Visual inspection system of automobile door panel assembly

Abstract

The invention discloses a visual detection system for an automobile door panel assembly, and relates to the technical field of automatic detection, and the visual detection system comprises the steps: installing a high-precision industrial camera, and carrying out the synchronous collection of the multi-view image data of an automobile door panel; preprocessing the acquired multi-view image data of the automobile door panel; carrying out detail enhancement on the preprocessed multi-view image data of the automobile door panel by adopting a super-resolution algorithm; integrating the enhanced multi-view image data of the automobile door panel into an automobile door panel panorama through binocular stereoscopic vision matching, and constructing an automobile door panel three-dimensional model; the automobile door panel three-dimensional model is compared with a standard automobile door panel three-dimensional model, and the buckle defect in the automobile door panel is recognized; and carrying out position marking and defect type analysis on the identified defect points, and generating an automobile door panel detection report according to the defect type analysis. According to the invention, details are enhanced by using a super-resolution algorithm, defects are accurately positioned by using a three-dimensional model, and the detection efficiency and accuracy are improved.

Description

Technical Field

[0001] The invention relates to the technical field of automated detection, in particular to a visual detection system for an automobile door panel assembly. Background Art

[0002] With the rapid development of the automobile manufacturing industry, the quality control requirements for automobile parts are becoming increasingly higher, especially for components such as automobile door panel components that are directly related to vehicle safety and comfort. Traditional inspection methods usually rely on manual inspection, which is not only inefficient but also easily affected by human factors, resulting in unstable inspection results. In recent years, with the advancement of machine vision technology, more and more manufacturers have begun to adopt automated visual inspection systems to improve production efficiency and product quality. Such systems can achieve fast and accurate inspection of automobile door panel components, greatly reducing labor costs and improving the accuracy and consistency of inspection.

[0003] However, the existing visual inspection system still has certain limitations. First, during the image acquisition stage, the traditional single-view image acquisition method is difficult to fully capture all the details of the door panel, which may cause some defects to not be discovered in time. Secondly, when processing multi-view image data, the existing 3D model construction technology often leads to large model reconstruction errors due to inaccurate feature point matching, which in turn affects the subsequent defect recognition accuracy. Finally, although the existing defect recognition and classification algorithms can identify some defects, the recognition rate and classification accuracy still need to be improved when facing subtle defects or specific types of defects (such as buckle defects). The existence of these problems limits the effectiveness of existing visual inspection systems in practical applications, especially in the field of modern automobile manufacturing that pursues high precision and high reliability. Summary of the invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a visual inspection system for automobile door panel components to solve the problems of incomplete image acquisition, large three-dimensional model reconstruction error and low defect recognition rate in the prior art.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] An embodiment of the present invention provides a visual inspection system for an automobile door panel assembly, which comprises:

[0008] Image acquisition module: used to install high-precision industrial cameras to synchronously acquire multi-view image data of automobile door panels;

[0009] Image preprocessing module: used to preprocess the collected multi-view image data of automobile door panels;

[0010] Image enhancement module: used to use super-resolution algorithm to enhance the details of pre-processed multi-view image data of automobile door panels;

[0011] 3D model building module: used to integrate the enhanced multi-view image data of the automobile door panel into a panoramic image of the automobile door panel through binocular stereo vision matching, and build a 3D model of the automobile door panel;

[0012] Defect recognition module: compares the 3D model of the car door panel with the 3D model of the standard car door panel to identify the buckle defects in the car door panel;

[0013] Defect type identification and report generation module: used to mark the location of identified defect points and analyze the defect type, and generate an automobile door panel inspection report based on the defect type analysis.

[0014] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, wherein: the high-precision industrial camera is installed to synchronously collect multi-view image data of the automobile door panel, and the specific steps are as follows:

[0015] Basler ace series high-precision industrial cameras are deployed at key locations of the door inspection station, including the edges of the door panels, the middle area, and the buckle installation positions.

[0016] A unified trigger mechanism is set up to control all cameras to be triggered at the same time. When the car door panel enters the inspection station, all cameras are started through the trigger mechanism to synchronously collect multi-view image data of the car door panel.

[0017] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, the preprocessing of the collected multi-view image data of the automobile door panel includes the following specific steps:

[0018] Use median filtering to remove random noise from multi-view image data of automobile door panels;

[0019] Use histogram equalization to adjust the image brightness in multi-view image data of automobile door panels;

[0020] Enhance image contrast in multi-view image data using gamma correction.

[0021] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, wherein: the super-resolution algorithm is used to enhance the details of the pre-processed multi-view image data of the automobile door panel, and the specific steps are as follows:

[0022] The super-resolution algorithm is used to convert the low-resolution image in the pre-processed multi-view image data of the automobile door panel into a high-resolution image. The Laplace operator is used to sharpen the super-resolution image to enhance the edge details. The expression is:

[0023] I s =I HR +α·(L*SuperResolution ( I LR ));

[0024] Among them, I s is the enhanced car door panel image data, I HR It is a high-resolution image processed by the super-resolution algorithm, α is the coefficient that controls the sharpening intensity, L is the Laplace operator used for edge detection and sharpening, and SuperResolution is a super-resolution algorithm.

[0025] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, the image stitching algorithm is used to integrate the enhanced multi-view image data of the automobile door panel into a panoramic image and construct a three-dimensional model of the automobile door panel. The specific steps are as follows:

[0026] The feature point detection algorithm SIFT is used to extract feature points from the enhanced image data. The feature points include corner points, edge points and screw holes.

[0027] Match the feature points of each pair of images, find the corresponding feature point pairs, use the RANSAC algorithm to align the images, and use the image stitching algorithm to stitch the aligned images into a panoramic image of the car door panel;

[0028] Through binocular stereo vision matching, each pixel in the panoramic image of the car door panel is mapped to three-dimensional space. We can construct point cloud data from the depth information and use the point cloud data to build a three-dimensional model of the car door panel.

[0029] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, the specific steps of comparing the automobile door panel three-dimensional model with the standard three-dimensional model are as follows:

[0030] Load standard 3D models constructed according to production standards from the database;

[0031] Use the point cloud alignment algorithm to iteratively minimize the distance between the two point clouds and align the 3D model of the car door panel with the standard 3D model;

[0032] The aligned detection model is compared with the standard model to calculate the relative position difference between the buckle point in the standard 3D model and the 3D model of the car door panel. The expression is:

[0033] Δd(p)= | d(p,M s )-d(p,M a )|;

[0034] Where Δd(p) represents the difference between the buckle point p in the standard 3D model and the 3D model M of the car door panel. a The relative position difference in , d is a distance function used to calculate the position difference between the relative buckle points of the automobile door panel 3D model and the standard 3D model.

[0035] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, the specific steps of identifying the buckle defects in the automobile door panel are as follows:

[0036] Based on the understanding of the manufacturing process and design specifications of the automobile door panel, a manufacturing tolerance threshold T is defined;

[0037] Compare the distance difference Δd(p) of each point with the manufacturing tolerance threshold T to find all buckle defects and form a defect point set, which is expressed as:

[0038] δ={p∈D|Δd ( p ) >T};

[0039] Among them, D is the point cloud dataset containing all the buckle points of the car door panel, and δ is the set of identified defect points.

[0040] As a preferred solution of the visual inspection system of the automobile door panel assembly of the present invention, the steps of marking the position of the identified defect points and analyzing the defect types are as follows:

[0041] The position of the defect point p in the identified defect point set δ is marked, and the expression is:

[0042] L(δ)=(x p ,y p ,z p );

[0043] Among them, x p represents the horizontal coordinate of the defect point p in the defect set δ, y p represents the ordinate of the defect point p in the defect set δ, z p represents the height coordinate of the defect point p in the defect set δ, and L represents the specific coordinate of the defect point p in δ in the three-dimensional space;

[0044] The specific coordinates of the defect point L(δ) are input into YOLOv and combined with the enhanced car door panel image data I s Identify the defect type, the expression is:

[0045] T d (p) = YOLOv5(I s ,L(δ));

[0046] Among them, T d (p) represents the defect type of defect point p.

[0047] As a preferred solution of the visual inspection system for automobile door panel components of the present invention, the specific steps of generating an automobile door panel inspection report according to the defect type analysis result are as follows:

[0048] Collect the defect type T of each car door panel defect point p d (p), defect type T d (p) includes missing, misaligned, deformed and damaged information, according to the defect type T d (p) Generate a monitoring report, which includes the defect location, defect type and defect quantity in the automobile door panel.

[0049] The beneficial effects of the present invention are as follows: by deploying high-precision industrial cameras at key positions and synchronously collecting multi-view image data, comprehensive coverage of automobile door panel components is achieved, ensuring the integrity and consistency of image acquisition; subsequently, preprocessing techniques such as median filtering, histogram equalization and gamma correction are used to improve image quality and enhance image contrast and clarity; then, a super-resolution algorithm is used to enhance the details of the image, making the edges and details in the image more prominent, thereby improving the sensitivity of defect recognition; finally, an accurate three-dimensional model is constructed through feature point detection and matching, image stitching and binocular stereo vision matching technology, which greatly improves the accuracy of defect positioning, thereby achieving the efficiency, accuracy and reliability of the detection system as a whole. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0051] Figure 1 This is a flow chart of the visual inspection system for the automobile door panel assembly in Example 1.

[0052] Figure 2 This is a flow chart for constructing a three-dimensional model of a car door panel in Example 1. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the accompanying drawings.

[0054] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0055] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0056] Example 1, reference Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a visual inspection system for an automobile door panel assembly, comprising the following steps:

[0057] Image acquisition module: used to install high-precision industrial cameras to synchronously acquire multi-view image data of automobile door panels;

[0058] Furthermore, Basler ace series high-precision industrial cameras are deployed at key locations of the door inspection station, including the edges of the car door panels, the middle area and the buckle installation position, to ensure that all important areas around the door are covered;

[0059] A unified trigger mechanism is set up to control all cameras to trigger at the same time. When the car door panel enters the inspection station, all cameras are started through the trigger mechanism to synchronously collect multi-view image data of the car door panel. Specifically: by installing sensors or using photoelectric switches at the inspection station, when the car door panel enters the inspection area, the sensor detects the arrival signal of the door panel and immediately triggers the preset synchronization trigger signal. This signal is sent synchronously to all Basler ace series high-precision industrial cameras deployed in key positions, so that all cameras are started at the same time and accurately and synchronously collect image data of the car door panel from different angles, thereby ensuring the synchronization and consistency of image acquisition, and providing high-quality raw data for subsequent image processing. This design not only ensures the consistency and integrity of image acquisition, but also greatly improves the detection efficiency and data reliability, providing high-quality basic data support for subsequent image processing and defect identification.

[0060] Image preprocessing module: used to preprocess the collected multi-view image data of automobile door panels.

[0061] Furthermore, median filtering is used to remove random noise in the multi-view image data of the car door panel. Specifically, by applying the median filtering algorithm, the system processes the multi-view image data of the car door panel pixel by pixel. In each sliding window, the algorithm will select a central pixel and form a set with the pixel values ​​around it, and then select the median value in the set as the new value of the central pixel. This process effectively eliminates salt and pepper noise and other random noise in the image, retains the main features of the image, and thus improves the clarity of the image and the accuracy of subsequent processing.

[0062] The image brightness in the multi-view image data of the car door panel is adjusted using histogram equalization. The purpose of histogram equalization is to enhance the contrast of the image by transforming the distribution of the brightness value of the image. The histogram equalization technology calculates the grayscale distribution frequency of the original image and constructs the cumulative distribution function (CDF) based on it. Then, the CDF is used to map the original grayscale value to the new grayscale, thereby expanding the dynamic range of the image and enhancing the overall contrast of the image. This process ensures that even in poor lighting conditions, the details in the image can be better presented, which is conducive to subsequent image analysis and feature extraction.

[0063] Gamma correction is used to enhance the image contrast in multi-view image data. Specifically, the image tone response curve is adjusted through a nonlinear mapping function. The appropriate gamma value can make the dark details of the image clearer and the bright parts not overexposed, thereby improving the overall image contrast and making the various features in the image more distinct, which is convenient for subsequent defect identification and classification.

[0064] Image enhancement module: used to use super-resolution algorithm to enhance the details of pre-processed multi-view image data of automobile door panels;

[0065] Furthermore, a super-resolution algorithm is used to convert the low-resolution image in the pre-processed multi-view image data of the automobile door panel into a high-resolution image, and the super-resolution image is sharpened using the Laplacian operator to enhance the edge details. The expression is:

[0066] I s =I HR +α·(L*SuperResolution(I LR ));

[0067] Among them, I s is the enhanced car door panel image data, I HRIt is a high-resolution image processed by the super-resolution algorithm, α is the coefficient that controls the sharpening intensity, L is the Laplace operator used for edge detection and sharpening, and SuperResolution is a super-resolution algorithm.

[0068] It should be noted that SuperResolution processing is to process the pre-processed low-resolution image I LR , using neural networks and other methods to extract key features, then using upsampling technology to increase image resolution, and then using a dedicated model to optimize details to enhance clarity, and finally through post-processing such as sharpening and contrast adjustment to generate high-quality high-resolution images. LR This process improves image resolution through a super-resolution algorithm and sharpens it using the Laplacian operator, significantly enhancing the edge details and clarity of the image, thereby improving the accuracy of defect detection.

[0069] 3D model building module: used to integrate the enhanced multi-view image data of the automobile door panel into a panoramic image of the automobile door panel through binocular stereo vision matching, and build a 3D model of the automobile door panel;

[0070] Furthermore, the feature point detection algorithm SIFT is used to extract feature points from the enhanced image data. The feature points include corner points, edge points, screw holes and other areas with unique structures. Specifically: this process identifies the position of key points through scale space extreme value detection, and uses the direction assignment algorithm to assign the main direction to each key point, thereby ensuring that the extracted feature points can remain invariant under different scales and rotation conditions.

[0071] Perform feature point matching on each pair of images to find the corresponding feature point pairs, use the RANSAC algorithm to align the images, and use the image stitching algorithm to stitch the aligned images into a panoramic image of the car door panel. For example, repeatedly test different feature point combinations through the RANSAC algorithm to eliminate incorrect matches and ensure that only correct corresponding points are used for image alignment, thereby generating a seamless panoramic image of the car door panel.

[0072] Through binocular stereo vision matching, each pixel in the panoramic image of the car door panel is mapped to three-dimensional space. We can construct point cloud data from the depth information and use the point cloud data to build a three-dimensional model of the car door panel.

[0073] It should also be noted that the specific process of constructing the three-dimensional model of the car door panel is as follows: through binocular stereo vision technology, we first obtain two images of the car door panel from different perspectives, and then extract and match the feature points of the two images to ensure that each pair of matched feature points represents the same physical point in the three-dimensional space. Next, the RANSAC (RANdomSAmple Consensus) algorithm is applied to optimize the matching results, remove mismatched points, and improve the matching accuracy. With reliable feature point matching, we can calculate the depth information between the matching points, that is, the distance from each point to the camera, through the principle of triangulation. Subsequently, this depth information is used to assign a three-dimensional coordinate to each image pixel point, thereby generating point cloud data. Finally, by processing and fusing the point cloud data, a three-dimensional model that accurately reflects the shape and structure of the car door panel can be constructed, so that we can observe the details of the car door panel from multiple angles, providing strong support for subsequent design, detection or repair work.

[0074] Defect recognition module: compares the 3D model of the car door panel with the 3D model of the standard car door panel to identify the buckle defects in the car door panel;

[0075] Furthermore, a standard 3D model constructed according to production standards is loaded from a database;

[0076] It should be noted that standard 3D models are usually accurately constructed by engineers based on the design drawings and production standards of the automotive door panels. These models contain all the geometric dimensions and structural features of the door panels to ensure consistency with actual production requirements. The standard 3D models are stored in the database as a benchmark for comparison during the inspection process to evaluate whether the actual produced automotive door panels meet the design and quality standards.

[0077] The point cloud alignment algorithm is used to iteratively minimize the distance between the two point clouds and align the 3D model of the car door panel with the standard 3D model. Specifically, the point cloud alignment algorithm is used to continuously adjust the position and posture of the 3D model of the car door panel through the iterative closest point (ICP) method, gradually reducing the distance between it and the standard 3D model until the two are optimally aligned. This process ensures that the 3D model of the car door panel accurately matches the standard model by repeatedly calculating and minimizing the point-to-point or point-to-plane distance between the two point clouds.

[0078] The aligned detection model is compared with the standard model to calculate the relative position difference between the buckle point in the standard 3D model and the 3D model of the car door panel. The expression is:

[0079] Δd(p)=|d(p,M s )-d(p,M a )|;

[0080] Where Δd(p) represents the difference between the buckle point p in the standard 3D model and the 3D model M of the car door panel. a The relative position difference in , d is a distance function used to calculate the position difference between the relative buckle points of the automobile door panel 3D model and the standard 3D model.

[0081] Based on the understanding of the manufacturing process and design specifications of the automobile door panel, a manufacturing tolerance threshold T is defined;

[0082] Compare the distance difference Δd(p) of each point with the manufacturing tolerance threshold T to find all buckle defects and form a defect point set, which is expressed as:

[0083] δ={p∈D|Δd(p)>T};

[0084] Among them, D is the point cloud dataset containing all the buckle points of the car door panel, and δ is the set of identified defect points.

[0085] It should also be noted that, based on the understanding of the manufacturing process and design specifications of the automobile door panel, a manufacturing tolerance threshold T is defined; then the distance difference Δd(p) of each point is compared with the manufacturing tolerance threshold T, and all snap points exceeding the threshold T are found, thereby identifying snap defects and forming a defect point set δ, where δ contains all snap points exceeding the tolerance range. This process ensures that only those points that do not really meet the manufacturing standards are marked as defects, thereby improving the accuracy and reliability of defect identification. For example, assume that this threshold T is set to 0.5 mm. Then the distance difference Δd(p) of each point is compared with the manufacturing tolerance threshold T, and all snap points exceeding this threshold are found to form a defect point set δ. Specifically, if the deviation Δd(p) of a snap point p is greater than 0.5 mm, it is marked as a defect point. In this way, through this process, all snap points exceeding the tolerance range are identified to form a defect point set δ, ensuring the accuracy and reliability of defect identification.

[0086] Defect type identification and report generation module: used to mark the location of identified defect points and analyze the defect type, and generate an automobile door panel inspection report based on the defect type analysis.

[0087] Furthermore, the position of the defect point p in the identified defect point set δ is marked, and the expression is:

[0088] L(δ)=(x p ,y p ,z p );

[0089] Among them, x p represents the horizontal coordinate of the defect point p in the defect set δ, y p represents the ordinate of the defect point p in the defect set δ, zp represents the height coordinate of the defect point p in the defect set δ, and L represents the specific coordinate of the defect point p in δ in the three-dimensional space;

[0090] The specific coordinates of the defect point L(δ) are input into YOLOv and combined with the enhanced car door panel image data I s Identify the defect type, the expression is:

[0091] T d (p) = YOLOv5(I s ,L(δ));

[0092] Among them, T d (p) represents the defect type of defect point p.

[0093] Collect the defect type T of each car door panel defect point p d (p), defect type T d (p) includes missing, misaligned, deformed and damaged information, according to the defect type T d (p) Generate a monitoring report, which includes the defect location, defect type and defect quantity in the automobile door panel.

[0094] It should also be noted that this process not only ensures that the location information of each defect point is accurately marked, but also further confirms the type of defect, such as missing, misaligned, deformed or damaged, through the YOLOv5 model. Finally, the system collects the type information of all defect points T d (p) and generate detailed monitoring reports, including the specific location, type and quantity of defects, thus providing precise data support for subsequent quality control and improvement.

[0095] In summary, the present invention achieves comprehensive coverage of automobile door panel components and ensures the integrity and consistency of image acquisition by: deploying high-precision industrial cameras at key positions and synchronously collecting multi-view image data; then, using preprocessing techniques such as median filtering, histogram equalization and gamma correction to improve image quality and enhance image contrast and clarity; then, using a super-resolution algorithm to enhance image details, making the edges and details in the image more prominent, thereby improving the sensitivity of defect recognition; finally, constructing an accurate three-dimensional model through feature point detection and matching, image stitching and binocular stereo vision matching technology, which greatly improves the accuracy of defect positioning, thereby achieving the efficiency, accuracy and reliability of the detection system as a whole.

[0096] Example 2, referring to Table 1, is the second example of the present invention. To further verify the technical solution of the present invention, experimental simulation data of a visual inspection system for an automobile door panel assembly are provided.

[0097] In order to verify the advantages of the technical solution of the present invention over the prior art, the following test was conducted. The test equipment includes a set of inspection stations equipped with Basler ace series high-precision industrial cameras, as well as a software platform for image preprocessing, enhancement, 3D model construction and defect identification. The test objects are 10 batches of 100 automobile door panels, of which 50 are manufactured by traditional processes and the other 50 are manufactured using the technical solution described in the present invention. All door panels are processed on the same production line to ensure the consistency of test conditions.

[0098] First, during the experimental preparation stage, an inspection station including a Basler ace series high-precision industrial camera was set up to ensure coverage of key positions of the car door panels, and synchronous triggering of image acquisition was achieved through photoelectric switches.

[0099] Secondly, in the image preprocessing stage, median filtering is used for denoising, histogram equalization is used to adjust brightness, and gamma correction is used to enhance contrast, thereby improving image quality; and super-resolution algorithm and Laplace operator are used to enhance image details.

[0100] Next, in the 3D model construction stage, SIFT is used to extract feature points, and binocular stereo vision matching technology is used to build a 3D model of the car door panel, which is then compared with the standard model to identify buckle defects.

[0101] Finally, in the defect type identification stage, the defect points are marked, the defect type is identified, and a monitoring report containing the defect location, type and quantity is generated.

[0102] The details are shown in the automobile door panel performance test table:

[0103] Table 1 Automobile door panel performance test table

[0104]

[0105] From the analysis of the above table, it can be seen that the scheme of the present invention is superior to the traditional process manufacturing in many aspects. For example, in terms of defect recognition accuracy, the door panels of the present invention reached 95%, 96%, and 97%, respectively, which is much higher than the 85%, 82%, and 84% of the traditional process door panels. This not only proves the high efficiency of the scheme of the present invention in defect recognition, but also shows its significant advantage in improving detection accuracy.

[0106] In summary, the technical solution of the present invention not only ensures the consistency and integrity of image acquisition, but also shows obvious technical progress in image processing, enhancement, three-dimensional model construction and defect identification. Through comparative analysis of test data, it can be clearly seen that the present invention has higher innovation and practicality than the prior art.

[0107] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A visual inspection system for automobile door panel components, characterized in that: include, Image acquisition module: used to install high-precision industrial cameras to synchronously acquire multi-view image data of automobile door panels; Image preprocessing module: used to preprocess the collected multi-view image data of automobile door panels; Image enhancement module: used to use super-resolution algorithm to enhance the details of pre-processed multi-view image data of automobile door panels; 3D model building module: used to integrate the enhanced multi-view image data of the automobile door panel into a panoramic image of the automobile door panel through binocular stereo vision matching, and build a 3D model of the automobile door panel; Defect recognition module: compares the 3D model of the car door panel with the 3D model of the standard car door panel to identify the buckle defects in the car door panel; Defect type identification and report generation module: used to mark the location of identified defect points and analyze the defect type, and generate an automobile door panel inspection report based on the defect type analysis.

2. The visual inspection system for automobile door panel components according to claim 1, characterized in that: The high-precision industrial camera is installed to synchronously collect multi-view image data of the automobile door panel. The specific steps are as follows: Basler ace series high-precision industrial cameras are deployed at key locations of the door inspection station, including the edges of the door panels, the middle area, and the buckle installation positions. A unified trigger mechanism is set up to control all cameras to be triggered at the same time. When the car door panel enters the inspection station, all cameras are started through the trigger mechanism to synchronously collect multi-view image data of the car door panel.

3. The visual inspection system for automobile door panel components according to claim 2, characterized in that: The preprocessing of the collected multi-view image data of the automobile door panel includes the following specific steps: Use median filtering to remove random noise from multi-view image data of automobile door panels; Use histogram equalization to adjust the image brightness in multi-view image data of automobile door panels; Enhance image contrast in multi-view image data using gamma correction.

4. The visual inspection system for automobile door panel components according to claim 3, characterized in that: The super-resolution algorithm is used to enhance the details of the pre-processed multi-view image data of the automobile door panel. The specific steps are as follows: The super-resolution algorithm is used to convert the low-resolution image in the pre-processed multi-view image data of the automobile door panel into a high-resolution image. The Laplace operator is used to sharpen the super-resolution image to enhance the edge details. The expression is: AND s =I HR +α (L*SuperResolution(I LR )); Among them, I s is the enhanced car door panel image data, I HR It is a high-resolution image processed by the super-resolution algorithm, α is the coefficient that controls the sharpening intensity, L is the Laplace operator used for edge detection and sharpening, and SuperResolution is a super-resolution algorithm.

5. The visual inspection system for automobile door panel components according to claim 4, characterized in that: The image stitching algorithm is used to integrate the enhanced multi-view image data of the automobile door panel into a panoramic image and construct a three-dimensional model of the automobile door panel. The specific steps are as follows: The feature point detection algorithm SIFT is used to extract feature points from the enhanced image data. The feature points include corner points, edge points and screw holes. Match the feature points of each pair of images, find the corresponding feature point pairs, use the RANSAC algorithm to align the images, and use the image stitching algorithm to stitch the aligned images into a panoramic image of the car door panel; Through binocular stereo vision matching, each pixel in the panoramic image of the car door panel is mapped to three-dimensional space. We can construct point cloud data from the depth information and use the point cloud data to build a three-dimensional model of the car door panel.

6. The visual inspection system for automobile door panel components according to claim 5, characterized in that: The specific steps of comparing the automobile door panel 3D model with the standard 3D model are as follows: Load standard 3D models constructed according to production standards from the database; Use the point cloud alignment algorithm to iteratively minimize the distance between the two point clouds and align the 3D model of the car door panel with the standard 3D model; The aligned detection model is compared with the standard model to calculate the relative position difference between the buckle point in the standard 3D model and the 3D model of the car door panel. The expression is: Δd(p)=|d(p,M s )-d(p,M a )|; Where Δd(p) represents the difference between the buckle point p in the standard 3D model and the 3D model M of the car door panel. a The relative position difference in , d is a distance function used to calculate the position difference between the relative buckle points of the automobile door panel 3D model and the standard 3D model.

7. The visual inspection system for automobile door panel components according to claim 6, characterized in that: The specific steps of identifying the buckle defect in the automobile door panel are as follows: Based on the understanding of the manufacturing process and design specifications of the automobile door panel, a manufacturing tolerance threshold T is defined; Compare the distance difference Δd(dp) of each point with the manufacturing tolerance threshold T to find all buckle defects and form a defect point set, which is expressed as: δ={p∈D|Δd ( p ) >T}; Among them, D is the point cloud dataset containing all the buckle points of the car door panel, and δ is the set of identified defect points.

8. The visual inspection system for automobile door panel components according to claim 7, characterized in that: The specific steps of marking the position of the identified defect points and analyzing the defect types are as follows: The position of the defect point p in the identified defect point set δ is marked, and the expression is: L ( δ ) (x p ,y p ,z p ): Among them, x p represents the horizontal coordinate of the defect point p in the defect set δ, y p represents the ordinate of the defect point p in the defect set δ, z p represents the height coordinate of the defect point p in the defect set δ, and L represents the specific coordinate of the defect point p in δ in the three-dimensional space; The specific coordinates of the defect point L(δ) are input into YOLOv and combined with the enhanced car door panel image data I s Identify the defect type, the expression is: T d (p)=YOLOv5(I s ,L(δ)); Among them, T d (p) represents the defect type of defect point p.

9. The visual inspection system for automobile door panel components according to claim 8, characterized in that: The specific steps of generating the automobile door panel inspection report according to the defect type analysis result are as follows: Collect the defect type T of each car door panel defect point p d (p), defect type T d (p) includes missing, misaligned, deformed and damaged information, according to the defect type T d (p) Generate a monitoring report, which includes the defect location, defect type and defect quantity in the automobile door panel.

Images