An automobile manufacturing part surface defect tracing positioning detection system

Abstract

The application discloses a surface defect tracing and positioning detection system for automobile manufacturing parts, and belongs to the technical field of industrial defect detection. The system comprises an isotropic scattering reference point module, the surface of which is provided with a specific micron-level three-dimensional geometric structure; a multi-view synchronous acquisition module is arranged around an automobile part detection station and comprises at least three industrial cameras arranged at different elevations, and a synchronous triggering acquisition device is used to acquire multi-angle original images with reference points; a scattering feature intelligent positioning module is connected with the multi-view synchronous acquisition module; a dynamic calibration compensation module receives initial registration results output by a multi-view geometric solution unit, corrects the scattering compensation coefficient of the reference points in a high-curvature area in real time according to the curvature distribution of the curved surface, and outputs corrected high-precision three-dimensional reconstruction data; and through a refraction compensation formula, material optical parameters and geometric curvature are fused to realize positioning error multiple compression in a high-distortion area such as a door ridge line.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of industrial defect detection, more particularly to a surface defect tracing and positioning detection system for automobile manufacturing parts. BACKGROUND

[0002] In the automated manufacturing and assembly process of automobile parts, especially for key appearance parts and high-precision structural parts such as door, hood, fender and other coverings, online and non-contact visual detection of surface quality and geometric dimensions is of great importance. Such detection often requires multiple industrial cameras to capture image information of the target area from different angles simultaneously or at different times to overcome the problem of local occlusion or information loss caused by the geometric complexity and high curvature of the parts themselves and the surface characteristics of high light reflection.

[0003] However, the existing multi-view visual positioning technology faces a key bottleneck problem that is severely underestimated on the surface of high-gloss, complex curved automobile parts such as galvanized steel stamping parts and high-gloss paint coverings: the projection imaging of key physical reference points used for camera external parameter calibration or point cloud stitching on high-gloss, complex curved surfaces is severely disturbed by multi-source, uncontrollable specular reflection and high-frequency "speckle effect". These reference points are not ideal diffuse point sources, especially under low-angle illumination or multi-camera observation angles, the imaging edges of which produce significant "halo diffusion" and "position blur" due to strong specular reflection, severely distorting the originally clear, high-contrast point contour geometric center imaging.

[0004] The existing solutions to this problem have the following obvious defects:

[0005] 1. Distortion of reference points under reflective curved surfaces: galvanized steel or high-gloss paint surfaces cause halo diffusion and speckle noise in the imaging of reference points. The existing marker points have a center positioning error of up to 0.35mm in a curvature >50m -1 region, which cannot meet the process requirement of a body gap difference of ±0.05mm.

[0006] 2. Systematic failure of dynamic interference: thermal deformation of metal parts due to temperature difference in the workshop, no compensation mechanism for traditional vision systems; oil film covering reference points, scattering characteristics decay leading to decreased algorithm accuracy; and sudden changes in ambient light causing image overexposure or underexposure, HDR fusion not coupled with feature extraction, and decreased positioning consistency. SUMMARY

[0007] Based on the above defects, the purpose of the present application is to provide a surface defect tracing and positioning detection system for automobile manufacturing parts, which solves the problems of distortion of reference points under reflective curved surfaces and systematic failure of dynamic interference.

[0008] In order to achieve the above object, the present application provides the following technical solutions:

[0009] A surface defect tracing and positioning detection system for automobile manufacturing parts comprises:

[0010] An isotropic scattering reference point module has a specific micron-level three-dimensional geometric structure on its surface, which is used to generate a highly uniform diffuse scattering light field within an incident angle range of 10°-80° and suppress mirror reflection;

[0011] A multi-view synchronous acquisition module is arranged circumferentially around a detection station for automobile parts and comprises at least three industrial cameras arranged at different elevations, which acquires multi-angle original images with reference points through a synchronous triggering acquisition device;

[0012] A scattering feature intelligent positioning module is connected with the multi-view synchronous acquisition module and is used to convert physical scattering characteristics into mathematically solvable spatial coordinate data, and specifically comprises:

[0013] A scattering schlieren feature extraction unit separates scattering areas from multi-view original images based on the optical characteristics of the reference point microstructure, calculates radial brightness gradient distribution, and generates a feature vector;

[0014] A sub-pixel registration unit matches the feature vector with a reference scattering schlieren template through a pre-trained triple-attention convolutional neural network and outputs reference point center coordinates with a precision better than 0.05 pixels;

[0015] A multi-view geometric solving unit establishes a matching point pair according to the sub-pixel center coordinates, iteratively solves a camera extrinsic matrix through Zhang Zhengyou calibration method, and performs weighted RANSAC point cloud initial registration;

[0016] A dynamic calibration compensation module receives the initial registration result output by the multi-view geometric solving unit, real-time corrects the scattering compensation coefficient of the reference point in a high-curvature area according to the curvature distribution of the curved surface, and outputs corrected high-precision three-dimensional reconstruction data.

[0017] Further, the isotropic scattering reference point module comprises:

[0018] A microcavity array unit: a conical microcavity with a side wall inclination angle θ=55±2° is formed by etching the surface of an alumina ceramic substrate, which is used to generate a uniform scattering field within an incident angle range of 10°-80°;

[0019] An anti-reflection coating unit: a silica nanoparticle layer coated on the surface of the microcavity, with titanium dioxide photocatalytic particles embedded on the surface, which is used to decompose adhering oil film and maintain scattering characteristics;

[0020] A mechanical positioning unit: an electromagnet and a flexible gimbal are built-in, so that the reference point can adaptively fit the curved surface of the workpiece.

[0021] Further, the convolutional neural network is trained by the following steps:

[0022] In the sample collection stage, the test area with a curvature value of 0-200m -1 is covered on the surface of the workpiece made of galvanized steel plate, high-gloss polyurethane paint and carbon fiber composite material, and a million-frame reference point scattering spot image sample is collected by a multi-angle industrial camera;

[0023] In the feature engineering stage, a 16-direction radial brightness gradient histogram with the reference point center as the origin is extracted from each frame of scattering spot image, the gradient module length average is calculated according to the equal angle interval of 0°-360°, and is combined into a 128-dimensional moire feature vector;

[0024] In the network optimization stage, a three-way attention mechanism is constructed by using a cascaded channel attention branch, a spatial attention branch and a curvature attention branch to optimize feature matching.

[0025] Further, the sub-pixel registration unit comprises:

[0026] A curvature response compensation mechanism: when the received real-time curvature K of the curved surface is greater than 50m -1 , the corresponding refractive index η of the material is called;

[0027] and the bilinear interpolation algorithm is activated to calculate the edge refraction distortion compensation amount, and the formula is:

[0028] δ = 0.02·|K|·(η-1);

[0029] Correct the reference point center coordinates: (x', y') = (x+δ x , y+δ y );

[0030] In the formula, (x, y) is the initial detected reference point center coordinates; δ x and δ y are the X-direction distortion compensation amount and the Y-direction compensation amount; (x', y') is the corrected real physical coordinates;

[0031] δ is the edge refraction distortion compensation amount, 0.02 is the curvature-refraction coupling coefficient, which is determined by ray tracing simulation and real vehicle data fitting; |K| is the absolute value of the surface curvature obtained by point cloud calculation, and the effective range is 50-250m -1 ; η is the relative refractive index of the workpiece, which is mapped through the pre-stored material database;

[0032] Further, the sub-pixel registration unit further comprises an illumination fluctuation response mechanism: when the ambient illumination sensor detects that the fluctuation amplitude is > 15%, control the industrial camera to switch to the HDR mode, to image in three exposure time layers, and extract the stria feature vector of each exposure layer image respectively, and output the anti-glare interference sub-pixel center coordinates by exposure time reciprocal weighted fusion.

[0033] Further, the multi-view geometry solving unit is further used for performing thermal expansion compensation.

[0034] Receive the workpiece surface temperature T collected by the infrared sensor, and apply the deformation compensation matrix:

[0035] [Δx,Δy]=[α·(T-T0)·L,β·(T-T0)·L]

[0036] In the formula: α and β are the thermal expansion coefficients of the workpiece material (α = 11.8 × 10 -6 / ℃, β = 9.5 × 10 -6 / ℃ for galvanized steel, pre-stored in the process database; T0 is the system calibration reference temperature (23℃±1℃, conforming to ISO 1 level tolerance); L is the Euclidean distance from the reference point to the nearest workpiece fixed edge, in mm;

[0037] The corrected initial registration coordinates are: (x', y') = (x+Δx, y+Δy)

[0038] In the formula: (x, y) is the center coordinates of the reference point initially detected, with thermal deformation error, Δx and Δy are the compensation displacement amounts caused by thermal expansion, calculated by the deformation matrix; (x', y') are the corrected real physical coordinates, i.e. eliminating temperature error.

[0039] Further, in the dynamic calibration compensation module, based on the workpiece material type and the local surface curvature, the formula for calculating the scattering compensation coefficient in real time is:

[0040] γ=1+λ×|K|

[0041] In the formula: λ is the dynamic material coefficient, determined by pre-stored database mapping, |K| is the absolute value of the surface curvature, obtained by point cloud normal vector difference calculation; 1 is the baseline, i.e. a plane;

[0042] The compensation execution is: correcting the initial registration point cloud coordinates (x0, y0, z0) to (x, y, z) = γ·(x0, y0, z0), wherein (x, y, z) are the corrected real coordinates.

[0043] Further, it further comprises a closed-loop error convergence module connected to the output end of the dynamic calibration compensation module, and the closed-loop error convergence module comprises:

[0044] A reference comparison unit is configured to perform ICP matching on the corrected three-dimensional reconstruction point cloud and the laser tracker reference point cloud, and calculate a Hausdorff distance ε;

[0045] A λ value dynamic optimization unit is configured to iteratively update a material coefficient λ of the weight 7 in a reverse direction when the ε is greater than 0.03 mm:

[0046] λ new = λ old + sign (Δε) * 0.0005

[0047] In the formula, λ old is a material coefficient used in a current batch; λ new is a material coefficient optimized in a next batch; Δε = a current error - a previous error, and sign is a sign function.

[0048] A historical error database construction unit is configured to store optimal λ values under different material and curvature combinations, and form a process knowledge graph.

[0049] A convergence verification unit is configured to lock a current λ value and output a calibration completion signal when the ε is less than or equal to 0.02 mm for three consecutive batches.

[0050] Compared with the prior art, the automobile manufacturing part surface defect tracing positioning detection system has the following beneficial effects:

[0051] 1. In use, the automobile manufacturing part surface defect tracing positioning detection system precisely suppresses low-angle mirror reflection light halo through a 55° inclination micro-cavity array design, reduces reflection intensity by 82% under a 80° incidence angle working condition, decomposes attached oil film in combination with an anti-reflection coating, compresses positioning error in an oil pollution environment from 3.2 pixels to 0.4 pixels, and makes a reference point self-adaptively fit a workpiece curved surface in cooperation with a mechanical positioning unit.

[0052] 2. In use, the automobile manufacturing part surface defect tracing positioning detection system realizes positioning error multiple compression in a high-distortion area such as a door ridge by fusing material optical parameters and geometric curvature through a refraction compensation formula, develops a triple attention convolution network to strengthen high-curvature area feature extraction, improves carbon fiber composite material detection accuracy, constructs a dynamic scattering compensation mechanism to solve point cloud distortion caused by material differences, and improves trans-material detection consistency of galvanized steel and aluminum alloy to more than 98%.

[0053] 3. In use, the automobile manufacturing part surface defect tracing positioning detection system offsets metal piece thermal expansion error in real time through a thermal deformation compensation matrix, adopts an HDR three-exposure reciprocal weighted fusion strategy to couple feature extraction and light adaptation, improves positioning stability in a strong light fluctuation environment, initiates a coefficient self-optimization engine to cope with material batch fluctuations, drives first-month detection accuracy of mass production to be further improved, and realizes continuous evolution through multiple process knowledge graphs. BRIEF DESCRIPTION OF DRAWINGS

[0054] Figure 1 It is a kind of automobile manufacturing parts surface defect trace positioning detection system overall flow chart;

[0055] Figure 2 It is a kind of automobile manufacturing parts surface defect trace positioning detection system scattering feature intelligent positioning module general diagram;

[0056] Figure 3 It is a kind of automobile manufacturing parts surface defect trace positioning detection system isotropic scattering reference point module general diagram. DETAILED DESCRIPTION

[0057] Please refer to Figure 1 - Figure 3 An automobile manufacturing parts surface defect trace positioning detection system, comprising: an isotropic scattering reference point module, which is configured with a specific micron-level three-dimensional geometric structure on the surface, for generating a highly uniform diffuse light field within an incident angle range of 10°-80°, and suppressing specular reflection.

[0058] Multi-view synchronous acquisition module, which is arranged circumferentially around the automobile parts detection station and comprises at least three industrial cameras arranged at different elevations, and obtains multi-angle original images with reference points through a synchronous trigger acquisition device.

[0059] Scattering feature intelligent positioning module, which is signal-connected with the multi-view synchronous acquisition module, and is used for converting physical scattering characteristics into mathematically solvable spatial coordinate data, and specifically comprises:

[0060] Scattering schlieren feature extraction unit: based on the optical characteristics of the reference point microstructure, separates the scattering area from the multi-view original images, calculates the radial brightness gradient distribution and generates a feature vector.

[0061] Sub-pixel registration unit: through a pre-trained triple-attention convolutional neural network, the feature vector is matched with a reference scattering schlieren template, and the center coordinates of the reference point with a precision better than 0.05 pixels are output.

[0062] Multi-view geometric solving unit: according to the sub-pixel center coordinates, a matching point pair is established, the camera extrinsic matrix is iteratively solved through Zhang Zhengyou calibration method, and weighted RANSAC point cloud initial registration is performed.

[0063] Dynamic calibration compensation module, which receives the initial registration result output by the multi-view geometric solving unit, and according to the curvature distribution of the curved surface, real-time corrects the scattering compensation coefficient of the reference point in the high-curvature area, and outputs the corrected high-precision three-dimensional reconstruction data.

[0064] Preferably, the isotropic scattering reference point module comprises:

[0065] Microcavity array unit: a conical microcavity with a diameter of 40-80 μm, a depth-diameter ratio of 0.5-1.2, and a side wall inclination angle θ = 55±2° is formed by etching the surface of an alumina ceramic substrate, which is used to generate a uniform scattering field in the range of 10°-80° incident angle; and the problem of mirror reflection halo in low-angle observation is solved;

[0066] Anti-reflection coating unit: a layer of silica nanoparticles coated on the surface of the microcavity, with titanium dioxide photocatalytic particles embedded on the surface, which is used to decompose the attached oil film and maintain the scattering properties; and the problem of scattering failure caused by stamping oil film covering is solved;

[0067] Mechanical positioning unit: an electromagnet and a flexible universal joint are built-in to make the reference point adaptively fit the curved surface of the workpiece.

[0068] Preferably, the convolutional neural network is trained by the following steps:

[0069] In the sample collection stage, a test area with a curvature value of 0-200 m -1 is covered on the surface of a galvanized steel plate, a high-gloss polyurethane paint surface, and a carbon fiber composite material workpiece, and a million-frame reference point scattering spot image sample is collected by a multi-angle industrial camera;

[0070] In the feature engineering stage, a 16-direction radial brightness gradient histogram with the reference point center as the origin is extracted for each frame of scattering spot image, the gradient module length mean value is calculated according to the equal angle interval of 0°-360°, and is combined into a 128-dimensional striae feature vector;

[0071] In the network optimization stage, a three-way attention mechanism is constructed by using a cascaded channel attention branch, a spatial attention branch, and a curvature attention branch to optimize feature matching.

[0072] Preferably, the sub-pixel registration unit comprises:

[0073] Curvature response compensation mechanism: when the received real-time curvature K of the curved surface is greater than 50 m -1 , the corresponding material refractive index η is called, for example, η = 2.42 for a galvanized steel plate and η = 1.55 for a high-gloss paint surface in this embodiment;

[0074] and a bilinear interpolation algorithm is activated to calculate the edge refraction distortion compensation amount, and the formula is:

[0075] δ = 0.02·|K|·(η-1);

[0076] Correct the reference point center coordinates: (x', y') = (x+δ x , y+δ y );

[0077] where (x, y) is the initial detected reference point center coordinates; δx and δ y are the X and Y direction distortion compensation amounts; (x', y') are the corrected real physical coordinates;

[0078] δ is the edge refraction distortion compensation amount, 0.02 is the curvature-refraction coupling coefficient, which is determined by ray tracing simulation and real vehicle data fitting; |K| is the absolute value of the surface curvature obtained by point cloud calculation, the effective range is 50-250m -1 ; η is the relative refractive index of the workpiece, which is mapped through a pre-stored material database: η of galvanized steel plate = 2.42, η of high-gloss polyurethane paint = 1.55;

[0079] In the curvature > 50m -1 region, such as the door edge line, the light after refraction by the high-reflective curved surface will cause the imaging position of the reference point to deviate, and the deviation δ is strongly correlated with the curvature K and the material refractive index η; the theoretical deviation δ is calculated by a physical model; the coordinates are pulled back to the real position by additive counter-compensation;

[0080] Preferably in this embodiment, the sub-pixel registration unit further comprises an illumination fluctuation response mechanism: when the environmental illumination sensor detects that the fluctuation amplitude is > 15%, control the industrial camera to switch to the HDR mode to image in three exposure times, and extract the stria feature vector for each exposure layer image respectively, and output the anti-glare interference sub-pixel center coordinates by inverse exposure time weighted fusion.

[0081] Preferably in this embodiment, the multi-view geometry solving unit is further used to perform thermal expansion compensation:

[0082] Receive the workpiece surface temperature T collected by the infrared sensor, and apply the deformation compensation matrix:

[0083] [Δx,Δy]=[α·(T-T0)·L,β·(T-T0)·L]

[0084] In the formula: α and β are the thermal expansion coefficients of the workpiece material (α of galvanized steel = 11.8×10 -6 / ℃, β = 9.5×10 -6 / ℃), which are pre-stored in the process database; T0 is the system calibration reference temperature (23℃±1℃, conforming to ISO 1 level tolerance); L is the Euclidean distance from the reference point to the nearest workpiece fixed edge, in mm;

[0085] Correct the initial registration coordinates as: (x', y') = (x+Δx, y+Δy)

[0086] In the formula: (x, y) is the initial detected reference point center coordinates, there is thermal deformation error, Δx and Δy are the compensation displacement caused by thermal expansion, which is calculated by the deformation matrix; (x', y') is the corrected real physical coordinates, that is, the temperature error is eliminated;

[0087] Due to the temperature difference of the automobile production line, the metal parts will expand and contract due to heat, for example, the body will shrink by 0.1mm / m from the baking room (60℃) to the assembly line (25℃). By real-time acquisition of the workpiece temperature T through infrared temperature measurement, using the pre-stored material parameters α, β (anisotropic expansion rate of galvanized steel), based on the lever effect (the larger the L value, the larger the compensation), temperature drift compensation is finally realized on the sub-pixel level positioning.

[0088] Preferably in the dynamic calibration compensation module, based on the workpiece material type and local surface curvature, the formula for real-time calculation of the scattering compensation coefficient is:

[0089] γ=1+λ×|K|

[0090] In the formula: λ is a dynamic material coefficient, which is determined by a pre-stored database mapping, for example,

[0091] λ=0.002 (carbon fiber composite material, low reflectivity weak compensation); λ=0.005 (aluminum alloy, normal reflectivity compensation); λ=0.007 (galvanized steel plate, high reflectivity needs strong compensation);

[0092] |K| is the absolute value of the surface curvature, which is obtained by the difference calculation of the point cloud normal vector; 1 is the baseline, that is, the plane;

[0093] The compensation execution is to correct the initial registration point cloud coordinates (x0, y0, z0) to (x, y, z)=γ·(x0, y0, z0).

[0094] In the formula, the absolute value |K| handles the asymmetric problem of convex and concave surface scattering distortion; λ, as a material coefficient, is not a fixed value, but is dynamically updated through transfer learning; the compensation mechanism is directly linked with the later point cloud reconstruction;

[0095] Preferably, the embodiment further comprises a closed-loop error convergence module connected to the output end of the dynamic calibration compensation module, and the closed-loop error convergence module comprises:

[0096] A reference comparison unit for ICP matching the corrected three-dimensional reconstruction point cloud with the laser tracker reference point cloud, and calculating the Hausdorff distance ε;

[0097] A λ value dynamic optimization unit for updating the material coefficient λ of the weight 7 in the reverse direction when ε>0.03mm:

[0098] λ new =λ old+ sign (Delta E) x 0.0005

[0099] In the formula: lambda old The material coefficient used in the current batch; lambda new The material coefficient optimized in the next batch; Delta E = this error - last error, sign is a sign function;

[0100] The historical error library construction unit stores the optimal lambda value under different material and curvature combinations to form a process knowledge graph.

[0101] The convergence verification unit locks the current lambda value and outputs a calibration completion signal when the error of three consecutive batches is less than or equal to 0.02 mm.

[0102] Due to the difference in reflectivity of galvanized sheets from different steel plants, the closed-loop error convergence module is used to solve the problem of lambda value drift caused by material batch fluctuation, form a self-learning process library, and continuously improve the compensation accuracy with the production time.

[0103] The above shows and describes the basic principles, main features and advantages of the present application. Those skilled in the art should understand that the present application is not limited by the above examples, and the above examples and descriptions in the specification are only preferred examples of the present application and are not intended to limit the present application. Without departing from the spirit and scope of the present application, various changes and improvements can be made to the present application, and these changes and improvements all fall within the scope of the claimed present application. The scope of protection of the present application is defined by the appended claims and their equivalents.

Claims

1. A system for tracing and locating surface defects in automotive manufacturing parts, characterized in that: include: The isotropic scattering reference point module includes a microcavity array unit: a conical microcavity with a sidewall tilt angle of θ=55±2° is formed by etching the surface of an alumina ceramic substrate, which is used to generate a uniform scattering field in the range of 10°-80° incident angle and suppress specular reflection; The multi-view synchronous acquisition module is deployed around the automotive parts inspection station and includes at least three industrial cameras arranged at different elevation angles. It acquires multi-angle raw images with reference points by synchronously triggering the acquisition device. A scattering feature intelligent positioning module, signal-connected to the multi-view synchronous acquisition module, is used to convert physical scattering characteristics into mathematically solvable spatial coordinate data, specifically including: Scattering schlieren feature extraction unit: Based on the optical properties of the microstructure of the reference point, it separates the scattering region from the original image from multiple views, calculates the radial brightness gradient distribution, and generates feature vectors; Sub-pixel registration unit: Through a pre-trained triple attention convolutional neural network, the feature vector is matched with the reference scattering schlieren template, and the center coordinates of the reference point with an accuracy better than 0.05 pixels are output. Multi-view geometry solving unit: Establishes matching point pairs based on sub-pixel center coordinates, iteratively solves the camera extrinsic matrix using the Zhang Zhengyou calibration method, and performs weighted RANSAC point cloud initial registration; The dynamic calibration and compensation module receives the initial registration results output by the multi-view geometry solution unit, corrects the scattering compensation coefficient of the reference point in the high curvature region in real time according to the surface curvature distribution, and outputs the corrected high-precision 3D reconstruction data.

2. The system for tracing and locating surface defects in automotive manufacturing parts according to claim 1, characterized in that: The isotropic scattering reference point module also includes: Anti-reflective coating unit: A layer of silica nanoparticles coated on the surface of the microcavity, with titanium dioxide photocatalytic particles embedded on its surface to decompose the attached oil film and maintain scattering characteristics; Mechanical positioning unit: It has a built-in electromagnet and flexible universal joint, which makes the reference point adaptively fit the curved surface of the workpiece.

3. The system for tracing and locating surface defects in automotive manufacturing parts according to claim 1, characterized in that: The convolutional neural network is trained through the following steps: During the sample collection phase, curvature values ​​of 0-200 μm were covered on the surfaces of galvanized steel sheets, high-gloss polyurethane paint surfaces, and carbon fiber composite materials. -1 In the test area, millions of frames of reference point scattered light spot image samples were collected using multi-angle industrial cameras; In the feature engineering stage, a 16-direction radial brightness gradient histogram with the reference point center as the origin is extracted for each frame of scattered light spot image. The average gradient magnitude is calculated by dividing the angle interval into 0°-360° equally and combining them into a 128-dimensional schlieren feature vector. During the network optimization phase, a triple attention mechanism is constructed by using cascaded channel attention branches, spatial attention branches, and curvature attention branches to optimize feature matching.

4. The surface defect tracing and location detection system for automotive manufacturing parts according to claim 1, characterized in that: The sub-pixel registration unit includes: Curvature-responsive compensation mechanism: When the received real-time curvature K of the surface is greater than 50m -1 At that time, retrieve the refractive index η of the corresponding material; And activate the bilinear interpolation algorithm to calculate the edge refraction distortion compensation amount, the formula is: δ=0.02·|K|·(η-1); Correct the coordinates of the reference point center: (x', y') = (x+δ) x y+δ y ); In the formula: (x, y) are the coordinates of the center of the initially detected reference point; δ x and δ y These represent the distortion compensation amounts in the X and Y directions, respectively; (x', y') are the corrected true physical coordinates. δ represents the edge refraction distortion compensation amount, 0.02 is the curvature-refraction coupling coefficient, determined through ray tracing simulation and fitting with actual vehicle data; |K| is the absolute value of the surface curvature, calculated from the point cloud normal vector difference, with an effective range of 50-250m. -1 η is the relative refractive index of the workpiece, which is mapped through a pre-stored material database.

5. The surface defect tracing and location detection system for automotive manufacturing parts according to claim 4, characterized in that: The subpixel registration unit also includes an illumination fluctuation response mechanism: when the ambient illumination sensor detects a fluctuation amplitude >15%, the industrial camera is controlled to switch to HDR mode, and three layers of imaging are performed with three exposure times. The schlieren feature vectors of each exposure layer image are extracted and fused by weighting the inverse of the exposure time to output the subpixel center coordinates that are resistant to glare interference.

6. The surface defect tracing and location detection system for automotive manufacturing parts according to claim 1, characterized in that: The multi-view geometry solving unit is also used to perform thermal expansion compensation: The workpiece surface temperature T is received from an infrared sensor, and a deformation compensation matrix is ​​applied: [Δx, Δy] = [α·(T-T0)·L,β·(T-T0)·L] In the formula: α and β are the coefficients of thermal expansion of the workpiece material, which are pre-stored in the process database; T0 is the system calibration reference temperature; L is the Euclidean distance from the reference point to the nearest fixed edge of the workpiece, in mm; The initial registration coordinates are corrected to: (x', y') = (x + Δx, y + Δy) In the formula: (x, y) are the initial detected coordinates of the reference point center, which have thermal deformation error; Δx and Δy are the compensation displacement caused by thermal expansion, calculated by the deformation matrix; (x', y') are the corrected real physical coordinates.

7. The system for tracing and locating surface defects in automotive manufacturing parts according to claim 1, characterized in that: In the dynamic calibration compensation module, the formula for calculating the scattering compensation coefficient in real time based on the workpiece material type and local surface curvature is as follows: In the formula: λ is the dynamic material coefficient, which is determined by mapping through a pre-stored database; |K| is the absolute value of the surface curvature, which is calculated by the difference of the point cloud normal vector; 1 is the baseline, i.e., the plane; The compensation is performed as follows: the initial registered point cloud coordinates (x0, y0, z0) are corrected to (x, y, z) = γ·(x0, y0, z0).

8. The surface defect tracing and location detection system for automotive manufacturing parts according to claim 7, characterized in that: It also includes a closed-loop error convergence module, which is connected to the output of the dynamic calibration compensation module. The closed-loop error convergence module includes: The benchmark comparison unit is used to perform ICP matching between the corrected 3D reconstructed point cloud and the laser tracker benchmark point cloud to calculate the Hausdorff distance ε. The λ value is dynamically optimized by the unit. When ε > 0.03 mm, the dynamic material coefficient λ is updated iteratively in reverse. In the formula: This refers to the material coefficient used in the current batch. The material coefficient for the next batch of optimization; Δε = current error - previous error, sign is the sign function; The historical error database construction unit stores the optimal λ value under different material and curvature combinations, forming a process knowledge graph; The convergence verification unit locks the current λ value and outputs a calibration completion signal when ε ≤ 0.02 mm for three consecutive batches.

Images