Vehicle body welding quality visual inspection system and method

Abstract

The application provides a vehicle body welding quality visual detection system and method, the device comprises: a transfer trolley, carrying a white body to be detected; a jacking positioning tool, adjusting the posture of the white body and locking the position; a multi-axis robot, the end of which is equipped with an industrial camera, moving according to a preset trajectory, collecting images of welding points on the white body; a control device, controlling the lifting and the motion trajectory of the multi-axis robot; an image processing device, receiving the welding point images collected by the industrial camera for defect identification and generating a detection result. The method comprises: entering a detection station; lifting and locking the white body to a first detection posture; moving and collecting all welding point images of the first side of the white body by the multi-axis robot; after the direction of the white body is adjusted, collecting all welding point images of the second side of the white body; the image processing device receives all welding point images, sequentially performs defect identification, panoramic image splicing and defect labeling, and outputs a detection result.

Description

Technical Field

[0001] This invention relates to the field of automotive body-in-white welding inspection technology, and more specifically, to a visual inspection system and method for body welding quality. Background Technology

[0002] In automobile manufacturing, the total number of weld points in a body-in-white typically ranges from 4,000 to 7,000. The vehicle prototyping stage is generally characterized by short production cycles, low production line pace, and poor component matching precision. The industry still primarily relies on manual welding combined with manual visual quality inspection. Manual welding is prone to incomplete welds and weak welds, and the consistency of manual spot welding is poor, resulting in a high incidence of defects such as porosity, cracks, and weld beads. Regarding quality inspection, a skilled operator needs approximately two hours to completely inspect a body-in-white, which is inefficient. Inspection results are highly dependent on the operator's experience, sense of responsibility, and mental state, making it difficult to standardize judgment criteria and leading to a high risk of missed or incorrect inspections. Furthermore, the confined interior space of the vehicle body and the sharp edges of sheet metal can easily injure workers, posing significant safety hazards.

[0003] Current weld quality inspection methods each have their limitations. Manual visual inspection is inefficient and insufficient for detecting minute defects. While ultrasonic testing can detect internal defects, it is difficult to automate, requires coupling agents, and has stringent requirements for weld surface flatness, making it unsuitable for the irregular appearance of welds in the prototyping stage. Metallographic destructive testing, which involves cutting and grinding for microscopic analysis, offers high precision but damages the workpiece, is costly and time-consuming, and cannot achieve real-time online inspection on the production line. In recent years, machine vision-based weld inspection solutions have received widespread attention, but most remain in the laboratory stage: constrained by factors such as camera field of view, robot accessibility, and the robustness of defect recognition algorithms in complex sheet metal environments, there is still no low-cost, fully automated vision inspection equipment suitable for prototyping production lines and covering the entire vehicle body for practical application. Summary of the Invention

[0004] This invention provides a visual inspection system and method for vehicle body welding quality, aiming to solve the problems of low efficiency, poor consistency and safety risks of manual weld inspection, and to improve inspection efficiency, accuracy and operational safety.

[0005] To achieve the above objectives, the present invention provides a visual inspection system for vehicle body welding quality, comprising: The testing station is equipped with floor rails inside. A transfer trolley is slidably mounted on the ground guide rail and carries the body-in-white to be inspected; A lifting and positioning fixture is installed in the inspection station. The lifting and positioning fixture includes a lifting mechanism and a locking mechanism. The lifting mechanism is vertically and vertically installed in the inspection station to adjust the posture of the body-in-white. The locking mechanism locks the position of the lifting mechanism after the posture of the body-in-white is adjusted to the correct position. A multi-axis robot is set up at the inspection station. The end of the multi-axis robot is equipped with an industrial camera. The multi-axis robot moves according to a preset trajectory and captures images of weld points on the white body through the industrial camera. The control device is communicatively connected to the lifting mechanism of the lifting and positioning fixture and the multi-axis robot, respectively, and outputs action commands to control the lifting and lowering of the lifting mechanism and the motion trajectory of the multi-axis robot. An image processing device is communicatively connected to the industrial camera, receives weld point images captured by the industrial camera for defect identification, and generates detection results.

[0006] In one embodiment, the lifting mechanism includes six sets of cylinders, which are arranged according to the support points at the bottom of the body-in-white; The control device is independently connected to the control valves of the six sets of cylinders to change the tilt posture of the body-in-white by controlling the extension and retraction stroke of each of the six sets of cylinders.

[0007] In one embodiment, the control device has preset pose adjustment logic, the pose adjustment logic including: Upon receiving a single-sided weld point detection command, the first designated combination of the six sets of cylinders is controlled to extend to lift the body-in-white to the first detection posture. Upon receiving a weld point detection command from the other side, the second designated combination of the six cylinders is controlled to extend, thereby lifting the body-in-white to the second detection posture.

[0008] In one embodiment, the bottom of the transfer trolley is provided with guide wheels and the side is provided with locking pins; The ground guide rail is provided with a guide groove that rolls with the guide wheel, and the detection station is provided with a locking hole that engages with the locking pin, so as to limit the transfer trolley to the detection origin.

[0009] In one embodiment, the multi-axis robot is an eight-axis collaborative robot.

[0010] In one embodiment, the image processing apparatus includes: The defect identification module calls the instance segmentation network model to identify defects in the weld point image; The image stitching module uses a feature matching algorithm to stitch together multiple continuously acquired local weld point images into a panoramic image of the white body; The annotation module performs coordinate matching between the panoramic image of the white body and the preset three-dimensional digital model of the white body, and maps and annotates the defect locations on the three-dimensional digital model.

[0011] In one embodiment, the instance segmentation network model is a YOLO-based instance segmentation model, and the feature matching algorithm is the SIFT algorithm.

[0012] In one embodiment, the image processing device further includes a data management module that stores defect detection data in layers according to the welding process of the body-in-white, and summarizes and generates a full-process quality inspection report.

[0013] A visual inspection method for vehicle body welding quality based on the system includes the following steps: S1, the transfer trolley carrying the white body enters the inspection station along the ground guide rail, and the transfer trolley is limited to the inspection origin; S2, the control device drives the lifting mechanism of the lifting and positioning fixture to extend according to the first detection posture parameters, lifts the white body and locks it to the first detection posture; S3, the control device controls the multi-axis robot to drive the industrial camera to move along the preset first spatial trajectory and collect images of all weld points on the first side of the white body; S4, after the direction of the white body is reversed, the control device drives the lifting mechanism to lift the white body and lock it to the second detection posture. S5, the control device controls the multi-axis robot to drive the industrial camera to move along the preset second spatial trajectory to collect images of all weld points on the second side of the white body; S6, the image processing device receives all weld point images from the first side and the second side of the body-in-white, performs defect identification and panoramic image stitching in sequence, and matches the stitching result with the three-dimensional digital model of the body-in-white to complete defect labeling and output the inspection result.

[0014] In one embodiment, the defect marking in step S6 specifically includes: registering the pixel coordinate system in the stitched panoramic image of the body-in-white with the spatial coordinate system of the three-dimensional digital model of the body-in-white, and automatically marking the location of the missing weld or defective weld at the corresponding coordinate point of the three-dimensional digital model.

[0015] The present invention has the following beneficial effects: 1. Reduce personnel barriers and operational risks: The system completes the inspection fully automatically, and operators only need basic operation training to start working, without the need for many years of welding quality inspection experience; no personnel need to enter the interior cavity of the vehicle body to work, completely eliminating the safety hazards of sheet metal scratches, greatly reducing the intensity of manual labor, and eliminating the inspection errors caused by subjective human judgment.

[0016] 2. Significantly improved inspection efficiency: Traditional manual vehicle inspection takes 2 hours, while this fully automated system only takes 40 minutes, greatly shortening the inspection cycle and adapting to the rapid iteration needs of prototype vehicles, eliminating the problem of missed inspections by humans.

[0017] 3. Comprehensive inspection coverage, enabling non-destructive online quality inspection. It can simultaneously inspect all spot welds and stud welds on the body-in-white, automatically count the total number of welds, and identify various welding defects. The entire set of equipment does not require damage to the workpiece, providing a stable, real-time, and non-destructive automated quality inspection platform for the trial production line. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of the structure of a visual inspection system for vehicle body welding quality, excluding the transfer trolley, according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the structure of the detection station and the transfer trolley according to an embodiment of the present invention; Figure 3 This is an exploded structural diagram of a transfer trolley according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the lifting and positioning fixture according to an embodiment of the present invention; Figure 5 This is a schematic diagram of the multi-angle structure of a lifting and positioning fixture according to an embodiment of the present invention; Figure 6 This is an exploded structural diagram of a lifting and positioning fixture according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of a multi-axis robot according to an embodiment of the present invention; Figure 8 This is a flowchart illustrating a visual inspection method for vehicle body welding quality according to an embodiment of the present invention.

[0019] 1 is the ground guide rail; 2 is the transfer trolley; 21 is the guide wheel; 22 is the locking pin; 23 is the slide rail; 24 is the first positioning sensor; 3 is the lifting and positioning fixture; 31 is the cylinder; 32 is the slide rail; 33 is the manual pin; 34 is the buffer; 35 is the second positioning sensor; 36 is the throttle valve; 37 is the lifting fulcrum; 4 is the multi-axis robot; 41 is the robot's seven-axis guide rail; 42 is the robot's seven-axis drive motor; 43 is the industrial camera; 44 is the robot's lifting driver. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the invention.

[0021] Figures 1 to 7This is a schematic diagram of the structure of a visual inspection system for vehicle body welding quality according to an embodiment of the present invention. The visual inspection system for vehicle body welding quality provided by the present invention consists of a hardware execution layer, an algorithm processing layer, and a data interaction layer. The diagram shows the relative positions of the lifting and positioning fixture 3, the ground guide rail 1, and the eight-axis collaborative robot within the workstation. The hardware execution layer of this application mainly includes a transfer trolley 2, the lifting and positioning fixture 3, and an eight-axis collaborative robot equipped with an industrial camera 43. The transfer trolley 2 moves along the ground guide rail 1 laid within the inspection workstation to carry and transport the vehicle body to be inspected. The lifting and positioning fixture 3 is fixedly installed within the workstation, responsible for lifting the vehicle body to different inspection postures and reliably locking it. The eight-axis collaborative robot is equipped with an industrial camera 43 at its end, which can acquire weld point images according to a preset trajectory. The control device communicates with the cylinder control valve of the lifting and positioning fixture 3 and the controller of the multi-axis robot 4, respectively, to coordinate the actions of each component. The image processing device receives the acquired images and, through its embedded defect recognition module, image stitching module, annotation module, and data management module, ultimately generates a full-process quality inspection report with defect location annotations.

[0022] In one embodiment, such as Figure 2 and Figure 3 As shown, the transfer trolley 2 is a specially designed white-body transport vehicle. Figure 2 This is a structural diagram of the transfer trolley 2. Figure 3 This is an exploded view of the assembly of the transfer trolley 2. The trolley has a slide rail 23 at its bottom and guide wheels 21 at its four corners. The slide rail 23 is slidably supported on the upper surface of the ground guide rail 1 to bear vertical loads. The guide wheels 21 can roll into guide grooves on the side of the ground guide rail 1 for lateral limiting and to ensure the trolley moves in a straight line. A retractable locking pin 22 is provided on the side of the trolley. When the trolley is pushed to the detection origin position within the workstation, the locking pin 22 can be inserted into the locking hole at the corresponding position of the workstation. Combined with the first positioning sensor 24, mechanical positioning is achieved, ensuring the repeatability accuracy of each placement of the body-in-white. The upper part of the transfer trolley 2 has a support platform that mates with the bottom of the body-in-white, directly supporting the body. The transfer trolley 2 adopts a welded frame structure, possessing sufficient rigidity and load-bearing capacity.

[0023] In one embodiment, such as Figure 1 , Figure 4 , Figure 5 , Figure 6 As shown, the lifting and positioning fixture 3 mainly consists of six lifting mechanisms, each including a cylinder 31. The six lifting mechanisms are as follows: Figure 1 As shown, it is vertically arranged on the load-bearing base plate of the testing station. Figure 4 , Figure 5 This is a 3D view of the overall tooling. Figure 6This is an assembly drawing. The distribution of the six lifting mechanisms matches the support points at the bottom of the body-in-white. Each cylinder 31 has a support plate at the top of its piston rod for direct contact with the positioning points at the bottom of the body. The intake and exhaust circuits of the cylinders 31 are independently controlled by electromagnetic control valves. The control device can individually adjust the extension and retraction stroke of each group of cylinders 31 by sending commands to each electromagnetic valve, thereby adjusting and locking the tilt posture of the body-in-white. The fixture also integrates a buffer 34, a second positioning sensor 35, and a slide rail 32 to ensure a smooth and controlled lifting process and to provide real-time feedback on the lifting height. The locking mechanism includes a lifting fulcrum 37 and a manual pin 33. The chassis of the body-in-white has a positioning hole. The upper end of the lifting fulcrum 37 can pass through the clearance notch of the transfer trolley 2 and be inserted into the positioning hole to fix the body-in-white to the lifting positioning fixture 3. When the cylinder 31 is lifted to the correct position, the manual pin 33 can be inserted to form a physical lock and maintain a stable posture. The second positioning sensor 35 is used to detect the presence or position of an object, confirm whether the vehicle body or tooling component has reached the predetermined working area, and provide a trigger signal to the system. The cylinder 31 is also equipped with a throttle valve 36, which can adjust the intake and exhaust flow of the cylinder 31, thereby precisely controlling the lifting and lowering speed, making the lifting action smooth and stable, and preventing sudden changes in vehicle body posture due to air pressure fluctuations.

[0024] Specifically, to achieve dual-side sequential detection of the body-in-white, the control device has a preset posture adjustment logic: when a single-side weld point detection command is received, the first designated combination of the six sets of cylinders 31 is controlled, for example, four sets of cylinders 31 extend simultaneously to lift the body-in-white to the first detection posture; when a weld point detection command is received on the other side, usually after the body has turned around, the second designated combination of the six sets of cylinders 31 is controlled, for example, four sets of cylinders 31 extend simultaneously to lift the body-in-white to the second detection posture. The first and second designated combinations are different and partially overlap to adapt to the center of gravity distribution when the body is facing different directions.

[0025] In one embodiment, such as Figure 7 As shown, the multi-axis robot 4 is equipped with a seven-axis guide rail 41, allowing it to move along this rail. The multi-axis robot 4 also includes a seven-axis drive motor 42 and a robot lifting actuator 44. The seven-axis drive motor 42 drives the multi-axis robot 4 to perform large-scale coarse positioning across the entire length of the robot body. The main body of the multi-axis robot 4 is a redundant degree-of-freedom robotic arm with at least seven rotary joints. Its end effector integrates a high-resolution industrial camera 43 and a ring light source via a tooling connection. The robot lifting actuator 44 drives the robotic arm to achieve multi-directional imaging. Including the linear motion axes along the seven-axis guide rail 41, the total number of motion axes in the entire module is no less than eight, thus it can be called an eight-axis collaborative robot.

[0026] When the multi-axis robot 4 moves along the seven-axis guide rail 41, it can be quickly transferred to different areas of the vehicle body. The eight-axis collaborative robot body can use its joints to adjust the end effector posture in narrow spaces, allowing the industrial camera 43 to reach into parts that are difficult for humans to access, such as crossbeams and cavities, at the optimal angle, so as to achieve clear images of every weld point without blind spots. During system debugging, the operator plans a shooting trajectory covering all weld points on the vehicle body for the robot through offline teaching, and sets the shooting posture and focal length parameters for each point. In the actual inspection process, the eight-axis collaborative robot first moves to the predetermined working area according to the program instructions, and then the eight-axis collaborative robot body arrives at each weld point position sequentially along the taught trajectory, driving the camera to automatically complete image acquisition.

[0027] In this application, the operator is only responsible for pushing and locking the transfer trolley 2 that carries the vehicle body, and adjusting the vehicle body direction according to the system prompts; all other actions are completed automatically by the system.

[0028] In one embodiment, the control device is centered around a programmable logic controller (PLC) and equipped with a human-machine interface touchscreen. The PLC establishes communication connections with each solenoid valve, position sensor, locking pin 22, and robot controller of the lifting and positioning fixture 3. The operator only needs to select the vehicle model to be inspected, the inspection mode, etc. through the interactive interface, and the control device can automatically call the corresponding cylinder 31 combination logic and robot motion program to sequentially complete all action sequences such as vehicle body lifting, robot start-up, and safety interlocking, without manual intervention.

[0029] In one embodiment, the image processing device is a high-performance industrial computer equipped with a GPU (Graphics Processing Unit) running pre-installed vision algorithm software. This device communicates with the industrial camera 43 via Gigabit Ethernet to receive weld joint images in real time. It mainly includes the following algorithm modules: The defect recognition module incorporates a weld joint defect recognition model trained using YOLO (You Only Look Once) instance segmentation networks, such as YOLOv8-seg (YOLO version 8 segmentation model). Before model training, LabelMe (an open-source image annotation tool) is used to perform pixel-level annotation on a massive number of weld joint sample images. The annotation categories cover typical defects such as cold welds, missing welds, porosity, and weld beads. The annotated data is then converted into the COCO (General Target Context) standard format dataset, and data augmentation processing such as image filtering, rotation, and scaling is applied. After training convergence, this module can accurately segment weld joint regions and automatically identify defect types against complex sheet metal texture backgrounds.

[0030] The image stitching module addresses the limitation of the industrial camera's single-frame field of view, which only captures a very small local area for each weld point. This module uses the Scale Invariant Feature Transform (SIFT) algorithm to extract feature points from each local image, then matches and fuses them to automatically stitch together hundreds of continuously acquired images into a complete panoramic image of the white body.

[0031] The annotation module performs coordinate registration between the stitched panoramic image and the pre-imported 3D digital model of the white body. By utilizing the mapping relationship between the image pixel coordinates and the spatial coordinates of the digital model, it automatically marks the information and type of missing weld points or specific defective weld points at the corresponding weld point coordinate positions on the 3D digital model, generating an intuitive 3D quality distribution map.

[0032] The data management module strictly follows the welding process of the body-in-white, storing original images, inspection data and defect records in sequence and layer according to the side panel assembly, floor assembly, door and hood assembly, etc., and finally summarizing and outputting a complete quality inspection report for the entire process, supporting historical traceability and statistical analysis.

[0033] Figure 8 This is a flowchart illustrating a visual inspection method for vehicle body welding quality according to an embodiment of the present invention. The visual inspection method for vehicle body welding quality specifically includes: First, the operator pushes the transfer trolley 2 carrying the white body into the inspection station along the ground guide rail 1. When the locking pin 22 on the transfer trolley 2 is inserted into the station locking hole and the first positioning sensor 24 is triggered, the indicator light lights up, confirming that the transfer trolley 2 has been limited to the inspection origin.

[0034] The operator closes the workstation door and presses the reset start button. After receiving the confirmation signal, the control device automatically controls the four sets of cylinders 31 (such as No. 1 to No. 4) of the first specified combination to lift up the body-in-white and mechanically lock it according to the single-sided detection instruction, such as detecting the left side first. At this time, the body is fixed in the first detection posture.

[0035] Subsequently, the multi-axis robot 4 returns to its zero position and moves to the initial working area. The eight-axis collaborative robot moves point by point along the preset first spatial trajectory. At each welding point, the industrial camera 43 automatically focuses and takes a picture, and transmits the image back to the image processing device for buffering in real time. After all points in a single area have been collected, the eight-axis collaborative robot moves to the next area to continue collecting pictures until all welding point images on the first side of the body-in-white have been collected.

[0036] The system displays a light and a pop-up window indicating that the first side data acquisition is complete and requests the operator to adjust the vehicle's posture. The operator pushes the transfer trolley 2 back into the workstation, unlocks the manual pin 33, and the control device lowers the cylinder 31, causing the vehicle to fall back onto the transfer trolley 2. The operator then removes the transfer trolley 2, rotates the vehicle 180° outside the workstation, pushes it back into the workstation along the guide rail, and locks the trolley. The control device then executes the other side detection command, controlling the four sets of cylinders 31 (e.g., numbers 3 to 6) of the second specified combination to lift and lock the vehicle, causing it to enter the second detection posture.

[0037] The robot restarted and collected images of all weld points on the second side of the white body along the preset second spatial trajectory.

[0038] After all data points are collected, the system automatically displays a pop-up notification indicating that the inspection is complete. The image processing device sequentially performs defect identification and panoramic image stitching on all collected images. It then matches the stitched panoramic image with the 3D digital model, automatically marking the type and location of any missing or defective solder joints at the corresponding coordinate points on the model. Finally, it outputs a quality inspection report with a 3D defect distribution, statistical charts, and a detailed defect list. Operators can view the inspection results at any time through the main interactive interface.

[0039] This invention effectively solves the problems of low efficiency, poor consistency and safety risks in existing manual weld point inspection by organically combining the above-mentioned hardware structure and algorithm process. It significantly improves inspection efficiency, accuracy and operational safety, and is especially suitable for the fast-paced, multi-model online quality inspection needs of body-in-white in the whole vehicle trial production stage.

[0040] In the description of this application, it should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. For ease of description, the dimensions of the various parts shown in the drawings are not drawn to actual scale. Techniques, methods, and devices known to those skilled in the art may not be discussed in detail, but where appropriate, such techniques, methods, and devices should be considered part of the specification. In all examples shown and discussed herein, any specific values ​​should be interpreted as merely exemplary and not as limitations. Therefore, other examples of exemplary embodiments may have different values. It should be noted that similar reference numerals and letters in the following drawings denote similar items; therefore, once an item is defined in one drawing, it need not be further discussed in subsequent drawings.

[0041] It should be noted that, in this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. It should also be noted that the scope of the methods and apparatuses in the embodiments of this application is not limited to performing functions in the order shown or discussed, but may also include performing functions substantially simultaneously or in the reverse order, depending on the functions involved. For example, the described methods may be performed in a different order than described, and various steps may be added, omitted, or combined. Furthermore, features described with reference to certain examples may be combined in other examples.

[0042] The embodiments described above are merely further illustrations of the present invention and are not intended to limit the present invention in any other way. The present invention may have many other embodiments. Without departing from the spirit and essence of the present invention, those skilled in the art can make various corresponding modifications and changes based on the present invention, but all such modifications and changes should fall within the protection scope of the present invention.

Claims

1. A visual inspection system for vehicle body welding quality, characterized in that, include: The testing station is equipped with floor rails inside. A transfer trolley is slidably mounted on the ground guide rail and carries the body-in-white to be inspected; A lifting and positioning fixture is installed in the inspection station. The lifting and positioning fixture includes a lifting mechanism and a locking mechanism. The lifting mechanism is vertically and vertically installed in the inspection station to adjust the posture of the body-in-white. The locking mechanism locks the position of the lifting mechanism after the posture of the body-in-white is adjusted to the correct position. A multi-axis robot is set up at the inspection station. The end of the multi-axis robot is equipped with an industrial camera. The multi-axis robot moves according to a preset trajectory and captures images of weld points on the white body through the industrial camera. The control device is communicatively connected to the lifting mechanism of the lifting and positioning fixture and the multi-axis robot, respectively, and outputs action commands to control the lifting and lowering of the lifting mechanism and the motion trajectory of the multi-axis robot. An image processing device is communicatively connected to the industrial camera, receives weld point images captured by the industrial camera for defect identification, and generates detection results.

2. The vehicle body welding quality visual inspection system according to claim 1, characterized in that, The lifting mechanism includes six sets of cylinders, which are arranged according to the support points at the bottom of the body-in-white. The control device is independently connected to the control valves of the six sets of cylinders to change the tilt posture of the body-in-white by controlling the extension and retraction stroke of each of the six sets of cylinders.

3. The vehicle body welding quality visual inspection system according to claim 2, characterized in that, The control device has preset pose adjustment logic, which includes: Upon receiving a single-sided weld point detection command, the first designated combination of the six sets of cylinders is controlled to extend to lift the body-in-white to the first detection posture. Upon receiving a weld point detection command from the other side, the second designated combination of the six cylinders is controlled to extend, thereby lifting the body-in-white to the second detection posture.

4. The vehicle body welding quality visual inspection system according to claim 1, characterized in that, The bottom of the transfer trolley is equipped with guide wheels, and the side is equipped with locking pins; The ground guide rail is provided with a guide groove that rolls with the guide wheel, and the detection station is provided with a locking hole that engages with the locking pin, so as to limit the transfer trolley to the detection origin.

5. The vehicle body welding quality visual inspection system according to claim 1, characterized in that, The multi-axis robot is an eight-axis collaborative robot.

6. The vehicle body welding quality visual inspection system according to claim 1, characterized in that, The image processing device includes: The defect identification module calls the instance segmentation network model to identify defects in the weld point image; The image stitching module uses a feature matching algorithm to stitch together multiple continuously acquired local weld point images into a panoramic image of the white body; The annotation module performs coordinate matching between the panoramic image of the white body and the preset three-dimensional digital model of the white body, and maps and annotates the defect locations on the three-dimensional digital model.

7. The vehicle body welding quality visual inspection system according to claim 6, characterized in that, The instance segmentation network model is a YOLO series instance segmentation model, and the feature matching algorithm is the SIFT algorithm.

8. The vehicle body welding quality visual inspection system according to claim 6, characterized in that, The image processing device also includes a data management module, which stores defect detection data in layers according to the welding process of the body-in-white, and summarizes and generates a full-process quality inspection report.

9. A visual inspection method for vehicle body welding quality based on the system described in any one of claims 1-8, characterized in that, Includes the following steps: S1, the transfer trolley carrying the white body enters the inspection station along the ground guide rail, and the transfer trolley is limited to the inspection origin; S2, the control device drives the lifting mechanism of the lifting and positioning fixture to extend according to the first detection posture parameters, lifts the white body and locks it to the first detection posture; S3, the control device controls the multi-axis robot to drive the industrial camera to move along the preset first spatial trajectory and collect images of all weld points on the first side of the white body; S4, after the direction of the white body is reversed, the control device drives the lifting mechanism to lift the white body and lock it to the second detection posture. S5, the control device controls the multi-axis robot to drive the industrial camera to move along the preset second spatial trajectory to collect images of all weld points on the second side of the white body; S6, the image processing device receives all weld point images from the first side and the second side of the body-in-white, performs defect identification and panoramic image stitching in sequence, and matches the stitching result with the three-dimensional digital model of the body-in-white to complete defect annotation and output the inspection result.

10. The visual inspection method for vehicle body welding quality according to claim 9, characterized in that, The defect annotation in step S6 specifically includes: registering the pixel coordinate system in the stitched panoramic image of the body-in-white with the spatial coordinate system of the three-dimensional digital model of the body-in-white, and automatically marking the location of the missing weld or defective weld at the corresponding coordinate point of the three-dimensional digital model.

Images