Battery pack automatic detection method and system

Abstract

The invention relates to the technical field of battery manufacturing, and particularly discloses a battery pack automatic detection method and system. The method comprises the following steps: positioning a carrier carrying a battery pack at a detection station; a robot is used for driving an image acquisition system to move to scan the battery pack, and a surface image of the battery pack is obtained; and comprehensively judging the appearance state of the battery pack based on the image. According to the invention, manual or semi-automatic detection is converted into a full-automatic integrated non-contact intelligent detection process based on machine vision, subjectivity and instability of manual vision are thoroughly eliminated through standardized mechanical positioning and image acquisition, and objectivity and traceability of a detection result are realized.

Description

Technical Field

[0001] This invention relates to the field of battery manufacturing technology, and in particular to an automatic detection method and system for battery packs. Background Technology

[0002] With the rapid development of the new energy vehicle industry, the appearance quality of the battery pack, as a core component, directly affects the safety and reliability of the entire vehicle. Traditional manual visual inspection methods are highly subjective, have inconsistent standards, are inefficient, and are prone to missed inspections due to fatigue, which has become a bottleneck for improving production capacity and quality control.

[0003] Currently, while rule-based traditional machine vision technology has achieved automated inspection to some extent, it relies on preset thresholds and templates. Its generalization ability for irregular defects such as complex scratches, dents, and assembly flaws on the surface of battery packs is insufficient. It is prone to misjudgment in variable production environments, making it difficult to meet the requirements of high-precision and high-robust full inspection, thus restricting the further improvement of production yield. Summary of the Invention

[0004] Based on this, the purpose of the present invention is to provide an automatic battery pack inspection method and system that utilizes machine vision surface defect detection technology to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention adopts the following technical solution:

[0006] This invention provides an automatic battery pack detection method, which includes the following steps: S1. Position the carrier carrying the battery pack at the testing station; S2. Use a robot to drive the image acquisition system to scan the battery pack and obtain an image of the battery pack surface; S3. Based on the image, make a comprehensive judgment on the appearance of the battery pack.

[0007] This invention transforms manual or semi-automatic inspection into a fully automated, integrated, non-contact intelligent inspection process based on machine vision. Through standardized mechanical positioning and image acquisition, it completely eliminates the subjectivity and instability of manual visual inspection, achieving objectivity and traceability of inspection results. This streamlined design lays the automated foundation for all subsequent refined inspections and is the core framework for improving overall inspection efficiency and consistency.

[0008] As a further improvement to the above-described solution of the present invention, step S1 includes: Initial positioning of the transport trolley transporting the aforementioned vehicle; The vehicle is separated from the transport trolley and lifted to achieve precise positioning of the vehicle.

[0009] This two-stage positioning strategy effectively solves the positioning challenges posed by the weight and size of the battery pack. Initial positioning is fast and efficient, while fine positioning ensures that the battery pack is in an absolutely stable and accurate position during subsequent high-precision scanning. This eliminates imaging blurring and detection errors caused by vehicle shaking or positional deviations, and is a prerequisite for ensuring the accuracy of subsequent visual inspection.

[0010] As a further improvement to the above-mentioned solution of the present invention, in step S2, two robots are configured, each driving an image acquisition system. The linear module drives the two robots to adjust their positions, achieving blind-spot-free inspection of the battery pack. This dual-robot collaborative working mode, combined with the linear module's wide-range position adjustment capability, enables efficient, blind-spot-free coverage scanning of the complex three-dimensional surface of the battery pack. This design significantly improves the flexibility of the inspection system, adapting to size variations of different battery pack models. Path optimization minimizes scanning time, thus meeting the full inspection requirements of high-speed production lines.

[0011] As a further improvement to the above-mentioned solution of the present invention, in step S2, the image acquisition system includes at least a 2D image acquisition system for visual positioning and two-dimensional feature detection, and a 3D image acquisition system for three-dimensional shape scanning; the 2D image acquisition system guides the robot to take pictures of the side of the battery pack to obtain a side image of the battery pack; the 3D image acquisition system guides the robot, in conjunction with the linear module, to scan the surface of the battery pack cover to obtain an image of the battery pack cover, and guides the robot to scan the side welds of the battery pack to obtain side weld images. The fusion of 2D and 3D vision constitutes a complementary, all-dimensional perception capability. The 2D system excels at quickly identifying features such as texture, color, and characters (e.g., the presence or absence of labels), while the 3D system can accurately acquire the depth, height, and three-dimensional contour information of objects, and is extremely sensitive to surface defects such as dents, scratches, and weld morphology. This fusion ensures that all dimensions of appearance quality, from flat to curved surfaces, from color to shape, can be effectively monitored.

[0012] As a further improvement to the above-mentioned solution of the present invention, in step S3, based on the side image, the presence or absence of the label, plug-in protective cover, vent valve and liquid cooling pipe port is detected.

[0013] As a further improvement to the above-described solution of the present invention, step S3, based on the image of the box lid, detects pits, scratches, and deformation defects on the surface of the box lid, specifically including: Based on the point cloud data acquired by the 3D image acquisition system, a height difference map is generated and potential defect areas are located using traditional image processing algorithms, including Gaussian filtering, reference surface difference, and threshold segmentation. The data of the potential defect area is input into a deep learning classification network to determine the authenticity and / or type of the defect; By combining the geometric feature information output from the primary detection stage with the classification confidence information output from the fine discrimination stage, a final defect determination is made.

[0014] As a further improvement to the above-mentioned solution of the present invention, step S3, which detects the quality of the side weld based on the side weld image, specifically includes: Based on the point cloud data acquired by the 3D image acquisition system, a height difference map is generated and potential defect areas are located using traditional image processing algorithms, including Gaussian filtering, reference surface difference, and threshold segmentation. The data of the potential defect area is input into a deep learning classification network to determine the authenticity and / or type of the defect; By combining the geometric feature information output from the primary detection stage with the classification confidence information output from the fine discrimination stage, a final defect determination is made.

[0015] The present invention also provides an automatic battery pack detection system for implementing the automatic battery pack detection method described above, comprising: frame; A positioning system, located below the frame, is used to position the vehicle carrying the battery pack. The robot is mounted on the frame via a linear module; Image acquisition system, an image acquisition device installed at the end of the robot; The host computer communicates with the image acquisition system and the robot to control the detection process and execute detection algorithms based on image data.

[0016] This hardware system constitutes a modular, highly integrated testing workstation. The functional modules are rationally laid out and rigidly supported on a mechanical frame. Unified scheduling and data fusion are achieved through a host computer, realizing a closed loop from physical positioning and image acquisition to intelligent analysis. This integrated design facilitates deployment and maintenance in production lines, providing a reliable physical foundation for transforming innovative methods into stable productivity.

[0017] As a further improvement to the above-described solution of the present invention, the positioning system includes at least two positioning platforms arranged opposite to each other; each positioning platform includes a positioning mechanism for initial positioning of the transport trolley and a lifting mechanism for lifting and precisely positioning the vehicle. The symmetrical dual-platform design provides balanced and stable clamping and lifting forces, making it particularly suitable for long or large battery packs, preventing tilting or deformation during positioning. The functional separation of the positioning mechanism and the lifting mechanism allows the system to adapt to different vehicles and transportation schemes, enhancing the system's versatility and production line compatibility.

[0018] As a further improvement of the above-mentioned solution of the present invention, the positioning mechanism includes a positioning cylinder and a positioning wheel driven therefrom, for engaging the positioning block of the mobile platform; the lifting mechanism includes a lifting cylinder, a support plate driven therefrom, and a lifting block disposed on the support plate, wherein the lifting block is adapted to engage in the limiting groove at the bottom of the carrier during lifting.

[0019] As a further improvement to the above-described solution of the present invention, the image acquisition system includes a 2D image acquisition system and a 3D image acquisition system; the 2D image acquisition system includes an area array camera, a fixed-focus lens and a ring light source; the 3D image acquisition system includes a 3D line laser camera and its controller.

[0020] Compared with the prior art, the present invention has the following beneficial effects: This application utilizes a non-contact appearance defect detection technology based on machine vision. It can inspect defects such as dents, scratches, and deformations on the surface of battery pack covers; check the location, content, and category of labels (nameplate barcodes, traceability codes, high-voltage markings, warning labels, etc.); inspect low-voltage connector pin retraction; inspect the appearance of high-voltage connectors; check for undamaged or missing connector protective covers and vent valves; and inspect the condition of liquid cooling pipe openings. This technology solves the problem of traditional machine vision technology's heavy reliance on preset thresholds and templates, resulting in insufficient generalization ability for complex scratches, dents, assembly defects, and other irregular defects on battery pack surfaces. It reduces the false positive rate in variable production environments, meets the requirements for high-precision and high-robust full inspection, and improves production yield. Attached Figure Description

[0021] Figure 1 A front view of an automatic battery pack detection system provided in an embodiment of the present invention; Figure 2 A side view of an automatic battery pack detection system provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the positioning and lifting mechanism in an automatic battery pack detection system according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of an image acquisition device in an automatic battery pack detection system provided by an embodiment of the present invention; Figure 5 A flowchart of an automatic battery pack detection method provided in this embodiment of the invention.

[0022] Reference numerals: 100, frame; 200, AGV (Automated Guided Vehicle); 300, pallet trolley; 400, pallet; 500, battery pack; 600, positioning and lifting mechanism; 601, pallet plate; 602, positioning block; 603, slider; 700, robot; 800, linear module; 900, image acquisition system; 901, area scan camera; 902, lens; 903, ring light source; 904, 3D line laser camera. Detailed Implementation

[0023] To facilitate understanding of the present invention, a more comprehensive description will be given below with reference to specific embodiments. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the disclosure of the present invention.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains. The terminology used herein in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.

[0025] like Figures 1 to 4 As shown in the figure, the automatic battery pack detection system provided in this embodiment of the invention mainly includes a frame 100, a hidden AGV 200, a pallet trolley 300, a pallet 400, a positioning platform 600, a collaborative robot 700, a linear module 800, an image acquisition system 900, and a host computer (not shown in the figure).

[0026] The AGV 200 is responsible for transporting the pallet cart 300 and pallet 400, which carry the battery pack 500. Two positioning platforms 600 are arranged opposite each other on the ground directly below the frame 100, forming an inspection station. Two six-axis collaborative robots 700 are mounted upside down on the same linear module 800, which serves as the seventh axis of the robot 700 and is horizontally mounted on top of the frame 100. Image acquisition systems 900 are installed at the end caps of both robots 700. A host computer communicates with the AGV 200, positioning platform 600, robot 700, linear module 800, and image acquisition system 900 to coordinate and control the entire inspection process.

[0027] like Figure 2 and Figure 3 As shown, the pallet 400 carries the battery pack 500 and is equipped with limiting components to secure the battery pack 500, preventing it from shifting during transportation and inspection. The pallet trolley 300 connects to corresponding interfaces on the bottom of the pallet 400 via four positioning posts. Each positioning post contains a positioning pin to ensure precise and error-free relative positioning between the pallet 400 and the pallet trolley 300 during transportation. Positioning blocks (not shown in the figure) are installed on all four wheels of the pallet trolley 300.

[0028] The positioning platform 600 is the core of this system for achieving high-precision positioning. Each positioning platform 600 mainly includes a tray 601, a lifting block 602, and a drive mechanism. The distance between the two positioning platforms 600 is carefully designed to allow the submersible AGV 200 to move freely and stop accurately.

[0029] It is understood that the carrier used to carry the battery pack 500 is not limited to the pallet 400, but can also be other carrier platforms with positioning structures. The transport trolley is not limited to the combination of the lurking AGV 200 and the pallet trolley 300, but can also be other automated guided vehicles or conveying mechanisms. The positioning system is not limited to the positioning platform 600 driven by cylinders, but can also use other linear drive mechanisms such as motor drive and hydraulic drive to achieve the lifting and positioning functions.

[0030] Based on the aforementioned automatic battery pack detection system, this embodiment also provides an automatic battery pack detection method, combined with... Figure 5 The specific implementation of the testing process includes steps S1-S3.

[0031] S1. Position the carrier carrying the battery pack at the testing station.

[0032] The 200-type stealth AGV carries a pallet cart 300 and a pallet 400. By reading the QR code labels laid on the ground, it autonomously navigates to the inspection station. When the 200-type stealth AGV accurately moves the pallet cart 300 between the two positioning platforms 600, it sends a positioning signal to the host computer via a communication protocol.

[0033] The host computer controls the positioning cylinder of the positioning platform 600 to extend the positioning wheel and firmly lock the positioning blocks on the four moving wheels of the pallet trolley 300, thus completing the initial positioning of the pallet trolley 300 and the entire transport unit.

[0034] Subsequently, the host computer controls the lifting cylinder to operate. A wedge block is provided at the bottom of the support plate 601 to cooperate with the drive slider 603. (See [link to relevant documentation]). Figure 3 The movement of the sliding block 603, in conjunction with the wedge-shaped inclined surface, causes the pallet 601 to rise smoothly. The four lifting blocks 602 on the pallet 601 rise accordingly and precisely engage with the limiting grooves of the four limiting blocks at the bottom of the pallet 400. During this process, the pallet 400 is lifted, separating the positioning pins and posts between it and the pallet trolley 300, leaving the pallet 400 and its battery pack 500 in a lifted and suspended state. This mechanical constraint achieves precise positioning of the pallet 400 (i.e., the carrier of the battery pack), completely eliminating potential positional errors in the AGV chassis and pallet trolley, laying a solid foundation for subsequent high-precision visual inspection.

[0035] S2. Use a robot to drive the image acquisition system to scan the battery pack and obtain an image of the battery pack surface.

[0036] After the pallet 400 is precisely positioned, the host computer controls the linear module 800 and the two robots 700 to begin collaborative operation. The image acquisition system 900 includes a 2D image acquisition system and a 3D image acquisition system.

[0037] First, the 2D image acquisition system (combined with) Figure 4 The 2D image acquisition system (consisting of a 12-megapixel area array camera 901, a fixed-focus lens 902, and a ring light source 903) quickly takes pictures of the battery pack 500 and guides the robot 700 to a preset precise observation position through image recognition technology.

[0038] Subsequently, the system performs comprehensive image acquisition: (1) 2D image acquisition: Two robots 700, in coordination with the linear module 800, move to the four sides of the battery pack 500 respectively. The 2D image acquisition system takes high-resolution pictures of the labeling areas (such as nameplate barcodes, traceability codes, high voltage markings, warning labels, etc.), low-voltage plugs, high-voltage plugs, plug protective covers, vent valves and liquid cooling pipe ports on each side to obtain side images for presence detection and appearance inspection.

[0039] (2) 3D Image Acquisition: The 3D image acquisition system (consisting of a 3D line laser camera 904 and its controller) begins operation. One robot 700, carrying the 3D line laser camera 904, scans the surface of the battery pack 500's lid at a uniform speed along a predetermined trajectory with the assistance of the seventh axis of the linear module 800, acquiring high-density three-dimensional point cloud data and generating a lid image (i.e., a height difference map) covering the entire lid. Another robot 700 performs 3D scanning on the weld seams on the side of the battery pack 500 to acquire side weld seam images.

[0040] S3. Based on the image, a comprehensive judgment is made on the appearance of the battery pack.

[0041] After acquiring all image data, the host computer calls its internal detection algorithm module for comprehensive analysis and judgment. The algorithm adopts a strategy that combines traditional image processing algorithms with deep learning algorithms, using different technical paths for different detection items to achieve highly robust and accurate judgment.

[0042] For the detection of labels, protective covers, etc.: a robust contour comparison algorithm is used. Specifically, clear images of qualified products are pre-acquired under standard lighting conditions, and their precise edge contours are extracted and stored as standard templates. During online detection, the real-time captured images of the area to be inspected undergo denoising, binarization, and other preprocessing, and their edge contours are extracted as real-time contours. The presence of the target object is comprehensively determined by calculating the matching degree between the real-time contour and the standard template and comparing it with a preset threshold. If the matching degree is higher than the threshold and the key features are complete, it is determined to be present; otherwise, it is determined to be absent or abnormal.

[0043] For the detection of defects such as pits, scratches, deformations, and weld quality on the lid surface: an innovative fusion algorithm process is adopted, which combines traditional algorithm for initial screening, deep learning for fine-tuning, and feature fusion for decision-making. The specific steps are as follows: (1) Preliminary Detection Stage (Traditional Algorithm): The point cloud data obtained from 3D scanning is preprocessed. First, Gaussian filtering is used to remove noise generated during the scanning process. Then, by using reference surface fitting and difference techniques, the overall height change caused by the natural curvature of the battery pack cover or slight installation tilt is eliminated to obtain a height difference map that truly reflects the surface unevenness. Then, a rule based on multi-level threshold segmentation is applied to this map for initial screening: a higher depth threshold is set for deeper pits, and a lower depth and continuity threshold is set for minor scratches. Through binarization segmentation, combined with morphological operations (such as opening and closing operations) and connected component analysis, all potential defect areas are quickly and initially located, and the geometric features of each area, such as maximum depth, average depth, area, and aspect ratio, are extracted.

[0044] (2) Fine-grained discrimination stage (deep learning): Each potential defect region image patch (ROI) initially identified by the traditional algorithm is input into a pre-trained lightweight deep learning classification network (e.g., a simplified convolutional neural network CNN). The core task of this network is to perform fine-grained discrimination: first, it performs a real / fake judgment, distinguishing between real defects (such as dents and scratches) and fake defects caused by dust, oil stains, reflections, or slight vibrations; second, it performs type recognition, accurately classifying the specific category of the defect, such as dents, scratches, protrusions, weld slag, weld undercut, weld porosity, etc.

[0045] (3) Fusion Decision Stage: The system integrates the geometric feature information provided in the initial detection stage with the classification results and confidence information provided in the fine discrimination stage, and makes a final decision according to the preset fusion rules. For example, for a region that is judged as a "scratch" by the deep learning model and has a high confidence level, the system will further verify the "depth" feature extracted by the traditional algorithm. If the depth value of the scratch is within the extremely small range allowed by the process, it can be finally judged as qualified or ignored; otherwise, if the depth exceeds the standard, it is judged as unqualified. This fusion mechanism effectively combines the accuracy of traditional algorithms in quantitative measurement with the powerful generalization ability of deep learning in pattern recognition, significantly reducing the false positive rate and false negative rate in complex production environments.

[0046] After all the testing items are completed, the host computer generates a comprehensive testing report and controls the positioning platform 600 to descend. The AGV 200 then transports the battery pack 500 away from the workstation, and one testing cycle ends.

[0047] It is understandable that the specific configurations of the 2D and 3D image acquisition systems in an image acquisition system can be adjusted according to the requirements of detection accuracy and speed. The core of the comprehensive judgment lies in the algorithmic idea of ​​integrating traditional and deep learning, and its specific network structure and feature fusion rules can be optimized according to the actual application scenario.

[0048] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0049] The embodiments described above are merely illustrative of several implementations of the present invention, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of the present invention, and these all fall within the protection scope of the present invention. Therefore, the protection scope of this invention patent should be determined by the appended claims.

Claims

1. An automatic detection method for battery packs, characterized in that, It includes the following steps: S1. Position the carrier carrying the battery pack at the testing station; S2. Use a robot to drive the image acquisition system to scan the battery pack and obtain an image of the battery pack surface; S3. Based on the image, make a comprehensive judgment on the appearance of the battery pack.

2. The automatic battery pack detection method according to claim 1, characterized in that, Step S1 includes: Initial positioning of the transport trolley transporting the aforementioned vehicle; The vehicle is separated from the transport trolley and lifted to achieve precise positioning of the vehicle.

3. The automatic battery pack detection method according to claim 1, characterized in that, In step S2, two robots are set up, and each robot drives an image acquisition system to move. The linear module drives the two robots to adjust their positions in order to achieve blind-spot-free detection of the battery pack.

4. The automatic battery pack detection method according to claim 3, characterized in that, In step S2, the image acquisition system includes at least a 2D image acquisition system for visual positioning and two-dimensional feature detection, and a 3D image acquisition system for three-dimensional shape scanning; the 2D image acquisition system guides the robot to take pictures of the side of the battery pack to obtain a side image of the battery pack; the 3D image acquisition system guides the robot to work with the linear module to scan the surface of the battery pack cover to obtain an image of the battery pack cover, and guides the robot to scan the side welds of the battery pack to obtain side weld images.

5. The automatic battery pack detection method according to claim 4, characterized in that, In step S3, based on the side image, the presence or absence of the label, plug-in protective cover, vent valve, and liquid cooling pipe opening is detected.

6. The automatic battery pack detection method according to claim 4, characterized in that, In step S3, based on the image of the box lid, dents, scratches, and deformation defects on the surface of the box lid are detected, specifically including: Based on the point cloud data acquired by the 3D image acquisition system, a height difference map is generated and potential defect areas are located using traditional image processing algorithms, including Gaussian filtering, reference surface difference, and threshold segmentation. The data of the potential defect area is input into a deep learning classification network to determine the authenticity and / or type of the defect; By combining the geometric feature information output from the primary detection stage with the classification confidence information output from the fine discrimination stage, a final defect determination is made.

7. The automatic battery pack detection method according to claim 4, characterized in that, In step S3, the quality of the side weld is detected based on the side weld image, specifically including: Based on the point cloud data acquired by the 3D image acquisition system, a height difference map is generated and potential defect areas are located using traditional image processing algorithms, including Gaussian filtering, reference surface difference, and threshold segmentation. The data of the potential defect area is input into a deep learning classification network to determine the authenticity and / or type of the defect; By combining the geometric feature information output from the primary detection stage with the classification confidence information output from the fine discrimination stage, a final defect determination is made.

8. An automatic battery pack detection system, characterized in that, It is used to implement the automatic battery pack detection method as described in any one of claims 1-7, comprising: Frame (100); A positioning system, located below the frame (100), is used to position the vehicle carrying the battery pack; The robot (700) is mounted on the frame (100) via a linear module (800); Image acquisition system (900), an image acquisition device installed at the end of the robot (700); The host computer is connected in communication with the image acquisition system (900) and the robot (700) to control the detection process and execute detection algorithms based on image data.

9. The automatic battery pack detection system according to claim 8, characterized in that, The positioning system includes at least two positioning platforms (600) arranged opposite to each other; the positioning platform (600) includes a positioning mechanism for initial positioning of the transport trolley and a lifting mechanism for lifting and precisely positioning the vehicle; the positioning mechanism includes a positioning cylinder and a positioning wheel driven therefrom, and a positioning block for engaging with the moving platform; the lifting mechanism includes a lifting cylinder, a pallet (601) driven therefrom and a lifting block (602) disposed on the pallet (601), the lifting block (602) being adapted to engage with a limiting groove at the bottom of the vehicle during lifting.

10. The automatic battery pack detection system according to claim 8, characterized in that, The image acquisition system (900) includes a 2D image acquisition system and a 3D image acquisition system; the 2D image acquisition system includes an area array camera (901), a fixed-focus lens (902), and a ring light source (903); the 3D image acquisition system includes a 3D line laser camera (904) and its controller.

Images