A machine vision-based radiator shower coating system and method

Abstract

The present application relates to the technical field of radiator curtain coating, and discloses a radiator curtain coating system and method based on machine vision, which comprises a curtain coating station, a conveying line, a curtain coating module, a visual detection module and a control module, the visual detection module is used for collecting the surface image of the radiator, the control module is used for controlling the curtain coating module to perform curtain coating, after the curtain coating is completed, the control module controls the visual detection module to collect the surface image and perform image preprocessing, a plurality of detection areas are divided, color features, gray features and texture features are extracted respectively, fusion analysis is performed, and an abnormal area is obtained, the type and position of the paint film defect are identified according to the abnormal area, when the type of the paint film defect is missing coating, local supplementary curtain coating is performed, when the type of the paint film defect is uneven paint film, sweeping supplementary curtain coating is performed, and when the type of the paint film defect is sagging, a subsequent curtain coating correction signal of the radiator is generated. The present application reduces the defect rate and improves the production efficiency.

Description

Technical Field

[0001] This invention relates to the field of radiator coating technology, and more specifically, to a radiator coating system and method based on machine vision. Background Technology

[0002] Radiators are widely used in automobiles, construction machinery, air conditioning and industrial heat exchange equipment. To improve the corrosion resistance, weather resistance and appearance quality of radiators, they are usually coated to form a uniform protective paint film on the surface of the radiator.

[0003] Existing patent CN112326669A discloses a coating defect detection and marking system and method. This system uses lighting devices, image acquisition devices, and robotic arms to acquire images of the workpiece surface. A computer system analyzes the acquired images to identify and classify coating defects. Simultaneously, a marking device marks the defect locations based on the defect type, providing a basis for subsequent manual repair and improving coating inspection efficiency and automation. However, the aforementioned machine vision inspection system is mainly suitable for coating inspection of large-sized planar or curved workpieces such as car bodies. Its primary purpose is to identify and mark the location of coating defects. Since radiators have multiple structural areas such as fins, manifolds, and frames, and the surface morphology of each area is complex with small fin spacing, shadows, reflections, and texture interference are easily generated. Traditional visual inspection methods struggle to accurately distinguish between different types of paint film defects such as missed coatings, uneven paint film, and runs, thus affecting detection accuracy.

[0004] Therefore, it is necessary to design a machine vision-based heat sink coating system and method to solve the problems existing in the current technology. Summary of the Invention

[0005] In view of this, the present invention proposes a radiator coating system and method based on machine vision, which aims to solve the problem that radiators have multiple structural areas such as fins, manifolds and frames, each with complex surface morphology, small fin spacing, and are prone to shadows, reflections and texture interference, making it difficult to accurately distinguish different types of paint film defects such as missed coating, uneven paint film and sagging.

[0006] In one aspect, the present invention proposes a machine vision-based heat sink coating system, including a coating station, a conveyor line, and a coating module, and further comprising: The vision inspection module is used to acquire surface images of the heat sink; The control module is used to control the coating module to apply coating to the heat sink; after the coating is completed, the control module is used to acquire the surface image; and the surface image is preprocessed. The control module is further configured to divide the radiator structure into several detection areas, extract color features, grayscale features, and texture features from each detection area, perform fusion analysis to obtain abnormal areas, identify the type and location of paint film defects based on the abnormal areas, control the coating module to perform local supplementary coating on the defect location when the paint film defect type is missed coating, control the coating module to perform sweeping supplementary coating on the defect location when the paint film defect type is uneven coating, and generate a subsequent coating correction signal for the radiator when the paint film defect type is sagging.

[0007] Furthermore, the control module performs image preprocessing on the surface image, including grayscale conversion, brightness compensation, image filtering, image enhancement, and edge enhancement.

[0008] Furthermore, when the control module divides the detection area according to the heat sink structure, it includes: The surface image is divided into a fin region, a manifold region, and a border region based on the heat sink structure.

[0009] Furthermore, when the control module extracts color features, grayscale features, and texture features, it includes: The image of the detection area is converted from the RGB color space (Red-Green-Blue Color Space) to the HSV color space (Hue-Saturation-Value Color Space). The first-order, second-order, and third-order color moments of the H, S, and V channels are determined respectively. The obtained color moments of each channel are combined to form the color feature. Determine the global grayscale mean and grayscale standard deviation of the grayscale image, and determine the grayscale gradient distribution of the grayscale image. Combine the global grayscale mean, grayscale standard deviation, and grayscale gradient distribution into the grayscale feature. A gray-level co-occurrence matrix is ​​constructed for the gray-level image. Energy, contrast, correlation, and entropy are extracted from the gray-level co-occurrence matrix as global texture features. Local micro-texture features of the gray-level image are extracted. The global texture features and local micro-texture features are combined to form the texture features.

[0010] Furthermore, when the control module performs fusion analysis, it includes: The color features, grayscale features, and texture features are mapped to a unified numerical range for normalization, and the normalized features are then weighted and fused to obtain a comprehensive feature evaluation value.

[0011] Furthermore, when the control module obtains an abnormal region, it includes: The comprehensive feature evaluation value is compared with the standard paint film feature threshold. When the comprehensive feature evaluation value deviates from the standard paint film feature threshold by more than the tolerance range, the detection area is determined to be the abnormal area.

[0012] Furthermore, when the control module identifies the type of paint film defect based on the abnormal area, it includes: When the similarity between the local micro-texture features in the texture features of the abnormal area and the texture features of the substrate is greater than the similarity threshold, and the global grayscale mean value in the grayscale features is greater than the brightness threshold, the paint film defect type is identified as missed coating. When the gray standard deviation in the gray features and the second-order color moment in the color features of the abnormal region are greater than the uniformity threshold, the type of paint film defect is identified as paint film unevenness. When the gray-level gradient distribution in the gray-level features of the abnormal region exhibits a continuous abrupt change along the direction of gravity, and the contrast in the texture features is greater than the contrast threshold, the paint film defect type is identified as sagging.

[0013] Furthermore, when the control module performs localized supplementary spraying or sweeping supplementary spraying, it includes: When performing the local supplementary coating, the center coordinates and area of ​​the defect location are obtained, and the coating module is controlled to move directly above the center coordinates to perform fixed-point coating; When performing the sweeping replenishment coating, the boundary contour of the defect location is obtained, and the coating module is controlled to move to the starting end of the boundary contour and perform reciprocating sweeping coating along the boundary contour.

[0014] Furthermore, when the control module generates the subsequent coating correction signal for the heat sink, it includes: When the paint film defect type is sagging, the detection area where the sagging occurs is obtained; when the sagging occurs in the manifold area, the coating pressure of the subsequent coating module is reduced; when the sagging occurs in the border area, the deflection angle of the subsequent coating module is adjusted; when the sagging occurs in the fin area, the moving speed of the subsequent coating module is increased.

[0015] Compared with existing technologies, the advantages of this invention are as follows: By acquiring surface images after the radiator has been coated, and dividing the detection area according to the radiator structure, different structural areas are detected separately. This allows for targeted image analysis of the different structural features of the fin area, manifold area, and frame area, reducing the impact of complex structures on the detection results and improving the accuracy of paint film defect detection. The color, grayscale, and texture features of the detection area are extracted and fused for analysis. Compared to single-color or grayscale detection methods, this comprehensively reflects changes in paint film color, brightness, and surface texture, improving the identification ability of different types of paint film defects such as missed coating, uneven paint film, and runs, and reducing false detection and false negative rates. Different processing strategies are adopted according to the identified paint film defect type: when a missed coating is identified, local supplementary coating is performed; when an uneven paint film is identified, sweeping supplementary coating is performed; and when a runs are identified, a subsequent coating correction signal is generated. This achieves linkage between the detection results and the coating control process, improving the targeting and automation of defect processing and reducing manual rework. The system generates a correction signal for subsequent coating to address sagging defects, and adjusts the coating process for the radiator. This allows visual inspection results to be fed back to the coating control process, achieving closed-loop control of coating quality inspection, defect compensation, and process parameter optimization. This improves the consistency of radiator coating quality and production efficiency.

[0016] On the other hand, this application also provides a machine vision-based heat sink coating method, applied to the aforementioned machine vision-based heat sink coating system, comprising: The control module applies a coating to the heat sink; after coating is completed, the control module acquires a surface image of the heat sink; and the surface image undergoes image preprocessing. The radiator structure is divided into several detection areas. Color features, grayscale features and texture features are extracted from each detection area. A fusion analysis is then performed to obtain abnormal areas. Based on the abnormal areas, the type and location of paint film defects are identified. When the paint film defect type is missed coating, the coating module is controlled to perform local supplementary coating on the defect location; when the paint film defect type is uneven paint film, the coating module is controlled to perform sweeping supplementary coating on the defect location; when the paint film defect type is sagging, a correction signal for subsequent coating of the radiator is generated.

[0017] It is understandable that the above-mentioned machine vision-based radiator coating system and method have the same beneficial effects, and will not be elaborated further here. Attached Figure Description

[0018] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A structural block diagram of a machine vision-based heat sink coating system provided in an embodiment of the present invention; Figure 2 A schematic diagram of the structure of a machine vision-based heat sink coating system provided in an embodiment of the present invention; Figure 3 This is a logic flowchart of heat sink coating provided in an embodiment of the present invention; Figure 4 A flowchart of a machine vision-based heat sink coating method provided in an embodiment of the present invention.

[0019] The components include: 1. Coating station; 2. Conveyor line; 3. Coating module; 4. Visual inspection module. Detailed Implementation

[0020] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the disclosure to those skilled in the art. It should be noted that, unless otherwise specified, embodiments and features in the embodiments of the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0021] In some embodiments of this application, see Figure 1-3 As shown, a machine vision-based heat sink coating system is proposed, including a coating station 1, a conveyor line 2, and a coating module 3, and also includes: Visual inspection module 4 is used to acquire surface images of the heat sink; The control module controls the coating module 3 to apply coating to the heat sink; after coating is completed, it controls the vision inspection module 4 to acquire surface images and performs image preprocessing on the surface images. The control module is also used to divide the radiator structure into several detection areas, extract color features, grayscale features and texture features for each detection area, and perform fusion analysis to obtain abnormal areas; identify the type and location of paint film defects based on the abnormal areas; when the paint film defect type is missed coating, control the coating module 3 to perform local supplementary coating on the defect location; when the paint film defect type is uneven paint film, control the coating module 3 to perform sweeping supplementary coating on the defect location; when the paint film defect type is sagging, generate a correction signal for subsequent coating of the radiator.

[0022] In this embodiment, the system includes a conveyor line 2, a coating station 1, a coating unit, a vision inspection module 4, and a control module.

[0023] The conveyor line 2 is used to transport the radiators that have completed the previous process to the coating station 1 and to transfer the workpieces between the stations. The conveyor line 2 can be one of a chain conveyor line 2, a roller conveyor line 2, or a clamping conveyor line 2. In order to ensure the stability of the radiator's posture during the transportation process, the conveyor line 2 can be used in conjunction with a limiting and guiding mechanism to ensure that the radiator enters the coating station 1 in a preset direction.

[0024] The coating unit is located at coating station 1 and is used for automatic coating of the radiator surface. The coating unit may include spray pipes, nozzle assemblies, flow control valves, pressure regulating valves, atomization control components, and moving actuators.

[0025] The vision inspection module 4 is used to acquire images of the radiator surface and transmit the images to the central controller for processing and analysis. The machine vision inspection module 4 includes an industrial camera, lens, supplementary lighting source, image acquisition card, and image processing unit.

[0026] The control module is used to coordinate and control the conveyor line 2, the coating unit and the vision inspection module 4, so as to realize the fully automated operation of the heat sink from feeding, coating, inspection, touch-up coating to diversion and discharge.

[0027] It may also include an automatic feeding mechanism located at the feeding end of conveyor line 2, used to transfer the radiator from conveyor line 2 to the bearing position of the coating station 1. The automatic feeding mechanism may employ one or more of the following: mechanical clamping arm, lifting and transferring device, pushing device, or clamp flipping mechanism, to ensure a smooth and accurate positioning process for the radiator feeding. The paint circulation system is used for storing, conveying, filtering, recirculating, and stabilizing the pressure of the paint, ensuring a continuous, stable, and recyclable paint supply. The drying unit is used to dry and cure the radiator after coating, so that the paint film reaches the specified dry state.

[0028] Specifically, after the radiator completes the previous process, it is conveyed to the coating station 1 by conveyor line 2. The automatic feeding mechanism transfers and fixes the radiator in the coating position. The control module controls the coating unit to apply the coating to the radiator. After coating is completed, the control module controls the vision inspection module 4 to acquire images of the radiator surface.

[0029] In a preferred embodiment, the vision inspection module 4 uses an industrial camera installed at the inspection position of the coating station 1. The industrial camera can have a resolution of 5 megapixels or higher, and the camera acquisition frame rate can be set to 15fps to 30fps. The exposure time is adjusted according to the ambient light. Image acquisition can be performed from multiple angles to ensure the completeness of the heat sink fins. The lens is preferably a telecentric lens or a low-distortion industrial lens to ensure the clarity of the imaging of the fin edges, manifold boundaries, and frame contours.

[0030] After image acquisition, image preprocessing is performed, including grayscale conversion, brightness compensation, image filtering, image enhancement, and edge enhancement.

[0031] The control module first converts the acquired color images into grayscale images. This is done to reduce the interference of color factors on subsequent processing, allowing image analysis to focus more on brightness, texture, and edge variations. In the heatsink coating scenario, grayscale images are better at highlighting abnormal structures such as paint film thickness differences, missed areas, and runners, facilitating subsequent feature extraction.

[0032] Due to the presence of shading, uneven structures, and localized shadows in the heat sink fin area, images are prone to localized overexposure or over-brightness, necessitating brightness compensation. The control module can adaptively correct the image brightness, resulting in a more uniform overall brightness distribution. In a preferred embodiment, brightness compensation prioritizes enhancing dark areas and suppressing highly reflective areas, thereby preventing coating boundaries from being obscured by reflections in the image. Brightness compensation is particularly important for aluminum heat sinks with strong surface reflectivity.

[0033] The image after brightness compensation still needs to undergo image filtering to remove random noise, shot noise, and local dirt interference from the acquisition process. Filtering can employ one or more of the following: median filtering, Gaussian filtering, mean filtering, or bilateral filtering. In a preferred embodiment, median filtering is first used to remove isolated noise points, followed by a light Gaussian filter to smooth image edges. The filtering intensity should not be too high to avoid erasing fine unpainted edges or uneven paint film textures. For heat sinks with densely packed fins, a smaller filtering window size is preferred to balance noise reduction and edge preservation.

[0034] After filtering, the image is enhanced to improve the contrast between normal and abnormal areas. Local contrast enhancement is preferred to make weaker defect outlines more prominent. For areas with missed coating, enhancement typically reveals a significant difference in brightness and texture compared to the surrounding paint film. For areas with uneven paint film, enhancement makes color differences, thickness variations, and surface undulations easier to identify. For areas with runs, enhancement highlights strip-like anomalies extending along the direction of gravity.

[0035] After image enhancement, edge enhancement processing is performed to highlight the contour information of coating boundaries, sagging edges, and areas of abrupt changes in paint film thickness. The purpose of edge enhancement is to help the control module more accurately identify defect boundaries, and it is particularly suitable for detecting the edges of heat sink fins and frame edges. In a preferred embodiment, edge enhancement is mainly used to strengthen the following types of areas: the first type is the transition boundary between normal and abnormal coatings; the second type is the sagging boundary formed by sagging; and the third type is the gap boundary formed by missed coating. After edge enhancement, the abnormal contours in the image are clearer, which is beneficial for subsequent zoning analysis and defect localization.

[0036] In this embodiment, the control module divides the surface image into regions based on the structural features of the heat sink to reduce the interference of complex structures on the detection results. Specifically, the control module divides the heat sink surface image into fin regions, manifold regions, and border regions based on a preset heat sink structure model or through edge contour recognition results.

[0037] Among them, the fin area is used to characterize the regularly arranged area of ​​the heat dissipation structure of the radiator, which usually has high-frequency periodic texture features; the manifold area is used to characterize the tubular structure area on both sides or in the middle of the radiator, which usually has strong curved surface reflective properties; the frame area is used to characterize the outer peripheral boundary area of ​​the radiator, which is prone to paint film accumulation, sagging or edge defects.

[0038] When the control module extracts color features from the fin area, manifold area, and border area, it first converts the RGB image to the HSV color space. After conversion to HSV space, it performs statistical analysis on the H, S, and V channels and extracts the distribution characteristic parameters of each channel to characterize the color change of the paint film.

[0039] Among them, the H channel is used to characterize hue changes and reflect the color shift of the paint film; the S channel is used to characterize saturation changes and reflect the purity and uniformity of the coating; and the V channel is used to characterize brightness changes and reflect the thickness and coverage of the paint film.

[0040] In a preferred embodiment, the control module performs multidimensional statistical descriptions on the three channels respectively, and combines the statistical results of each channel to form a color feature vector, which is used to characterize the overall color distribution state of the current detection area. This color feature can effectively reflect the following abnormalities: when there is missed coating, the V channel increases significantly; when there is uneven paint film, the S channel fluctuates significantly; when there is local accumulation or sagging, both the H and V channels show abnormal shifts.

[0041] The control module extracts grayscale features from the image of the detection area to reflect the brightness distribution and surface uniformity of the paint film. Specifically, the image is first converted to grayscale, transforming the color image into a single-channel grayscale image. Then, the overall brightness of the grayscale image is statistically analyzed to obtain the overall brightness distribution of the detection area. Simultaneously, the local changes in the grayscale image are analyzed to reflect variations in paint film thickness and local abnormal areas. Furthermore, by analyzing the spatial trend of grayscale values, the presence of abrupt change areas is determined, thereby assisting in the identification of defects such as drips, accumulations, or local missing areas.

[0042] The control module extracts texture features from the detection area to characterize the continuity and uniformity of the paint film surface structure. Specifically, firstly, a gray-level co-occurrence relation matrix is ​​constructed based on the gray-level image to describe the spatial distribution relationship between pixels; then, multiple texture description parameters are extracted from this matrix to characterize surface roughness, structural uniformity, and local variation features; in a preferred embodiment, the texture features include at least surface uniformity, local fluctuation degree, structural continuity, and texture complexity.

[0043] The texture features described above can effectively distinguish between normal and abnormal areas of the paint film. For example, the texture distribution in normal areas is stable; the texture continuity in areas with missed coatings is disrupted; the texture in areas with sagging shows obvious directional changes; and the texture in areas with uneven paint film shows localized enhanced fluctuations.

[0044] The control module fuses and analyzes color features, grayscale features, and texture features to obtain a comprehensive feature evaluation result. Specifically, the control module first performs uniform scaling on different types of features to ensure that features of different dimensions are within a comparable range; then, it performs comprehensive fusion processing on various features to form a unified comprehensive feature expression result; the control module assigns different weights to different features and performs comprehensive calculations to obtain the comprehensive feature evaluation value of the detection area. Color features reflect the material coverage state, grayscale features reflect changes in brightness and thickness, and texture features reflect the continuity of the surface structure. The weights of different features can be configured according to the radiator model, paint type, and production process parameters. For example, in high-reflectivity aluminum radiators, the weight of grayscale features can be appropriately increased; in color difference-sensitive coatings, the weight of color features can be increased; and in complex textured fin structures, the weight of texture features can be increased. The control module sets initial weight parameters according to the type of radiator structure area, prioritizing texture features in fin areas, grayscale features in manifold areas, and color features in border areas. Furthermore, during production operation, the control module adaptively adjusts the weight parameters based on historical detection results. When a certain type of feature has a high false detection rate in defect identification, its weight is automatically reduced; when a certain type of feature contributes significantly to defect identification, its weight is increased.

[0045] In a preferred embodiment, the fusion process is based on the combined effect of color information, brightness information and texture information, and can simultaneously consider the "color abnormality", "brightness abnormality" and "structural abnormality" of the paint film, thereby improving the ability to identify complex defects.

[0046] In this embodiment, the control module determines anomalies in the detection area based on the fused comprehensive feature evaluation results. Specifically, the control module pre-establishes a standard paint film feature reference model, which characterizes the comprehensive distribution characteristics of a normal paint film in terms of color, grayscale, and texture. During the detection process, the comprehensive features of the current detection area are compared with the standard features: when the comprehensive features of the detection area deviate significantly from the standard paint film features (the comprehensive feature evaluation value deviates from the standard paint film feature threshold by more than the tolerance range), the detection area is determined to be an abnormal area; when the deviation is small or within the allowable range (the comprehensive feature evaluation value deviates from the standard paint film feature threshold by less than the tolerance range), the detection area is determined to be a normal area. An abnormal area indicates that the area may have defects such as missed coating, uneven paint film, or sagging.

[0047] The control module pre-acquires multiple known qualified radiator samples, covering normal coating conditions of different batches, materials, and paint types. It then performs image acquisition and feature extraction on the samples to obtain a standard feature dataset. This standard feature dataset includes comprehensive distribution information of color, grayscale, and texture features. Further, the control module statistically summarizes the standard feature dataset to form a standard paint film feature model, and determines the standard threshold and tolerance range based on the model's statistical distribution range.

[0048] In this embodiment, when the local micro-texture structure of the abnormal area is highly similar to the texture structure of the substrate itself, it indicates that the paint film coverage in the area is incomplete and the structural features of the substrate are exposed; secondly, the overall brightness level of the area is significantly higher than that of the normal paint film area, indicating that there is a coating missing or insufficient coverage in the area; when the above two characteristics are met at the same time, the control module determines that the area is a missing coating defect.

[0049] Before the heat sink enters the coating process, the control module controls the vision inspection module 4 to acquire images of the uncoated heat sink, obtaining the original texture image of the substrate. Further, the original texture image is preprocessed to extract its local texture structure features, which are then stored in the feature database as substrate texture features. In subsequent inspections, the local micro-texture features of the current inspection area are matched and compared with the substrate texture features to determine if there are any gaps in coating coverage.

[0050] When the grayscale distribution within the abnormal area fluctuates significantly, resulting in localized inconsistencies in brightness and darkness, it indicates a significant difference in the thickness of the paint film. Secondly, the color distribution in this area is uneven, resulting in localized color shifts or inconsistencies in shade. When both grayscale fluctuation and color unevenness characteristics are present, the control module determines that the area is a paint film unevenness defect.

[0051] When the grayscale change in the abnormal area shows a clear directional extension in space, and the direction of extension is consistent with or approximately consistent with the direction of gravity, it indicates that the paint is drooping and flowing under the action of gravity; secondly, the texture contrast of the area is significantly enhanced, showing strip-like or wavy structural features; when the above two features exist at the same time, the control module determines that the area is a sagging defect.

[0052] When the identification result indicates a missed coating, the control module performs a localized supplementary coating operation. Specifically, the control module first calculates the center point of the defect location based on the image of the abnormal area and estimates the area of ​​the defect, thereby determining the area to be coated. Subsequently, the control module controls the coating module 3 to move directly above the defect area, aligning the nozzle center with the defect center point and maintaining a stable distance. During the coating process, the coating action adopts a fixed-point or small-area localized coating method, ensuring that the paint concentrates on covering the missed area while avoiding secondary accumulation on the surrounding normal paint film. The coating duration is automatically adjusted according to the defect area; when the defect area is large, the coating time is appropriately extended to ensure complete coverage.

[0053] When the identification result indicates uneven paint film, the control module performs a sweeping replenishment coating. Specifically, the control module first extracts the boundaries of the abnormal area to determine the outer contour range of the uneven paint film area. Then, it controls the coating module 3 to move to the starting boundary position of the area, causing the nozzle to move continuously along the boundary or internal path of the defect area. During the coating process, the coating path adopts a reciprocating sweeping motion, allowing the paint to evenly cover the uneven areas, thereby improving the local paint film consistency. The sweeping speed is related to the difference in paint film thickness; when the thickness difference is large, the moving speed is reduced to increase the paint coverage per unit area.

[0054] When a sagging defect is identified, the control module no longer simply performs localized recoating on the current workpiece. Instead, it generates a subsequent coating correction signal to adjust the coating process parameters for subsequent workpieces, reducing the recurrence of similar defects from the source. Specifically, the control module first determines the type of detection area where sagging occurs, including the manifold area, the border area, or the fin area, and executes different correction strategies based on the characteristics of different areas. When sagging occurs in the manifold area, the control module reduces the coating pressure of the subsequent coating module 3, making the coating adhesion more stable in the curved area, thereby reducing the accumulation phenomenon caused by gravity flow. When sagging occurs in the border area, the control module adjusts the spray deflection angle of the coating module, making the coating direction more in line with the changes in the border structure, thereby reducing the risk of edge accumulation and sagging. When sagging occurs in the fin area, the control module increases the moving speed of the coating module, reducing the residence time of the coating per unit area, thereby reducing the probability of coating sagging due to gravity.

[0055] Based on another preferred embodiment described above, see [link to preferred embodiment]. Figure 4 As shown, this embodiment provides a machine vision-based heatsink coating method, applied to the aforementioned machine vision-based heatsink coating system, including: S100: Controls the coating module to apply coating to the heat sink; after coating is completed, controls the vision inspection module to acquire surface images of the heat sink; and performs image preprocessing on the surface images. S200: Divide the radiator structure into several detection areas, extract color features, grayscale features and texture features for each detection area, perform fusion analysis to obtain abnormal areas, and identify the type and location of paint film defects based on the abnormal areas. S300: When the paint film defect type is missed coating, the control module performs local supplementary coating on the defect location; when the paint film defect type is uneven paint film, the control module performs sweeping supplementary coating on the defect location; when the paint film defect type is sagging, a correction signal for subsequent coating of the radiator is generated.

[0056] In summary, by acquiring surface images after the radiator has completed coating, and dividing the detection area according to the radiator structure, different structural areas can be detected separately. This allows for targeted image analysis of the different structural features of the fin area, manifold area, and frame area, reducing the impact of complex structures on the detection results and improving the accuracy of paint film defect detection. Extracting color, grayscale, and texture features of the detection area and fusing these features, compared to single-color or grayscale detection methods, comprehensively reflects changes in paint film color, brightness, and surface texture. This improves the ability to identify different types of paint film defects such as missed coating, uneven paint film, and runs, reducing false positive and false negative rates. Different processing strategies are adopted based on the identified paint film defect type: local supplementary coating is performed when a missed coating is identified; sweeping supplementary coating is performed when an uneven paint film is identified; and subsequent coating correction signals are generated when a runs are identified. This achieves linkage between the detection results and the coating control process, improving the targeting and automation of defect processing and reducing manual rework. The system generates a correction signal for subsequent coating to address sagging defects, and adjusts the coating process for the radiator. This allows visual inspection results to be fed back to the coating control process, achieving closed-loop control of coating quality inspection, defect compensation, and process parameter optimization. This improves the consistency of radiator coating quality and production efficiency.

[0057] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the protection scope of the present invention.

Claims

1. A machine vision-based heat sink coating system, comprising a coating station, a conveyor line, and a coating module, characterized in that, Also includes: The vision inspection module is used to acquire surface images of the heat sink; The control module is used to control the coating module to apply a coating to the heat sink; After the coating is completed, the vision detection module is controlled to acquire the surface image; and the surface image is preprocessed. The control module is also used to divide the heat sink structure into several detection areas, and extract color features, grayscale features and texture features for each detection area respectively. The system performs fusion analysis to identify abnormal areas; it then identifies the type and location of paint film defects based on these abnormal areas; when the defect type is missed coating, it controls the coating module to perform localized supplementary coating on the defect location; when the defect type is uneven paint film, it controls the coating module to perform sweeping supplementary coating on the defect location; and when the defect type is sagging, it generates a correction signal for subsequent coating of the radiator.

2. The machine vision-based radiator coating system according to claim 1, characterized in that, The control module performs image preprocessing on the surface image, including grayscale conversion, brightness compensation, image filtering, image enhancement, and edge enhancement.

3. The machine vision-based radiator coating system according to claim 2, characterized in that, When the control module divides the radiator structure into several detection areas, it includes: The surface image is divided into a fin region, a manifold region, and a border region based on the heat sink structure.

4. The machine vision-based radiator coating system according to claim 3, characterized in that, When the control module extracts color features, grayscale features, and texture features, it includes: The image of the detection area is converted from the RGB color space to the HSV color space, and the first-order color moments, second-order color moments, and third-order color moments of the H channel, S channel, and V channel are determined respectively. The obtained color moments of each channel are combined to form the color feature. Determine the global grayscale mean and grayscale standard deviation of the grayscale image, and determine the grayscale gradient distribution of the grayscale image. Combine the global grayscale mean, grayscale standard deviation, and grayscale gradient distribution into the grayscale feature. A gray-level co-occurrence matrix is ​​constructed for the gray-level image. Energy, contrast, correlation, and entropy are extracted from the gray-level co-occurrence matrix as global texture features. Local micro-texture features of the gray-level image are extracted. The global texture features and local micro-texture features are combined to form the texture features.

5. The machine vision-based radiator coating system according to claim 4, characterized in that, When the control module performs fusion analysis, it includes: The color features, grayscale features, and texture features are mapped to a unified numerical range for normalization, and the normalized features are then weighted and fused to obtain a comprehensive feature evaluation value.

6. The machine vision-based radiator coating system according to claim 5, characterized in that, When the control module obtains an abnormal region, it includes: The comprehensive feature evaluation value is compared with the standard paint film feature threshold. When the comprehensive feature evaluation value deviates from the standard paint film feature threshold by more than the tolerance range, the detection area is determined to be the abnormal area.

7. The machine vision-based radiator coating system according to claim 6, characterized in that, When the control module identifies the type of paint film defect based on the abnormal area, it includes: When the similarity between the local micro-texture features in the texture features of the abnormal area and the texture features of the substrate is greater than the similarity threshold, and the global grayscale mean value in the grayscale features is greater than the brightness threshold, the paint film defect type is identified as a missed coating. When the gray standard deviation in the gray features and the second-order color moment in the color features of the abnormal region are greater than the uniformity threshold, the type of paint film defect is identified as paint film unevenness. When the gray-level gradient distribution in the gray-level features of the abnormal region exhibits a continuous abrupt change along the direction of gravity, and the contrast in the texture features is greater than the contrast threshold, the paint film defect type is identified as sagging.

8. The machine vision-based radiator coating system according to claim 7, characterized in that, When the control module performs localized supplementary spraying or sweeping supplementary spraying, it includes: When performing the local supplementary coating, the center coordinates and area of ​​the defect location are obtained, and the coating module is controlled to move directly above the center coordinates to perform fixed-point coating; When performing the sweeping replenishment coating, the boundary contour of the defect location is obtained, and the coating module is controlled to move to the starting end of the boundary contour and perform reciprocating sweeping coating along the boundary contour.

9. The machine vision-based radiator coating system according to claim 8, characterized in that, When the control module generates the subsequent coating correction signal for the heat sink, it includes: When the paint film defect type is sagging, the detection area where the sagging occurs is obtained; when the sagging occurs in the manifold area, the coating pressure of the subsequent coating module is reduced; when the sagging occurs in the border area, the deflection angle of the subsequent coating module is adjusted; when the sagging occurs in the fin area, the moving speed of the subsequent coating module is increased.

10. A machine vision-based heat sink coating method, applied to the machine vision-based heat sink coating system as described in any one of claims 1-9, characterized in that, include: Control the coating module to apply coating to the heat sink; After the coating is completed, the visual inspection module is controlled to acquire surface images of the heat sink; and the surface images are preprocessed. The heat sink structure is divided into several detection areas, and color features, grayscale features and texture features are extracted from each detection area. Then, a fusion analysis is performed to obtain abnormal areas; based on the abnormal areas, the type and location of paint film defects are identified. When the paint film defect type is missed coating, the coating module is controlled to perform local supplementary coating on the defect location; when the paint film defect type is uneven paint film, the coating module is controlled to perform sweeping supplementary coating on the defect location; when the paint film defect type is sagging, a correction signal for subsequent coating of the radiator is generated.

Images