightness is moderate, the grayscale background is consistent, and the edges of the pads and welds are clearly visible.

[0139] Record the camera, lens, and light source parameters at this time as hardware parameters for subsequent device replication.

[0140] 2. Algorithm processing part

[0141] 2.1 Free Rectangular Target Detection and Positioning Module

[0142] Use the free rectangle target detection AI algorithm to accurately locate the weld mark from the complex background and obtain the minimum circumscribed rectangle of each weld mark, including the coordinates of the center point of the rectangle and its length and width.

[0143] 2.2 Measurement module

[0144] Battery measurement: Use traditional algorithms to extract the center coordinates and radius of the battery.

[0145] Pad measurement: Extract the position coordinates and angle of the pad.

[0146] Weld mark measurement: Calculate the measurement value of the pad relative to the battery and the measurement value of the weld mark relative to the pad based on the coordinates and radius of the battery, the position and angle of the pad, and the center and length and width of the weld mark.

[0147] 2.3 Measurement and judgment module

[0148] Based on the size specifications and relative position requirements of the battery, pads and weld marks, comprehensively judge whether the measurement results meet the standards.

[0149] 2.4 Instance Segmentation Module

[0150] The local weld pattern is extracted based on the minimum circumscribed rectangle of the weld pattern to reduce the interference of complex background.

[0151] Use the trained instance segmentation AI model to extract weld defects and obtain defect category, bounding rectangle, defect mask and confidence score.

[0152] 2.5 Feature Extraction and Analysis Module

[0153] Based on the information obtained from instance segmentation and defect masks, more defect features are extracted from the image.

[0154] The type of characteristics required for each defect are selected through analysis and testing.

[0155] 2.6 Defect Detection and Judgment Module

[0156] Design inspection specification requirements, post-processing rules and judgment logic for each defect to facilitate fine-tuning of the inspection criteria for each defect.

[0157] Comprehensively judge whether the test results meet the standards and form a judgment process.

[0158] 3. Final judgment

[0159] Only battery pads that have passed measurement and defect detection will be considered qualified products, otherwise they will be judged as unqualified products.

[0160] The battery pad post-welding measurement and detection device based on deep learning technology disclosed in an embodiment of the present invention also includes:

[0161] Free rectangle target detection and positioning module, used to locate the weld mark from the background;

[0162] The measurement module is used to extract the center coordinates and radius of the battery through an algorithm; extract the position coordinates and angle of the pad through an algorithm; calculate the measurement value of the pad relative to the battery and the measurement value of each weld line relative to the pad based on the center coordinates and radius of the battery, the position coordinates and angle of the pad, and the center and length and width of the weld line;

[0163] The measurement judgment module is used to comprehensively judge whether the measured part meets the standards according to the respective size specification requirements of the battery, pad and weld mark and the specification requirements of the relative position;

[0164] The instance segmentation module is used to extract the weld mark locally according to the weld mark's minimum circumscribed rectangle;

[0165] The feature extraction and analysis module is used to extract defect features from the image based on the information obtained by the instance segmentation module and the defect mask, and select the required feature type for each defect through analysis and testing;

[0166] The defect detection and judgment module is used to design detection specification requirements, post-processing rules and judgment logic for each defect, and adjust the detection standard for each defect.

[0167] In order to ensure the welding quality of the pad, the present invention needs to measure the size of the weld mark and detect the appearance defects of the weld mark to meet the appearance inspection standard of the welding. However, the existing lithium battery welding appearance inspection mainly focuses on detection, and less on accurate measurement, especially accurate measurement under complex imaging background. In addition, for point weld marks and slight weld marks, and in the case of serious interference and pollution near the weld marks, it is difficult to accurately detect with the existing methods. Therefore, the device and method involved in the present invention are proposed, which can well solve the above problems.

[0168] A local image of a complex weld pattern Figure 2 shown.

[0169] The battery pad post-welding measurement and detection method based on deep learning technology in an embodiment of the present invention mainly includes an imaging part and an algorithm processing part.

[0170] The imaging part includes an image acquisition system formed by an industrial array camera, lens, light source, etc. A clear and stable image is the basis for accurate measurement and detection. In order to more accurately adjust the imaging parameters and achieve better image effects, the adjustment process is as follows: Figure 3 .

[0171] First, select N products after soldering of pads, including typical OK products and typical NG products, which can represent the product set for inspection and measurement as samples for imaging refinement adjustment.

[0172] Then, adjust the light controller to make the light as bright as possible. Adjust the lens aperture to make the aperture as small as possible to increase the depth of field. Adjust the focus ring and exposure to make the image clear and the brightness and darkness appropriate. This is the initial adjustment of exposure.

[0173] Then, the exposure is adjusted in detail so that the images of N products are clear, the brightness of the images is appropriate, the gray background is relatively consistent, and the edges of the pads and welds are relatively clearer. The camera, lens and light source parameters at this time are recorded as the hardware parameters of the replication machine. Because the imaging of N products has been adjusted in detail many times, the hardware parameters of the imaging are more accurate, and the image effect is relatively better.

[0174] The algorithm processing part mainly includes free rectangular target detection and positioning module, measurement module, measurement judgment module, instance segmentation module, feature extraction and analysis module, and defect detection and judgment module. Among them, the free rectangular target detection and positioning module and the instance segmentation module are different types of AI algorithm modules. Products judged as NG are eliminated, and products judged as OK flow into the next process.

[0175] refer to Figure 4 For the free rectangle target detection and positioning module, the free rectangle target detection AI algorithm can accurately locate the weld mark from the complex background and obtain the minimum enclosing rectangle of each weld mark, including the coordinates of the center point of the rectangle and the length and width of the rectangle. Compared with the axis-parallel target detection AI algorithm, the free rectangle target detection AI algorithm can locate the weld mark in a smaller range, and the minimum enclosing rectangle of the weld mark can be accurately obtained through a large number of accurately labeled trained models.

[0176] For the measurement module, first, the center coordinates and radius of the battery are extracted through the traditional algorithm; then, the position coordinates and angle of the pad are extracted through the traditional algorithm; finally, based on the center coordinates and radius values ​​of the battery, the position coordinates and angle of the pad, and the center and length and width of the weld mark, the measurement value of the pad relative to the battery and the measurement value of each weld mark relative to the pad are calculated.

[0177] For the measurement and judgment module, according to the size and relative position requirements of the battery, pad and solder mark, it is comprehensively judged whether the measurement part meets the standards. The specific execution process is as follows: Figure 5 shown.

[0178] For the instance segmentation module, first, the weld mark is locally extracted based on the minimum enclosing rectangle of the weld mark obtained by the positioning module, which can greatly reduce the interference of complex background; then, based on the AI ​​model trained with a large number of accurately labeled weld mark defects, the instance segmentation AI algorithm is used to extract the weld mark defects to obtain the defect category, enclosing rectangle, defect mask and confidence score.

[0179] For the feature extraction and analysis module, first, based on the information obtained from instance segmentation, more defect features are extracted in combination with the defect mask and image; then, based on the characteristics of each defect, the feature type to be used for each defect is selected through analysis and testing.

[0180] For the defect detection and judgment module, according to the detection specification requirements of each defect, the post-processing rules and judgment logic of each defect are designed to facilitate fine-tuning of the overkill of each defect. Through the detection and judgment of each defect, it is comprehensively judged whether the detected part meets the standard. The specific post-processing rules and judgment logic of different defects are different. The judgment process is as follows: Figure 6 shown.

[0181] Finally, only battery pads that are judged as qualified by measurement and defect detection are qualified products, otherwise they are unqualified products.

[0182] The method and device involved in the present invention not only have good application effects in the measurement and detection of battery pads after welding, but also have good applicability and reference in other welding measurement and detection aspects.

[0183] The existing methods are difficult to accurately locate weld marks in complex backgrounds. If the semantic segmentation algorithm is simply relied on, there will be a large number of incorrect segmentations when the weld mark background interference is serious. Using axis-parallel target detection for weld mark positioning will result in a lot of background interference information. The current measurement algorithm for weld marks on pads is a traditional algorithm, which has poor adaptability to complex backgrounds and low contrast situations and cannot meet the needs of industrial field detection.

[0184] In the case of complex welding pattern imaging background and various welding pattern shapes, based on the above difficulties and shortcomings, the innovative points and advantages of the method and device disclosed in the present invention mainly include:

[0185] 1. The free rectangle target detection AI algorithm is used for positioning. Compared with the axis-parallel target detection AI algorithm, it can locate the weld mark within a smaller range. The model trained with a large number of accurate annotations can accurately obtain the minimum circumscribed rectangle of the weld mark, which can greatly reduce the interference of complex background textures.

[0186] 2. The measurement of weld marks is combined with the accurate positioning information of the free rectangle target detection AI algorithm, which can better adapt to complex texture backgrounds. For the measurement information of pads and batteries, the traditional algorithm is used because of the good imaging edges. Finally, the combination of AI algorithm and traditional algorithm makes the measurement of weld marks more stable and accurate. The on-site application meets the requirements.

[0187] 3. The detection of weld marks is combined with the accurate positioning information of the free rectangle target detection AI algorithm, which can greatly reduce the influence of complex interference background. Then, combined with the instance segmentation AI algorithm, a number of feature extraction and analysis are performed to ultimately achieve a more accurate detection purpose. Compared with the direct use of the semantic segmentation AI algorithm, the method of the present invention has obvious advantages.

[0188] On the other hand, an embodiment of the present invention further provides a computer-readable storage medium, which stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the above method.

[0189] On the other hand, an embodiment of the present invention also provides a battery pad post-weld measurement and detection system based on deep learning technology, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the above method is implemented.

[0190] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc.

[0191] The memory may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage. Among them, ROM can store static data or instructions required by the processor or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at run time. In addition, the memory may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (e.g., DVD-ROM, double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (e.g., SD card, mini SD card, Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not include carrier waves and transient electronic signals transmitted wirelessly or wired.

[0192] The memory stores executable codes, and when the executable codes are processed by the processor, the processor can execute part or all of the methods described above.

[0193] In the above description of this specification, the description with reference to the terms "one embodiment / example", "another embodiment / example" or "certain embodiments / examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0194] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the claims and their equivalents.

[0195] The above is a specific description of the preferred implementation of the present invention, but the present invention is not limited to the embodiments. Those skilled in the art may make various equivalent modifications or substitutions without violating the spirit of the present invention. These equivalent modifications or substitutions are all included in the scope defined by the claims of this application.

Claims

1. A battery pad post-welding measurement and detection method based on deep learning technology, characterized in that: The method comprises the following steps: Get the image of the battery pad after welding; According to the battery pad after welding image, obtaining the minimum circumscribed rectangle of the welding mark; Obtaining a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes an offset of the welding pad relative to the battery and an offset of each welding mark relative to the welding pad; Obtaining weld mark defect segmentation data according to the weld mark minimum circumscribed rectangle; Obtaining defect analysis data according to the weld defect segmentation data; According to the measurement values ​​and defect analysis data, the post-welding measurement and detection results of the battery pad are obtained.

2. The method according to claim 1, characterized in that The method of obtaining the battery pad after welding comprises the following steps: Obtain several pad soldering products; Adjusting imaging parameters; the imaging parameters include light source adjustment data, lens adjustment data, focus and exposure adjustment data; According to the post-welding product of the welding pad and the imaging parameters, a post-welding image of the battery welding pad is obtained.

3. The method according to claim 1, characterized in that The method of obtaining the minimum circumscribed rectangle of the weld mark according to the image of the battery pad after welding comprises the following steps: Using a free rectangle target detection AI algorithm, the weld lines are located from the image of the battery pad after welding, and the minimum circumscribed rectangle of each weld line is obtained; The minimum circumscribed rectangle of the weld pattern includes the coordinates of the center point of the rectangle and its length and width.

4. The method according to claim 1, characterized in that: The method of obtaining the measured value according to the minimum circumscribed rectangle of the weld mark comprises the following steps: According to the minimum circumscribed rectangle of the weld mark, the center coordinates and radius of the battery are extracted using a traditional algorithm; Extracting the position coordinates and angle of the welding pad according to the minimum circumscribed rectangle of the welding pattern; According to the center coordinates and radius of the battery, the position coordinates and angle of the welding pad, and the center and length and width of the welding mark, the offset of the welding pad relative to the battery and the offset of each welding mark relative to the welding pad are calculated; The offset of the pad relative to the battery and the offset of each weld line relative to the pad are taken as measured values.

5. The method according to claim 1, characterized in that The method of obtaining weld mark defect segmentation data according to the weld mark minimum circumscribed rectangle comprises the following steps: Extracting a local part of the weld pattern based on the minimum circumscribed rectangle of the weld pattern; According to the local weld line, a segmentation module is used to extract weld line defects; According to the weld defect, a defect category, a bounding rectangle, a defect mask and a confidence score are obtained; The defect category, the circumscribed rectangle, the defect mask and the confidence score are used as weld defect segmentation data.

6. The method according to claim 1, characterized in that The method of obtaining defect analysis data according to the weld defect segmentation data comprises the following steps: Extracting defect features based on the weld defect segmentation data and in combination with the defect mask to obtain defect feature data; The defect characteristic data is analyzed and tested to obtain defect analysis data.

7. The method according to claim 1, characterized in that The method of obtaining the post-welding measurement and detection result of the battery pad according to the measured value and the defect analysis data comprises the following steps: According to the defect analysis data, a plurality of defect characteristic values ​​are obtained; According to the defect characteristic value and the measurement value, combined with the defect category and defect detection requirements, the post-welding measurement and detection result of the battery pad is obtained.

8. A battery pad post-welding measurement and detection device based on deep learning technology, characterized in that: The device comprises: The first module is used to obtain the image of the battery pad after welding; The second module is used to obtain the minimum circumscribed rectangle of the welding mark according to the image of the battery pad after welding; The third module is used to obtain a measurement value according to the minimum circumscribed rectangle of the weld mark; the measurement value includes an offset of the pad relative to the battery and an offset of each weld mark relative to the pad; The fourth module is used to obtain weld mark defect segmentation data according to the minimum circumscribed rectangle of the weld mark; A fifth module is used to obtain defect analysis data according to the weld defect segmentation data; The sixth module is used to obtain the post-welding measurement and detection results of the battery pad according to the measurement values ​​and defect analysis data.

9. The device according to claim 8, characterized in that The device also includes: Free rectangle target detection and positioning module, used to locate the weld mark from the background; A measurement module, used to extract the center coordinates and radius of the battery through an algorithm; extract the position coordinates and angle of the pad through an algorithm; calculate the measurement value of the pad relative to the battery and the measurement value of each weld line relative to the pad according to the center coordinates and radius of the battery, the position coordinates and angle of the pad, and the center and length and width of the weld line; The measurement judgment module is used to comprehensively judge whether the measured part meets the standards according to the respective size specification requirements of the battery, pad and weld mark and the specification requirements of the relative position; An instance segmentation module, used for extracting a part of the weld mark according to the minimum circumscribed rectangle of the weld mark; A feature extraction and analysis module, used to extract defect features from the image based on the information obtained by the instance segmentation module and the defect mask, and select the required feature type for each defect through analysis and testing; The defect detection and judgment module is used to design detection specification requirements, post-processing rules and judgment logic for each defect, and adjust the detection standard for each defect.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to enable a computer to execute the method according to any one of claims 1 to 7.

Images