Automobile gluing quality detection system based on machine vision algorithm

Abstract

The application relates to the technical field of automobile gluing quality detection, and discloses a vehicle gluing quality detection system based on a machine vision algorithm, which comprises an image acquisition unit, an image preprocessing unit, a gluing feature extraction unit, a defect identification and analysis unit, a result output and early warning unit, a system calibration and feedback unit and a data management unit. The vehicle gluing quality detection system based on the machine vision algorithm extracts the geometric features and the gray features of gluing by adopting a multispectral camera combined with a machine vision algorithm, and classifies and identifies the gluing defects by using a support vector machine or a convolutional neural network, so that various defect types such as broken glue, offset glue, stacked glue and bubbles can be effectively distinguished; meanwhile, the defect grades are divided through a severity evaluation module, the detection precision and reliability are improved, and the problems of high false detection rate and poor adaptability of artificial visual inspection and traditional algorithms are overcome.

Description

Technical Field

[0001] This invention relates to the field of automotive adhesive quality inspection technology, specifically an automotive adhesive quality inspection system based on machine vision algorithms. Background Technology

[0002] Automotive adhesive application is a crucial step in the vehicle body manufacturing process, widely used for sealing, vibration damping, corrosion protection, and structural connections of components such as doors, sunroofs, hoods, and floor panels. The quality of adhesive application directly affects the overall vehicle's sealing performance, durability, NVH (noise, vibration, and ride comfort), and crash safety. As automotive manufacturing moves towards higher precision, efficiency, and automation, higher demands are placed on the real-time performance, accuracy, and traceability of adhesive application quality inspection. Currently, machine vision technology, due to its advantages of non-contact operation, speed, and rich information, has been gradually introduced into the industrial inspection field, providing a technological foundation for the automated inspection of adhesive application quality.

[0003] However, existing methods for inspecting the quality of automotive adhesive application still have significant shortcomings. Traditional manual visual inspection is not only inefficient and highly subjective, but also difficult to adapt to the pace of high-speed continuous production lines, easily leading to missed inspections or misjudgments. Some production lines use contact sensors (such as laser displacement sensors or probes) for spot checks, which can only obtain local dispensing information and cannot fully reflect the continuity, width uniformity, and cross-sectional shape of the adhesive line. The few machine vision-based systems usually use fixed grayscale thresholds or simple edge detection algorithms, which are poorly adaptable to changes in lighting, differences in adhesive color, and different vehicle models' adhesive application trajectories, making it difficult to accurately identify complex defects such as broken adhesive, off-center adhesive, adhesive buildup, and bubbles. In addition, existing systems generally lack closed-loop feedback mechanisms and data management functions for inspection results, making it impossible to automatically optimize the inspection model based on historical inspection results, and also making it difficult to trace and analyze defect samples, resulting in insufficient long-term stability and adaptability of the inspection system.

[0004] Therefore, it is necessary to invent a vehicle adhesive quality inspection system based on machine vision algorithms to solve the above problems. Summary of the Invention

[0005] The purpose of this invention is to provide an automotive adhesive quality inspection system based on machine vision algorithms to solve the problems mentioned in the background art.

[0006] To achieve the above objectives, the present invention provides the following technical solution: an automotive adhesive coating quality inspection system based on machine vision algorithms, comprising an image acquisition unit, an image preprocessing unit, an adhesive coating feature extraction unit, a defect identification and analysis unit, a result output and early warning unit, a system calibration and feedback unit, and a data management unit. The image acquisition unit and the image preprocessing unit are signal-connected; the image preprocessing unit and the adhesive coating feature extraction unit are signal-connected; the adhesive coating feature extraction unit and the defect identification and analysis unit are signal-connected; the defect identification and analysis unit and the result output and early warning unit are signal-connected; the defect identification and analysis unit and the system calibration and feedback unit are signal-connected; and the system calibration and feedback unit and the data management unit are signal-connected. The image acquisition unit is used to acquire original images of the adhesive-coated area of ​​the vehicle. The image preprocessing unit is used to perform noise reduction, enhancement and binarization preprocessing on the original image; The adhesive coating feature extraction unit is used to extract the geometric and grayscale features of the adhesive coating using machine vision algorithms; The defect identification and analysis unit is used to identify the type of adhesive coating defect and analyze the severity of the defect based on a classifier model. The result output and early warning unit is used to output the detection results and issue an early warning when the defects exceed the standard. The system calibration and feedback unit is used to periodically calibrate visual inspection parameters and receive feedback to optimize the inspection model. The data management unit is used to store historical detection data and defect sample data.

[0007] Preferably, the image acquisition unit includes a multispectral camera module, a light source control module, and a trigger synchronization module. The output terminal of the multispectral camera module is signal-connected to the light source control module, and the output terminal of the light source control module is signal-connected to the trigger synchronization module. The multispectral camera module is used to acquire images of the adhesive coating in different spectral bands; The light source control module is used to automatically adjust the intensity of the lighting source according to the ambient brightness. The trigger synchronization module is used to trigger the camera to capture images based on the vehicle's transmission position and ensure that the images are synchronized with the glue application position.

[0008] Preferably, the image preprocessing unit includes a filtering and denoising module, a contrast enhancement module, and an image segmentation module. The output of the filtering and denoising module is connected to the contrast enhancement module, and the output of the contrast enhancement module is connected to the image segmentation module. The filtering and denoising module is used to remove image noise using median filtering or Gaussian filtering. The contrast enhancement module is used to enhance the contrast between the coated area and the background using a histogram equalization method. The image segmentation module is used to extract the region of interest in the coated area using threshold segmentation or edge detection algorithms.

[0009] Preferably, the adhesive coating feature extraction unit includes a geometric feature calculation module, a grayscale feature statistics module, and a feature dimensionality reduction module. The output of the geometric feature calculation module is connected to the feature dimensionality reduction module, and the output of the grayscale feature statistics module is connected to the feature dimensionality reduction module. The geometric feature calculation module is used to calculate the width, continuity, offset, and area of ​​the adhesive line. The grayscale feature statistics module is used to statistically analyze the mean grayscale value, variance, and texture parameters of the glued area. The feature dimensionality reduction module is used to reduce feature redundancy using principal component analysis.

[0010] Preferably, the defect identification and analysis unit includes a classifier model module, a defect type determination module, and a severity assessment module. The output of the classifier model module is signal-connected to the defect type determination module, and the output of the defect type determination module is signal-connected to the severity assessment module. The classifier model module is used to classify the adhesive coating features using a support vector machine or a convolutional neural network; The defect type determination module is used to identify defect types including glue breakage, glue deviation, glue buildup, and air bubbles. The severity assessment module is used to classify defect levels based on defect size or area percentage.

[0011] Preferably, the result output and early warning unit includes a detection result display module, an audible and visual alarm module, and a report generation module. The output terminal of the detection result display module is signal-connected to the audible and visual alarm module, and the output terminal of the detection result display module is signal-connected to the report generation module. The test result display module is used to display the glue coating quality test results and defect location markings in real time on the monitoring interface; The audible and visual alarm module is used to issue an audible and visual alarm when the detected defect level exceeds a set threshold. The report generation module is used to generate statistical reports on adhesive coating quality inspection by shift or batch.

[0012] Preferably, the system calibration and feedback unit includes a parameter calibration module, a feedback optimization module, and a model update module. The output of the parameter calibration module is signal-connected to the feedback optimization module, and the output of the feedback optimization module is signal-connected to the model update module. The parameter calibration module is used to periodically calibrate camera intrinsic parameters, light source brightness, and detection thresholds. The feedback optimization module is used to adjust the threshold of the feedback signal optimization classifier based on the results of manual review or the parameters of the glue applicator. The model update module is used to incrementally train and update the classifier model using newly collected defect samples.

[0013] Preferably, the data management unit includes a historical data storage module, a defect sample library module, and a query and traceability module. The output of the historical data storage module is signal-connected to the defect sample library module, and the output of the defect sample library module is signal-connected to the query and traceability module. The historical data storage module is used to save all original detection images, processing results, and early warning records; The defect sample library module is used to classify and store various defect images and feature data. The query and traceability module is used to retrieve historical test data based on time, vehicle model, or glue application area.

[0014] Preferably, an image encoding and compression module is further provided between the image acquisition unit and the adhesive application feature extraction unit. The image encoding and compression module is used to compress and encode the preprocessed image to reduce data transmission bandwidth.

[0015] Preferably, a defect re-inspection module is provided between the defect identification and analysis unit and the result output and early warning unit. The defect re-inspection module is used to perform a second image acquisition and feature comparison on the glued area that was initially determined to be defective in order to reduce the false detection rate.

[0016] Compared with existing technologies, this invention provides an automotive adhesive coating quality inspection system based on machine vision algorithms, which has the following advantages: 1. This automotive adhesive coating quality inspection system based on machine vision algorithms uses a multispectral camera combined with machine vision algorithms to extract the geometric and grayscale features of the adhesive coating. It then uses support vector machines or convolutional neural networks to classify and identify adhesive coating defects, effectively distinguishing various defect types such as broken adhesive, off-center adhesive, adhesive buildup, and bubbles. At the same time, it classifies defect levels through a severity assessment module, improving detection accuracy and reliability, and overcoming the problems of high false detection rate and poor adaptability of manual visual inspection and traditional algorithms.

[0017] 2. This automotive adhesive coating quality inspection system based on machine vision algorithms, through the setting of result output and early warning units, realizes real-time visualization of inspection results, defect location marking, and audible and visual alarms for defects exceeding the standard. It can also generate quality statistical reports by batch, which makes it convenient for production line managers to grasp the adhesive coating quality status in a timely manner. At the same time, the system calibration and feedback unit can automatically optimize the classifier threshold and incrementally update the detection model based on the results of manual review or the adjustment information of adhesive coating machine parameters, forming a closed-loop mechanism of detection-feedback-optimization, which improves the system's adaptability and detection consistency under different production conditions.

[0018] 3. This automotive adhesive coating quality inspection system based on machine vision algorithms integrates a data management unit, which can completely save original inspection images, processing results, early warning records, and defect sample library. It supports multi-dimensional query and traceability by time, vehicle model, adhesive coating area, etc., providing data support for quality analysis and process improvement. In addition, the introduction of image encoding compression module and defect re-inspection module reduces data transmission bandwidth and lowers the false detection rate, enabling the system to operate stably in high-speed continuous production lines. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort: Figure 1 This is a schematic diagram of the system structure of the present invention; Figure 2 This is a schematic diagram of the image acquisition unit structure of the present invention; Figure 3 This is a schematic diagram of the image preprocessing unit structure of the present invention; Figure 4 This is a schematic diagram of the adhesive coating feature extraction unit structure of the present invention; Figure 5 This is a schematic diagram of the defect identification and analysis unit structure of the present invention; Figure 6 This is a schematic diagram of the result output and early warning unit structure of the present invention; Figure 7 This is a schematic diagram of the calibration and feedback unit structure of the system of the present invention; Figure 8 This is a schematic diagram of the data management unit structure of the present invention.

[0020] The diagram shows: 1. Image acquisition unit; 11. Multispectral camera module; 12. Light source control module; 13. Trigger synchronization module; 2. Image preprocessing unit; 21. Filtering and noise reduction module; 22. Contrast enhancement module; 23. Image segmentation module; 3. Adhesive coating feature extraction unit; 31. Geometric feature calculation module; 32. Gray-scale feature statistics module; 33. Feature dimensionality reduction module; 4. Defect identification and analysis unit; 41. Classifier model module; 42. Defect type determination module; 43. Severity assessment module; 5. Result output and early warning unit; 51. Detection result display module; 52. Audible and visual alarm module; 53. Report generation module; 6. System calibration and feedback unit; 61. Parameter calibration module; 62. Feedback optimization module; 63. Model update module; 7. Data management unit; 71. Historical data storage module; 72. Defect sample library module; 73. Query and traceability module; 8. Image encoding and compression module; 9. Defect re-inspection module. Detailed Implementation

[0021] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; 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; they can refer to the internal communication of two components or the interaction between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0023] Please see Figure 1-8 This invention provides a technical solution: an automotive adhesive coating quality inspection system based on machine vision algorithms, comprising an image acquisition unit 1, an image preprocessing unit 2, an adhesive coating feature extraction unit 3, a defect identification and analysis unit 4, a result output and early warning unit 5, a system calibration and feedback unit 6, and a data management unit 7. The image acquisition unit 1 and the image preprocessing unit 2 are connected by a signal; the image preprocessing unit 2 and the adhesive coating feature extraction unit 3 are connected by a signal; the adhesive coating feature extraction unit 3 and the defect identification and analysis unit 4 are connected by a signal; the defect identification and analysis unit 4 and the result output and early warning unit 5 are connected by a signal; the defect identification and analysis unit 4 and the system calibration and feedback unit 6 are connected by a signal; and the system calibration and feedback unit 6 and the data management unit 7 are connected by a signal. Image acquisition unit 1 is used to acquire raw images of the adhesive-coated area of ​​the car. Image preprocessing unit 2 is used to perform denoising, enhancement and binarization preprocessing on the original image; The adhesive coating feature extraction unit 3 is used to extract the geometric and grayscale features of the adhesive coating using machine vision algorithms; Defect identification and analysis unit 4 is used to identify the type of adhesive coating defect and analyze the severity of the defect based on a classifier model; The result output and early warning unit 5 is used to output the detection results and issue an early warning when the defects exceed the standard. System calibration and feedback unit 6 is used to periodically calibrate visual inspection parameters and receive feedback to optimize the inspection model; Data management unit 7 is used to store historical detection data and defect sample data.

[0024] Furthermore, the image acquisition unit 1 includes a multispectral camera module 11, a light source control module 12, and a trigger synchronization module 13. The output terminal of the multispectral camera module 11 is connected to the light source control module 12, and the output terminal of the light source control module 12 is connected to the trigger synchronization module 13. The multispectral camera module 11 is used to acquire images of the adhesive coating in different spectral bands; The light source control module 12 is used to automatically adjust the intensity of the lighting source according to the ambient brightness; The trigger synchronization module 13 is used to trigger the camera to capture images based on the vehicle's transmission position and ensure that the images are synchronized with the glue application position.

[0025] Furthermore, the image preprocessing unit 2 includes a filtering and denoising module 21, a contrast enhancement module 22, and an image segmentation module 23. The output of the filtering and denoising module 21 is connected to the contrast enhancement module 22, and the output of the contrast enhancement module 22 is connected to the image segmentation module 23. The filtering and denoising module 21 is used to remove image noise using median filtering or Gaussian filtering; Contrast enhancement module 22 is used to enhance the contrast between the coated area and the background using a histogram equalization method; Image segmentation module 23 is used to extract the region of interest in the coated area using threshold segmentation or edge detection algorithms.

[0026] Furthermore, the adhesive coating feature extraction unit 3 includes a geometric feature calculation module 31, a grayscale feature statistics module 32, and a feature dimensionality reduction module 33. The output of the geometric feature calculation module 31 is connected to the feature dimensionality reduction module 33 by signal, and the output of the grayscale feature statistics module 32 is connected to the feature dimensionality reduction module 33 by signal. The geometric feature calculation module 31 is used to calculate the width, continuity, offset, and area of ​​the adhesive line. The grayscale feature statistics module 32 is used to statistically analyze the grayscale mean, variance, and texture parameters of the glued area. The feature dimensionality reduction module 33 is used to reduce feature redundancy by employing principal component analysis.

[0027] The formula for calculating the adhesive application width using the geometric feature calculation module 31 is as follows:

[0028] Where W is the average adhesive width, and n is the total number of detection points selected along the adhesive line direction. and These are the x-coordinates of the right and left edges of the adhesive line at the i-th detection point, respectively.

[0029] Furthermore, the defect identification and analysis unit 4 includes a classifier model module 41, a defect type determination module 42, and a severity assessment module 43. The output of the classifier model module 41 is connected to the defect type determination module 42, and the output of the defect type determination module 42 is connected to the severity assessment module 43. Classifier model module 41 is used to classify the adhesive application features using support vector machines or convolutional neural networks; The defect type determination module 42 is used to identify defect types including glue breakage, glue deviation, glue buildup, and air bubbles. The severity assessment module 43 is used to classify defect levels based on defect size or area percentage.

[0030] The formula for calculating the defect level score S in the severity assessment module 43 is as follows:

[0031] in, The total length of the defect along the glue line direction. For the total length of the adhesive applied, The area of ​​the defect region. This represents the total area of ​​the adhesive region. and The weight coefficients are given and satisfy the following conditions: + =1.

[0032] Furthermore, the result output and early warning unit 5 includes a detection result display module 51, an audible and visual alarm module 52, and a report generation module 53. The output terminal of the detection result display module 51 is connected to the audible and visual alarm module 52, and the output terminal of the detection result display module 51 is connected to the report generation module 53. The test result display module 51 is used to display the glue coating quality test results and defect location markings in real time on the monitoring interface; The audible and visual alarm module 52 is used to issue an audible and visual alarm when the detected defect level exceeds a set threshold. The report generation module 53 is used to generate statistical reports on adhesive coating quality inspection by shift or batch.

[0033] Furthermore, the system calibration and feedback unit 6 includes a parameter calibration module 61, a feedback optimization module 62, and a model update module 63. The output of the parameter calibration module 61 is connected to the feedback optimization module 62 by a signal, and the output of the feedback optimization module 62 is connected to the model update module 63 by a signal. The parameter calibration module 61 is used to periodically calibrate the camera's intrinsic parameters, light source brightness, and detection threshold. Feedback optimization module 62 is used to adjust the feedback signal optimization classifier threshold based on the results of manual review or the parameters of the glue applicator; The model update module 63 is used to incrementally train and update the classifier model using newly acquired defect samples.

[0034] The model update module 63 uses the following formula to update the classifier model parameters through incremental learning: in, Let be the current model parameter vector. For learning rate, Let the loss function L be applied to the current new sample ( gradient at ) For feedback adjustment coefficient, These are the target parameters obtained based on manual verification results or after calibration of the glue applicator parameters.

[0035] Furthermore, the data management unit 7 includes a historical data storage module 71, a defect sample library module 72, and a query and traceability module 73. The output of the historical data storage module 71 is connected to the defect sample library module 72 by a signal, and the output of the defect sample library module 72 is connected to the query and traceability module 73 by a signal. Historical data storage module 71 is used to save all original detection images, processing results, and early warning records; The defect sample library module 72 is used to classify and store various defect images and feature data; The query and traceability module 73 is used to retrieve historical test data based on time, vehicle model, or glue application area.

[0036] Furthermore, an image encoding and compression module 8 is provided between the image acquisition unit 1 and the adhesive application feature extraction unit 3. The image encoding and compression module 8 is used to compress and encode the preprocessed image to reduce the data transmission bandwidth.

[0037] Furthermore, a defect re-inspection module 9 is provided between the defect identification and analysis unit 4 and the result output and early warning unit 5. The defect re-inspection module 9 is used to perform a second image acquisition and feature comparison on the glued area that was initially determined to be defective in order to reduce the false detection rate.

[0038] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A machine vision algorithm-based automotive adhesive coating quality inspection system, comprising an image acquisition unit (1), an image preprocessing unit (2), an adhesive coating feature extraction unit (3), a defect identification and analysis unit (4), a result output and early warning unit (5), a system calibration and feedback unit (6), and a data management unit (7), characterized in that: The image acquisition unit (1) is connected to the image preprocessing unit (2) by signal, the image preprocessing unit (2) is connected to the glue application feature extraction unit (3) by signal, the glue application feature extraction unit (3) is connected to the defect identification and analysis unit (4) by signal, the defect identification and analysis unit (4) is connected to the result output and early warning unit (5) by signal, the defect identification and analysis unit (4) is connected to the system calibration and feedback unit (6) by signal, and the system calibration and feedback unit (6) is connected to the data management unit (7). The image acquisition unit (1) is used to acquire the original image of the glued area of ​​the car; The image preprocessing unit (2) is used to perform noise reduction, enhancement and binarization preprocessing on the original image; The adhesive coating feature extraction unit (3) is used to extract the geometric and grayscale features of the adhesive coating using machine vision algorithms; The defect identification and analysis unit (4) is used to identify the type of adhesive coating defect and analyze the severity of the defect based on the classifier model; The result output and early warning unit (5) is used to output the detection results and issue an early warning when the defects exceed the standard; The system calibration and feedback unit (6) is used to periodically calibrate visual inspection parameters and receive feedback to optimize the inspection model; The data management unit (7) is used to store historical detection data and defect sample data.

2. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The image acquisition unit (1) includes a multispectral camera module (11), a light source control module (12), and a trigger synchronization module (13). The output terminal of the multispectral camera module (11) is connected to the light source control module (12), and the output terminal of the light source control module (12) is connected to the trigger synchronization module (13). The multispectral camera module (11) is used to acquire adhesive coating images in different spectral bands; The light source control module (12) is used to automatically adjust the intensity of the lighting source according to the ambient brightness; The trigger synchronization module (13) is used to trigger the camera to capture images based on the vehicle's transmission position and ensure that the images are synchronized with the glue application position.

3. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The image preprocessing unit (2) includes a filtering and denoising module (21), a contrast enhancement module (22), and an image segmentation module (23). The output of the filtering and denoising module (21) is connected to the contrast enhancement module (22), and the output of the contrast enhancement module (22) is connected to the image segmentation module (23). The filtering and denoising module (21) is used to remove image noise using median filtering or Gaussian filtering; The contrast enhancement module (22) is used to enhance the contrast between the coated area and the background using a histogram equalization method; The image segmentation module (23) is used to extract the region of interest in the coated area using threshold segmentation or edge detection algorithms.

4. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The adhesive coating feature extraction unit (3) includes a geometric feature calculation module (31), a gray-scale feature statistics module (32), and a feature dimensionality reduction module (33). The output of the geometric feature calculation module (31) is connected to the feature dimensionality reduction module (33), and the output of the gray-scale feature statistics module (32) is connected to the feature dimensionality reduction module (33). The geometric feature calculation module (31) is used to calculate the width, continuity, offset and area of ​​the adhesive line; The grayscale feature statistics module (32) is used to statistically analyze the grayscale mean, variance and texture parameters of the coated area; The feature dimensionality reduction module (33) is used to reduce feature redundancy by employing principal component analysis.

5. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The defect identification and analysis unit (4) includes a classifier model module (41), a defect type determination module (42), and a severity assessment module (43). The output of the classifier model module (41) is connected to the defect type determination module (42), and the output of the defect type determination module (42) is connected to the severity assessment module (43). The classifier model module (41) is used to classify the adhesive coating features using a support vector machine or a convolutional neural network; The defect type determination module (42) is used to identify defect types including glue breakage, glue deviation, glue buildup, and bubbles; The severity assessment module (43) is used to classify defect levels based on defect size or area ratio.

6. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The result output and early warning unit (5) includes a detection result display module (51), an audible and visual alarm module (52), and a report generation module (53). The output end of the detection result display module (51) is connected to the audible and visual alarm module (52), and the output end of the detection result display module (51) is connected to the report generation module (53). The test result display module (51) is used to display the glue coating quality test results and defect location markings on the monitoring interface in real time; The audible and visual alarm module (52) is used to issue an audible and visual alarm when the detected defect level exceeds a set threshold. The report generation module (53) is used to generate adhesive coating quality inspection statistical reports by shift or batch.

7. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The system calibration and feedback unit (6) includes a parameter calibration module (61), a feedback optimization module (62) and a model update module (63). The output of the parameter calibration module (61) is connected to the feedback optimization module (62), and the output of the feedback optimization module (62) is connected to the model update module (63). The parameter calibration module (61) is used to periodically calibrate the camera intrinsic parameters, light source brightness, and detection threshold. The feedback optimization module (62) is used to adjust the threshold of the feedback signal optimization classifier according to the manual review results or the parameters of the glue applicator; The model update module (63) is used to incrementally train and update the classifier model using newly collected defect samples.

8. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: The data management unit (7) includes a historical data storage module (71), a defect sample library module (72), and a query and traceability module (73). The output of the historical data storage module (71) is connected to the defect sample library module (72), and the output of the defect sample library module (72) is connected to the query and traceability module (73). The historical data storage module (71) is used to save all original detection images, processing results and early warning records; The defect sample library module (72) is used to classify and store various defect images and feature data; The query and traceability module (73) is used to retrieve historical test data based on time, vehicle model, or glue application area.

9. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: An image encoding and compression module (8) is also provided between the image acquisition unit (1) and the glue application feature extraction unit (3). The image encoding and compression module (8) is used to compress and encode the preprocessed image to reduce the data transmission bandwidth.

10. The automotive adhesive coating quality inspection system based on machine vision algorithm according to claim 1, characterized in that: A defect re-inspection module (9) is also provided between the defect identification and analysis unit (4) and the result output and early warning unit (5). The defect re-inspection module (9) is used to perform a second image acquisition and feature comparison on the glued area that was initially determined to be defective in order to reduce the false detection rate.

Images