Weld seam detection method and apparatus, electronic device, and storage medium

Abstract

A weld seam detection method and apparatus, an electronic device, and a storage medium, relating to the technical field of battery production. The method comprises: on the basis of a first detection image acquired by an image acquisition device for a weld seam to be detected, determining defects and defect parameters of said weld seam; obtaining scores of the defects on the basis of the defect parameters; and determining a detection result of said weld seam on the basis of the scores of the defects.

Description

Weld detection method, device, electronic device and storage medium

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on January 9, 2024, with application number 202410028441.6 and invention name “Defect Detection Method, Device, Electronic Device and Storage Medium”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The present application relates to the field of battery production technology, and in particular to a weld detection method, device, electronic device, and storage medium. Background Art

[0003] Welding is a commonly used processing method in battery production. Due to different environmental conditions and welding techniques during the welding process, various defects will occur during the welding process, affecting product quality. For example, laser welding is an efficient and precise welding method that uses a high-energy-density laser beam as a heat source. During the welding process of the battery cell top cover and the aluminum shell, due to improper laser power adjustment or the presence of impurities at the joint of the top cover and aluminum shell, defects such as bumps and pinholes are easily caused to appear in the weld. In related technologies, AI models are used to process weld photos to identify defects such as bumps and pinholes that cause unqualified welds. When AI identifies defects that cause unqualified welds, it only considers the size of the defect, but does not consider the impact of other factors such as the location dimension of the defect on the quality of the weld. It may result in unqualified welds being determined as qualified welds, affecting the detection accuracy of the welds. Technical issues

[0004] In view of the above problems, the present application provides a weld detection method, device, electronic device and storage medium to improve the detection accuracy of weld defect detection. Technical Solutions

[0005] The technical solution adopted in this application is:

[0006] In a first aspect, a weld detection method is provided, comprising: determining a defect of the weld to be detected and defect parameters of the defect based on a first detection image captured by an image acquisition device for the weld to be detected; obtaining a score of the defect based on the defect parameters; and determining a detection result of the weld to be detected based on the defect score.

[0007] In this embodiment, the defects and defect parameters of the weld to be inspected are obtained, and a comprehensive score is performed based on the defect parameters to determine the inspection result of the weld to be inspected; the defect parameter factor is taken into account when identifying the defects that cause the weld to be unqualified. The defect score can quantify the impact of the defect on the quality, improve the inspection accuracy of the weld, avoid the situation where an unqualified weld is determined to be a qualified weld, and ensure product quality.

[0008] In some embodiments, the defect parameter includes defect size; obtaining the defect score based on the defect parameter includes setting the defect score to be greater than or equal to a defect score threshold when the defect size exceeds a defect size threshold.

[0009] In this embodiment, a defect whose size exceeds the defect size threshold is a defect that causes the weld to be unqualified. The score of this defect can be set to be greater than or equal to the defect score threshold, and the weld to be inspected can be directly determined as unqualified. There is no need to determine the authority score based on other scoring dimensions, which can improve the efficiency and accuracy of weld detection.

[0010] In some embodiments, the defect parameters include: defect location; obtaining the score of the defect based on the defect parameters includes: determining the scoring dimension information of the defect based on the defect location when the defect size does not exceed the defect size threshold; obtaining the score of the defect based on the defect size and the scoring dimension information.

[0011] In this embodiment, the scoring dimension information of the defect is determined based on the defect location, and the defect score is calculated based on the defect size and the scoring dimension information. When identifying defects that cause unqualified welds, multiple dimensional information can be considered, and the impact of defects on weld quality can be more accurately measured, thereby improving the detection accuracy of welds.

[0012] In some embodiments, the scoring dimension information includes: defect location dimension information; the scoring dimension information of determining the defect based on the defect location includes: determining the distance between the defect and the position detection benchmark based on the position information of the defect location and the position detection benchmark; and generating the defect location dimension information of the defect based on the distance.

[0013] In this embodiment, defect scoring dimension information is generated based on the distance between the defect and the position detection benchmark. When identifying defects that cause unqualified welds, the impact of the defect location on the weld quality can be considered, thereby improving the detection accuracy of the weld.

[0014] In some embodiments, the scoring dimension information includes: defect overlay dimension information; the scoring dimension information for determining the defect based on the defect position includes: when there are other defects in the weld to be inspected, determining the overlay defect of the defect among the other defects based on the defect position, the defect position of the other defects and the overlay detection benchmark; wherein the overlay detection benchmark includes: an overlay detection line or an overlay detection area; the defect and the overlay defect are both located on the same overlay detection line or within the same overlay detection area; based on the defect and the overlay defect, the defect overlay dimension information of the defect is generated.

[0015] In this embodiment, defect superposition dimension information of the defect is generated based on the position of the defect and other defects and the superposition detection benchmark. When identifying defects that cause unqualified welds, the impact of multiple defects on the weld quality on or within the same superposition detection benchmark can be considered, thereby improving the detection accuracy of the weld.

[0016] In some embodiments, obtaining the score of the defect based on the defect size and the scoring dimension information includes: obtaining the size weight of the defect size and the dimension weight of the scoring dimension information; determining a defect scoring processing network for the defect in the defect comprehensive scoring model; inputting the defect size, the size weight, the scoring dimension information and the dimension weight into the defect scoring processing network to obtain the score of the defect output by the defect scoring processing network.

[0017] In this embodiment, by setting weights for defect size and scoring dimension information, the impact of defect size and different scoring dimensions on weld quality can be more accurately reflected, which can improve the effectiveness and accuracy of defect detection; the defect score is obtained through the defect scoring processing network, which can improve the efficiency and accuracy of scoring.

[0018] In some embodiments, determining the defect of the weld to be detected and the defect parameters of the defect based on the first detection image captured by the image acquisition device for the weld to be detected includes: preprocessing the first detection image to obtain the image to be identified; inputting the image to be identified into the defect detection model to obtain the defect, the defect parameters of the defect and the confidence level output by the defect detection model; wherein the confidence level is greater than a preset confidence threshold.

[0019] In this embodiment, defects, defect parameters and confidence levels are obtained through a defect detection model, which can improve the efficiency and accuracy of defect detection. Defects can be re-screened based on the confidence level, which can improve the accuracy of detection.

[0020] In some embodiments, determining the inspection result of the weld to be inspected based on the score of the defects includes: when the score of at least one defect is greater than or equal to the defect score threshold, determining that the inspection result is unqualified for the weld to be inspected; when the scores of all defects are less than the defect score threshold, determining that the inspection result is qualified for the weld to be inspected.

[0021] In this embodiment, the impact of defects on quality can be quantified through defect scoring and scoring thresholds. When the score is greater than or equal to the scoring threshold, the weld to be inspected is determined to be unqualified. The defect causing the weld to be unqualified can be determined, thereby improving the detection accuracy of the weld and avoiding situations such as determining an unqualified weld as a qualified weld.

[0022] In some embodiments, shooting posture test information of the image acquisition device is generated; a second detection image of the weld to be inspected captured by the image acquisition device under a shooting posture corresponding to the shooting posture test information is obtained; and shooting posture optimization information of the image acquisition device is determined based on image loss data of the second detection image and the shooting posture test information; wherein, the image acquisition device performs image acquisition on the weld to be inspected under the shooting posture corresponding to the shooting posture optimization information to obtain the first detection image.

[0023] In this embodiment, based on the shooting posture test information and the image loss data of the detection image collected based on the shooting posture test information, the shooting posture optimization information of the image acquisition device is determined, which is used to adjust the installation angle of the image acquisition device to obtain a qualified weld detection image, which can reduce the number of adjustments, reduce the difficulty of adjusting the installation angle, and improve work efficiency.

[0024] In some embodiments, determining the shooting posture optimization information of the image acquisition device based on the image loss data of the second detection image and the shooting posture test information includes: determining the correlation between the image loss data and the shooting posture test information; using prediction software and performing prediction processing based on the correlation to determine the shooting posture optimization information.

[0025] In this embodiment, prediction software is used and based on the correlation between image loss data and shooting posture test information, the shooting posture optimization information of the image acquisition device is determined, the installation angle of the image acquisition device after adjustment can be automatically determined, and qualified weld inspection images can be obtained. The number of adjustments can be reduced, work efficiency can be improved, the difficulty of adjusting the installation angle can be reduced, and the detection accuracy of defect detection can be guaranteed.

[0026] In some embodiments, generating shooting posture test information of multiple image acquisition devices includes: obtaining a valid range of the shooting posture test information; and generating the shooting posture test information based on the valid range using the prediction software.

[0027] In this embodiment, the shooting posture test information is generated by using prediction software based on the effective range, which can reduce the number of adjustments, improve work efficiency, and reduce the difficulty of adjusting the installation angle.

[0028] In some embodiments, the image loss data includes: the number of missing pixels; the method further includes: inputting the second detection image into a target detection model, obtaining the number of pixels of the weld to be detected output by the target detection model; and determining the number of missing pixels based on the number of pixels of the weld to be detected and the required number of pixels.

[0029] In this embodiment, the number of pixels of the detection target is obtained through the target detection model, which can improve the efficiency and accuracy of the detection and also improve the accuracy of the number of missing pixels.

[0030] In some embodiments, the shooting posture test information includes: a first installation angle of the image acquisition device; and the shooting posture optimization information includes: a second installation angle of the image acquisition device.

[0031] According to a second aspect of the present disclosure, a weld detection device is provided, comprising: a defect determination module for determining the defect of the weld to be detected and defect parameters of the defect based on a first detection image captured by an image acquisition device for the weld to be detected; a defect scoring module for obtaining a score of the defect based on the defect parameters; and a detection module for determining a detection result of the weld to be detected based on the defect score.

[0032] In this embodiment, the defects and defect parameters of the weld to be inspected are obtained, and a comprehensive score is performed based on the defect parameters to determine the inspection result of the weld to be inspected; the defect parameter factor is taken into account when identifying the defects that cause the weld to be unqualified. The defect score can quantify the impact of the defect on the quality, improve the inspection accuracy of the weld, avoid the situation where an unqualified weld is determined to be a qualified weld, and ensure product quality.

[0033] According to a third aspect of the present disclosure, an electronic device is provided, comprising: a memory; and a processor coupled to the memory, wherein the processor is configured to execute the method described above based on instructions stored in the memory.

[0034] According to a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores computer instructions, and the instructions are executed by a processor to perform any of the methods described above.

[0035] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the embodiments of the present application. Obviously, the drawings described on the lower wall are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on the drawings without paying any creative work.

[0037] FIG1 is a schematic flow chart of some embodiments of the defect detection method disclosed herein;

[0038] FIG2 is a schematic diagram of a process for determining defects and defect parameters in some embodiments of the defect detection method disclosed herein;

[0039] FIG3 is a schematic diagram of a process for obtaining a defect score in some embodiments of the defect detection method disclosed herein;

[0040] FIG4 is a schematic diagram of a process for generating defect location dimension information in some embodiments of the defect detection method disclosed herein;

[0041] FIG5A is a schematic diagram of a process for generating defect overlay dimension information in some embodiments of the defect detection method disclosed herein; FIG5B-FIG5D are schematic diagrams of overlay detection benchmarks;

[0042] FIG6A is a schematic diagram of a process for obtaining a defect score using a defect score processing network in some embodiments of the defect detection method of the present disclosure; FIG6B is a schematic diagram of a defect comprehensive scoring model;

[0043] FIG7 is a schematic diagram of a process for optimizing the shooting posture of an image acquisition device in some embodiments of the defect detection method disclosed herein;

[0044] FIG8A is a schematic diagram of determining shooting posture optimization information in some embodiments of the defect detection method disclosed herein; FIG8B is a diagram of angle data prediction generated by JMP software;

[0045] FIG9A is a schematic diagram of modules of some embodiments of the defect detection apparatus of the present disclosure; FIG9B is a schematic diagram of modules of a defect scoring module of some embodiments of the defect detection apparatus of the present disclosure; FIG9C is a schematic diagram of modules of other embodiments of the defect detection apparatus of the present disclosure;

[0046] FIG10 is a schematic diagram of modules of some embodiments of the electronic device disclosed herein. Modes for Carrying Out the Invention

[0047] The following embodiments of the technical solution of the present application will be described in detail with reference to the accompanying drawings. The following embodiments are only used to more clearly illustrate the technical solution of the present application and are therefore only examples and are not intended to limit the scope of protection of the present application.

[0048] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application belongs; the terms used herein are only for the purpose of describing specific embodiments and are not intended to limit this application; the terms "including" and "having" and any variations thereof in the specification and claims of this application and the above-mentioned figure descriptions are intended to cover non-exclusive inclusions.

[0049] In the description of the embodiments of this application, the technical terms "first" and "second" are used only to distinguish different objects and should not be understood to indicate or imply relative importance or implicitly specify the quantity, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, the meaning of "plurality" is more than two, unless otherwise clearly and specifically defined.

[0050] References herein to "embodiments" mean that a particular feature, structure, or characteristic described in connection with the embodiments may be included in at least some embodiments of the present application. The appearance of this phrase in various places in the specification does not necessarily refer to the same embodiment, nor does it constitute an independent or alternative embodiment that is mutually exclusive of other embodiments. It is understood, both explicitly and implicitly, by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0051] In the description of the embodiments of this application, the term "and / or" is simply a description of the association relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent the following three situations: A exists alone, A and B exist simultaneously, and B exists alone. In addition, the character " / " in this document generally indicates that the associated objects are in an "or" relationship.

[0052] In the description of the embodiments of the present application, the term "multiple" refers to more than two (including two). Similarly, "multiple groups" refers to more than two groups (including two groups), and "multiple pieces" refers to more than two pieces (including two pieces).

[0053] In the description of the embodiments of the present application, the technical terms "center", "longitudinal", "lateral", "length", "width", "thickness", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", "clockwise", "counterclockwise", "axial", "radial", "circumferential", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present application and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the embodiments of the present application.

[0054] In the description of the embodiments of the present application, unless otherwise expressly specified or limited, technical terms such as "installed," "connected," "connected," and "fixed" should be understood in a broad sense. For example, they can refer to fixed connections, detachable connections, or integration; mechanical connections or electrical connections; direct connections or indirect connections through an intermediate medium; internal connections between two components or interactions between two components. Those skilled in the art can understand the specific meanings of the above terms in the embodiments of the present application based on specific circumstances.

[0055] According to the inventors' knowledge, laser welding is used to weld the cell top cover and aluminum shell together during the production of power lithium-ion battery cells. Due to the different material compositions of the cell top cover and aluminum shell, as well as the divergence of the welding laser, defects may occur in the weld between the cell top cover and aluminum shell. Some defects may result in unqualified welds. Therefore, it is necessary to inspect the weld between the cell top cover and aluminum shell to improve product yield and enhance battery cell safety.

[0056] At present, cameras are installed in the production line of power lithium battery cells. When inspecting the weld between the top cover and the aluminum shell of the battery cell, a camera is used to take a picture of the weld to obtain an inspection image containing the weld. The inspection image is input into the AI ​​model, and the AI ​​model is used to identify defects such as bumps and pinholes that cause unqualified welds. Defects in different positions may have different effects on the quality of the weld. When AI identifies defects that cause unqualified welds, it only considers the size of the defect, but does not consider other factors such as the location of the defect. It may be possible to determine unqualified welds as qualified welds, etc., which affects the inspection accuracy of the weld, reduces the product yield, and brings safety hazards to the use of battery cells. In view of this, the present disclosure provides a technical solution for defect detection to solve the above technical problems.

[0057] FIG1 is a flow chart of some embodiments of the defect detection method disclosed herein, as shown in FIG1 :

[0058] In step S101, defects of the weld to be inspected and defect parameters of the defects are determined based on a first inspection image captured by an image acquisition device for the weld to be inspected.

[0059] The weld to be inspected can be of various types, such as the weld between the battery cell top cover and the aluminum shell. The image acquisition device can be of various types, such as a 2D camera, a 3D camera, etc. The weld to be inspected is photographed using a 2D camera, a 3D camera, etc., to obtain a first inspection image containing the weld to be inspected. Various processing methods can be used on the first inspection image to obtain defects and defect parameters of the weld to be inspected in the first inspection image. Defects include various defects such as weld bumps, pinholes, and polarization. Defect parameters include parameters such as defect size and location.

[0060] In step S102, a defect score is obtained according to the defect parameters.

[0061] The size, location, etc. of the defect can be used as scoring dimensions, and a comprehensive score of the defect can be obtained based on the scoring dimensions as the score of the defect.

[0062] In step S103, based on the defect scores, the inspection results of the weld to be inspected are determined.

[0063] The defect detection method in the above embodiment detects the inspection image of the weld to be inspected, obtains defects and defect parameters, and performs a comprehensive score based on the defect parameters to determine the inspection result of the weld to be inspected, thereby improving the inspection accuracy of the weld, avoiding the situation where unqualified welds are determined to be qualified welds, and improving the yield rate of the product.

[0064] In some embodiments, a defect score threshold can be set. When the score of at least one defect is greater than or equal to the defect score threshold, the inspection result is determined to be unqualified for the weld to be inspected; when the scores of all defects are less than the defect score threshold, the inspection result is determined to be qualified for the weld to be inspected.

[0065] For example, the defect score threshold is set to 100. When there are multiple defects in the weld to be inspected, the score of each defect is obtained separately. When the score of any defect is greater than 100, the inspection result of the weld to be inspected is determined to be unqualified; if the scores of all defects are less than 100, the inspection result of the weld to be inspected is determined to be qualified.

[0066] The impact of defects on quality can be quantified through defect scoring and scoring thresholds. When the score is greater than or equal to the scoring threshold, the weld to be inspected is determined to be unqualified. The defects that cause the weld to be unqualified can be determined, thereby improving the inspection accuracy of the weld and avoiding situations such as determining an unqualified weld as a qualified weld.

[0067] FIG2 is a schematic diagram of a process for determining defects and defect parameters in some embodiments of the defect detection method disclosed herein, as shown in FIG2 :

[0068] In step S201, the first detection image is preprocessed to obtain an image to be recognized.

[0069] In some embodiments, a variety of preprocessing methods can be used for the first detection image. For example, the first detection image can be preprocessed by size normalization, denoising, and enhancement to obtain an image to be identified, thereby improving the accuracy of image-based defect detection.

[0070] In step S202, the image to be identified is input into the defect detection model to obtain the defect, defect parameters and confidence of the defect output by the defect detection model; wherein the confidence of the defect is greater than a preset confidence threshold.

[0071] Construct a defect detection model. The defect detection model can be a variety of neural network models, such as a convolutional neural network model. The defect detection model can include multiple convolutional layers, fully connected layers, etc. Obtain image samples containing welds, annotate the image samples with defects, and use the annotated image samples to train the defect detection model.

[0072] During inspection, the image to be identified is fed into the defect detection model, which identifies the defects and their parameters within the image. It can also calculate the confidence level of each defect using functions such as activation functions. If the confidence level of a defect is greater than a confidence threshold, the defect detection model outputs the defect, its parameters, and its confidence level. If the confidence level of a defect is less than or equal to the confidence threshold, the defect detection model does not output the defect, its parameters, or its confidence level.

[0073] For example, the defects output by the defect detection model include defects such as bumps and pinholes, the defect parameters include the size and position of defects such as bumps and pinholes, and the confidence of the defects includes the confidence of defects such as bumps and pinholes.

[0074] After the defect detection model outputs the defect, its parameters, and confidence level, you can set a confidence level tolerance to further screen the defects. For example, if the confidence level of a defect is less than the tolerance, the defect is filtered and removed. This rescreening improves detection accuracy and stability.

[0075] In some embodiments, different defects may have different defect sizes. For example, for a pinhole defect, the defect size may be one or more of the pinhole diameter, area, and depth; for a pit defect, the defect size may be one or more of the pit diameter, area, and depth; and for a bump defect, the defect size may be one or more of the bump area and height.

[0076] Defect size can be used as a scoring dimension. If the defect size exceeds the defect size threshold, the defect score is set to be greater than or equal to the defect score threshold. For example, a defect detection model can be used to determine that the defects in a weld to be inspected include pinholes, bumps, and other defects. Pinhole defect parameters include a diameter of 5 mm, and the pinhole defect size threshold includes a diameter of 4 mm. The defect score threshold can be 100.

[0077] If it is determined that the pinhole diameter of 5 mm is greater than the pinhole diameter threshold of 4 mm, the pinhole score is set to be greater than or equal to the defect score threshold of 100. The defect parameters of the bump include a bump height of 3 mm, and the bump defect size threshold includes a bump height threshold of 2 mm. If, for example, the bump height of 3 mm is greater than the bump height threshold of 2 mm, the bump score is set to be greater than or equal to the defect score threshold of 100.

[0078] Defects whose size exceeds the defect size threshold are considered defects that cause weld failure. The score of this defect can be set to be greater than or equal to the defect score threshold. The weld to be inspected can be directly determined as unqualified without the need to determine the authority score based on other scoring dimensions. This can improve the efficiency and accuracy of weld inspection.

[0079] FIG3 is a schematic diagram of a process for obtaining a defect score in some embodiments of the defect detection method disclosed herein, as shown in FIG3 :

[0080] In step S301, when the defect size does not exceed the defect size threshold, the scoring dimension information of the defect is determined according to the defect position.

[0081] Defect parameters include parameters such as defect location. The defect location can be the location of a defect such as a pinhole or convex point in the first detection image as output by the defect detection model. For example, the defect detection model can determine the coordinates of the pinhole or convex point in the coordinate system corresponding to the first detection image as the defect location of the pinhole or convex point.

[0082] When the defect size of defects such as pinholes and convex spots does not exceed the corresponding defect size threshold, the scoring dimension information of defects such as pinholes and convex spots is determined according to the defect positions of the defects such as pinholes and convex spots; the scoring dimension information can be the defect position dimension information, defect superposition dimension information, etc. of defects such as pinholes and convex spots.

[0083] In step S302, the defect score is obtained based on the defect size and score dimension information.

[0084] The scoring dimension information of the defect is determined based on the defect location, and the defect score is calculated based on the defect size and scoring dimension information. Multiple dimensions can be considered when identifying defects that cause unqualified welds, which can more accurately measure the impact of defects on weld quality and improve the detection accuracy of welds.

[0085] FIG4 is a schematic diagram of a process for generating defect location dimension information in some embodiments of the defect detection method disclosed herein, as shown in FIG4 :

[0086] In step S401 , the distance between the defect and the position detection reference is determined based on the position information of the defect position and the position detection reference.

[0087] In step S402, defect position dimension information of the defect is generated based on the distance. The defect position dimension information includes information such as the distance between the defect and the position detection reference.

[0088] In some embodiments, a position detection reference is set according to the detection requirements of the weld quality. The position detection reference can be a location and position that has a greater impact on the quality of the weld to be detected. For example, the position detection reference can be the fusion line between the weld between the battery cell top cover and the aluminum shell and the battery cell top cover, the fusion line between the weld between the battery cell top cover and the aluminum shell and the aluminum shell, the fusion zone between the weld between the battery cell top cover and the aluminum shell and the battery cell top cover, the fusion zone between the weld between the battery cell top cover and the aluminum shell and the aluminum shell, etc.; the fusion line refers to the boundary line between the weld metal and the battery cell top cover or the aluminum shell body, and the fusion zone refers to the boundary area between the weld metal and the battery cell top cover or the aluminum shell body.

[0089] The position information of the position detection reference can be obtained using a variety of methods. For example, when a defect detection model processes an image, it detects the position detection reference and its position information in the image. When the defect detection model outputs a defect, its defect parameters, and its confidence level, it also outputs the position detection reference and its position information.

[0090] In some embodiments, the position detection reference is set to the weld between the battery cell top cover and the aluminum shell and the fusion line between the battery cell top cover, and the defect detection model outputs this fusion line and the position information of the fusion line. The position information of the fusion line is the coordinate information of the fusion line in the first detection image.

[0091] The closer the defect is to the fusion line, the greater its impact on weld quality. For example, taking a pinhole defect as an example, the distance between the pinhole and the fusion line is determined based on the coordinates of the pinhole in the first inspection image and the fusion line in the first inspection image. This distance can be the shortest distance between the pinhole and the fusion line. Based on this distance, defect location dimensional information for the bump is generated. This defect location dimensional information includes information such as the distance between the pinhole and the fusion line.

[0092] The defect scoring dimension information is generated based on the distance between the defect and the position detection benchmark to calculate the defect score. When identifying defects that cause unqualified welds, the influence of the defect location on the weld quality can be considered, thereby improving the detection accuracy of the welds, avoiding the situation where unqualified welds are determined to be qualified welds, and improving the product yield.

[0093] FIG5A is a schematic diagram of a process for generating defect superposition dimension information in some embodiments of the defect detection method disclosed herein, as shown in FIG5A :

[0094] In step S501, when other defects exist in the weld to be inspected, a superimposed defect of the defect is determined among the other defects based on the defect location, the defect locations of the other defects, and the superimposed inspection datum; wherein the defect and the superimposed defect are both located on or within the same superimposed inspection datum;

[0095] In step S502, defect superposition dimension information of the defect is generated based on the defect and the superimposed defect.

[0096] In some embodiments, multiple defects on or within the same superimposed detection reference may have a greater impact on the quality of the weld than multiple defects located at different positions in the weld.

[0097] Depending on the weld quality inspection requirements, multiple superimposed inspection benchmarks can be set, including superimposed inspection lines and superimposed inspection areas. As shown in Figure 5B, the superimposed inspection line can be a longitudinal straight line 504 on the surface of weld 505, which can be the longitudinal centerline of the weld or parallel to the longitudinal centerline of the weld. As shown in Figure 5C, the superimposed inspection benchmark can be a longitudinal straight line 506 on the surface of weld 507, which is perpendicular to the longitudinal centerline of the weld. As shown in Figure 5D, the superimposed inspection benchmark can be a superimposed inspection area 509 set on the weld surface 508.

[0098] When the defect detection model outputs the defect, its parameters, and confidence level, it can also output the weld's longitudinal centerline position information or overlay the position information of the inspection area. For example, consider a pinhole defect, A. The weld to be inspected also contains two other pinhole defects, B and C, and two convex defects, D and E.

[0099] The superposition inspection benchmark is a longitudinal straight line on the surface of the weld to be inspected in the first inspection image. Based on the coordinate information of pinhole defects A, B, C, convex defect D, and E in the first inspection image, it is determined that pinhole defects A, C, and D are located on the same longitudinal straight line (this longitudinal straight line is parallel to the longitudinal centerline of the weld to be inspected). In this case, pinhole defects C and D are determined to be superimposed defects of pinhole defect A.

[0100] According to the pinhole defect A and the superimposed defect of pinhole defect A, the defect superposition dimension information of pinhole defect A is generated, and the defect superposition dimension information includes: pinhole defect A, pinhole defect C and convex defect D, as well as the position information of pinhole defect A, pinhole defect C and convex defect D and other information.

[0101] Based on the position of the defect and other defects and the superposition detection benchmark, the defect superposition dimension information of the defect is generated to calculate the defect score. When identifying the defects that cause unqualified welds, the impact of multiple defects on the weld quality on or within the same superposition detection benchmark can be considered, thereby improving the detection accuracy of the welds, avoiding the situation where unqualified welds are determined to be qualified welds, and improving the product yield.

[0102] FIG6A is a schematic diagram of a process for obtaining a defect score using a defect score processing network in some embodiments of the defect detection method of the present disclosure, as shown in FIG6A :

[0103] In step S601, the dimension weight of the defect size and the dimension weight of the scoring dimension information are obtained.

[0104] In step S602, a defect scoring processing network for the defect is determined in the defect comprehensive scoring model.

[0105] In step S603, the defect size, size weight, scoring dimension information and dimension weight are input into the defect scoring processing network to obtain the defect score output by the defect scoring processing network.

[0106] In some embodiments, weights can be assigned to defect size and scoring dimension information. These weights represent the weight of these dimensions in the scoring. By assigning appropriate weights, the importance of these dimensions in the overall scoring can be reflected. For example, since defect size has a greater impact on weld quality, a higher weight may be assigned to the weld size.

[0107] For each defect, the defect scoring processing network can score based on the scoring dimension information of the defect, and sum the scores of each scoring dimension information according to the weight to obtain a comprehensive score of the defect. The comprehensive score can measure the severity of the impact of the defect on the weld quality.

[0108] A comprehensive defect scoring model is constructed, in which a defect scoring processing network is constructed for each defect. The defect scoring processing network can be a variety of neural network models, such as a convolutional neural network. Samples of defect size, size weights, scoring dimension information, and dimension weights are obtained for each defect to generate training samples. These training samples are labeled with information such as the defect score. The labeled training samples are used to train the defect scoring processing network of each model, resulting in a trained defect scoring processing network.

[0109] As shown in FIG6B , the defect comprehensive scoring model includes defect scoring processing networks such as a pinhole defect scoring processing network and a convex point defect scoring processing network constructed for pinhole defects, convex points and other defects; the defect comprehensive scoring model also includes a data input module and a defect scoring output module.

[0110] For example, the data input module receives data on a pinhole defect, including information such as the pinhole defect size, pinhole defect location dimensional information, and pinhole defect overlay dimensional information. The module then obtains a preset size weight of 0.4, a defect location dimensional weight of 0.3, and a defect overlay dimensional weight of 0.3. The module then determines the pinhole defect scoring processing network within the comprehensive defect scoring model and inputs data such as the pinhole defect size and size weight, the pinhole defect location dimensional information and defect location dimensional weight, and the pinhole defect overlay dimensional information and defect overlay dimensional weight into the pinhole defect scoring processing network.

[0111] The defect scoring output module obtains the pinhole defect score output by the pinhole defect scoring processing network. If the pinhole defect score is greater than or equal to the defect scoring threshold, the inspection result is determined to be unqualified for the weld to be inspected, and the product is determined to be an NG (No Good) product.

[0112] The data input module receives data on a convex defect, including the convex defect size, convex defect location dimensional information, and convex defect stacking dimensional information. The module also obtains a preset size weight of 0.4, a defect location dimensional weight of 0.3, and a defect stacking dimensional weight of 0.3. The module then determines a convex defect scoring processing network for the convex defect within the comprehensive defect scoring model and inputs the convex defect size and size weight, convex defect location dimensional information and defect location dimensional weight, and convex defect stacking dimensional information and defect stacking dimensional weight into the convex defect scoring processing network.

[0113] The defect scoring output module receives the bump defect scores output by the bump defect scoring processing network. If the bump defect score is greater than or equal to the defect scoring threshold, the weld under test is determined to be unqualified and the product is classified as NG (No Good). If the scores of all defects are less than the defect scoring threshold, the weld under test is determined to be qualified and the product is classified as OK.

[0114] By setting weights for defect size and scoring dimension information, the impact of defect size and different scoring dimensions on weld quality can be more accurately reflected, which can improve the effectiveness and accuracy of defect detection; obtaining defect scores through the defect scoring processing network can improve the efficiency and accuracy of scoring.

[0115] In some embodiments, when the detection result is that the detection target is unqualified, all defects with scores greater than or equal to the defect score threshold are classified and processed to obtain defect type information, and a defect processing task is generated based on the defect type information.

[0116] There are many ways to classify defects. For example, all defects can be classified based on the percentage of defects. When the inspection result is that the weld is unqualified, the defects with a score greater than or equal to the defect score threshold include four pinhole defects and one convex defect. Since the repair of pinhole defects and convex defects requires different processes, they are classified based on the percentage of pinhole defects and convex defects. If the number of pinhole defects accounts for 80% and the number of convex defects accounts for 20%, the defect type is determined to be a pinhole defect type, a pinhole defect processing task is generated, and sent to the pinhole defect repair process.

[0117] In some embodiments, when using a camera to photograph the weld, the camera's mounting angle needs to be adjusted for each type of battery cell. The camera's mounting angle is adjusted to a suitable angle so that the camera's shooting surface covers the entire weld and a qualified weld photograph is obtained. Currently, the camera's mounting angle is adjusted based on the worker's experience. After each adjustment, the camera needs to photograph the weld to obtain a test image and determine whether the test image is qualified. If the test image is unqualified, the camera's mounting angle needs to be readjusted. Manually determining the camera's adjustment angle requires many adjustments, making the camera's mounting angle adjustment difficult, resulting in low work efficiency and difficulty in ensuring the quality of the weld photograph.

[0118] FIG7 is a schematic diagram of a process for optimizing the shooting posture of an image acquisition device in some embodiments of the defect detection method disclosed herein, as shown in FIG7 :

[0119] In step S701 , shooting posture test information of the image acquisition device is generated.

[0120] In some embodiments, the image acquisition device can be a variety of cameras, such as a 2D camera, a 3D camera, etc. The image acquisition device has an effective working distance specified at the factory, and imaging cannot be performed beyond the working distance. In an actual production environment, the installation position of the image acquisition device is determined based on the effective working distance of the image acquisition device. The installation position of the image acquisition device is usually fixed. In order for the image acquisition device's shooting area to cover the entire weld, the installation angle of the image acquisition device needs to be adjusted.

[0121] The shooting posture test information includes a first installation angle of the image capture device. The first installation angle can include multiple angles, such as a first horizontal tilt angle and a first vertical tilt angle. The horizontal tilt angle is the tilt or rotation angle of the image capture device in the horizontal direction, and the vertical tilt angle is the tilt or rotation angle of the image capture device in the vertical direction.

[0122] When using an image acquisition device to take pictures of the weld, the installation angle of the image acquisition device needs to be adjusted for each model of battery cell. The installation angle of the image acquisition device is adjusted to an appropriate angle so that the shooting surface of the image acquisition device covers the entire weld to obtain a qualified weld image.

[0123] In step S702, a second detection image of the weld to be detected is acquired by the image acquisition device in a shooting posture corresponding to the shooting posture test information.

[0124] For example, when adjusting the installation angle of the image acquisition device, a first installation angle of the image acquisition device is generated, and the image acquisition device is adjusted according to the first installation angle; after the adjustment is completed, the image acquisition device is in a shooting posture corresponding to the first installation angle, and the image acquisition device is used to take a picture of the weld to be inspected to obtain a second inspection image.

[0125] In step S703, shooting posture optimization information of the image acquisition device is determined according to the image loss data of the second detection image and the shooting posture test information.

[0126] The shooting posture optimization information includes a second installation angle of the image acquisition device, which includes a second horizontal inclination angle, a second vertical inclination angle, etc. After the second installation angle is determined, the image acquisition device is adjusted according to the second installation angle. After the adjustment is completed, the image acquisition device is in a shooting posture corresponding to the second installation angle, and the image acquisition device is used to photograph the weld to be inspected to obtain a first inspection image. The defect in the weld to be inspected and the defect parameters of the defect are determined based on the first inspection image.

[0127] Based on the shooting posture test information and the image loss data of the detection image collected based on the shooting posture test information, the shooting posture optimization information of the image acquisition device is determined, which is used to adjust the installation angle of the image acquisition device to obtain qualified weld detection images. This can reduce the number of adjustments, reduce the difficulty of adjusting the installation angle, and improve work efficiency.

[0128] In some embodiments, various methods can be used to generate the shooting posture test information. For example, the valid range of the shooting posture test information can be obtained, and the shooting posture test information can be generated based on the valid range using prediction software. The valid range of the shooting posture test information can be determined based on factors such as the scene of the weld photography and the installation requirements of the image acquisition equipment. For example, the valid range of the shooting posture test information includes a valid range of 0°-90° for the first vertical inclination angle and a valid range of 0°-45° for the first horizontal inclination angle.

[0129] Using prediction software and based on the valid range, generate shooting posture test information. Prediction software can be a variety of software, such as JMP software, etc. JMP software is an interactive visualization statistical discovery software. Shooting posture test information can be generated within the valid range through prediction software such as JMP software. For example, the valid range of the first vertical inclination angle is 0°-90°, and the valid range of the first horizontal inclination angle is 0°-45°. Enter the JMP software, JMP software can filter random sampling points, and the generated part of the shooting posture test information is shown in Table 1 below:

[0130] Table 1 - Shooting posture test information table

[0131] In some embodiments, the image capture device is sequentially adjusted based on the first vertical tilt angle and the first horizontal tilt angle in Table 1 above. After each adjustment, the image capture device is used to photograph the weld to be inspected to obtain a second inspection image. After the second inspection image is captured for the weld to be inspected, image loss data for the second inspection image is determined. The image loss data may include various data such as pixel loss.

[0132] The image loss data of the second inspection image can be determined using a variety of methods. For example, the second inspection image can be input into an object detection model, the number of pixels of the weld to be inspected, as output by the object detection model, can be obtained, and the number of missing pixels can be determined based on the number of pixels of the weld to be inspected and the required number of pixels.

[0133] Construct a target detection model. The target detection model can be a variety of neural network models, such as a convolutional neural network model. Obtain a sample of the detection image, and annotate the weld area and the number of weld pixels on the detection image sample. Use the annotated detection image sample to train the target detection model. For welds, the required number of weld pixels can be set. When determining the number of missing pixels, input the second detection image into the target detection model, obtain the number of weld pixels output by the target detection model, calculate the difference between the number of weld pixels and the required number of weld pixels, and determine the number of missing pixels.

[0134] FIG8A is a schematic diagram of determining shooting posture optimization information in some embodiments of the defect detection method disclosed herein, as shown in FIG8A :

[0135] In step S801 , the association relationship between image loss data and shooting posture test information is determined.

[0136] In step S802, prediction software is used to perform prediction processing based on the association relationship to determine shooting posture optimization information. The shooting posture optimization information includes a second installation angle of the image acquisition device; the second installation angle includes: a second horizontal tilt angle, a second vertical tilt angle, etc.

[0137] In some embodiments, the association between image loss data and shooting posture test information may include: a first association between the first vertical inclination angle of the image acquisition device and the number of missing pixels, and a second association between the first horizontal inclination angle of the image acquisition device and the number of missing pixels.

[0138] By inputting the first and second correlation relationships into the JMP software, the response surface analysis method, desirability function, and prediction profiler built into the JMP software can be used to perform predictions based on the first and second correlation relationships through the JMP software to obtain prediction data between the first vertical inclination angle and the number of missing pixels, and between the first horizontal inclination angle and the number of missing pixels.

[0139] The prediction graph generated by JMP software based on the first and second associations is shown in Figure 8B. Based on the prediction graph generated by JMP software, it can be determined that the first vertical tilt angle that minimizes the number of missing pixels is 30.29°, and the first horizontal tilt angle that minimizes the number of missing pixels is 12.8°. There are various methods for determining the second installation angle of the image acquisition device. For example, the second horizontal tilt angle can be set to 30.29° and the second vertical tilt angle can be set to 12.8°. Alternatively, an angle adjustment threshold can be set, such that the second horizontal tilt angle is set to 30.29° + the angle adjustment threshold, or 30.29° - the angle adjustment threshold; and the second vertical tilt angle is set to 12.8° + the angle adjustment threshold, or 12.8° - the angle adjustment threshold.

[0140] After determining the shooting posture optimization information, that is, after determining the second horizontal inclination angle, the second vertical inclination angle, etc., the image acquisition device is adjusted according to the second horizontal inclination angle, the second vertical inclination angle, etc. After the adjustment is completed, the image acquisition device used is used to take a picture of the weld to be inspected, obtain a first inspection image, and determine the defects of the weld to be inspected and the defect parameters of the defects.

[0141] By using prediction software and based on the correlation between image loss data and shooting posture test information, the shooting posture optimization information of the image acquisition device is determined, and the installation adjustment angle of the image acquisition device can be automatically determined. Qualified weld inspection images can be obtained, which can reduce the number of adjustments, improve work efficiency, and ensure the detection accuracy of defect detection.

[0142] In some embodiments, the present disclosure provides a defect detection device, as shown in FIG9A . The defect detection device includes a defect determination module 910, a defect scoring module 920, and a detection processing module 930. The defect determination module 910 determines the defect and defect parameters of the weld to be detected based on a first detection image captured by an image acquisition device of the weld to be detected. For example, the defect determination module 910 preprocesses the first detection image to obtain an image to be identified; the defect determination module 910 inputs the image to be identified into a defect detection model to obtain the defect, defect parameters, and confidence level output by the defect detection model; wherein the confidence level is greater than a preset confidence threshold.

[0143] The detection processing module 930 determines the inspection result of the weld to be inspected based on the defect scores. For example, if the score of at least one defect is greater than or equal to the defect score threshold, the detection processing module 930 determines that the weld to be inspected is unqualified. If the scores of all defects are less than the defect score threshold, the detection processing module 930 determines that the weld to be inspected is qualified.

[0144] The defect scoring module 920 obtains a defect score based on defect parameters, including defect size. If the defect size exceeds a defect size threshold, the defect scoring module 920 sets the defect score to be greater than or equal to a defect score threshold.

[0145] In one embodiment, as shown in FIG9B , the defect scoring module 920 includes a scoring dimension determination unit 921 and a score obtaining unit 922. Defect parameters include defect location. If the defect size does not exceed a defect size threshold, the scoring dimension determination unit 921 determines defect scoring dimension information based on the defect location. The score obtaining unit 922 obtains a defect score based on the defect size and the scoring dimension information.

[0146] The scoring dimension information includes defect location dimension information. The scoring dimension determination unit 921 determines the distance between the defect and the position detection benchmark based on the position information of the defect location and the position detection benchmark; the scoring dimension determination unit 921 generates the defect location dimension information of the defect based on the distance.

[0147] The scoring dimension information includes defect superposition dimension information. When there are other defects in the weld to be inspected, the scoring dimension determination unit 921 determines the superposition defect of the defect among other defects based on the defect position, the defect position of other defects and the superposition detection benchmark; wherein the defect and the superposition defect are both located above or within the same superposition detection benchmark; the scoring dimension determination unit 921 generates the defect superposition dimension information of the defect based on the defect and the superposition defect.

[0148] The scoring acquisition unit 922 obtains the size weight of the defect size and the dimension weight of the scoring dimension information; the scoring acquisition unit 922 determines the defect scoring processing network of the defect in the defect comprehensive scoring model, inputs the defect size, size weight, scoring dimension information and dimension weight into the defect scoring processing network, and obtains the defect score output by the defect scoring processing network.

[0149] In one embodiment, as shown in FIG9C , the defect detection apparatus further includes an equipment adjustment module 940. Equipment adjustment module 940 generates shooting posture test information for the image acquisition device and obtains a second inspection image of the weld to be inspected, captured by the image acquisition device in a shooting posture corresponding to the shooting posture test information. Equipment adjustment module 940 determines shooting posture optimization information for the image acquisition device based on the image loss data of the second inspection image and the shooting posture test information. The image acquisition device captures an image of the weld to be inspected in the shooting posture corresponding to the shooting posture optimization information to obtain the first inspection image.

[0150] Equipment Adjustment Module 940 obtains the valid range of the shooting posture test information and generates the shooting posture test information based on the valid range using prediction software. Image loss data includes the number of missing pixels. Equipment Adjustment Module 940 inputs the second test image into the target detection model and obtains the target detection model's output of the number of pixels in the weld to be inspected. Equipment Adjustment Module 940 determines the number of missing pixels based on the number of pixels in the weld to be inspected and the required number of pixels.

[0151] The equipment adjustment module 940 determines the correlation between the image loss data and the shooting posture test information, uses prediction software and performs prediction processing based on the correlation to determine the shooting posture optimization information.

[0152] Figure 10 is a block diagram of some embodiments of an electronic device according to the present disclosure. As shown in Figure 10 , the electronic device may include a memory 1001, a processor 1002, a communication interface 1003, and a bus 1004. Memory 1001 is used to store instructions, and processor 1002 is coupled to memory 1001. Processor 1002 is configured to execute the above-described defect detection method based on the instructions stored in memory 1001.

[0153] Memory 1001 can be high-speed RAM, non-volatile memory, or a memory array. Memory 1001 can also be divided into blocks, and the blocks can be combined into virtual volumes according to certain rules. Processor 1002 can be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the defect detection method of the present disclosure.

[0154] In some embodiments, the present disclosure provides a computer-readable storage medium storing computer instructions, and the instructions are executed by a processor to execute the method in any of the above embodiments.

[0155] Computer readable storage media can adopt any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can, for example, include but is not limited to a system, device or component of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive enumeration) of readable storage media can include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof.

[0156] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram and the combination of the processes and / or boxes in the flowchart and / or block diagram can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate a device for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0157] The methods and systems of the present disclosure may be implemented in many ways. For example, the methods and systems of the present disclosure may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present disclosure are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present disclosure may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the methods according to the present disclosure. Therefore, the present disclosure also covers recording media that store programs for executing the methods according to the present disclosure.

[0158] Although the present application has been described with reference to preferred embodiments, various modifications may be made thereto and components may be substituted with equivalents without departing from the scope of the present application. In particular, the various technical features described in the various embodiments may be combined in any manner as long as there are no structural conflicts. The present application is not limited to the specific embodiments disclosed herein, but encompasses all technical solutions within the scope of the claims.

Claims

1. A weld detection method, characterized in that, Including: Determine the defects of the weld to be detected and the defect parameters of the defects according to the first detection image collected by the image acquisition device for the weld to be detected. Obtain the score of the defect according to the defect parameters. Wherein, the defect parameters include: defect location; the obtaining the score of the defect according to the defect parameters includes: when the defect size does not exceed the defect size threshold, determining the score dimension information of the defect according to the defect location; obtaining the score of the defect according to the defect size and the score dimension information. Based on the score of the defect, determine the detection result of the weld to be detected.

2. The method according to claim 1, characterized in that The defect parameters include: defect size; the obtaining the score of the defect according to the defect parameters includes: When the defect size exceeds the defect size threshold, set the score of the defect to be greater than or equal to the defect score threshold.

3. The method according to claim 2, wherein The score dimension information includes: defect location dimension information; the determining the score dimension information of the defect according to the defect location includes: Determine the distance between the defect and the position detection reference according to the defect location and the position information of the position detection reference. Generate the defect location dimension information of the defect based on the distance.

4. The method according to claim 2, wherein The score dimension information includes: defect superposition dimension information; the determining the score dimension information of the defect according to the defect location includes: When there are other defects in the weld to be detected, determine the superposition defect of the defect among the other defects according to the defect location, the defect locations of the other defects, and the superposition detection reference. Wherein, the superposition detection reference includes: a superposition detection line or a superposition detection area; both the defect and the superposition defect are located on the same superposition detection line or within the same superposition detection area. Generate the defect superposition dimension information of the defect according to the defect and the superposition defect.

5. The method according to claim 2, characterized in that, The obtaining the score of the defect according to the defect size and the score dimension information includes: Obtain the size weight of the defect size and the dimension weight of the score dimension information. Determine the defect score processing network of the defect in the defect comprehensive scoring model. Input the defect size, the size weight, the score dimension information, and the dimension weight into the defect score processing network, and obtain the score of the defect output by the defect score processing network.

6. The method according to claim 1, wherein The determining the defects of the weld to be detected and the defect parameters of the defects according to the first detection image collected by the image acquisition device for the weld to be detected includes: Preprocess the first detection image to obtain an image to be recognized. Input the image to be recognized into the defect detection model, and obtain the defects, the defect parameters of the defects, and the confidence level output by the defect detection model. Wherein, the confidence level is greater than a preset confidence level threshold.

7. The method according to any one of claims 1 to 6, characterized in that The determining the detection result of the weld to be detected based on the score of the defect includes: When the scores of at least one defect are greater than or equal to the defect score threshold, determine that the detection result is that the weld to be detected is unqualified. When the scores of all defects are less than the defect score threshold, it is determined that the detection result of the weld to be detected is qualified.

8. The method according to any one of claims 1 to 6, characterized in that The method further includes: Generating the shooting attitude test information of the image acquisition device; Obtaining a second detection image collected by the image acquisition device for the weld to be detected in the shooting attitude corresponding to the shooting attitude test information; Determining the shooting attitude optimization information of the image acquisition device according to the image loss data of the second detection image and the shooting attitude test information; Wherein, the image acquisition device acquires an image of the weld to be detected in the shooting attitude corresponding to the shooting attitude optimization information to obtain the first detection image.

9. The method according to claim 8, wherein The determining the shooting attitude optimization information of the image acquisition device according to the image loss data of the second detection image and the shooting attitude test information includes: Determining the correlation between the image loss data and the shooting attitude test information; Using prediction software and performing prediction processing based on the correlation to determine the shooting attitude optimization information.

10. The method according to claim 9, wherein The generating the shooting attitude test information of the image acquisition device includes: Obtaining the effective range of the shooting attitude test information; Using the prediction software and generating the shooting attitude test information based on the effective range.

11. The method according to claim 9, characterized in that The image loss data includes: the number of missing pixel points; the method further includes: Inputting the second detection image into a target detection model to obtain the number of pixel points of the weld to be detected output by the target detection model; Determining the number of missing pixel points according to the number of pixel points of the weld to be detected and the required number of pixel points.

12. The method according to claim 8, wherein The shooting attitude test information includes: the first installation angle of the image acquisition device; The shooting attitude optimization information includes: the second installation angle of the image acquisition device.

13. The method according to claim 6, wherein After obtaining the defects, the defect parameters and the confidence levels output by the defect detection model, the method further includes: When the confidence level of the defect is less than the set confidence level allowable value, filtering the defects with the confidence level less than the confidence level allowable value.

14. The method according to claim 6, wherein After preprocessing the first detection image to obtain an image to be recognized, the method further includes: Inputting the image to be recognized into a defect detection model to obtain the position detection reference output by the defect detection model and the position information of the position detection reference.

15. The method according to claim 3 or 14, characterized in that, The position detection reference includes: the fusion line between the weld between the battery cell top cover and the aluminum shell and the battery cell top cover, the fusion line between the weld between the battery cell top cover and the aluminum shell and the aluminum shell, the fusion zone between the weld between the battery cell top cover and the aluminum shell and the battery cell top cover, the fusion zone between the weld between the battery cell top cover and the aluminum shell and the aluminum shell; wherein, the fusion line refers to the boundary line between the weld metal and the battery cell top cover or the aluminum shell body, and the fusion zone refers to the boundary area between the weld metal and the battery cell top cover or the aluminum shell body.

16. The method according to claim 7, wherein After determining the detection result of the weld to be detected based on the score of the defect, the method further includes: In the case where the detection result shows that the weld to be detected is unqualified, classify all defects with a score greater than or equal to the defect score threshold to obtain defect type information; Generate a defect handling task according to the defect type information.

17. The method according to claim 11, wherein Before inputting the second detection image into the target detection model, the method further includes: Obtain a detection image sample, and label the weld area and the number of weld pixel points in the detection image sample; Use the labeled detection image sample to train the target detection model.

18. The method according to claim 9, wherein The determination of the correlation between the image loss data and the shooting posture test information includes: Determine a first correlation between the first vertical direction inclination angle of the image acquisition device and the number of missing pixel points, and a second correlation between the first horizontal direction inclination angle of the image acquisition device and the number of missing pixel points; Among them, the correlation includes the first correlation and the second correlation.

19. The method according to claim 18, wherein The shooting posture optimization information includes the second installation angle of the image acquisition device, and the second installation angle includes: a second horizontal direction inclination angle and a second vertical direction inclination angle; the use of the prediction software and based on the correlation to perform prediction processing to determine the shooting posture optimization information includes: Use the prediction software and perform prediction processing based on the first correlation and the second correlation to obtain prediction data between the first vertical direction inclination angle and the number of missing pixel points, and prediction data between the first horizontal direction inclination angle and the number of missing pixel points; Based on the prediction data between the first vertical direction inclination angle and the number of missing pixel points, determine the second vertical direction inclination angle that minimizes the number of missing pixel points; Based on the prediction data between the first horizontal direction inclination angle and the number of missing pixel points, determine the second horizontal direction inclination angle that minimizes the number of missing pixel points.

20. A weld detection device, characterized in that, Include: A defect determination module, configured to determine the defects of the weld to be detected and the defect parameters of the defects according to the first detection image collected by the image acquisition device for the weld to be detected; A defect scoring module, configured to obtain the score of the defect according to the defect parameters; Among them, the defect parameters include the defect position, and the defect scoring module includes: A scoring dimension determination unit, configured to determine the scoring dimension information of the defect according to the defect position in the case where the defect size does not exceed the defect size threshold; A scoring obtaining unit, configured to obtain the score of the defect according to the defect size and the scoring dimension information; A detection module, configured to determine the detection result of the weld to be detected based on the score of the defect.

21. An electronic device, characterized in that, Include: A memory; And a processor coupled to the memory, the processor is configured to execute the method according to any one of claims 1 to 19 based on instructions stored in the memory.

22. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions, and the computer instructions are executed by the processor to perform the method according to any one of claims 1 to 19.

Images