Method for detecting gluing quality of sealant

Abstract

The invention discloses a sealant gluing quality detection method, which comprises the following steps: S1, acquiring images of a vehicle body sealant gluing area, and matching the images; s2, carrying out binarization processing on the corrected image, setting a target point and a width threshold value in the binarized image according to the vehicle model, measuring the gluing widths of the two sides of the target point, and comparing and judging the gluing centering degree; s3, performing gridding processing on the corrected image, equally dividing a gluing area in the gridding image along the length direction, setting the area of a target gluing area and a gluing coverage rate threshold value, calculating the area of an actual gluing area, and calculating the area of the actual gluing area; according to the area of the actual gluing area, the area of the target gluing area and the gluing coverage rate, whether missed gluing and glue breaking occur or not is judged; and S4, collecting image information in the step S2 and the step S3, importing the image information into a visual deep learning system, and establishing a deep learning model. The detection method disclosed by the invention can be used for comprehensively judging the quality according to the centering degree of gluing and whether missed gluing and broken gluing occur or not.

Description

Technical Field

[0001] The present invention relates to the technical field of glue coating quality detection, and more specifically, to a sealant glue coating quality detection method. Background Art

[0002] In the process of vehicle manufacturing, each process in the paint shop not only needs to ensure the commercial quality of the paint surface, but also plays a vital role in ensuring the quality of the vehicle's anti-rust, waterproof and other functional properties. Among them, the sealant process is a key link to ensure the waterproof function. The main work content is to spray the sealant on the white body seam by manual or robot, and then adhere to the body seam after drying to achieve the seam sealing. In the process of vehicle manufacturing, the glue coating process is a key link and plays an important role in the waterproof function of the whole vehicle. It needs to be sealed and covered on the white body seam. The quality management requirements include but are not limited to: no virtual holes in the seal; no deviation from the seam; no splashing after spraying, and no obstruction to the installation of parts; the visible surface sealant needs to ensure the commercial quality of the appearance.

[0003] With the widespread application of automated spraying of sealants, not only has the spraying efficiency been improved and the labor cost been reduced, but compared with manual spraying, the stability, uniformity and high appearance of the sealant spraying have been significantly improved, thereby improving the waterproof function of the car. The subsequent technical improvement topic is to improve the quality confirmation and detection and monitoring capabilities of the sealant after spraying. Traditional quality inspection methods often rely on manual inspection, which is not only inefficient, but also easily affected by subjective factors, resulting in insufficient accuracy and consistency of the test results. With the widespread application of automated spraying, quality inspection after spraying also needs to introduce more efficient and intelligent quality inspection solutions with the development of intelligent manufacturing and artificial intelligence technology.

[0004] The prior art CN201810962290.6 discloses a method for visually following and detecting automobile gluing. The camera follows the glue gun to obtain the gluing image in sections, and the image processing center evaluates the gluing quality in sections. The gluing image acquisition includes: obtaining the workpiece digital model and the distribution of the glue strip; determining the movement trajectory and posture of the glue gun; determining the number and installation position of the camera; following and obtaining each section of the gluing image; image processing judgment includes: rasterizing and marking the glue strip segment image and performing width detection; if it meets the error range, the quality is qualified. This technical solution can only perform width detection, the detection type is single, and the detection reliability is low. Summary of the invention

[0005] The purpose of the present invention is to overcome the shortcomings of the prior art of single detection type and low detection reliability, and to propose a sealant coating quality detection method, which can judge the centering degree of the glue coating and whether there is any missing coating or glue breakage, and has high detection reliability.

[0006] To achieve the above object, a sealant coating quality detection method of the present invention comprises the following steps:

[0007] S1: collecting images of the sealant coating area of ​​the vehicle body and matching the images;

[0008] S2: Binarize the corrected image, set a target point and a width threshold in the binarized image according to the vehicle model, measure the glue width on both sides of the target point, and compare the width of the glue area on both sides according to the width threshold to determine the glue centering degree;

[0009] S3: performing grid processing on the corrected image, dividing the glue-coated area equally along the length direction in the grid-coated image, setting the area of ​​the target glue-coated area and the glue-coating coverage threshold, calculating the area of ​​the actual glue-coated area, and judging whether there is missing glue and glue breakage according to the area of ​​the actual glue-coated area, the area of ​​the target glue-coated area and the glue-coating coverage;

[0010] S4: Collect the image information in step S2 and step S3 and import it into the visual deep learning system to establish a deep learning model.

[0011] In this technical solution, the vehicle body is photographed by an industrial camera next to the vehicle body conveyor line, and then the image is imported into a computer, and the image is processed by visual technology. Step S1 uses a template matching function to select specific image features as a template, and preset parameters to determine the search space, and search for targets similar to the template in the image to achieve business needs such as positioning, counting, and judging the presence or absence. The matched image is used as the basis for subsequent judgment of the glue quality. Step S2 can check the centering of the glue. Step S3 uses a segmented full inspection method for segmented inspection. This step can detect the coverage of the glue and detect whether there is any missing or broken coating. Step S4 can facilitate the traceability of the glue quality, and use deep learning to reduce the false detection rate of visual recognition, thereby improving the detection accuracy.

[0012] As a preferred solution, the image matching in step S1 includes: creating a template based on specific image features, searching for a target matching the template in the image through the template, and correcting the motion offset of the target based on the template matching algorithm of edge points, which can correct the motion offset of the target and achieve precise positioning.

[0013] As a preferred solution, the motion offset correction of the target includes: obtaining position information of template matching, creating a reference point in the reference image based on the position information, calculating information such as the offset and rotation angle relative to the reference point in the image, correcting the image according to the offset and rotation angle and other information, marking the reference point in the image to be corrected, the reference point in the image to be corrected has the same coordinates as the reference point in the reference image, finding the running point in the image to be corrected based on the target matched in the image to be corrected by template matching, the position of the running point relative to the target is consistent with the position of the reference point in the reference image relative to the target, and in the image to be corrected, by comparing the positions of the running point and the reference point, the posture change information of the target in the image to be corrected relative to the reference image can be calculated.

[0014] As a preferred solution, the template matching algorithm based on edge points is:

[0015]

[0016] Where: F m is the edge point feature of the template; F s is the edge point feature of the search image; x is the horizontal coordinate of the template in the search image; y is the vertical coordinate of the template in the search image; θ is the rotation angle of the template; scaleX is the scale ratio of the template in the horizontal direction; scaleY is the scale ratio of the template in the vertical direction. This operator is used to find a set of parameters (F m , F s ; x, y, θ, scaleX, scaleY), these parameters define the position and posture of the template in the image. By maximizing this similarity function, the best matching position and posture of the template in the search image can be found.

[0017] As a preferred solution, the target point includes: a positioning hole, a process hole and a gap, wherein when the target point is a gap, the distance to the gap can be detected.

[0018] As a preferred solution, in order to determine the centering degree of the glue coating, the measurement of the width of the glue coating area on both sides of the target point in step S2 includes: setting a reference point, measuring the distance from the reference point to the edge of the glue coating area on the left side of the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L2, measuring the distance from the reference point to the edge of the glue coating area on the right side of the target point and recording it as L3, and the width of the glue coating area on both sides of the target point is calculated by the following formula:

[0019] LA=L1-L2

[0020] LB=L2-L3

[0021] Where: LA is the left side glue coating width; LB is the right side glue coating width of the target point;

[0022] Set the glue coating width to LC and the width threshold to 1 / 3LC. If LA and LB are both greater than or equal to 1 / 3LC, it is considered qualified, otherwise it is unqualified.

[0023] As a preferred solution, in order to detect whether there is any missing or broken glue coating, the area of ​​the actual glue coating area is calculated in step S3 and recorded as SA, the area of ​​the target glue coating area is set and recorded as SB, and the glue coating coverage threshold is 80%. If SA is greater than or equal to 80% SB, it is judged as qualified, otherwise it is unqualified.

[0024] As a preferred solution, in order to improve the accuracy of detection, ensure data diversity, and cover various possible defect situations, collecting image information in step S4 includes: collecting qualified and unqualified image data, marking the defect location, preprocessing the image, and dividing the image data set into a training set, a verification set, and a test set.

[0025] As a preferred solution, the process of establishing a deep learning model in step S4 includes: selecting a suitable deep learning model, designing a network structure, training the model using the training set, optimizing model parameters, evaluating model performance on the validation set, adjusting hyperparameters to prevent overfitting, deploying the trained model to the AI ​​tool system, testing the model using the test set, evaluating its detection rate in actual applications, and optimizing the model based on the test results to improve the detection rate.

[0026] As a preferred solution, preprocessing the image includes scaling, cropping and rotating the image to increase the generalization ability of the data, and designing the network structure includes determining the number of layers, the number of neurons and the activation function.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] 1. The present invention can detect the centering degree and coverage degree of the glue coating through visual recognition, and comprehensively judge the quality of the glue coating. It will not be affected by the subjective factors and fatigue of the operator, and the detection stability is high.

[0029] 2. The present invention can detect the centering degree of glue coating by setting the target point and the reference point, calculating the glue coating width on both sides of the target point and then comparing the width threshold. It is suitable for positioning holes, process holes and gaps of the vehicle body, has a wide range of applications and accurate detection.

[0030] 3. The present invention performs binarization processing on the image to facilitate the distinction and identification of the glue-coated area and the non-glue-coated area, and then compares the glue coverage threshold to determine whether there is any missing coating or glue breakage.

[0031] 4. The sealant coating quality detection method of the present invention is a contactless detection method, which avoids potential damage to the coating and is suitable for delicate and sensitive coating processes.

[0032] 5. By importing image information into the visual deep learning system, a deep learning model is established to improve the detection rate, reduce false detections, and facilitate quality traceability. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 Flow chart of the sealant coating quality detection method of the present invention

[0034] Figure 2 is a flowchart of image matching in step S1;

[0035] Figure 3 is a reference image for calculating the glue centering degree in step S2;

[0036] Figure 4 is the reference image of the normal glued image in step S3;

[0037] Figure 5 It is a reference image of the missed coating and broken glue images in step S3.

[0038] Figure 6 is the reference image for the centering degree detection in step S2;

[0039] Figure 7 This is a reference diagram of the normal installation of the sealing rubber plug in Example 3;

[0040] Figure 8 This is a reference diagram of the sealing rubber plug not being installed in Example 3. DETAILED DESCRIPTION

[0041] The drawings are only for illustrative purposes and cannot be construed as limiting the present invention. To better illustrate the present embodiment, some parts of the drawings may be omitted, enlarged, or reduced, and do not represent the size of the actual product. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the drawings. The positional relationships described in the drawings are only for illustrative purposes and cannot be construed as limiting the present invention.

[0042] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "long", "short" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limitations on this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0043] The technical solution of the present invention is further described in detail below through specific embodiments and in conjunction with the accompanying drawings:

[0044] Embodiment 1:

[0045] like Figure 1 As shown, this embodiment provides a sealant coating quality detection method, comprising the following steps:

[0046] S1: collecting images of the sealant coating area of ​​the vehicle body and matching the images;

[0047] S2: Binarize the corrected image, set a target point and a width threshold in the binarized image according to the vehicle model, measure the glue width on both sides of the target point, and compare the width of the glue area on both sides according to the width threshold to determine the glue centering degree;

[0048] S3: performing grid processing on the corrected image, dividing the glue-coated area equally along the length direction in the grid-coated image, setting the area of ​​the target glue-coated area and the glue-coating coverage threshold, calculating the area of ​​the actual glue-coated area, and judging whether there is missing glue and glue breakage according to the area of ​​the actual glue-coated area, the area of ​​the target glue-coated area and the glue-coating coverage;

[0049] S4: Collect the image information in step S2 and step S3 and import it into the visual deep learning system to establish a deep learning model.

[0050] In this embodiment, in the binarization processing of the image, the actual coated area is set to white and the uncoated area is set to black, so as to facilitate the distinction and identification of the coated area and the uncoated area. Rasterization can convert vector graphics into pixel images, so that the user can precisely control each pixel, which is convenient for equal division, distance measurement, width calculation and area calculation.

[0051] Specifically, Figure 2As shown, in step S1, the image matching includes: creating a template based on specific image features, searching for a target matching the template in the image through the template, and correcting the motion offset of the target based on the template matching algorithm of edge points.

[0052] In this embodiment, a reference point is marked in the image to be corrected, and the coordinates of the reference point in the image to be corrected are the same as those of the reference point in the reference image. Based on the target matched in the image to be corrected by template matching, the running point is found in the image to be corrected, and the position of the running point relative to the target is consistent with the position of the reference point in the reference image relative to the target. In the image to be corrected, the positions of the running point and the reference point are compared, so as to calculate the posture change information of the target in the image to be corrected relative to the reference image.

[0053] Specifically, the motion offset correction of the target includes: obtaining position information of template matching, creating a reference point in the reference image based on the position information, calculating information such as the offset and rotation angle in the image relative to the reference point, and correcting the image according to the information such as the offset and rotation angle.

[0054] In this embodiment, the template matching algorithm based on edge points is:

[0055]

[0056] Where: F m is the edge point feature of the template; F s is the edge point feature of the search image; x is the horizontal coordinate of the template in the search image; y is the vertical coordinate of the template in the search image; θ is the rotation angle of the template; scaleX is the scaling ratio of the template in the horizontal direction; scaleY is the scaling ratio of the template in the vertical direction.

[0057] In this embodiment, the target points include: positioning holes, process holes and gaps.

[0058] Specifically, Figure 3 , 6 As shown, in step S2, measuring the width of the glued area on both sides of the target point includes: setting a reference point, measuring the distance from the reference point to the edge of the glued area on the left side of the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L2, measuring the distance from the reference point to the edge of the glued area on the right side of the target point and recording it as L3, and the width of the glued area on both sides of the target point is calculated by the following formula:

[0059] LA=L1-L2

[0060] LB=L2-L3

[0061] Where: LA is the left side glue coating width; LB is the right side glue coating width of the target point;

[0062] Set the glue coating width to LC and the width threshold to 1 / 3LC. If LA and LB are both greater than or equal to 1 / 3LC, it is considered qualified, otherwise it is unqualified.

[0063] Specifically, collecting image information in step S4 includes: collecting qualified and unqualified image data, marking defect locations, preprocessing the images, and dividing the image data set into a training set, a verification set, and a test set.

[0064] Specifically, the process of establishing a deep learning model in step S4 includes: selecting a suitable deep learning model, designing a network structure, training the model using the training set, evaluating the model performance on the validation set, deploying the trained model to the AI ​​tool system, testing the model using the test set, evaluating its detection rate in actual applications and optimizing the model based on the test results.

[0065] In this embodiment, preprocessing the image includes scaling, cropping and rotating the image, and designing the network structure includes determining the number of layers, the number of neurons and the activation function.

[0066] In this embodiment, the AI ​​tool system is integrated with the PLC control system on the production line, and a detection threshold is set. When the detection rate is abnormal, the PLC alarm is triggered, and the alarm signal can be transmitted to the terminal device of the on-site manager in real time. New pictures of good and defective products are collected regularly, the data set is updated, and the model is continuously trained and optimized to improve the detection rate and ensure that the detection effect is stable and reliable.

[0067] Embodiment 2:

[0068] This embodiment is similar to Embodiment 1, except that, in this embodiment, Figure 4 , 5 As shown, in step S3, the actual glue coating area is calculated and recorded as SA, the target glue coating area is set and recorded as SB, and the glue coating coverage threshold is 80%. If SA is greater than or equal to 80% SB, it is judged as qualified, otherwise it is unqualified.

[0069] In this embodiment, the white part is the actual glue-coated area, and the black part is the un-glue-coated area. Figure 4 It is the reference image of the normal glued image; Figure 5 Reference image for leaks and glue breaks.

[0070] In this embodiment, the collected images are transmitted to the visual system via a cable or wireless network. Subsequently, these analog signals are converted into digital signals for processing and analysis by the visual system. After receiving the digital images, the visual system processes them. Through feature extraction technology, the visual system identifies the key features of the target object, such as the shape and color of the rubber strip. These features help to distinguish different objects and provide a basis for subsequent area calculations. Once the features of the target object are accurately identified, the visual system can calculate its area. This process usually involves geometric principles and polygon filling algorithms. Finally, the detection results are output in the form of data.

[0071] Embodiment 3:

[0072] This embodiment is similar to Embodiment 1, except that, in this embodiment, after the image is binarized, the visual system can identify the holes on the bottom of the vehicle where the sealing plugs are missing. Through a preset image processing algorithm, the system can detect the features of the holes and confirm whether the sealing plugs are present. The missing holes may present color or texture features different from the surrounding areas. The visual system can identify these differences through comparative analysis, thereby determining whether the sealing plugs are missing.

[0073] In this embodiment, if Figure 7 The figure shown is a reference diagram for the normal installation of the sealing plug. Figure 8 The figure shown is a reference diagram for leaking the sealing plug.

[0074] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

[0075] Obviously, the above embodiments of the present invention are only examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A sealant coating quality detection method, characterized in that: The following steps are involved: S1: collecting images of the sealant coating area of ​​the vehicle body and matching the images; S2: Binarize the corrected image, set a target point and a width threshold in the binarized image according to the vehicle model, measure the glue width on both sides of the target point, and compare the width of the glue area on both sides according to the width threshold to determine the glue centering degree; S3: performing grid processing on the corrected image, dividing the glue-coated area equally along the length direction in the grid-coated image, setting the area of ​​the target glue-coated area and the glue-coating coverage threshold, calculating the area of ​​the actual glue-coated area, and judging whether there is missing glue and glue breakage according to the area of ​​the actual glue-coated area, the area of ​​the target glue-coated area and the glue-coating coverage; S4: Collect the image information in step S2 and step S3 and import it into the visual deep learning system to establish a deep learning model.

2. A sealant coating quality detection method according to claim 1, characterized in that: In step S1, the image matching includes: creating a template based on specific image features, searching for a target matching the template in the image through the template, and correcting the motion offset of the target based on the template matching algorithm of edge points.

3. A sealant coating quality detection method according to claim 2, characterized in that: The motion offset correction of the target includes: obtaining position information of template matching, creating a reference point in the reference image based on the position information, calculating information such as the offset and rotation angle relative to the reference point in the image, and correcting the image according to the information such as the offset and rotation angle.

4. A sealant coating quality detection method according to claim 2, characterized in that: The template matching algorithm based on edge points is: Where: F m is the edge point feature of the template; F s is the edge point feature of the search image; x is the horizontal coordinate of the template in the search image; y is the vertical coordinate of the template in the search image; θ is the rotation angle of the template; scaleX is the scaling ratio of the template in the horizontal direction; scaleY is the scaling ratio of the template in the vertical direction.

5. A sealant coating quality detection method according to claim 1, characterized in that: The target points include: positioning holes, process holes and gaps.

6. A sealant coating quality detection method according to claim 5, characterized in that: Measuring the width of the glued area on both sides of the target point in step S2 includes: setting a reference point, measuring the distance from the reference point to the edge of the glued area on the left side of the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L1, measuring the distance from the reference point to the target point and recording it as L2, measuring the distance from the reference point to the edge of the glued area on the right side of the target point and recording it as L3, and the width of the glued area on both sides of the target point is calculated by the following formula: LA=L1-L2 LB=L2-L3 Where: LA is the left side glue coating width; LB is the right side glue coating width of the target point; Set the glue coating width to LC and the width threshold to 1 / 3LC. If LA and LB are both greater than or equal to 1 / 3LC, it is considered qualified, otherwise it is unqualified.

7. A sealant coating quality detection method according to claim 1, characterized in that: In step S3, the actual glue coating area is calculated and recorded as SA, the target glue coating area is set and recorded as SB, and the glue coating coverage threshold is 80%. If SA is greater than or equal to 80% SB, it is judged as qualified, otherwise it is unqualified.

8. A sealant coating quality detection method according to claim 1, characterized in that: Collecting image information in step S4 includes: collecting qualified and unqualified image data, marking defect locations, preprocessing the images, and dividing the image data set into a training set, a verification set, and a test set.

9. A sealant coating quality detection method according to claim 8, characterized in that: The process of establishing a deep learning model in step S4 includes: selecting a suitable deep learning model, designing a network structure, training the model using the training set, evaluating the model performance on the validation set, deploying the trained model to the AI ​​tool system, testing the model using the test set, evaluating its detection rate in actual applications and optimizing the model based on the test results.

10. A sealant coating quality detection method according to claim 9, characterized in that: Image preprocessing includes scaling, cropping, and rotating the image, and network structure design includes determining the number of layers, the number of neurons, and the activation function.

Images