Self-adaptive automatic paint spraying production line based on visual recognition

Abstract

The invention discloses a self-adaptive automatic paint spraying production line based on visual recognition. The production line comprises a bearing device, a driving device and a paint spraying device, wherein the bearing device is used for bearing and driving workpieces to advance in a preset X direction; the visual identification system is arranged at an upstream station of the servo drive transmission line and comprises a plurality of high-definition cameras and a strong light directional projection light source, the high-definition cameras are annularly arranged in a YZ plane perpendicular to the X direction, the strong light directional projection light source is used for projecting directional light beams to the surface of a workpiece, and a non-zero included angle is formed between the optical axis direction of the strong light directional projection light source and the X direction; shadow is generated on the surface of the workpiece due to the shape of the workpiece; and the plurality of high-definition cameras are configured to collect a workpiece surface image containing the shadow, and high-precision three-dimensional reconstruction is realized through complementation of visual shadow analysis and laser radar three-dimensional scanning in combination with servo motion data. The two-dimensional contour provided by the visual system provides spatial constraint for point cloud data processing, and the problems of point cloud noise and missing are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of industrial automation technology, and in particular to an adaptive automatic painting production line based on visual recognition. Background Technology

[0002] In the field of industrial painting, especially in manufacturing industries such as steel structures and heavy machinery, automated painting of workpieces has long faced challenges. Currently, mainstream technologies include manual spraying, automated painting equipment with fixed trajectories, and adaptive painting systems based on vision recognition. For example, prior art document CN120620188A discloses a robot painting system based on machine vision. This system acquires workpiece images through a camera, uses image processing algorithms (such as edge detection and threshold segmentation) to identify the two-dimensional contour of the workpiece, and generates a painting path. This technology improves basic adaptability by replacing manual intervention with visual sensing, but its technical solution still has the following drawbacks: it relies on two-dimensional visual images to analyze the workpiece contour and lacks accurate acquisition of three-dimensional spatial data. For workpieces with complex curved surfaces, irregular structures, or depth variations (such as steel structure welds and uneven surfaces), two-dimensional images cannot reproduce the true three-dimensional shape, resulting in a mismatch between the generated painting trajectory and the actual surface of the workpiece, which can easily lead to missed paint or wasted paint. At the same time, this solution only uses ordinary vision sensors, which are easily affected by ambient light, workpiece surface reflection, or paint color. For example, dark-colored workpieces or highly reflective surfaces can lead to distortion in image feature extraction, and the system lacks robustness due to the absence of complementary sensing technologies such as multispectral or lidar.

[0003] Therefore, there is an urgent need in this field for an automated painting production line that can integrate multi-sensor data for accurate modeling, adapt to complex workpiece shapes, and have closed-loop quality inspection and automatic compensation functions, in order to solve the shortcomings of existing technologies in terms of accuracy, robustness and full-process automation. Summary of the Invention

[0004] To address the above shortcomings, this invention provides a vision-based adaptive automatic painting production line. Through multi-sensor data fusion and AI algorithms, it achieves automatic generation and adjustment of the workpiece painting trajectory, improving painting quality and efficiency. Specific technical solutions include: Servo drive transmission line, used to carry and drive the workpiece to travel along a preset X direction; A visual recognition system, located at the upstream station of the servo drive transmission line, includes multiple high-definition cameras arranged in a ring in the YZ plane perpendicular to the X direction, and a high-intensity directional projection light source for projecting directional beams onto the workpiece surface. The optical axis of the high-intensity directional projection light source forms a non-zero angle with the X direction, so that the workpiece surface casts a shadow due to its own shape. The multiple high-definition cameras are configured to capture images of the workpiece surface containing the shadow. A lidar system, located downstream of the visual recognition system, includes multiple lidars arranged in a ring within the YZ plane, used to acquire three-dimensional dot cloud data of the workpiece. The data processing unit, communicatively connected to the visual recognition system, the lidar system, and the servo drive transmission line, is configured as follows: Receive the workpiece surface image and extract the two-dimensional contour data of the workpiece in the YZ plane by analyzing the shape of the shadow; The three-dimensional point cloud data is received, and the two-dimensional contour data and the three-dimensional point cloud data are spatially registered and fused. The three-dimensional point cloud data is optimized using the two-dimensional contour data as a spatial constraint to generate a corrected three-dimensional point cloud. Based on the travel data fed back from the servo drive transmission line, the length information of the workpiece in the X direction is obtained through integration calculation; By combining the corrected 3D dot cloud with the length information, a complete 3D digital model of the workpiece is constructed. A painting robot system is located downstream of the lidar system and is communicatively connected to the data processing unit. The data processing unit is further configured to: analyze the complete three-dimensional digital model using AI algorithms, automatically generate a paint trajectory line that matches the surface morphology of the workpiece, and control the paint robot system to perform paint spraying operations on the moving workpiece according to the paint trajectory line.

[0005] Preferably, the strong light directional projection light source is a multispectral light source capable of emitting at least two different wavelengths of visible light; the data processing unit is configured to: control the strong light directional projection light source to switch the emitted wavelengths, and based on the images acquired by the multiple high-definition cameras under different wavelength illuminations, select the image sequence with the highest imaging contrast for the extraction of the two-dimensional contour data.

[0006] Preferably, the number of high-definition cameras arranged in a ring is no less than three, and they are distributed at equal angular intervals to synchronously acquire panoramic images of the workpiece in the YZ plane; the number of lidars arranged in a ring is no less than three. In scenarios where the painting robot system can only handle a 180-degree working surface or the workpiece is only suitable for single-sided painting due to the limitations of the fixed fixture, the multiple high-definition cameras and lidars arranged in a ring can cover only a 180-degree angle, process only one side of the workpiece at a time, and after the single-sided operation is completed, the workpiece is flipped over and the operation is repeated.

[0007] Preferably, the data processing unit is configured to optimize the three-dimensional point cloud data in the following manner: Filtering: Identify and remove three-dimensional point cloud data points located outside the area defined by the two-dimensional contour data as outliers; Denoising: Statistical filtering is performed on the three-dimensional point cloud data within the area defined by the two-dimensional contour data to remove discrete noise points; Completeness: For areas where point cloud data is missing due to occlusion, triangular facet interpolation is performed using the boundaries defined by the two-dimensional contour data to generate a complete workpiece surface model.

[0008] Preferably, the drive end of the servo drive transmission line integrates a high-precision encoder for real-time feedback of its travel speed and position; the length information of the workpiece in the X direction is calculated using the following formula: L = ∫v(t)dt Where L is the workpiece length, v(t) is the travel speed time function of the servo drive transmission line, and the integration interval is the entire detection period of the workpiece passing through the vision recognition system and the lidar system.

[0009] Preferably, the AI ​​algorithm is a supervised learning deep learning neural network model, and the training data of the model includes a large number of workpiece 3D model samples and their corresponding optimal spraying path labels; the process of generating the spraying trajectory line includes: taking the complete 3D digital model and the real-time travel speed of the servo drive transmission line as input to the model, and the output of the model is a series of control command sequences that define the spatial motion path, switching timing and attitude angle of the spray gun.

[0010] Preferably, the paint spray trajectory is specifically defined as including a three-dimensional spatial coordinate point sequence P={P1, P2,..., Pn}, and a set of control parameters Ci associated with each coordinate point Pi in the sequence, wherein Ci includes at least: the on / off state of the spray gun, the TCP attitude angle of the spray gun, and the paint flow rate value.

[0011] Preferably, the visual recognition system is further configured to perform a quality re-inspection of the painted workpiece after the painting robot system has completed the painting operation; The servo drive transmission line is configured to reverse the flow of the painted workpiece from the painting station back to the detection area of ​​the vision recognition system. The high-intensity directional projection light source is switched to uniform diffuse illumination mode during re-inspection; The data processing unit is further configured to: The multiple high-definition cameras are used to capture re-inspection images of the painted workpiece surface; The re-inspection image is compared with the preset qualified paint surface standard by an image recognition algorithm to identify areas of missed spraying, thin spraying or paint film defects, and the defect areas are mapped onto the complete three-dimensional digital model to generate three-dimensional defect location information.

[0012] Preferably, the data processing unit is further configured to perform one or a combination of the following two processing strategies based on the defect three-dimensional location information: Strategy 1: Automatically generate local touch-up spraying trajectories for the defective areas and control the painting robot system to perform precise touch-up spraying on the workpieces returned to the painting station; Strategy 2: When the defect area exceeds a preset threshold or is a specific defect type, an alarm signal is triggered and the three-dimensional location information of the defect is highlighted on the human-machine interface, awaiting manual processing.

[0013] Compared with existing technologies, the advantages of this invention are: upgrading from two-dimensional vision to precise three-dimensional modeling; achieving high-precision three-dimensional reconstruction through the complementarity of visual shadow analysis and LiDAR three-dimensional scanning, combined with servo motion data. The two-dimensional contours provided by the vision system provide spatial constraints for point cloud data processing, effectively solving the problems of point cloud noise and missing data, realizing closed-loop control of the entire process from recognition to quality inspection, greatly improving the system's intelligence and practicality, and effectively controlling system costs while enhancing functionality, making it more suitable for industrial applications. Attached Figure Description

[0014] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0015] Figure 1 This is a schematic diagram of the overall system layout of the present invention; Figure 2 This is a schematic diagram showing the layout of the visual recognition system and the lidar system; Figure 3 This is a flowchart of data processing and paint trajectory generation; Figure 4 This is a flowchart of the point cloud data correction algorithm; Figure 5 This is a flowchart of closed-loop quality inspection and respray control. Figure 6 This is a schematic diagram of the data structure for the paint spray trajectory line.

[0016] In the diagram: 1. Servo drive transmission line; 2. Vision recognition system; 21. High-definition camera; 22. High-intensity directional projection light source; 3. LiDAR system; 31. LiDAR; 4. Data processing unit; 5. Painting robot system; 6. Workpiece. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0018] In the description of this invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0019] In the description of this invention, "several" means one or more, "multiple" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. Where the terms "first," "second," and "third" are used for descriptive purposes and to distinguish technical features, they should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the sequential relationship of the indicated technical features.

[0020] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. Furthermore, the technical features involved in the different embodiments of this invention described below can be combined with each other as long as they do not conflict with each other.

[0021] Example 1 This embodiment provides a complete production line system that can automatically identify workpiece 6 and generate paint spraying trajectories.

[0022] refer to Figure 1The automated painting production line, along the flow direction (X direction) of workpiece 6, sequentially includes: a feeding area, a vision recognition system 2, a lidar system 3, and a painting robot system 5. The entire system is uniformly coordinated and controlled by a data processing unit 4. In this embodiment, workpiece 6 is a welded pipe with a branch structure. Due to the limitations of the fixed fixture, workpiece 6 is only suitable for single-sided painting. The multiple high-definition cameras and lidar arranged in a ring can cover only a 180-degree angle, processing only one side of workpiece 6 at a time. After single-sided operation is completed, workpiece 6 is flipped over for further operation.

[0023] The servo-driven transmission line 1 employs a high-precision servo motor drive, coupled with a ball screw or precision gear rack mechanism. Servo motor drive is existing technology and will not be discussed further. This ensures stable travel speed and precise positioning of the workpiece 6 in the X direction. The servo motor has a built-in encoder, which can provide real-time feedback to the data processing unit 4 on the motor's speed, angle, and cumulative displacement. Precision positioning fixtures are installed on the transmission line to fix various steel workpieces 6, preventing shaking during travel and painting.

[0024] The visual recognition system 2, located after the loading area, includes a high-intensity directional projection light source 22 and a ring-shaped high-definition camera 21. In this embodiment, the high-intensity directional projection light source 22 uses a high-brightness multi-color LED array, capable of emitting high-intensity parallel light of four colors: red, green, blue, and white. It is mounted above the workpiece 6, with its optical axis along the X direction and projecting onto the surface of the workpiece 6 at a 15-30 degree downward angle. The ring-shaped high-definition camera 21 consists of three 20-megapixel high-definition industrial cameras, which are evenly distributed in the YZ plane surrounding the transmission line, forming a ring-shaped shooting area.

[0025] The workflow of the visual recognition system 2 is as follows: Workpiece 6 enters the visual recognition area at a constant speed along the conveyor line. After the trigger sensor detects the front end of workpiece 6, it sends a signal to the data processing unit 4. The data processing unit 4 controls the high-intensity light source 22 to light up and simultaneously triggers all the ring cameras 21 to take pictures, acquiring a set of shadow images covering 180 degrees of workpiece 6. This set of images is transmitted to the data processing unit 4 in real time via gigabit Ethernet. The data processing unit 4 runs an image processing algorithm, including three steps: image preprocessing, shadow extraction, and contour synthesis, and finally obtains a 180-degree closed contour line of workpiece 6 on the YZ plane.

[0026] See Figure 2The lidar system 3, located immediately after the visual recognition system 2, comprises three lidars 31 arranged in a ring. In this embodiment, the three lidars 31 form a ring scanning unit. The lidars 31 employ a time-of-flight principle scanner, featuring a high scanning frequency and high point cloud density. When the workpiece 6 enters the lidar area, the system triggers all lidars 31 to synchronously scan the surface of the workpiece 6, obtaining dense, raw point cloud data with three-dimensional coordinates.

[0027] Reference Figure 3 and Figure 4 The data processing unit 4 receives data from the vision recognition system 2 and the lidar system 3, and performs data fusion and correction. First, the YZ plane contour obtained by the vision system is registered with the original point cloud obtained by the lidar system 31 in the same world coordinate system. Then, using the visual contour as a spatial constraint, the original point cloud data is optimized, including filtering, denoising, and completion operations. During the entire process of the workpiece 6 passing through the detection area, the vision and lidar systems 3 perform multiple, equally spaced samplings. All these corrected YZ cross-sectional point clouds at different X positions are sequentially stitched together. Combined with the X-direction position coordinates provided by the servo system, a complete and accurate three-dimensional digital model of the workpiece 6 can be constructed.

[0028] See Figure 1 and Figure 6 The painting robot system 5 includes a suspended painting robot located inside the paint booth. After the data processing unit 4 completes the 3D modeling, it activates the AI ​​trajectory planning module. This embodiment uses a trained deep convolutional neural network model, whose training data comes from a large number of known 3D models and their optimal painting paths designed by experienced engineers. The 3D model of the workpiece 6 constructed in this embodiment is input into the model, and the model outputs an optimal painting trajectory. This trajectory includes a series of spatial coordinate points, as well as parameters such as the spray gun TCP posture and spray gun on / off status corresponding to each coordinate point. After the workpiece 6 enters the paint booth, the painting robot strictly follows this program, combined with the real-time travel speed of the workpiece 6, to synchronously track and paint.

[0029] Example 2 Reference Figure 5 This embodiment adds a key function of paint quality detection and automatic processing based on the same vision system, based on Embodiment 1.

[0030] The hardware layout of this system is exactly the same as that of Embodiment 1, but its control logic and software functions have been greatly enhanced. The core innovation lies in reusing the "visual recognition system 2" located in front of the paint booth entrance to complete the quality inspection task after painting.

[0031] The entire system's operation process forms a complete closed loop of "modeling-planning-painting-quality inspection-decision-execution". Workpiece 6 first enters the vision recognition system 2 and the lidar system 3 along the conveyor line to complete 3D modeling and the first painting, which is exactly the same as in Example 1.

[0032] After the first painting is completed, the workpiece does not leave the production line immediately. Instead, the data processing unit 4 controls the servo drive transmission line 1 to make the workpiece 6 precisely back out of the paint booth exit and move in the opposite direction along the X direction, so that it can fully enter the detection area of ​​the vision recognition system 2 again.

[0033] At this point, the visual recognition system 2 switches its operating mode from "modeling mode" to "quality inspection mode". In "quality inspection mode", the strong directional projection light source 22 is turned off, and the system turns on a uniform diffuse white light source. The ring-shaped high-definition camera 21 uses a higher resolution sampling setting to acquire 180-degree high-definition color images of the returned workpiece 6.

[0034] The acquired high-definition color images are sent to the quality inspection module of data processing unit 4. This module runs an image segmentation network based on deep learning, which can accurately identify every pixel in the image that belongs to defects such as "missed spraying" or "uneven spraying". Subsequently, the system maps the identified two-dimensional defect areas to the constructed accurate three-dimensional digital model of workpiece 6, and calculates the precise location, area and shape of each defect area on the three-dimensional model of workpiece 6.

[0035] Data processing unit 4 automatically makes decisions based on the identification results according to preset program logic. If the defect type is clear, the location is reachable, and the area is less than a set threshold, the system automatically enters the repainting process. The system will activate the repainting trajectory planner to extract the defect area from the complete 3D model and generate an efficient and accurate local painting trajectory for it. After generating the repainting program, the conveyor line sends the workpiece 6 back into the paint booth in a forward direction. The painting robot loads and executes the repainting program to perform precise and localized repair painting on the defect location.

[0036] If the defect area is too large, the type is complex, or the system determines that automatic re-spraying cannot meet the quality requirements, the system will immediately issue an alarm and clearly mark the defect location on the host computer. At the same time, it will control the transmission line to transport workpiece 6 to the designated waiting area for manual intervention.

[0037] After the touch-up spraying is completed, workpiece 6 can be re-inspected through the "backtracking-quality inspection" process to ensure that the problem has been resolved. All quality inspection results, decisions, and handling records are stored in a database for quality traceability and process optimization.

[0038] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A vision-recognition-based adaptive automatic spray painting production line, characterized in that, include: Servo drive transmission line (1) is used to carry and drive the workpiece (6) to travel along a preset X direction; The visual recognition system (2) is located at the upstream station of the servo drive transmission line (1), and includes a plurality of high-definition cameras (21) arranged in a ring in the YZ plane perpendicular to the X direction, and a strong light directional projection light source (22) for projecting a directional beam onto the surface of the workpiece (6). The optical axis of the strong light directional projection light source (22) forms a non-zero angle with the X direction so that the surface of the workpiece (6) is shadowed by its own shape. The plurality of high-definition cameras (21) are configured to acquire images of the surface of the workpiece (6) containing the shadow. The lidar system (3) is located downstream of the visual recognition system (2) and includes multiple lidars (31) arranged in a ring in the YZ plane for collecting three-dimensional dot cloud data of the workpiece (6). The data processing unit (4) is communicatively connected to the visual recognition system (2), the lidar system (3), and the servo drive transmission line (1), and is configured as follows: Receive the surface image of the workpiece (6) and extract the two-dimensional contour data of the workpiece (6) in the YZ plane by analyzing the shape of the shadow; The three-dimensional point cloud data is received, and the two-dimensional contour data and the three-dimensional point cloud data are spatially registered and fused. The three-dimensional point cloud data is optimized using the two-dimensional contour data as a spatial constraint to generate a corrected three-dimensional point cloud. Based on the travel data fed back by the servo drive transmission line (1), the length information of the workpiece (6) in the X direction is obtained by integral calculation; By combining the corrected three-dimensional dot cloud with the length information, a complete three-dimensional digital model of the workpiece (6) is constructed. The painting robot system (5) is located downstream of the lidar system (3) and is communicatively connected to the data processing unit (4). The data processing unit (4) is further configured to: analyze the complete three-dimensional digital model through AI algorithm, automatically generate a paint trajectory line that matches the surface morphology of the workpiece (6), and control the paint robot system (5) to perform paint spraying operation on the moving workpiece (6) according to the paint trajectory line.

2. The production line according to claim 1, characterized in that, The strong light directional projection light source (22) is a multispectral light source capable of emitting at least two different wavelengths of visible light; the data processing unit (4) is configured to: control the strong light directional projection light source (22) to switch the emitted wavelength, and select the image sequence with the highest imaging contrast based on the images acquired by the multiple high-definition cameras (21) under different wavelength illumination for the extraction of the two-dimensional contour data.

3. The production line according to claim 1, characterized in that, The number of the multiple high-definition cameras (21) arranged in the ring is no less than three, and they are distributed at equal angular intervals to synchronously acquire panoramic images of the workpiece (6) in the YZ plane; the number of the multiple lidars (31) arranged in the ring is no less than three.

4. The production line according to claim 1, characterized in that, The data processing unit (4) is configured to optimize the three-dimensional point cloud data in the following manner: Filtering: Identify and remove three-dimensional point cloud data points located outside the area defined by the two-dimensional contour data as outliers; Denoising: Statistical filtering is performed on the three-dimensional point cloud data within the area defined by the two-dimensional contour data to remove discrete noise points; Completeness: For areas where point cloud data is missing due to occlusion, triangular facet interpolation is performed using the boundaries defined by the two-dimensional contour data to generate a complete workpiece (6) surface model.

5. The production line according to claim 1, characterized in that, The servo drive transmission line (1) has a high-precision encoder integrated at its drive end for real-time feedback of its travel speed and position; the length information of the workpiece (6) in the X direction is calculated by the following formula: L = ∫v(t)dt Where L is the length of the workpiece (6), v(t) is the travel speed time function of the servo drive transmission line (1), and the integration interval is the entire detection period of the workpiece (6) passing through the visual recognition system (2) and the lidar system (3).

6. The production line according to claim 1, characterized in that, The AI ​​algorithm is a supervised learning deep learning neural network model. The training data of the model includes a large number of workpiece (6) three-dimensional model samples and their corresponding optimal spraying path labels. The process of generating the paint trajectory line includes: using the complete three-dimensional digital model and the real-time travel speed of the servo drive transmission line (1) as inputs to the model, and the output of the model is a series of control command sequences that define the spatial motion path, switching timing and attitude angle of the spray gun.

7. The production line according to claim 1, characterized in that, The paint spray trajectory is specifically defined as a sequence of three-dimensional spatial coordinate points P={P1, P2, ..., Pn}, and a set of control parameters Ci associated with each coordinate point Pi in the sequence, wherein Ci includes at least: the on / off state of the spray gun, the TCP attitude angle of the spray gun, and the paint flow rate.

8. The production line according to any one of claims 1 to 7, characterized in that, The visual recognition system (2) is also configured to perform a quality re-inspection on the painted workpiece (6) after the painting robot system (5) has completed the painting operation; The servo drive transmission line (1) is configured to reverse the flow of the painted workpiece (6) from the painting station back to the detection area of ​​the vision recognition system (2); The high-intensity directional projection light source (22) is switched to uniform diffuse illumination mode during re-inspection; The data processing unit (4) is further configured to: The multiple high-definition cameras (21) are used to collect re-inspection images of the painted workpiece (6) surface; The re-inspection image is compared with the preset qualified paint surface standard by an image recognition algorithm to identify areas of missed spraying, thin spraying or paint film defects, and the defect areas are mapped onto the complete three-dimensional digital model to generate three-dimensional defect location information.

9. The production line according to claim 8, characterized in that, The data processing unit (4) is also configured to perform one or a combination of the following two processing strategies based on the three-dimensional location information of the defect: Strategy 1: Automatically generate local touch-up spraying trajectories for the defective areas and control the painting robot system (5) to perform precise touch-up spraying on the workpieces (6) that have been returned to the painting station; Strategy 2: When the defect area exceeds a preset threshold or is a specific defect type, an alarm signal is triggered and the three-dimensional location information of the defect is highlighted on the human-machine interface, awaiting manual processing.

Images