Machine Vision-Based New Energy Vehicle Wiring Harness Assembly Quality Inspection System

Abstract

This invention relates to the field of image analysis technology, specifically to a machine vision-based quality inspection system for new energy vehicle wiring harnesses and cable ties assembly. The system includes: an image acquisition module that uses polarimetric stereo technology to acquire multi-angle images; a geometric reconstruction and reference extraction module that calculates the surface normal field of the insulation layer and separates the background field of the cable's cylindrical curvature and the micro-indentation residual field, thereby extracting the global axial and radial curvature moduli of the cable; a feature decoupling analysis module that calculates the contour-shaping migration flow field of the indentation micro-residual field relative to the standard contour, and performs anisotropic decomposition of the flow field based on the global axial direction of the cable to obtain the axial slip index; and a quality inspection module that determines an adaptive slip judgment threshold based on the cable's radial curvature modulus, thereby generating inspection results. This invention solves the problems of difficulty in sensing micro-deformation on reflective cylindrical surfaces and the difficulty in decoupling slip risk from assembly tolerances, thus improving the accuracy of wiring harness and cable tie assembly quality inspection.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and more specifically to a machine vision-based system for inspecting the assembly quality of wiring harnesses and cable ties in new energy vehicles. Background Technology

[0002] In the field of new energy vehicle manufacturing, wiring harnesses are the "blood vessels" and "nerves" of electrical systems. Their assembly quality is directly related to the safety and reliability of the entire vehicle. To fix and organize wiring harnesses, cable ties are widely used in the bundling and fixing process. High-quality cable tie assembly requires that the locking head and the cable insulation layer form a tight and stable mechanical engagement to ensure that the cable does not loosen or shift under the long-term vibration environment of vehicle operation. High-voltage wiring harnesses in new energy vehicles have relatively thick wire diameters and are wrapped with high-molecular insulation materials such as cross-linked polyethylene. These materials have smooth surfaces, low coefficients of friction, and significant cylindrical curved surface geometry. When automatic cable tie guns are used for locking operations, two types of difficult-to-detect quality problems often occur: First, false locking, where although the cable tie lock head appears to be engaged, in reality, due to insufficient friction or oil contamination on the cable surface, the lock head slips slightly along the cable axis, causing the locking rack to not fully engage to the designed depth, resulting in severely insufficient pull-out resistance. Second, hidden material defects, where the insulation layer hardens (leading to shallow indentations) or softens (leading to excessive compression) due to batch differences, making it impossible to form effective contact indentations under the same locking torque, affecting the final holding force.

[0003] The current mainstream method for inspecting the assembly quality of cable ties in new energy vehicle wiring harnesses is two-dimensional machine vision inspection, which analyzes the assembly quality of high-voltage wiring harnesses by extracting indentation contours. However, in reality, cable insulation layers typically have high reflectivity, resulting in poor image quality and difficulty in extracting micron-level micro-indentation features. Secondly, it is difficult to distinguish between normal assembly tolerances and dangerous axial slippage. Rotational displacement of the locking head around the cable circumference in the cable tie is a normal process tolerance, while displacement along the cable axial direction indicates locking failure. Due to the lack of depth information and three-dimensional geometric reference in two-dimensional images, existing methods easily confuse these two types of displacement. Either standards are relaxed to tolerate tolerances, leading to missed slippage, or standards are tightened to intercept slippage, leading to misjudgment (over-judgment) of normal tolerances, resulting in inaccurate detection of wiring harness cable tie assembly quality. Therefore, existing two-dimensional machine vision inspection methods cannot accurately perceive three-dimensional contact deformation on highly reflective cylindrical surfaces and effectively decouple the risk of axial slippage from circumferential assembly tolerances, thus affecting the safety and reliability of new energy vehicle electrical systems under long-term vibration environments. Summary of the Invention

[0004] To address the technical problem that existing two-dimensional machine vision inspection methods cannot accurately perceive three-dimensional contact deformation on highly reflective cylindrical surfaces and effectively decouple axial slippage risk from circumferential assembly tolerances, thus failing to accurately detect the assembly quality of wiring harness cable ties in new energy vehicles, the present invention aims to provide a machine vision-based assembly quality inspection system for wiring harness cable ties in new energy vehicles. The specific technical solution adopted is as follows:

[0005] This invention provides a machine vision-based quality inspection system for the assembly of wiring harnesses and cable ties in new energy vehicles. The system includes:

[0006] The image acquisition module is used to acquire images of the cable surface from multiple angles using polarimetric stereoscopic technology;

[0007] The geometric reconstruction and benchmark extraction module is used to calculate the normal field of the insulation layer surface based on images from multiple angles, separate the normal field of the insulation layer surface, and obtain the background field of the cable cylindrical curvature and the micro-indentation residual field; based on the normal gradient variation characteristics of the background field of the cable cylindrical curvature, the global axial direction and the radial curvature modulus of the cable are obtained.

[0008] The feature decoupling analysis module is used to obtain the contour shaping migration flow field based on the topological distribution difference between the indentation micro residual field and the preset standard lock head reference contour; it performs anisotropic decomposition of the contour shaping migration flow field based on the global axial direction of the cable to obtain the axial slip component and calculate the axial slip index;

[0009] The quality inspection module is used to determine the adaptive slip judgment threshold based on the radial curvature modulus of the cable, and to obtain the axial slip quality inspection result based on the comparison result between the axial slip index and the adaptive slip judgment threshold.

[0010] Furthermore, the method for obtaining the normal field of the insulating layer surface is as follows:

[0011] For any pixel, extract the gray value of that pixel in each image and construct the brightness observation vector of that pixel;

[0012] Obtain the light source direction corresponding to each image and construct a light source direction matrix;

[0013] A system of linear equations is constructed using the least squares method to establish the brightness observation vector and the light source direction matrix. Solving the system of linear equations yields the initial normal vector of the pixel.

[0014] The initial normal vector is normalized to obtain the unit normal analysis vector of the pixel.

[0015] The set of unit normal analysis vectors of all pixels is taken as the normal field of the insulating layer surface.

[0016] Furthermore, the method for obtaining the background field of the cable cylindrical curvature and the micro-indentation residual field is as follows:

[0017] For any unit normal analysis vector in the normal field of the insulating layer surface, the unit normal analysis vector is decomposed into multiple dimensional components, and each dimensional component is smoothed by Gaussian low-pass filtering to obtain the background value of each dimensional component.

[0018] Combine all the background values ​​corresponding to the unit normal analysis vector into a single vector, which serves as the background vector of the unit normal analysis vector;

[0019] The set of background vectors of all unit normal analysis vectors in the normal field of the insulation layer surface is taken as the background field of the cable cylindrical curvature.

[0020] The difference vector between each unit normal analysis vector and its background vector is used as the micro residual vector;

[0021] The set of microscopic residual vectors of all unit normal analysis vectors in the normal field of the insulating layer surface is taken as the microscopic indentation residual field.

[0022] Furthermore, the method for obtaining the global axial direction and radial curvature modulus of the cable is as follows:

[0023] For any background vector in the background field of the cylindrical curvature of the cable, obtain the structure tensor of the background vector and perform eigenvalue decomposition. The direction of the eigenvector corresponding to the smallest eigenvalue is taken as the local axial analysis direction of the background vector.

[0024] The probability density distribution of the local axial direction of all background vectors in the background field of the cylindrical curvature of the cable is statistically analyzed, and the direction with the highest probability density is taken as the global axial direction of the cable.

[0025] The magnitude of the directional derivative of the background vector along the direction perpendicular to the global axial direction of the cable is used as the radial variation analysis value of the background vector;

[0026] The mean value of the radial variation analysis of all background vectors in the background field of the cylindrical curvature of the cable is taken as the radial curvature modulus of the cable.

[0027] Furthermore, the method for obtaining the contour shaping and migration flow field is as follows:

[0028] Divergence calculation and Poisson reconstruction are performed on the micro residual field of the indentation to obtain the measured indentation depth distribution map. Then, the pixels with values ​​greater than the preset noise threshold in the measured indentation depth distribution map are normalized to generate the measured indentation quality distribution.

[0029] Calculate the centroid coordinates and spindle rotation angle of the measured indentation mass distribution, translate the centroid of the preset standard lock head reference profile to coincide with the centroid coordinates, and perform rotation alignment based on the spindle rotation angle to obtain the aligned target distribution;

[0030] Establish source domain index space and target domain index space, calculate the squared Euclidean distance from non-zero pixels in the measured indentation quality distribution to non-zero pixels in the target distribution, and construct the cost matrix;

[0031] An entropy regularization term is introduced, and the optimal transmission coupling matrix corresponding to the cost matrix is ​​solved using the Sinkhorn iterative algorithm.

[0032] Based on the optimal transmission coupling matrix, calculate the weighted average displacement vector required for each pixel in the source domain to match the target distribution;

[0033] The set of weighted average displacement vectors of all pixels in the entire field is used as the contour shaping migration flow field.

[0034] Furthermore, the method for obtaining the axial slip component is as follows:

[0035] For any topology migration vector in the contour-shaping migration flow field, calculate the projection modulus of the topology migration vector in the global axial direction of the cable, which is taken as the axial slip component of the topology migration vector.

[0036] Furthermore, the method for obtaining the axial slip index is as follows:

[0037] For any topology migration vector in the contour shaping migration flow field, obtain the value of the corresponding pixel in the measured indentation quality distribution of the topology migration vector, and use it as the weight analysis value of the topology migration vector;

[0038] The product of the axial slip component of the topology migration vector and the weighted analysis value is used as the weighted slip analysis value of the topology migration vector.

[0039] The weighted sum of the slip analysis values ​​of all topology migration vectors in the contour-shaping migration flow field is used as the axial slip index.

[0040] Furthermore, the method for obtaining the adaptive slip determination threshold is as follows:

[0041] The product of the radial curvature modulus of the cable and the curvature sensitivity coefficient is used as the first characteristic value;

[0042] The sum of the first eigenvalue and the preset constant is used as the attenuation factor; where the preset constant is greater than 0.

[0043] The ratio of the baseline slip threshold to the attenuation factor is used as the dynamically calculated threshold.

[0044] The dynamic calculation threshold is compared with the preset system noise floor tolerance limit, and the larger of the two values ​​is used as the adaptive sliding judgment threshold.

[0045] Furthermore, the method for obtaining the axial slip mass detection result is as follows:

[0046] When the axial slip index is greater than the adaptive slip judgment threshold, an unqualified result is generated, indicating that the axial slip is too large.

[0047] When the axial slip index is less than or equal to the adaptive slip judgment threshold, a qualified result representing normal axial slip is generated.

[0048] Furthermore, the machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system also includes a material performance verification module, which is used to take the product of the peak torque, effective stroke angle and preset transmission coefficient after the cable tie contacts the insulation layer during the cable tie assembly process as the effective crimping work.

[0049] The volume integral of the micro-indentation residual field is taken as the total displaced volume of the insulating layer.

[0050] The normalized ratio of the total displaced volume of the insulation layer to the effective crimping work is used as the relative efficiency index.

[0051] When the relative efficiency index is less than the preset lower limit of energy efficiency, an unqualified result is generated, indicating that the material is too hard or the energy efficiency is too low.

[0052] When the relative efficiency index is greater than the preset energy efficiency limit, an unqualified result is generated, indicating that the material is too soft or the energy efficiency is too high.

[0053] When the relative efficiency index is greater than or equal to the preset lower limit of energy efficiency and less than or equal to the preset upper limit of energy efficiency, a qualified result for characterizing the material is generated.

[0054] The present invention has the following beneficial effects:

[0055] This invention first calculates the normal field of the insulation layer surface based on images from multiple angles, which is beneficial for reconstructing the microscopic three-dimensional morphology of the cable surface using a two-dimensional image sequence without relying on a three-dimensional scanner, overcoming imaging interference caused by highly reflective materials. To extract pure indentation features and eliminate the influence of the cable's body shape, the normal field of the insulation layer surface is further separated to obtain the cable's cylindrical curvature background field and the microscopic indentation residual field, facilitating independent quantitative analysis of micrometer-level indentation depth. Furthermore, based on the normal gradient variation characteristics of the cable's cylindrical curvature background field, the global axial direction and radial curvature modulus of the cable are obtained, accurately reflecting the physical extension posture and geometric thickness characteristics of the cable in the image. To quantify the complex topological distortion of the measured indentation relative to a standard shape, the contour shaping migration flow field is obtained based on the difference in topological distribution between the indentation microscopic residual field and the preset standard lock head reference contour, which is beneficial for comprehensively evaluating the microscopic deformation trend of the indentation. Finally, the contour shaping migration flow field is further analyzed based on the global axial direction of the cable. Anisotropic decomposition is used to obtain the axial slip component, accurately eliminating circumferential tolerance interference caused by normal assembly rotation, and retaining only the axial drag feature representing failure risk. This allows for accurate calculation of the axial slip index, accurately reflecting the cumulative degree of axial slip risk throughout the indentation area, which helps focus on true failure hazards and significantly reduces the false positive rate. To address the issue of different slip tolerances for cables of varying thicknesses, an adaptive slip judgment threshold is determined based on the cable's radial curvature modulus. This allows for dynamic adjustment of the judgment scale according to the cable's geometric properties, preventing missed detection of thin cables or over-detection of thick cables. Furthermore, the comparison between the axial slip index and the adaptive slip judgment threshold accurately obtains the axial slip quality inspection results, effectively intercepting defective products with false locking risks. In addition, by introducing a relative efficiency index, a correlation between mechanical input and geometric deformation is established, enabling the identification of hidden defects such as hardening or softening of insulation materials, filling the blind spots of single visual inspection, and comprehensively ensuring the vibration resistance reliability of the electrical system of new energy vehicles. Attached Figure Description

[0056] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0057] Figure 1 This is a structural block diagram of a machine vision-based new energy vehicle wiring harness and cable tie assembly quality inspection system provided in one embodiment of the present invention.

[0058] Figure 2 This is a schematic diagram of a computer device provided according to an embodiment of the present invention. Detailed Implementation

[0059] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of the machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0061] The specific solution of the machine vision-based new energy vehicle wiring harness and cable tie assembly quality inspection system provided by the present invention will be described in detail below with reference to the accompanying drawings.

[0062] Example 1:

[0063] This invention proposes a machine vision-based system for inspecting the assembly quality of wiring harnesses and cable ties in new energy vehicles. Please refer to [link / reference]. Figure 1 The diagram shows a structural block diagram of a machine vision-based new energy vehicle wiring harness assembly quality inspection system provided by an embodiment of the present invention. The system includes: an image acquisition module 10, a geometric reconstruction and benchmark extraction module 20, a feature decoupling analysis module 30, and a quality inspection module 40.

[0064] Image acquisition module 10 is used to acquire images of the cable surface from multiple angles using polarimetric stereoscopic technology.

[0065] Specifically, in order to enable the testing system to adapt to the physical characteristics of wire harnesses of different specifications and to avoid the risk of insufficient disclosure due to unknown parameter sources, this embodiment first performs an offline calibration process before the system is officially put into operation (e.g., before the system goes online or when the wire harness specifications change). The calibration process mainly includes the following three aspects:

[0066] First, the calibration of slippage judgment parameters. In this embodiment, no fewer than 50 sets of standard wire harness samples covering the target wire diameter range (e.g., 5mm to 20mm) were selected. Each set of samples was assembled using a standard cable tie gun, and an axial pull-out failure test was conducted using a high-precision tensile tester. The geometric deformation characteristic values ​​of each set of samples at the physical slippage failure critical point were recorded as the original sample data. Subsequently, the least squares method was used to fit the sample data to establish a functional relationship between the cable radial curvature modulus and the slippage failure critical value: In the formula, T is the critical value for slip failure, and K is the radial curvature modulus of the cable. The baseline sliding threshold; The curvature sensitivity coefficient is used as the reference slip threshold. Two empirical constants, the reference slip threshold and the curvature sensitivity coefficient, are obtained through fitting. To improve the calibration process and prevent parameters from becoming unusable due to poor data quality in extreme cases, the system checks the goodness of fit. If the goodness of fit is lower than a preset threshold (e.g., 0.8), it indicates fitting failure, and the system will automatically use the preset factory default parameters for subsequent testing, prompting the user to check the calibration data. The reference slip threshold and curvature sensitivity coefficient are stored in the system's non-volatile memory for accurate subsequent data processing. It should be noted that the implementer can set the number of standard harness samples according to actual conditions; this is not limited here. The least squares method and goodness of fit are well-known techniques and will not be elaborated further.

[0067] Second, calibration of material performance parameters. A set of qualified insulation layer samples with material hardness conforming to technical specifications (e.g., Shore hardness D45-55) were selected, and cable ties were assembled under standard torque settings. The average mechanical work output by the cable tie gun was recorded. Simultaneously, the average volume of the indentation on the samples was measured using a high-precision 3D scanner. The average mechanical work and average volume were stored as reference performance parameters for subsequent analysis.

[0068] Third, the standard reference profile is preset. The system preloads the CAD design drawing of the standard cable tie lock head, extracts the two-dimensional profile shape of the contact area between the bottom surface of the lock head and the insulation layer, converts the profile into a grayscale image, sets the grayscale value inside the profile to be non-zero and the grayscale value outside to be zero, and then normalizes the grayscale value in the grayscale image so that the sum of the grayscale values ​​of all pixels is equal to 1. The normalized grayscale data is defined as the standard lock head reference profile.

[0069] After completing the above calibration, the system enters the online testing phase. Considering the cylindrical surface and strong specular reflection characteristics of the insulation layer of new energy wiring harnesses (such as cross-linked polyethylene XLPE), this embodiment employs a polarized light illumination scheme to eliminate high-light interference and obtain pure diffuse reflection data. The specific hardware configuration is as follows: The system is equipped with four strip LED light sources, located above, below, left, and right of the industrial camera lens, respectively. The angle between the principal optical axis of the light source and the camera optical axis is 45°. A polarizer is installed in front of each LED light source, and an analyzer is installed in front of the camera lens. The transmission axes of the polarizer and analyzer are adjusted to be perpendicular to each other (orthogonal). This utilizes the principle of cross-polarization to block specular reflection light that maintains its polarization state from entering the lens, allowing only diffuse reflection light that has undergone depolarization to pass through. In addition, the system monitors the action signal of the cable tie gun in real time during the image acquisition process. When the system detects that the cable tie gun has completed the cutting action and is in the rebound stabilization period (e.g., within the 50ms to 100ms window after the cutting signal is triggered), the system triggers the acquisition command. That is, the system controls the LED light sources in four directions to light up in sequence. During each light source lighting period, the camera synchronously exposes and acquires a grayscale image, and finally obtains a grayscale image sequence containing four different lighting angles.

[0070] It should be noted that this embodiment operates under fixed optomechanical imaging conditions, and the calculation of all geometric parameters (such as the radial curvature modulus of the cable and the indentation volume) is based on a unified pixel equivalent or a calibrated physical coordinate system to ensure the applicability of the calibration parameters (such as the reference slip threshold).

[0071] The geometric reconstruction and benchmark extraction module 20 is used to solve the normal field of the insulation layer surface based on images from multiple angles, separate the normal field of the insulation layer surface, and obtain the background field of the cylindrical curvature of the cable and the micro-indentation residual field; based on the normal gradient change characteristics of the background field of the cylindrical curvature of the cable, the global axial direction and the radial curvature modulus of the cable are obtained.

[0072] Specifically, to recover the three-dimensional geometry of the cable surface from two-dimensional grayscale information, this embodiment calculates the surface normal field of the insulation layer based on images from multiple angles. This facilitates the accurate perception of minute surface tilts using illumination changes without relying on expensive 3D sensors. Considering that the actual calculated surface normal field of the insulation layer contains two types of geometric information superimposed: one is the low-frequency normal variation (macroscopic shape) caused by the cylindrical surface of the cable itself, and the other is the high-frequency normal abrupt change (microscopic defect) caused by the clamping and compression of the cable tie; in order to extract pure indentation features, the two must be separated to ensure that subsequent feature analysis is not interfered with by the cable's own curvature. Therefore, this embodiment separates the surface normal field of the insulation layer to obtain the background field of the cable's cylindrical curvature and the microscopic indentation residual field. This facilitates the independent quantification of micrometer-level indentation depth and provides a pure background reference for the subsequent establishment of the cable's physical coordinate system.

[0073] It is known that in the cylindrical surface model of a cable, the normal vector changes drastically along the circumferential direction (large curvature) while changing only slightly along the axial direction (curvature approaching 0). Therefore, utilizing this anisotropic characteristic, the system calculates the principal direction of curvature of the cable's cylindrical curvature background field, accurately identifying the physical extension posture of the cable in the image. Furthermore, this embodiment obtains the global axial direction and radial curvature modulus of the cable based on the normal gradient change characteristics of the cable's cylindrical curvature background field. The obtained global axial direction provides a unique physical coordinate reference for subsequently distinguishing between axial slip and circumferential tolerance; the obtained radial curvature modulus quantifies the cable's thickness, providing a crucial geometric basis for subsequent adaptive adjustment of the discrimination threshold. A larger radial curvature modulus indicates a smaller cable diameter (thinner cable), a more drastic rate of change of its surface normal in the radial direction, and a smaller frictional resistance per unit surface area. Therefore, the risk of axial slip failure during cable tie assembly is higher, and the system's tolerance for slip determination should be lower.

[0074] Preferably, in one feasible embodiment, the method for obtaining the normal field of the insulating layer surface is as follows: First, extract the region of interest on the cable surface, for example, limit it to the area within ±60° on both sides of the cable axis to avoid the influence of edge occlusion and shadow on the normal calculation; because the cable is a cylinder, the side light source is easily blocked by the cable itself, resulting in shadow on the backlight surface. By limiting the acquisition area to the upper surface sector of the cable facing the camera (such as ±60°), it is ensured that all pixels involved in the calculation can be effectively illuminated by at least 3 light sources, satisfying the physical premise of photometric stereo calculation. For any pixel within the region of interest, its grayscale value in each image is extracted to construct a brightness observation vector, accurately reflecting the surface reflection intensity information of that pixel under different lighting conditions. To establish a physical mapping relationship between lighting intensity and surface geometry, the light source direction corresponding to each image is obtained and constructed as a light source direction matrix, accurately reflecting the incident geometric parameters of the lighting at each acquisition. To calculate a unique surface normal from multi-angle lighting observations, a linear equation system of the brightness observation vector and the light source direction matrix is ​​constructed using the least squares method. A strategy for minimizing the error of the overdetermined equation system (a well-known technique, not elaborated upon here) is then used to solve the problem. The process involves solving a system of linear equations to obtain the initial normal vector of the pixel, accurately reflecting the tilt of the pixel's surface relative to the optical axis. Considering that the magnitude of the initial normal vector may be inconsistent due to the influence of surface albedo, the magnitude of the initial normal vector is normalized to obtain the unit normal analysis vector of the pixel, accurately reflecting the pure geometric orientation of the pixel and eliminating the interference of material color depth. In order to characterize the three-dimensional morphology of the cable insulation layer surface as a whole, the set of unit normal analysis vectors of all pixels is used as the normal field of the insulation layer surface, providing a complete vector field data foundation for subsequent separation of macroscopic cylindrical background and microscopic indentation features.

[0075] Preferably, in one feasible embodiment of this invention, the method for obtaining the background field of the cable cylindrical curvature and the residual field of the micro-indentation is as follows: For any unit normal analysis vector in the normal field of the insulation layer surface, the unit normal analysis vector is first decomposed into multiple dimensional components (e.g., components in the x, y, and z directions), which is beneficial for independent analysis of the frequency of change in each spatial dimension. In order to filter out the high-frequency local abrupt change information caused by cable tie compression and retain the low-frequency macroscopic morphology of the cable surface, each dimensional component is then smoothed using Gaussian low-pass filtering to obtain the background value of each dimensional component, which accurately reflects the normal trend of the cylindrical surface of the cable body at that point; wherein, Gaussian low-pass filtering is a known technique and will not be described in detail. In order to reconstruct the normal field of the cable body, all background values ​​corresponding to the unit normal analysis vector are combined into a vector as the background vector of the unit normal analysis vector, which accurately reflects the theoretical cylindrical normal of the unit normal analysis vector after removing indentation interference; in order to obtain the overall macroscopic geometric reference of the cable, the set of background vectors of all unit normal analysis vectors in the normal field of the insulation layer surface is used as the background field of the cable cylindrical curvature.

[0076] Considering that the original normal field is the superposition of the background field and the micro deformation field, in order to extract pure indentation features, the difference vector between each unit normal analysis vector and its background vector is taken as the micro residual vector, which accurately reflects the local normal deflection of each unit normal analysis vector caused by the compression of the cable tie; in order to quantify the micro deformation distribution of the entire field, the set of micro residual vectors of all unit normal analysis vectors in the normal field of the insulation layer surface is taken as the micro indentation residual field.

[0077] Preferably, in one feasible embodiment of this method, the method for obtaining the global axial direction and radial curvature modulus of the cable is as follows: For any background vector in the cylindrical curvature background field of the cable, in order to accurately identify the principal curvature direction of the cable cylinder at that point, the structural tensor of the background vector is obtained and eigenvalue decomposition is performed. Utilizing the cylindrical geometric properties, i.e., the normal change along the generatrix (axial direction) is minimal, the direction of the eigenvector corresponding to the minimum eigenvalue is taken as the local axial analysis direction of the background vector, accurately reflecting the direction with the gentlest normal change at that point, i.e., the potential cable extension direction; In order to eliminate local noise interference and obtain a unified physical reference for the entire field, the probability density distribution of the local axial analysis directions of all background vectors in the cylindrical curvature background field of the cable is statistically analyzed. The larger the probability density, the higher the consistency of the corresponding local axial analysis direction in the entire field, and the closer it is to the true physical axis of the cable; The direction with the highest probability density is then taken as the global axial direction of the cable.

[0078] Considering that the thickness of the cable directly determines the rate of change of the normal in the radial direction, in order to quantify the geometric dimensional characteristics of the cable, the modulus of the directional derivative of the background vector along the direction perpendicular to the global axial direction of the cable is used as the radial variation analysis value of the background vector. This accurately reflects the degree of change of the normal in the circumferential tangential direction at that point. The larger the radial variation analysis value, the greater the surface curvature at that point, and the smaller the corresponding local diameter of the cable. In order to comprehensively evaluate the average thickness properties of the cable and eliminate local measurement errors, the mean of the radial variation analysis values ​​of all background vectors in the cylindrical curvature background field of the cable is used as the radial curvature modulus of the cable.

[0079] The feature decoupling analysis module 30 is used to obtain the contour shaping migration flow field based on the topological distribution difference between the micro residual field of the indentation and the preset standard lock head reference contour; and to perform anisotropic decomposition of the contour shaping migration flow field based on the global axial direction of the cable to obtain the axial slip component and calculate the axial slip index.

[0080] Specifically, considering that the difference between the measured indentation and the standard shape includes not only positional deviation but also complex topological stretching deformation, traditional rigid registration methods are difficult to quantify this non-rigid distortion. In order to comprehensively evaluate the microscopic deformation trend of the indentation and enable subsequent quality judgment to be based on fine deformation characteristics rather than simple positional errors, this embodiment obtains the contour shaping migration flow field based on the topological distribution difference between the microscopic residual field of the indentation and the preset standard lock head reference contour (i.e., the standard lock head reference contour in step S1). This accurately reflects the optimal migration path and cost of each micro-element point required to move the measured indentation to the standard shape, which is beneficial for subsequent independent analysis of deformation in specific physical directions.

[0081] It is known that the deformation on the cylindrical surface of a cable has a clear directional physical meaning: axial drag represents the risk of slip failure, while circumferential offset represents normal assembly rotation tolerance. To achieve precise decoupling of slip risk and eliminate interference from normal assembly tolerance, the profile shaping migration flow field is anisotropically decomposed based on the global axial direction of the cable to obtain the axial slip component. This accurately reflects the dangerous component in the indentation deformation that is only related to axial drag, which helps to focus on the true failure characteristics and avoid misjudgment. To obtain a uniform quantitative index across the entire field, the axial slip index is calculated based on the axial slip component. This accurately reflects the cumulative degree of axial slip risk in the entire indentation area, which is conducive to achieving automated and standardized quality grading. Among them, the larger the axial slip index, the more severe the axial relative displacement of the cable tie during the locking process, and the higher the corresponding risk of false locking.

[0082] Preferably, in one feasible embodiment of this method, the method for obtaining the contour shaping and migration flow field is as follows: First, the divergence calculation and Poisson reconstruction are performed on the microscopic residual field of the indentation. When performing Poisson reconstruction, a zero boundary condition is adopted, that is, the depth value at the edge of the region of interest is assumed to be 0, thereby solving for a unique measured indentation depth distribution map. This is beneficial for converting the vector form of normal deviation into an intuitive scalar depth energy distribution, which is convenient for mass transfer analysis. In order to prevent background noise from interfering with the flow field calculation and avoid division by zero error, the total energy of the pixels with values ​​greater than the preset noise floor threshold in the measured indentation depth distribution map is normalized to generate the measured indentation mass distribution, ensuring that the sum of the distributions is 1, satisfying the mass conservation premise of the optimal transfer algorithm, which is beneficial for ensuring the convergence of the algorithm calculation and the correctness of the physical meaning. In this embodiment, the preset noise floor threshold is set to 3 times the average value of the camera dark current noise to filter out invalid small fluctuations. The implementer can set the size of the preset noise floor threshold according to the actual situation, which is not limited here.

[0083] To eliminate the influence of circumferential positional deviation of the indentation (i.e., normal assembly tolerance) on deformation analysis, while retaining the axial misalignment characteristics, the centroid coordinates and principal axis rotation angle of the measured indentation mass distribution are obtained through image moment analysis (i.e., first-order moment and second-order central moment). To focus on the topological deformation characteristics of the indentation caused by frictional drag, the centroid of a preset standard lock head reference profile is translated to coincide with the above centroid coordinates, and then rotated and aligned based on the principal axis rotation angle to obtain the aligned target distribution. It should be noted that axial slippage manifests in microscopic visual characteristics as topological deformation caused by asymmetric stretching, dragging, or smearing of the indentation along the cable axis. By aligning the centroid of the measured indentation mass distribution with the standard lock head reference profile, rigid displacement interference is eliminated. Thus, the flow field vector is used to capture this plastic deformation trend caused by frictional failure, ensuring that the subsequent flow field calculations only reflect the plastic deformation trend caused by extrusion and slip drag.

[0084] To quantify the topological mapping relationship between the measured indentation quality distribution and the target distribution, a source domain index space (measured distribution) and a target domain index space (standard distribution) are further established, and a discretized computation grid is defined to ensure that the computation process covers the entire indentation area. In order to measure the migration cost, the squared Euclidean distance from the non-zero pixel in the measured indentation quality distribution to the non-zero pixel in the target distribution is calculated, and a cost matrix is ​​constructed, which is helpful to characterize the movement distance in physical space.

[0085] To improve solution efficiency and obtain a smooth flow field, an entropy regularization term is introduced. The Sinkhorn iterative algorithm is then used to solve for the optimal transmission coupling matrix corresponding to the cost matrix, accurately reflecting the optimal probability distribution of mass transfer between the source and target domains. This helps to solve the computational complexity problem of large-scale pixel matching. To meet the cycle time requirements of real-time production line detection, this embodiment performs downsampling processing (e.g., reducing to a certain value) on the measured indentation mass distribution and target distribution before constructing the cost matrix. or (Mesh), or only perform calculations within a defined region of interest in the indentation, thereby keeping the computational scale within the millisecond range;

[0086] Then, based on the optimal transmission coupling matrix, the weighted average displacement vector required for each pixel in the source domain to match the target distribution is calculated, i.e., the trend vector in which the mass at that point should move. It should be noted that the slip in this embodiment not only includes macroscopic rigid displacement, but also, in the context of flow field analysis with centroid alignment, refers more deeply to the axial displacement potential energy of the indentation centroid, i.e., the geometric tendency to move along the axial direction implied by the topological shape of the measured indentation (such as asymmetric tailing or stretching). The Sinkhorn flow field accurately quantifies this microscopic motion trend caused by frictional failure. In addition, when introducing the entropy regularization term, a regularization coefficient is set (e.g., a value of 0.01-0.1) to control the smoothness of the transmission scheme and prevent calculation divergence. In order to describe the deformation trend of the entire field as a whole, the set of weighted average displacement vectors of all pixels in the entire field is used as the contour shaping migration flow field. The divergence calculation, Poisson reconstruction, index space establishment, Euclidean distance, entropy regularization term, and Sinkhorn iterative algorithm are all well-known and will not be elaborated upon further.

[0087] Preferably, in one feasible embodiment of this invention, the axial slip component is obtained as follows: For any topology migration vector in the contour-shaping migration flow field, the absolute value of the inner product of the topology migration vector and the unit vector in the global axial direction of the cable is calculated through vector dot product operation. The projection modulus of the topology migration vector in the global axial direction of the cable is then obtained as the axial slip component of the topology migration vector. It should be noted that, considering the mathematical characteristics of the Sinkhorn flow field vector, this axial slip component not only represents the physical drag displacement but also covers the axial asymmetric distribution of indentations caused by uneven force (such as eccentric indentations). In engineering quality evaluation, both actual physical slip and severe axial asymmetric deformation are characteristics of cable tie locking failure. Therefore, this embodiment refers to both as generalized slip risk for unified management without distinction. Thus, the axial slip component is used to determine the risk of false locking.

[0088] Preferably, in one feasible embodiment of this invention, the axial slip index is obtained as follows: For any topological migration vector in the contour-shaping migration flow field, the value of the corresponding pixel point (i.e., the normalized depth value) of the topological migration vector in the measured indentation mass distribution is obtained as the weighted analysis value of the topological migration vector; the larger the weighted analysis value, the deeper the indentation depth corresponding to the topological migration vector, the more it belongs to the critical area of ​​contact stress concentration, and the higher its slip characteristics contribute to the locking failure; in order to focus on the effective deformation of the main indentation area and suppress the noise interference of the shallow edge area, the product of the axial slip component of the topological migration vector and the weighted analysis value is used as the weighted slip analysis value of the topological migration vector; the larger the weighted slip analysis value, the more accurately it indicates that the location of the point corresponding to the topological migration vector is more likely to have significant and dangerous axial slip; in order to obtain a uniform slip quantification index across the entire field, the sum of the weighted slip analysis values ​​of all topological migration vectors in the contour-shaping migration flow field is used as the axial slip index, which comprehensively reflects the axial failure risk of the entire indentation area.

[0089] The quality inspection module 40 is used to determine the adaptive slip judgment threshold based on the radial curvature modulus of the cable, and to obtain the axial slip quality inspection result based on the comparison result between the axial slip index and the adaptive slip judgment threshold.

[0090] Specifically, it is known that thin-diameter cables, due to their large curvature and small contact area, have significantly weaker resistance to axial slippage than thick-diameter cables. Using a fixed threshold can easily lead to missed detections of high-risk cables. To avoid a one-size-fits-all judgment standard and achieve precise control over wire harnesses of different specifications, this embodiment determines an adaptive slippage judgment threshold based on the radial curvature modulus of the cable. This helps to dynamically tighten the judgment criteria for high-risk cables, ensuring that alarms are triggered when thin cables experience even minor slippage. Considering that the axial slippage index directly reflects the current slippage risk level, the axial slippage quality detection result is obtained by comparing the axial slippage index with the adaptive slippage judgment threshold. This accurately determines whether the cable tie assembly has experienced a false locking failure exceeding the safety tolerance limit, which helps to ensure the vibration resistance reliability of the wire harness product.

[0091] Preferably, in one feasible embodiment, the method for obtaining the adaptive slippage judgment threshold is as follows: the product of the radial curvature modulus of the cable and the curvature sensitivity coefficient calibrated in step S1 is used as the first characteristic value. The larger the first characteristic value, the higher the slippage risk gain brought about by the cable's geometric properties. The radial curvature modulus of the cable is multiplied by the curvature sensitivity coefficient because it is considered that the thinner the cable (the larger the radial curvature modulus of the cable), the smaller its corresponding physical diameter, and the weaker the frictional resistance per unit contact area. Therefore, the slippage risk is higher, and it is necessary to amplify this geometric risk gain through multiplication so as to significantly tighten the judgment threshold in the future. It is known that the threshold correction should be smooth and convergent in actual practice. In order to avoid the calculation collapse caused by the denominator being zero or negative, the sum of the first characteristic value and the preset constant is used as the attenuation factor. The preset constant is greater than 0. In this embodiment, the preset constant is set to 1 to ensure that the threshold does not attenuate when the radial curvature modulus of the cable is 0 (theoretical plane). The implementer can set the size of the preset constant according to the actual situation, which is not limited here.

[0092] To achieve a negative correlation control logic where a thinner cable results in a stricter adaptive slip threshold, the ratio of the baseline slip threshold to the attenuation factor is used as the dynamically calculated threshold. It should be noted that, to prevent program crashes and enhance the engineering robustness of the solution during calculation, if the attenuation factor in the denominator is less than a preset small positive number, it is set to this preset small positive number. In this embodiment, the preset small positive number is set to... The implementer can set the preset small positive number according to the actual situation, which is not limited here. The larger the dynamic calculation threshold, the greater the allowable slip tolerance of the current detection object. Considering that in the case of extremely thin cables, the attenuation factor may be too large, causing the dynamic calculation threshold to approach 0, thus misjudging normal sensor noise as slip failure, in order to ensure the basic robustness of the system, the dynamic calculation threshold and the preset system noise tolerance lower limit are compared, and the larger value of the two is used as the adaptive slip judgment threshold to ensure that the threshold is always higher than the background noise level of the system. In this embodiment, the preset system noise tolerance lower limit is set to 0.05 (dimensionless exponent value), which corresponds to the extreme fluctuation of the sensor when there is no deformation. The implementer can set the size of the preset system noise tolerance lower limit according to the actual situation, which is not limited here.

[0093] Preferably, in one feasible method of this embodiment, the method for obtaining the axial slip quality detection result is as follows: when the axial slip index is greater than the adaptive slip judgment threshold, it indicates that the cable tie has undergone significant axial relative displacement during the locking process, and there is a risk of false locking failure. At this time, generating an unqualified result characterized by excessive axial slip is beneficial to timely intercept defective products and prevent wire harnesses with insufficient vibration resistance from flowing into the next process; when the axial slip index is less than or equal to the adaptive slip judgment threshold, it indicates that the current micro deformation is mainly caused by normal assembly tolerances and no dangerous axial drag has occurred. At this time, generating a qualified result characterized by normal axial slip is beneficial to avoid excessive damage to normal tolerances while ensuring quality and improving the production line first pass rate.

[0094] Assuming the axial slippage quality test result is qualified by the system, further material performance verification is performed. Considering that the hardness of the insulation material may fluctuate in batches in actual situations, slippage detection alone cannot identify hidden defects such as excessively hard material (leading to shallow indentations and reduced anti-slippage ability) or excessively soft material (leading to excessive compression and damage to the wire core). In order to establish the performance correlation between mechanical input energy and geometric deformation output, this embodiment uses the product of the peak torque after the cable tie contacts the insulation layer, the effective stroke angle (i.e., the angle after excluding the idle stroke), and the preset transmission coefficient as the effective crimping work, which accurately reflects the total energy input applied by the equipment. The greater the effective crimping work, the higher the fastening energy output by the equipment. In this embodiment, the preset transmission coefficient is set to the comprehensive coefficient of the internal reduction ratio and winding radius of the cable tie gun (e.g., 0.5) to ensure the conversion of physical dimensions between the angle and linear displacement. The implementer can set the size of the preset transmission coefficient according to the actual situation, which is not limited here. It should be noted that this embodiment reads the feedback data of the cable tie gun controller in real time through the industrial fieldbus to obtain the peak torque and effective stroke angle. In order to quantify the actual deformation response of the insulating layer after being subjected to force, the result of volume integration of the micro-indentation residual field, that is, the sum of the micro-residual values ​​of all pixels, is used as the total displaced volume of the insulating layer, which accurately reflects the total amount of material displaced by the insulating layer; the larger the total displaced volume of the insulating layer, the more significant the plastic deformation is.

[0095] To eliminate the influence of individual equipment differences and measurement dimensions, and to achieve standardized performance evaluation, the ratio of the total insulation layer displacement volume to the effective crimping work is normalized based on reference performance parameters and used as a relative performance index, specifically: In the formula, It is a relative efficiency index; This refers to the total volume of the insulation layer that has been partially displaced. The average volume; For effective crimping; This is the average mechanical work. To prevent program crashes during calculation, if the effective pressing work (or the normalized effective pressing work term) in the denominator is less than a preset small positive number, it will be set to the preset small positive number, or the step will be skipped and a default alarm signal (such as abnormal mechanical work data) will be output.

[0096] It should be noted that the relative efficiency index is not intended to precisely measure the absolute physical modulus of a material, but rather serves as a statistical characteristic indicator used to screen outliers in the same batch of processes whose material properties significantly deviate from the mean. The relative efficiency index accurately reflects the deformation efficiency per unit of energy generated relative to the standard process. When the relative efficiency index is less than the preset lower limit, it indicates that sufficient mechanical work has been input, but the resulting deformation is too small. In this case, an unqualified result is generated, indicating that the material is too hard or has too low energy efficiency, which helps identify the risk of insulation layer aging and hardening. When the relative efficiency index is greater than the preset upper limit, it indicates that a small amount of mechanical work has produced excessive deformation. In this case, an unqualified result is generated, indicating that the material is too soft or has too high energy efficiency, which helps prevent core damage due to excessively soft insulation. When the relative efficiency index is greater than or equal to the preset lower limit and less than or equal to the preset upper limit, it indicates that the material properties are normal and the energy conversion meets expectations, resulting in a qualified result. This embodiment sets a preset lower limit of energy efficiency of 0.8 and a preset upper limit of energy efficiency of 1.2. This is because, according to the statistical law of normal distribution, the efficiency fluctuation of qualified products is usually concentrated within ±20% of the mean. Exceeding this range likely corresponds to an anomaly. Implementers can set the preset lower and upper limits of energy efficiency according to actual conditions; no specific limit is imposed here. This multi-dimensional judgment logic fills the blind spots of single visual inspection, achieving comprehensive control over assembly quality.

[0097] In summary, this embodiment includes: an image acquisition module that uses polarimetric stereo technology to acquire multi-angle images; a geometric reconstruction and reference extraction module that calculates the surface normal field of the insulation layer and separates the background field of the cable's cylindrical curvature and the micro-indentation residual field, thereby extracting the global axial and radial curvature moduli of the cable; a feature decoupling analysis module that calculates the contour-shaping migration flow field of the indentation micro-residual field relative to the standard profile, and performs anisotropic decomposition of the flow field based on the global axial direction of the cable to obtain the axial slip index; and a quality inspection module that determines the adaptive slip judgment threshold based on the cable's radial curvature modulus, thereby generating the inspection result. This invention solves the problems of difficulty in sensing micro-deformation on reflective cylindrical surfaces and difficulty in decoupling slip risk from assembly tolerances, improving the accuracy of wire harness and cable tie assembly quality inspection.

[0098] Example 2:

[0099] This invention also proposes a machine vision-based new energy vehicle wiring harness and cable tie assembly quality inspection device. This device includes a memory and a processor. The memory stores executable program code, and the processor calls and executes the executable program code to perform the machine vision-based new energy vehicle wiring harness and cable tie assembly quality inspection system provided in this application embodiment. Specifically, the device may be a chip, component, or module. The chip may include a connected processor and memory; the memory stores instructions, and when the processor calls and executes the instructions, the chip can execute the machine vision-based new energy vehicle wiring harness and cable tie assembly quality inspection system provided in the above embodiment.

[0100] Furthermore, this application also protects a computer device; please refer to [link to relevant documentation]. Figure 2 The computer device includes a memory 401, a processor 402, and a computer program 403 stored in the memory 401 and running on the processor 402. When the processor 402 executes the computer program 403, the computer device can execute any of the machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection systems described above.

[0101] Example 3:

[0102] The present invention also provides a computer-readable storage medium storing computer program code, which, when executed on a computer, causes the computer to perform the aforementioned method steps to implement the machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system provided in the above embodiments.

[0103] Example 4:

[0104] The present invention also provides a computer program product that, when run on a computer, causes the computer to perform the aforementioned related steps to realize the machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system provided in the above embodiments.

[0105] In this embodiment, the device, computer-readable storage medium, computer program product, or chip are all used to execute the corresponding methods provided above. Therefore, the beneficial effects they can achieve can be referred to the beneficial effects in the corresponding methods provided above, and will not be repeated here.

[0106] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0107] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. A machine vision-based quality inspection system for the assembly of wiring harnesses and cable ties in new energy vehicles, characterized in that, The system includes: The image acquisition module is used to acquire images of the cable surface from multiple angles using polarimetric stereoscopic technology; The geometric reconstruction and benchmark extraction module is used to calculate the normal field of the insulation layer surface based on images from multiple angles, separate the normal field of the insulation layer surface, and obtain the background field of the cable cylindrical curvature and the micro-indentation residual field; based on the normal gradient variation characteristics of the background field of the cable cylindrical curvature, the global axial direction and the radial curvature modulus of the cable are obtained. The feature decoupling analysis module is used to obtain the contour shaping migration flow field based on the topological distribution difference between the indentation micro residual field and the preset standard lock head reference contour; it performs anisotropic decomposition of the contour shaping migration flow field based on the global axial direction of the cable to obtain the axial slip component and calculate the axial slip index; The quality inspection module is used to determine the adaptive slip judgment threshold based on the radial curvature modulus of the cable, and to obtain the axial slip quality inspection result based on the comparison result between the axial slip index and the adaptive slip judgment threshold.

2. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The method for obtaining the normal field of the insulating layer surface is as follows: For any pixel, extract the gray value of that pixel in each image and construct the brightness observation vector of that pixel; Obtain the light source direction corresponding to each image and construct a light source direction matrix; A system of linear equations is constructed using the least squares method to establish the brightness observation vector and the light source direction matrix. Solving the system of linear equations yields the initial normal vector of the pixel. The initial normal vector is normalized to obtain the unit normal analysis vector of the pixel. The set of unit normal analysis vectors of all pixels is taken as the normal field of the insulating layer surface.

3. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 2, characterized in that, The method for obtaining the background field of the cylindrical curvature of the cable and the residual field of the micro-indentation is as follows: For any unit normal analysis vector in the normal field of the insulating layer surface, the unit normal analysis vector is decomposed into multiple dimensional components, and each dimensional component is smoothed by Gaussian low-pass filtering to obtain the background value of each dimensional component. Combine all the background values ​​corresponding to the unit normal analysis vector into a single vector, which serves as the background vector of the unit normal analysis vector; The set of background vectors of all unit normal analysis vectors in the normal field of the insulation layer surface is taken as the background field of the cable cylindrical curvature. The difference vector between each unit normal analysis vector and its background vector is used as the micro residual vector; The set of microscopic residual vectors of all unit normal analysis vectors in the normal field of the insulating layer surface is taken as the microscopic indentation residual field.

4. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 3, characterized in that, The method for obtaining the global axial and radial curvature moduli of the cable is as follows: For any background vector in the background field of the cylindrical curvature of the cable, obtain the structure tensor of the background vector and perform eigenvalue decomposition. The direction of the eigenvector corresponding to the smallest eigenvalue is taken as the local axial analysis direction of the background vector. The probability density distribution of the local axial direction of all background vectors in the background field of the cylindrical curvature of the cable is statistically analyzed, and the direction with the highest probability density is taken as the global axial direction of the cable. The magnitude of the directional derivative of the background vector along the direction perpendicular to the global axial direction of the cable is used as the radial variation analysis value of the background vector; The mean value of the radial variation analysis of all background vectors in the background field of the cylindrical curvature of the cable is taken as the radial curvature modulus of the cable.

5. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The method for obtaining the contour shaping and migration flow field is as follows: Divergence calculation and Poisson reconstruction are performed on the micro residual field of the indentation to obtain the measured indentation depth distribution map. Then, the pixels with values ​​greater than the preset noise threshold in the measured indentation depth distribution map are normalized to generate the measured indentation quality distribution. Calculate the centroid coordinates and spindle rotation angle of the measured indentation mass distribution, translate the centroid of the preset standard lock head reference profile to coincide with the centroid coordinates, and perform rotation alignment based on the spindle rotation angle to obtain the aligned target distribution; Establish source domain index space and target domain index space, calculate the squared Euclidean distance from non-zero pixels in the measured indentation quality distribution to non-zero pixels in the target distribution, and construct the cost matrix; An entropy regularization term is introduced, and the optimal transmission coupling matrix corresponding to the cost matrix is ​​solved using the Sinkhorn iterative algorithm. Based on the optimal transmission coupling matrix, calculate the weighted average displacement vector required for each pixel in the source domain to match the target distribution; The set of weighted average displacement vectors of all pixels in the entire field is used as the contour shaping migration flow field.

6. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The method for obtaining the axial slip component is as follows: For any topology migration vector in the contour-shaping migration flow field, calculate the projection modulus of the topology migration vector in the global axial direction of the cable, which is taken as the axial slip component of the topology migration vector.

7. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 5, characterized in that, The method for obtaining the axial slip index is as follows: For any topology migration vector in the contour shaping migration flow field, obtain the value of the corresponding pixel in the measured indentation quality distribution of the topology migration vector, and use it as the weight analysis value of the topology migration vector; The product of the axial slip component of the topology migration vector and the weighted analysis value is used as the weighted slip analysis value of the topology migration vector. The weighted sum of the slip analysis values ​​of all topology migration vectors in the contour-shaping migration flow field is used as the axial slip index.

8. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The method for obtaining the adaptive slip determination threshold is as follows: The product of the radial curvature modulus of the cable and the curvature sensitivity coefficient is used as the first characteristic value; The sum of the first eigenvalue and the preset constant is used as the attenuation factor; where the preset constant is greater than 0. The ratio of the baseline slip threshold to the attenuation factor is used as the dynamically calculated threshold. The dynamic calculation threshold is compared with the preset system noise floor tolerance limit, and the larger of the two values ​​is used as the adaptive sliding judgment threshold.

9. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The method for obtaining the axial slip mass detection results is as follows: When the axial slip index is greater than the adaptive slip judgment threshold, an unqualified result is generated, indicating that the axial slip is too large. When the axial slip index is less than or equal to the adaptive slip judgment threshold, a qualified result representing normal axial slip is generated.

10. The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system as described in claim 1, characterized in that, The machine vision-based new energy vehicle wiring harness cable tie assembly quality inspection system also includes a material performance verification module, which is used to take the product of the peak torque, effective stroke angle and preset transmission coefficient after the cable tie contacts the insulation layer during the cable tie assembly process as the effective crimping work. The volume integral of the micro-indentation residual field is taken as the total displaced volume of the insulating layer. The normalized ratio of the total displaced volume of the insulation layer to the effective crimping work is used as the relative efficiency index. When the relative efficiency index is less than the preset lower limit of energy efficiency, an unqualified result is generated, indicating that the material is too hard or the energy efficiency is too low. When the relative efficiency index is greater than the preset energy efficiency limit, an unqualified result is generated, indicating that the material is too soft or the energy efficiency is too high. When the relative efficiency index is greater than or equal to the preset lower limit of energy efficiency and less than or equal to the preset upper limit of energy efficiency, a qualified result for characterizing the material is generated.

Images