A Machine Vision-Based Method and System for Quality Inspection of Cable Accessories

Abstract

This invention discloses a machine vision-based method and system for inspecting the manufacturing quality of cable accessories. The method includes: acquiring information and extracting edge contour information using an RGB three-color separation edge detection algorithm; constructing an initial 3D point cloud dataset, merging it, and matching it with a reference coordinate system after coordinate transformation; dividing the region and calculating feature parameters based on process parameters; and comparing the process parameters with standard parameters. The method extends to include a model formula incorporating process parameters. The system comprises six units, including edge contour analysis, which are sequentially connected. This method and system, through multi-dimensional processing—from image acquisition and point cloud construction to parameter comparison and output—achieves a complete transformation from visual information to manufacturing quality assessment, ensuring the systematic and efficient nature of the inspection process and enabling effective inspection of the manufacturing quality of cable accessories.

Description

Technical Field

[0001] This invention relates to the field of cable manufacturing quality inspection, and in particular to a method and system for inspecting the manufacturing quality of cable accessories based on machine vision. Background Technology

[0002] In the operation of power systems, cable accessories, as key components connecting cables and equipment, directly impact the stability of power transmission due to their manufacturing quality. With the power industry's increasing demands for equipment reliability, the need for testing process parameters such as insulation thickness, surface contour accuracy, and geometric dimensional matching of cable accessories is becoming increasingly urgent. Traditional testing methods rely heavily on manual visual inspection and contact measuring tools, making it difficult to comprehensively capture the complex surface features and three-dimensional spatial parameters of cable accessories. Machine vision-based inspection technology, through image acquisition and processing, provides a new technical path for the quality inspection of cable accessories. Among these, edge detection using RGB three-color separation and the combination of 3D point cloud modeling with process parameters have become important directions for improving inspection efficiency and accuracy.

[0003] Existing technologies have significant limitations in the quality inspection of cable accessories. On the one hand, traditional edge detection methods rely solely on single-channel image information, failing to effectively distinguish the complex color edge features on the surface of cable accessories caused by material differences and varying process zones. This results in insufficient completeness of the extracted contour information, making it difficult to accurately reflect the actual morphology of critical areas such as insulation layer overlaps and shielding layer transition zones. On the other hand, the specific process parameters of cable accessories are not fully integrated during 3D modeling, leading to deviations between the constructed model and the actual process design benchmark. Consequently, the geometric dimension detection results cannot be directly correlated with process indicators such as material shrinkage rate and wall thickness tolerance, resulting in reduced matching between the detection data and process requirements. Summary of the Invention

[0004] In order to overcome the shortcomings and deficiencies of the existing technology, the present invention provides a method and system for inspecting the process quality of cable accessories based on machine vision.

[0005] The technical solution adopted in this invention is a machine vision-based method for inspecting the manufacturing quality of cable accessories, comprising:

[0006] Step S1: Collect RGB three-color image information of the cable accessory surface, and use the RGB three-color separation edge detection algorithm to perform multi-dimensional analysis on the gray value distribution characteristics of different color gamut channels in the image to extract the edge contour information of the calibrated area on the cable accessory surface;

[0007] Step S2: Based on the extracted edge contour information, construct an initial three-dimensional point cloud data set for the surface of the cable accessory, which includes the coordinate parameters of the edge contour feature points in three-dimensional space;

[0008] Step S3: Perform spatial coordinate transformation on the initial 3D point cloud data set to match the point cloud data with the reference coordinate system of the cable accessory process design;

[0009] Step S4: Combine the preset geometric parameters of the cable accessory process, perform region division on the matched 3D point cloud data to obtain sub-point cloud datasets corresponding to each process partition of the cable accessory.

[0010] Step S5: For each sub-point cloud dataset, calculate the spatial topology parameters and surface curvature distribution characteristics, and correlate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process;

[0011] Step S6: Compare the spatial topology parameters, surface curvature distribution characteristics, and cable accessory process standard parameters of each sub-point cloud dataset, and output the detection results.

[0012] Furthermore, in step S1, when using the RGB three-color separation edge detection algorithm, an edge contour feature extraction model is constructed, the expression of which is:

[0013]

[0014] in, E i,j For the image ( i , j The edge intensity value of a pixel; k This is the index for the RGB three-color channels, with values ​​1, 2, and 3 corresponding to the R, G, and B channels, respectively. oh k The edge detection weight coefficient for the k-th channel; R k ( i , j ), G k ( i , j ), B k ( i , j ) are respectively the kth channel in ( i , j The pixel value of a pixel; α , β , c These are the reflectance parameters of the cable accessory insulation material for red, green, and blue light, respectively. For gradient operators; i i, j for( i , jThe surface texture orientation angle parameter of the cable accessory at the pixel point; in step S2, when constructing the initial three-dimensional point cloud data set, the two-dimensional image coordinates of the edge contour feature points are converted into three-dimensional spatial coordinates, and the radial dimension parameter of the cable accessory process is introduced during the conversion process. D and axial length parameters L The spatial scale of the point cloud data is consistent with the actual dimensions of the cable accessories.

[0015] Furthermore, in step S3, when performing spatial coordinate transformation on the initial 3D point cloud data set, a coordinate transformation model is established, the expression of which is:

[0016]

[0017] in,( X , Y , Z () represents the original coordinates of the feature points in the initial 3D point cloud data; To match the coordinates of the reference coordinate system after transformation; Let be the coordinate transformation matrix, and:

[0018]

[0019] in, i This is a rotation angle parameter about the X-axis, which is related to the bending angle process parameters of the cable accessories; The rotation angle parameter around the Y-axis is associated with the torsion process parameters of the cable accessories; a , b , c The translation parameters in the X, Y, and Z axes are determined by the installation and positioning process parameters of the cable accessories; in step S4, when performing region division on the matched 3D point cloud data, the insulation layer layering parameters of the cable accessory process are used. h n , and the overlap width parameter of the shielding layer W s , where n is the number of insulation layers, the sub-point cloud dataset is divided according to the insulation layer interface and the shielding layer boundary.

[0020] Furthermore, in step S5, when calculating the spatial topology parameters of the sub-point cloud dataset, a spatial neighborhood relationship model is constructed, expressed as:

[0021]

[0022] in, T p For point p Spatial topological correlation degree; N p For point pThe number of neighboring points; N ( p ) is a point p The set of neighborhood points; X p , Y p , Z p ), ( X q , Y q , Z q ) are points p and its neighboring points q 3D coordinates; d std Standard distance between adjacent feature points specified for cable accessory manufacturing process; l The elastic deformation coefficient parameter of the cable accessory material; when calculating the surface curvature distribution characteristics, the allowable deviation parameter of the curvature radius of the cable accessory manufacturing process is also considered. The curvature values ​​of each sub-point cloud dataset are divided into intervals, and the proportion of point clouds in different curvature intervals is statistically analyzed.

[0023] Furthermore, in step S1, the RGB three-color separation edge detection algorithm also includes quantitative analysis of color differences in different process areas on the surface of the cable accessory, constructing a color difference edge detection model, the expression of which is:

[0024]

[0025] in, E color The color edge intensity between different process areas; k , For indexing different process areas; For the region k With the region Color difference weighting coefficient; R k , G k , B k They are respectively regions k The average pixel values ​​of the R, G, and B channels; They are respectively regions The average pixel values ​​of the R, G, and B channels; The R, G, and B channels are located in the region. k and The color attenuation coefficient is related to the surface coating process parameters of the cable accessories; in step S2, when constructing the initial three-dimensional point cloud data set, the wall thickness parameters of the cable accessories are introduced.t The thickness direction is corrected for the three-dimensional coordinates of the edge contour feature points.

[0026] Furthermore, in step S4, when performing region segmentation on the matched 3D point cloud data, a region segmentation model based on process parameters is established, expressed as:

[0027]

[0028] in, S m For the first m Sub-point cloud datasets for each process partition; P Points in a 3D point cloud; x 0m , y 0m , z 0m ) is the first m The center coordinates of each process zone; r m For the first m The reference radius of each process zone; Δ r m For the first m Permissible radius deviation parameters for each process zone; t m For the first m The reference wall thickness for each process zone; Δ t m For the first m Permissible wall thickness deviation parameters for each process zone; t P For point P The actual wall thickness of the cable accessory at the location; in step S5, when relating the material shrinkage rate of the cable accessory process, the correlation parameter between the material shrinkage rate and temperature change is used. Temperature compensation correction is applied to the spatial topology parameters.

[0029] Further, in step S1, when extracting the edge contour information of the calibration area on the surface of the cable accessory, an edge contour continuity detection model is constructed, the expression of which is:

[0030]

[0031] in, C It is the continuity index of the edge contour; n The number of feature points on the edge contour; d i,i+1 For the first i The and the first i +1 distance between edge feature points; d0 represents the baseline distance parameter between adjacent edge feature points; s d The standard deviation of the edge point spacing is related to the forming process parameters of the cable accessories; i i,i+1 For the first i The and the first i +1 direction angle of the line connecting edge feature points; i 0 is the reference direction angle of the edge profile; in step S3, when performing spatial coordinate transformation, the origin and coordinate axis direction of the reference coordinate system are determined by using the flange positioning hole parameters of the cable accessory.

[0032] Furthermore, in step S5, when calculating the surface curvature distribution characteristics of the sub-point cloud dataset, a correlation model between curvature and process parameters is established, expressed as:

[0033]

[0034] in, K p For point p The average curvature of the region; A p For point p The area where it is located; R q For point q The radius of curvature; or This is the temperature influence coefficient, which is related to the coefficient of thermal expansion of the cable accessory materials; Δ T This is the difference between the actual measured temperature and the standard temperature. T 0 represents the standard temperature parameter; This is the pressure influence coefficient, which is related to the installation pressure parameters of the cable accessories; Δ P This is the difference between the actual installation pressure and the standard installation pressure. P 0 represents the standard installation pressure parameter; in step S6, when comparing the parameters of each sub-point cloud dataset with the standard parameters, the insulation resistance parameter of the cable accessory process is used. r The contrast threshold for the surface curvature distribution characteristics is adjusted.

[0035] Furthermore, in step S2, when constructing the initial three-dimensional point cloud data set of the cable accessory surface, a correlation model between point cloud density and process parameters is established, expressed as:

[0036]

[0037] in, r cloud ( x , y , z ) is in three-dimensional spacex , y , z Point cloud density at location ) r 0 is the baseline point cloud density; α The surface roughness influence coefficient of cable accessories; s ( x , y , z )for( x , y , z Surface roughness parameters at location ) β The coefficient representing the influence of the structural complexity of cable accessories; c ( x , y , z )for( x , y , z The structural complexity parameter at the location is related to the number of process corners at that location; in step S3, when matching point cloud data with the reference coordinate system, the position parameter of the positioning pin hole of the cable accessory is used. x p , y p , z p (), which serves as the reference feature point for coordinate matching.

[0038] A machine vision-based cable accessory manufacturing quality inspection system includes:

[0039] The edge contour extraction and parsing unit is used to acquire RGB three-color image information of the cable accessory surface and extract edge contour information through the RGB three-color separation edge detection algorithm. Its output end is connected to the input end of the data set construction and coordinate parameter integration unit.

[0040] The data set construction and coordinate parameter integration unit is used to construct an initial three-dimensional point cloud data set including the three-dimensional spatial coordinate parameters of the edge contour feature points based on the edge contour information. Its output end is connected to the input end of the data space coordinate transformation and reference coordinate system matching unit.

[0041] The data spatial coordinate transformation and reference coordinate system matching unit is used to transform the initial three-dimensional point cloud data set into spatial coordinates and match it with the reference coordinate system of cable accessory process design. Its output end is connected to the input end of the data area division and sub-point cloud dataset generation unit.

[0042] The data region division and sub-point cloud dataset generation unit is used to perform region division on the matched three-dimensional point cloud data in combination with the preset geometric parameters of the cable accessory process and obtain the sub-point cloud dataset corresponding to each process partition. Its output end is connected to the input end of the distribution feature calculation and process parameter association unit.

[0043] The distribution feature calculation and process parameter association unit is used to calculate the spatial topology parameters and surface curvature distribution characteristics of each sub-point cloud dataset and associate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process. Its output end is connected to the input end of the cable accessory process standard parameter comparison and analysis unit.

[0044] The cable accessory process standard parameter comparison and analysis unit is used to receive the spatial topology parameters and surface curvature distribution characteristics output by the distribution characteristic calculation and process parameter correlation unit, compare them with the cable accessory process standard parameters, and output the test results.

[0045] Beneficial Effects: This invention proposes a machine vision-based method and system for inspecting the process quality of cable accessories. It employs an RGB three-color separation edge detection algorithm to perform multi-dimensional analysis of the grayscale distribution characteristics of different color gamut channels in cable accessory images, extracting edge contour information of key surface areas. This overcomes the problem of insufficient contour information completeness caused by traditional edge detection methods relying solely on single-channel image information, accurately reflecting the actual shape of key parts such as insulation layer overlaps and shielding layer transition areas. By using 3D point cloud modeling to construct an initial dataset and merging it for spatial coordinate transformation and matching with a reference coordinate system, and combining cable accessory process parameters to divide regions and calculate feature parameters, this invention addresses the shortcomings of insufficient integration of process parameters in 3D modeling, leading to deviations between the model and design benchmarks, and low matching between inspection data and process requirements. This allows the inspection results to be directly correlated with process indicators such as material shrinkage rate and wall thickness tolerance. The system's units are sequentially connected, collaboratively completing the entire process from image acquisition to result output, achieving comprehensive and accurate inspection of cable accessory process quality, improving the targeting and effectiveness of inspection, and meeting the technical requirements of the power industry for cable accessory quality inspection. Attached Figure Description

[0046] Figure 1 This is a flowchart of the method steps of the present invention;

[0047] Figure 2 This is a diagram showing the system unit composition of the present invention. Detailed Implementation

[0048] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. The application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0049] like Figure 1As shown, the machine vision-based method for inspecting the manufacturing quality of cable accessories includes:

[0050] Step S1: Collect RGB three-color image information of the cable accessory surface, and use the RGB three-color separation edge detection algorithm to perform multi-dimensional analysis on the gray value distribution characteristics of different color gamut channels in the image to extract the edge contour information of the calibrated area on the cable accessory surface;

[0051] Specifically, the technical parameters in this step cover the grayscale value ranges of each of the RGB three color channels, the sharpness index of the edge contours of key areas on the surface of cable accessories, and the feature overlap after resolution of different color gamut channels. The grayscale value range determines the ability to distinguish color differences on the surface of cable accessories, the edge contour sharpness index affects the basic accuracy of subsequent 3D modeling, and the feature overlap reflects the consistency of multi-channel resolution results. The significance of the above design is that by performing multi-dimensional analysis of the grayscale value distribution characteristics of the three RGB color gamut channels respectively, it is possible to capture edge details that are difficult to identify under a single channel, providing a more accurate two-dimensional contour basis for subsequent 3D point cloud construction, and ensuring accurate identification of the surface process features of cable accessories during subsequent inspection.

[0052] In practice, an industrial camera with a resolution of at least 2048×1536 is used to acquire images of cable accessories. The camera's white balance parameters are set to 5000K-6500K to ensure the stability of the grayscale values ​​of the RGB three-color channels. The acquired images are decomposed into three single-channel images: R, G, and B. The grayscale value range of each channel is controlled between 0-255. For the insulation layer area of ​​the cable accessories, the grayscale values ​​of the R channel must be between 80-150, the G channel between 70-140, and the B channel between 60-130 to highlight the grayscale difference between this area and other parts. Grayscale gradient calculation is performed on the image of each channel. Pixels with gradient values ​​greater than 30 are marked as edge candidate points. Then, morphological operations are used to remove isolated candidate points (groups of consecutive points with fewer than 5 points). The edge candidate points of the three channels are spatially compared. Candidate points that appear in all three channels or whose positional deviation in two channels is less than 2 pixels are retained to form the final edge contour information. The length of the continuous segment of the contour line is not less than 10 pixels to ensure the integrity of the contour.

[0053] Step S2: Based on the extracted edge contour information, construct an initial three-dimensional point cloud data set for the surface of the cable accessory, which includes the coordinate parameters of the edge contour feature points in three-dimensional space;

[0054] Specifically, the technical parameters for this step include the sampling density of edge contour feature points, the measurement error range of three-dimensional coordinates, and the spatial distribution uniformity index of the point cloud data. Sampling density determines the point cloud data's ability to represent contour details, the measurement error range directly relates to the degree of agreement between the point cloud data and the actual object, and the spatial distribution uniformity index affects the stability of subsequent data processing. The significance of this design lies in transforming two-dimensional edge contour information into a point cloud data set including three-dimensional spatial coordinate parameters, realizing the transformation from planar detection to three-dimensional analysis. This provides data support for comprehensively reflecting the spatial structural characteristics of cable accessories, enabling subsequent evaluation of the morphological parameters of various process components from a three-dimensional perspective.

[0055] In specific implementation, based on the edge contour information obtained in step S1, feature points are extracted at a density of 3-5 points per millimeter. For curved contour segments with a radius of curvature less than 50mm, the sampling density is increased to 5-8 points per millimeter. A laser tracker is used to calibrate the image acquisition system, obtaining the camera's intrinsic and extrinsic parameters, with the calibration error controlled within 0.02mm. Using the principle of triangulation, the two-dimensional image coordinates are converted into three-dimensional spatial coordinates, with the coordinate conversion errors in the X, Y, and Z axes controlled within ±0.015mm, ±0.015mm, and ±0.02mm, respectively. The converted three-dimensional coordinates are sorted according to spatial location to form an initial three-dimensional point cloud dataset. The distribution of the point cloud data in space must satisfy the following condition: the number of points within any 10mm×10mm×10mm cube region must not deviate from the average value by more than 15%. Noise points are removed from the generated point cloud data, eliminating points whose average distance to the 20 nearest points is greater than 0.1mm, ensuring the validity of the point cloud data.

[0056] Step S3: Perform spatial coordinate transformation on the initial 3D point cloud data set to match the point cloud data with the reference coordinate system of the cable accessory process design;

[0057] Specifically, the technical parameters for this step include the adjustment range of the rotation angle for coordinate transformation, the adjustment accuracy of the translation amount, and the overlap index after coordinate system matching. The adjustment range of the rotation angle determines the ability of the point cloud data to correct its spatial attitude, the adjustment accuracy of the translation amount affects the accuracy of the benchmark alignment, and the overlap index is used to quantify the matching effect. The significance of the above design is that it ensures that the point cloud data is consistent with the process design benchmark coordinate system through spatial coordinate transformation, eliminating systematic deviations caused by acquisition angle and placement position, providing a unified spatial benchmark for subsequent testing by process zone, and ensuring the comparability of test data from different batches.

[0058] In practice, the reference coordinate system parameters for the cable accessory process design are obtained, including the origin position (set as the intersection of the cable accessory's central axis and end face) and the coordinate axis directions (X-axis along the central axis, Y-axis and Z-axis perpendicular to the central axis and mutually perpendicular). An iterative nearest-point algorithm is used to register the initial 3D point cloud data with the reference coordinate system. First, 3-5 feature points located at key parts of the cable accessory (such as the center of the flange bolt holes) are selected as registration references, and the initial transformation matrix is ​​calculated. After transforming the point cloud data according to the initial transformation matrix, the distance from all points to their corresponding theoretical positions in the reference coordinate system is calculated. Points with a distance greater than 0.1 mm are marked as outliers. The transformation matrix is ​​optimized through weighted iterative calculation (outlier weight set to 0.3) to ensure that the average distance of non-outliers is less than 0.05 mm and the maximum distance is less than 0.1 mm. Finally, the distribution range of the matched point cloud data in the X, Y, and Z axes of the reference coordinate system must deviate from the design model by less than 1 mm to ensure the accuracy of coordinate matching.

[0059] Step S4: Combine the preset geometric parameters of the cable accessory process, perform region division on the matched 3D point cloud data to obtain sub-point cloud datasets corresponding to each process partition of the cable accessory.

[0060] Specifically, the technical parameters for this step include the number of process partitions, the spatial boundary accuracy of each partition, and the integrity index of the sub-point cloud dataset. The number of process partitions is determined based on the structural complexity of the cable accessories; the spatial boundary accuracy determines the accuracy of the correspondence between the partitions and the actual process locations; and the integrity index ensures that the point cloud data of each partition can reflect the overall characteristics of that location. The significance of this design is that dividing the point cloud data according to process partitions allows for separate inspections targeting locations with different process requirements. This facilitates comparison of the inspection results of each region with the corresponding process parameters, improving the targeting and accuracy of quality judgment.

[0061] In practice, based on the process design documents of the cable accessories, the three-dimensional spatial boundaries of each process zone are determined. For example, the end area of ​​the insulation layer is a cylindrical area with an X-axis distance of 0-50mm and a radial distance of 20-30mm from the central axis; the overlapping area of ​​the shielding layer is a cylindrical area with an X-axis distance of 50-100mm and a radial distance of 15-25mm from the central axis. The setting error of the boundary coordinates is less than 0.2mm. For the matched three-dimensional point cloud data, the spatial coordinates of each point are checked to see if they fall within the boundary range of each process zone. Point cloud data belonging to the same zone are grouped into a sub-point cloud dataset. The boundary integrity of each sub-point cloud dataset is checked to ensure that it includes feature points on all edge contours of the process zone (coverage rate not less than 95%), and the number of internal points meets the requirement of not less than 50 points per square centimeter of projected area. For feature points that cross zones (belonging to the boundaries of two zones simultaneously), they are classified according to their main process location (closer to the center of that location), with a classification deviation of less than 0.5mm.

[0062] Step S5: For each sub-point cloud dataset, calculate the spatial topology parameters and surface curvature distribution characteristics, and correlate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process;

[0063] Specifically, the technical parameters in this step include the computational resolution of spatial topology parameters, the accuracy of surface curvature assessment, and the correlation strength coefficient of process parameters. The computational resolution determines the level of detail in the topology analysis, the accuracy of curvature assessment affects the accuracy of surface smoothness judgment, and the correlation strength coefficient reflects the degree of matching between feature parameters and process parameters. The significance of this design lies in realizing the transformation of point cloud data into process quality indicators by calculating the spatial topology parameters and surface curvature distribution characteristics of the sub-point cloud dataset and establishing a correlation with process parameters, thus providing directly applicable quantitative parameters for subsequent comparative analysis.

[0064] In practice, neighborhood analysis is performed on each sub-point cloud dataset, with a neighborhood radius of 5mm. The average distance and standard deviation between each point and all points in its neighborhood are calculated. The average distance is used as the local density parameter for that point, and the standard deviation is used as the distribution uniformity parameter. The least squares method is used to fit a surface to the neighborhood points of each point, calculating the principal curvature and Gaussian curvature of that point. The number of sampling points for curvature calculation is no less than 15, and the fitting error is less than 0.03mm. The cable accessory process parameter database is accessed to obtain the material shrinkage rate (range 0.5%-3%) and insulation layer thickness (range 2-10mm) corresponding to each process zone. The local density parameter is linearly correlated with the material shrinkage rate (the density parameter correction coefficient is adjusted by 0.02 for every 1% increase in shrinkage rate), and the principal curvature is correlated with the insulation layer thickness (the curvature assessment threshold is relaxed by 0.01 for every 1mm increase in thickness). This forms a set of associated feature parameters, and the error of the association calculation is controlled within 5%.

[0065] Step S6: Compare the spatial topology parameters, surface curvature distribution characteristics, and cable accessory process standard parameters of each sub-point cloud dataset, and output the detection results.

[0066] Specifically, the technical parameters for this step include the comparison threshold between characteristic parameters and standard parameters, the weighting coefficients for each process zone, and the judgment threshold for the final test result. The comparison threshold determines the acceptable range of parameter deviation, the weighting coefficients reflect the importance of different process zones in the overall quality, and the judgment threshold is used to comprehensively evaluate the overall quality status. The significance of the above design is that by comparing the characteristic parameters after associating process parameters with standard parameters, the quality status of each process zone can be obtained, providing a quantitative basis for the objective evaluation of the process quality of cable accessories, and ensuring that the test results can directly reflect the compliance of process execution.

[0067] In practice, the standard ranges of characteristic parameters for each process zone are retrieved from the process standard database. For example, the standard value for the local density parameter in the insulation layer area is 8-12 mm, with a standard deviation of no more than 1.5 mm; the standard value for the Gaussian curvature in the shielding layer area is -0.05 to 0.05. A comparison threshold is set at 10% of the difference between the upper and lower limits of the standard range. When a characteristic parameter exceeds the standard range and the deviation is greater than the comparison threshold, the parameter is marked as unqualified. Weighting coefficients are set according to the importance of each process zone (0.2-0.3 for critical zones, 0.05-0.15 for general zones), and the weighted percentage of unqualified parameters in each zone is calculated. When the weighted percentage of any critical zone exceeds 15%, or the total weighted percentage of all zones exceeds 10%, the cable accessory's process quality is deemed unqualified; otherwise, it is deemed compliant. All calculation values ​​are rounded to two decimal places to ensure the accuracy of the results.

[0068] Preferably, in step S1, when using the RGB three-color separation edge detection algorithm, an edge contour feature extraction model is constructed, the expression of which is:

[0069]

[0070] in, E i,j For the image ( i , j The edge intensity value of a pixel; k This is the index for the RGB three-color channels, with values ​​1, 2, and 3 corresponding to the R, G, and B channels, respectively. oh k The edge detection weight coefficient for the k-th channel; R k (i , j ), G k ( i , j ), B k ( i , j ) are respectively the kth channel in ( i , j The pixel value of a pixel; α , β , c These are the reflectance parameters of the cable accessory insulation material for red, green, and blue light, respectively. For gradient operators; i i, j for( i , j The surface texture orientation angle parameter of the cable accessory at the pixel point; in step S2, when constructing the initial three-dimensional point cloud data set, the two-dimensional image coordinates of the edge contour feature points are converted into three-dimensional spatial coordinates, and the radial dimension parameter of the cable accessory process is introduced during the conversion process. D and axial length parameters L The spatial scale of the point cloud data is consistent with the actual dimensions of the cable accessories.

[0071] Specifically, by combining an RGB three-color separation edge detection algorithm with 3D point cloud modeling, the accuracy control of the analysis of edge features of cable accessories and the construction of point clouds is further expanded. The technical parameters involved include edge detection weight coefficients for different channels, color reflectance coefficients, surface texture direction angles, radial dimension parameters, and axial length parameters. The setting of these parameters directly affects the accuracy of edge intensity calculation and the matching degree between the point cloud spatial scale and the actual size. The significance of the above design lies in improving the completeness of edge contour extraction and the realism of the point cloud model through multi-dimensional parameter integration, providing more reliable basic data for subsequent detection. In specific implementation, the reflectance coefficients of the R, G, and B channels are first determined based on the characteristics of the cable accessory insulation material. Typically, the R channel is 0.6-0.8, the G channel is 0.5-0.7, and the B channel is 0.4-0.6. Based on the direction of the surface texture, the texture direction angle is set to a range of 0-360 degrees with a step size of 1 degree. When calculating edge strength values, weighting coefficients are assigned to the R channel (0.35-0.45), G channel (0.3-0.4), and B channel (0.2-0.3) based on the contribution of each channel to edge features. When constructing the 3D point cloud, the radial dimensions of the cable accessories are measured with an accuracy of ±0.1mm, and the axial length is measured with an accuracy of ±0.5mm. These parameters are used to correct the 3D coordinates of the feature points, ensuring that the radial and axial scale errors of the point cloud data are both less than 0.2mm, thus enabling the point cloud model to more accurately reflect the actual shape of the cable accessories.

[0072] Preferably, in step S3, when performing spatial coordinate transformation on the initial 3D point cloud data set, a coordinate transformation model is established, the expression of which is:

[0073]

[0074] in,( X , Y , Z () represents the original coordinates of the feature points in the initial 3D point cloud data; The transformed coordinates match the reference coordinate system. M Let be the coordinate transformation matrix, and:

[0075]

[0076] in, i This is a rotation angle parameter about the X-axis, which is related to the bending angle process parameters of the cable accessories; The rotation angle parameter around the Y-axis is associated with the torsion process parameters of the cable accessories; a , b , cThe translation parameters in the X, Y, and Z axes are determined by the installation and positioning process parameters of the cable accessories; in step S4, when performing region division on the matched 3D point cloud data, the insulation layer layering parameters of the cable accessory process are used. h n , and the overlap width parameter of the shielding layer W s , where n is the number of insulation layers, the sub-point cloud dataset is divided according to the insulation layer interface and the shielding layer boundary.

[0077] Specifically, the accuracy of the bending angle, torsion degree, installation positioning, and structural layering of the accessories directly affects the matching accuracy between the point cloud and the reference coordinate system, as well as the rationality of the region division. The significance of the above design lies in integrating process parameters into the coordinate transformation and region division process, making the spatial position and partitioning of the point cloud data more closely match the actual process design, laying the foundation for accurate detection of each process partition. In specific implementation, based on the bending process requirements of the cable accessories, the adjustment range of the rotation angle around the X-axis is determined to be -5 degrees to 5 degrees, with a step size of 0.1 degrees; based on the torsion process parameters, the rotation angle range around the Y-axis is set to -3 degrees to 3 degrees, with a step size of 0.1 degrees. The translation parameters are set according to the allowable deviation of the installation positioning, with the translation range in the X, Y, and Z axes set to -10mm to 10mm, with an accuracy of 0.01mm. When dividing the region, the thickness of each layer of the insulation layer is measured with an accuracy of ±0.05mm to determine the layering parameters; the overlap width of the shielding layer is measured with an accuracy of ±0.2mm, and the partition boundary is determined based on this to ensure that each sub-point cloud dataset can accurately correspond to each layer of the insulation layer and the overlap area of ​​the shielding layer, and the position error of the partition boundary is controlled within ±0.3mm.

[0078] Preferably, in step S5, when calculating the spatial topology parameters of the sub-point cloud dataset, a spatial neighborhood relationship model is constructed, expressed as:

[0079]

[0080] in, T p For point p Spatial topological correlation degree; N p For point p The number of neighboring points; N ( p ) is a point p The set of neighborhood points; X p , Y p , Z p ), ( X q ,Y q , Z q ) are points p and its neighboring points q 3D coordinates; d std Standard distance between adjacent feature points specified for cable accessory manufacturing process; l The elastic deformation coefficient parameter of the cable accessory material; when calculating the surface curvature distribution characteristics, the allowable deviation parameter of the curvature radius of the cable accessory manufacturing process is also considered. The curvature values ​​of each sub-point cloud dataset are divided into intervals, and the proportion of point clouds in different curvature intervals is statistically analyzed.

[0081] Specifically, the calculation of spatial topology parameters and curvature distribution characteristics of the sub-point cloud dataset involves technical parameters such as the number of neighboring points, standard distance, elastic deformation coefficient, and allowable deviation of curvature radius. The appropriate selection of these parameters is crucial to the accuracy of spatial topology correlation calculation and the scientific validity of curvature interval division. The significance of the above design lies in integrating material properties and process allowable deviations into the characteristic parameter calculation, enabling the calculation results to more realistically reflect the process quality of cable accessories. In practical implementation, based on the structural complexity of the cable accessories, the number of neighboring points for each point is determined to be 15-25; based on the distance requirements between adjacent feature points in the process design, a standard distance is set with an accuracy of ±0.05mm; and based on the elastic performance test data of the material, the elastic deformation coefficient is determined to be 0.001-0.003. When calculating spatial topology correlation, these parameters are used for quantitative analysis to ensure that the error in the correlation calculation is less than 5%. At the same time, the radius of curvature of each part of the cable accessories is measured to determine the allowable deviation range of ±0.5mm. Based on this, the curvature intervals are divided, and the proportion of point cloud in each interval is statistically analyzed. The statistical error is controlled within ±2%, so that the curvature distribution characteristics can accurately reflect the process compliance of the surface morphology.

[0082] Preferably, in step S1, the RGB three-color separation edge detection algorithm further includes quantitative analysis of color differences in different process areas on the surface of the cable accessory, and constructing a color difference edge detection model, the expression of which is:

[0083]

[0084] in, E color The color edge intensity between different process areas; k , For indexing different process areas; For the region k With the region Color difference weighting coefficient; Rk , G k , B k They are respectively regions k The average pixel values ​​of the R, G, and B channels; They are respectively regions The average pixel values ​​of the R, G, and B channels; The R, G, and B channels are located in the region. k and The color attenuation coefficient is related to the surface coating process parameters of the cable accessories; in step S2, when constructing the initial three-dimensional point cloud data set, the wall thickness parameters of the cable accessories are introduced. t The thickness direction is corrected for the three-dimensional coordinates of the edge contour feature points.

[0085] Specifically, the quantitative analysis of color differences and the thickness correction of point cloud coordinates involve technical parameters such as color difference weighting coefficients, color attenuation coefficients, and wall thickness parameters. These parameters are related to the surface coating process and structural thickness requirements of cable accessories, directly affecting the accuracy of color edge intensity calculation and the effectiveness of point cloud coordinate correction. The significance of the above design lies in improving the targeting of edge detection and the accuracy of the point cloud model by quantifying the color differences in different process areas and considering the influence of wall thickness on coordinates. In specific implementation, the color difference weighting coefficient between regions is determined according to the characteristics of the coating materials in different process areas, with a value range of 0.6-0.9; based on the thickness and uniformity of the coating layer, the color attenuation coefficients for the R, G, and B channels are set: R channel 0.02-0.04, G channel 0.015-0.035, and B channel 0.01-0.03. When calculating the color edge intensity, these parameters are substituted into the calculation to ensure that the intensity value accurately reflects the color transition characteristics between regions. When correcting the point cloud coordinates, the wall thickness of the cable accessory is measured with an accuracy of ±0.05mm. Based on the wall thickness value, the three-dimensional coordinates of the feature points are adjusted in the thickness direction. The calculation error of the correction amount is less than 0.1mm, so that the point cloud data can more accurately reflect the actual thickness of the cable accessory.

[0086] Preferably, in step S4, when performing region partitioning on the matched 3D point cloud data, a region partitioning model based on process parameters is established, with the expression:

[0087]

[0088] in, S m For the first m Sub-point cloud datasets for each process partition; P Points in a 3D point cloud; x 0m , y0m , z 0m ) is the first m The center coordinates of each process zone; r m For the first m The reference radius of each process zone; Δ r m For the first m Permissible radius deviation parameters for each process zone; t m For the first m The reference wall thickness for each process zone; Δ t m For the first m Permissible wall thickness deviation parameters for each process zone; t P For point P The actual wall thickness of the cable accessory at the location; in step S5, when relating the material shrinkage rate of the cable accessory process, the correlation parameter between the material shrinkage rate and temperature change is used. Temperature compensation correction is applied to the spatial topology parameters.

[0089] Specifically, based on the application of the region division model and material shrinkage rate correlation based on process parameters, technical parameters such as zone center coordinates, reference radius, radius allowable deviation, reference wall thickness, wall thickness allowable deviation, and temperature compensation coefficient are introduced. These parameters comprehensively reflect the structural characteristics of each process zone and the temperature sensitivity of the material, significantly affecting the accuracy of region division and the temperature adaptability of characteristic parameters. The significance of the above design lies in making the region division results and characteristic parameters more consistent with actual process conditions by constructing a more refined region division model and considering the influence of temperature on material shrinkage. In specific implementation, high-precision measuring instruments are used to determine the center coordinates of each process zone with an accuracy of ±0.1mm; the reference radius and reference wall thickness are obtained according to the design drawings with accuracies of ±0.05mm and ±0.02mm, respectively; based on the process allowable error, the radius allowable deviation is set to ±0.2mm, and the wall thickness allowable deviation is set to ±0.1mm. When dividing the region, these parameters are used to construct the model to ensure that the boundary error of each sub-point cloud dataset is less than 0.3mm. Meanwhile, the temperature compensation coefficient is obtained through material testing, with a value of 0.0005-0.0015 / ℃. Based on the difference between the ambient temperature and the standard temperature, the spatial topology parameters are corrected, and the error of the corrected parameters is controlled within ±3%.

[0090] Preferably, in step S1, when extracting the edge contour information of the calibration area on the surface of the cable accessory, an edge contour continuity detection model is constructed, the expression of which is:

[0091]

[0092] in, C It is the continuity index of the edge contour; n The number of feature points on the edge contour; d i,i+1 For the first i The and the first i +1 distance between edge feature points; d 0 represents the baseline distance parameter between adjacent edge feature points; s d The standard deviation of the edge point spacing is related to the forming process parameters of the cable accessories; i i,i+1 For the first i The and the first i +1 direction angle of the line connecting edge feature points; i 0 is the reference direction angle of the edge profile; in step S3, when performing spatial coordinate transformation, the origin and coordinate axis direction of the reference coordinate system are determined by using the flange positioning hole parameters of the cable accessory.

[0093] Specifically, the benchmark features for edge contour continuity detection and coordinate matching involve technical parameters such as the distance between adjacent edge points, benchmark distance, distance standard deviation, orientation angle deviation, mold precision parameters, and positioning hole position parameters. These parameters are closely related to the molding process precision of cable accessories and the accuracy of the positioning benchmark, directly affecting the reliability of edge continuity judgment and the accuracy of coordinate matching. The significance of the above design lies in improving the integrity of edge detection and the accuracy of point cloud coordinate matching by quantifying the continuity of the edge contour and utilizing precise positioning benchmarks. In specific implementation, the distance between adjacent feature points on the edge is measured with an accuracy of ±0.05mm, and the benchmark distance and distance standard deviation are determined, with the standard deviation generally being 0.05-0.15mm; based on the manufacturing precision of the mold, the standard deviation of the orientation angle deviation is set to 0.5-1 degree. When calculating the continuity index, these parameters are substituted to ensure that the index accurately reflects the continuity state of the edge, with a calculation error of less than 0.05. At the same time, the center position of the flange positioning hole is measured with an accuracy of ±0.02mm. This is used as the origin of the reference coordinate system and the reference for the coordinate axis direction, so that the matching error between the point cloud data and the reference coordinate system is controlled within ±0.05mm.

[0094] Preferably, in step S5, when calculating the surface curvature distribution characteristics of the sub-point cloud dataset, a correlation model between curvature and process parameters is established, expressed as:

[0095]

[0096] in, K p For point p The average curvature of the region; Ap For point p The area where it is located; R q For point q The radius of curvature; or This is the temperature influence coefficient, which is related to the coefficient of thermal expansion of the cable accessory materials; Δ T This is the difference between the actual measured temperature and the standard temperature. T 0 represents the standard temperature parameter; This is the pressure influence coefficient, which is related to the installation pressure parameters of the cable accessories; Δ P This is the difference between the actual installation pressure and the standard installation pressure. P 0 represents the standard installation pressure parameter; in step S6, when comparing the parameters of each sub-point cloud dataset with the standard parameters, the insulation resistance parameter of the cable accessory process is used. r The contrast threshold for the surface curvature distribution characteristics is adjusted.

[0097] Specifically, the curvature calculation and parameter comparison threshold adjustment for the sub-point cloud dataset incorporates technical parameters such as temperature influence coefficient, standard temperature, pressure influence coefficient, standard pressure, and insulation resistance parameters. These parameters are related to the thermal expansion characteristics of the material, installation pressure requirements, and insulation performance indicators, respectively, affecting the accuracy of curvature calculation and the rationality of the comparison threshold. The significance of this design lies in making the curvature feature detection more consistent with actual working conditions and improving the accuracy of quality judgment by considering the influence of temperature and pressure on curvature and adjusting the threshold using insulation resistance parameters. In specific implementation, the temperature influence coefficient is determined to be 0.001-0.003 / ℃ through material testing, and the standard temperature is set to 25℃±2℃; according to the installation process requirements, the standard pressure and pressure influence coefficient are set, with the pressure influence coefficient generally being 0.002-0.004 / MPa. When calculating the average curvature, corrections are made based on the difference between the actual temperature and pressure of the detection environment and the standard values, and the corrected curvature error is less than 0.01. Simultaneously, the insulation resistance of cable accessories is measured, and the comparison threshold of surface curvature distribution characteristics is adjusted according to the resistance value. The larger the resistance value, the threshold can be relaxed by 5%-10% to ensure that the threshold can meet the quality judgment requirements under different insulation performance.

[0098] Preferably, in step S2, when constructing the initial three-dimensional point cloud data set of the cable accessory surface, a correlation model between point cloud density and process parameters is established, expressed as:

[0099]

[0100] in, r cloud ( x , y ,z ) is in three-dimensional space x , y , z Point cloud density at location ) r 0 is the baseline point cloud density; α The surface roughness influence coefficient of cable accessories; s ( x , y , z )for( x , y , z Surface roughness parameters at location ) β The coefficient representing the influence of the structural complexity of cable accessories; c ( x , y , z )for( x , y , z The structural complexity parameter at the location is related to the number of process corners at that location; in step S3, when matching point cloud data with the reference coordinate system, the position parameter of the positioning pin hole of the cable accessory is used. x p , y p , z p (), which serves as the reference feature point for coordinate matching.

[0101] Specifically, the selection of reference feature points for density control and coordinate matching of 3D point clouds involves technical parameters such as surface roughness influence coefficient, structural complexity influence coefficient, surface roughness parameters, structural complexity parameters, and positioning pin hole position parameters. These parameters are related to the surface processing quality, structural complexity, and positioning accuracy requirements of cable accessories, directly affecting the ability of point cloud data to express surface details and the accuracy of coordinate matching. The significance of the above design lies in improving the detail representation and coordinate matching accuracy of the point cloud model by constructing a correlation model between point cloud density and process parameters and using precise positioning pin holes as a reference. In specific implementation, the surface roughness of the cable accessories is measured, generally Ra 0.8-3.2μm, and the surface roughness influence coefficient is determined to be 0.1-0.3; according to the number of structural corners, the structural complexity parameter is set, with the parameter increasing by 0.1-0.2 for each additional corner, resulting in a structural complexity influence coefficient of 0.2-0.4. When constructing the point cloud, the point cloud density at different locations is adjusted based on these parameters to ensure that the point cloud density in areas with high surface roughness and complex structures is 30%-50% higher than the reference density. At the same time, the position parameters of the positioning pin holes are measured with an accuracy of ±0.01mm, which are used as the reference feature points for coordinate matching, so that the matching error between the point cloud data and the reference coordinate system is controlled within ±0.03mm.

[0102] like Figure 2 As shown, the machine vision-based cable accessory process quality inspection system includes:

[0103] The edge contour extraction and parsing unit is used to acquire RGB three-color image information of the cable accessory surface and extract edge contour information through the RGB three-color separation edge detection algorithm. Its output end is connected to the input end of the data set construction and coordinate parameter integration unit.

[0104] The data set construction and coordinate parameter integration unit is used to construct an initial three-dimensional point cloud data set including the three-dimensional spatial coordinate parameters of the edge contour feature points based on the edge contour information. Its output end is connected to the input end of the data space coordinate transformation and reference coordinate system matching unit.

[0105] The data spatial coordinate transformation and reference coordinate system matching unit is used to transform the initial three-dimensional point cloud data set into spatial coordinates and match it with the reference coordinate system of cable accessory process design. Its output end is connected to the input end of the data area division and sub-point cloud dataset generation unit.

[0106] The data region division and sub-point cloud dataset generation unit is used to perform region division on the matched three-dimensional point cloud data in combination with the preset geometric parameters of the cable accessory process and obtain the sub-point cloud dataset corresponding to each process partition. Its output end is connected to the input end of the distribution feature calculation and process parameter association unit.

[0107] The distribution feature calculation and process parameter association unit is used to calculate the spatial topology parameters and surface curvature distribution characteristics of each sub-point cloud dataset and associate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process. Its output end is connected to the input end of the cable accessory process standard parameter comparison and analysis unit.

[0108] The cable accessory process standard parameter comparison and analysis unit is used to receive the spatial topology parameters and surface curvature distribution characteristics output by the distribution characteristic calculation and process parameter correlation unit, compare them with the cable accessory process standard parameters, and output the test results.

[0109] This machine vision-based method for inspecting the manufacturing quality of cable accessories utilizes an RGB three-color separation edge detection algorithm to perform multi-dimensional analysis of the grayscale value distribution characteristics of different color gamut channels in cable accessory images, extracting edge contour information of key surface areas. This approach overcomes the limitations of traditional single-channel image analysis, enabling a more comprehensive capture of the morphological features of key areas such as insulation layer overlaps and shielding layer transition zones. It compensates for the insufficient integrity of contour information in traditional edge detection methods, making the extracted edge contours more closely match the actual manufacturing process of cable accessories.

[0110] This method constructs an initial dataset through 3D point cloud modeling, precisely matches it with the process design reference coordinate system through spatial coordinate transformation, and divides the region and calculates characteristic parameters in conjunction with process parameters. This process closely correlates the 3D point cloud data with process parameters such as the material shrinkage rate and insulation layer thickness of cable accessories, solving the problem of low matching between test data and process requirements caused by the disconnect between traditional 3D modeling and process parameters. This allows the test results to directly reflect the compliance of process indicators, enhancing the practical value of test data for process quality assessment.

[0111] In the corresponding inspection system, six units, including the edge contour extraction and analysis unit, the data set construction and coordinate parameter integration unit, are connected sequentially to form a coherent inspection process. These units work collaboratively, from image acquisition and point cloud construction to parameter comparison and output, achieving a complete transformation from visual information to process quality assessment. This overcomes the problem of poor coordination between different stages in traditional inspection, ensuring the systematic and efficient nature of the inspection process and fully meeting the technical requirements for the process quality inspection of cable accessories.

[0112] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "set," "install," "connect," "link," and "fix" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal communication between two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0113] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various equivalent changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A machine vision-based method for inspecting the manufacturing quality of cable accessories, characterized in that, include: Step S1: Collect RGB three-color image information of the cable accessory surface, and use the RGB three-color separation edge detection algorithm to perform multi-dimensional analysis on the gray value distribution characteristics of different color gamut channels in the image to extract the edge contour information of the calibrated area on the cable accessory surface; Step S2: Based on the extracted edge contour information, construct an initial three-dimensional point cloud data set for the surface of the cable accessory, which includes the coordinate parameters of the edge contour feature points in three-dimensional space; Step S3: Perform spatial coordinate transformation on the initial 3D point cloud data set to match the point cloud data with the reference coordinate system of the cable accessory process design; Step S4: Combine the preset geometric parameters of the cable accessory process, perform region division on the matched 3D point cloud data to obtain sub-point cloud datasets corresponding to each process partition of the cable accessory. Step S5: For each sub-point cloud dataset, calculate the spatial topology parameters and surface curvature distribution characteristics, and correlate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process; When calculating the spatial topology parameters of the sub-point cloud dataset, a spatial neighborhood relationship model is constructed, expressed as: in, T p For point p Spatial topological correlation degree; N p For point p The number of neighboring points; N ( p ) is a point p The set of neighborhood points; X p , Y p , Z p ), ( X q , Y q , Z q ) are points p and its neighboring points q 3D coordinates; d std Standard distance between adjacent feature points specified for cable accessory manufacturing process; λ The elastic deformation coefficient parameter of the cable accessory material; when calculating the surface curvature distribution characteristics, the allowable deviation parameter of the curvature radius of the cable accessory manufacturing process is also considered. The curvature values ​​of each sub-point cloud dataset are divided into intervals, and the proportion of point clouds in different curvature intervals is counted. When calculating the surface curvature distribution characteristics of the sub-point cloud dataset, a correlation model between curvature and process parameters is established, expressed as: in, K p For point p The average curvature of the region; A p For point p The area where it is located; R q For point q The radius of curvature; η This is the temperature influence coefficient, which is related to the coefficient of thermal expansion of the cable accessory materials; Δ T This is the difference between the actual measured temperature and the standard temperature. T 0 represents the standard temperature parameter; This is the pressure influence coefficient, which is related to the installation pressure parameters of the cable accessories; Δ P This is the difference between the actual installation pressure and the standard installation pressure. P 0 represents the standard installation pressure parameter; Step S6: Compare the spatial topology parameters and surface curvature distribution characteristics of each sub-point cloud dataset with the standard parameters of cable accessory manufacturing processes, and output the detection results; among them, when comparing the parameters of each sub-point cloud dataset with the standard parameters, the insulation resistance parameters of the cable accessory manufacturing process are used. ρ The contrast threshold for the surface curvature distribution characteristics is adjusted.

2. The method according to claim 1, characterized in that, In step S1, when using the RGB three-color separation edge detection algorithm, an edge contour feature extraction model is constructed, the expression of which is: in, E i,j For the image ( i , j The edge intensity value of a pixel; k This is the index for the RGB three-color channels, with values ​​1, 2, and 3 corresponding to the R, G, and B channels, respectively. ω k The edge detection weight coefficient for the k-th channel; R k ( i , j ), G k ( i , j ), B k ( i , j ) are respectively the kth channel in ( i , j The pixel value of a pixel; α , β , γ These are the reflectance parameters of the cable accessory insulation material for red, green, and blue light, respectively. For gradient operators; θ i, j for( i , j The surface texture orientation angle parameter of the cable accessory at the pixel point; in step S2, when constructing the initial three-dimensional point cloud data set, the two-dimensional image coordinates of the edge contour feature points are converted into three-dimensional spatial coordinates, and the radial dimension parameter of the cable accessory process is introduced during the conversion process. D and axial length parameters L The spatial scale of the point cloud data is consistent with the actual dimensions of the cable accessories.

3. The method according to claim 1, characterized in that, In step S3, when performing spatial coordinate transformation on the initial 3D point cloud data set, a coordinate transformation model is established, and its expression is: in,( X , Y , Z () represents the original coordinates of the feature points in the initial 3D point cloud data; The transformed coordinates match the reference coordinate system. M Let be the coordinate transformation matrix, and: in, θ This is a rotation angle parameter about the X-axis, which is related to the bending angle process parameters of the cable accessories; The rotation angle parameter around the Y-axis is associated with the torsion process parameters of the cable accessories; a , b , c The translation parameters in the X, Y, and Z axes are determined by the installation and positioning process parameters of the cable accessories; in step S4, when performing region division on the matched 3D point cloud data, the insulation layer layering parameters of the cable accessory process are used. h n , and the overlap width parameter of the shielding layer W s , where n is the number of insulation layers, the sub-point cloud dataset is divided according to the insulation layer interface and the shielding layer boundary.

4. The method according to claim 1, characterized in that, In step S1, the RGB three-color separation edge detection algorithm further includes quantitative analysis of color differences in different process areas on the surface of the cable accessory, and constructing a color difference edge detection model, the expression of which is: in, E color The color edge intensity between different process areas; k , For indexing different process areas; For the region k With the region Color difference weighting coefficient; R k , G k , B k They are respectively regions k The average pixel values ​​of the R, G, and B channels; They are respectively regions The average pixel values ​​of the R, G, and B channels; The R, G, and B channels are located in the region. k and The color attenuation coefficient is related to the surface coating process parameters of the cable accessories; in step S2, when constructing the initial three-dimensional point cloud data set, the wall thickness parameters of the cable accessories are introduced. t The thickness direction is corrected for the three-dimensional coordinates of the edge contour feature points.

5. The method according to claim 1, characterized in that, In step S4, when performing region partitioning on the matched 3D point cloud data, a region partitioning model based on process parameters is established, expressed as: in, S m For the first m Sub-point cloud datasets for each process partition; P Points in a 3D point cloud; x 0m , y 0m , z 0m ) is the first m The center coordinates of each process zone; r m For the first m The reference radius of each process zone; Δ r m For the first m Permissible radius deviation parameters for each process zone; t m For the first m The reference wall thickness for each process zone; Δ t m For the first m Permissible wall thickness deviation parameters for each process zone; t P For point P The actual wall thickness of the cable accessory at the location; in step S5, when relating the material shrinkage rate of the cable accessory process, the correlation parameter between the material shrinkage rate and temperature change is used. Temperature compensation correction is applied to the spatial topology parameters.

6. The method according to claim 1, characterized in that, In step S1, when extracting the edge contour information of the calibration area on the surface of the cable accessory, an edge contour continuity detection model is constructed, and its expression is: in, C It is the continuity index of the edge contour; n The number of feature points on the edge contour; d i,i+1 For the first i The and the first i +1 distance between edge feature points; d 0 represents the baseline distance parameter between adjacent edge feature points; σ d The standard deviation of the edge point spacing is related to the forming process parameters of the cable accessories; θ i,i+1 For the first i The and the first i +1 direction angle of the line connecting edge feature points; θ 0 is the reference direction angle of the edge profile; in step S3, when performing spatial coordinate transformation, the origin and coordinate axis direction of the reference coordinate system are determined by using the flange positioning hole parameters of the cable accessory.

7. The method according to claim 1, characterized in that, In step S2, when constructing the initial three-dimensional point cloud data set of the cable accessory surface, a correlation model between point cloud density and process parameters is established, expressed as: in, ρ cloud ( x , y , z ) is in three-dimensional space x , y , z Point cloud density at location ) ρ 0 is the baseline point cloud density; α The surface roughness influence coefficient of cable accessories; s ( x , y , z )for( x , y , z Surface roughness parameters at location ) β The coefficient representing the influence of the structural complexity of cable accessories; c ( x , y , z )for( x , y , z The structural complexity parameter at the location is related to the number of process corners at that location; in step S3, when matching point cloud data with the reference coordinate system, the position parameter of the positioning pin hole of the cable accessory is used. x p , y p , z p (), which serves as the reference feature point for coordinate matching.

8. A machine vision-based process quality inspection system for cable accessories, characterized in that, include: The edge contour extraction and parsing unit is used to acquire RGB three-color image information of the cable accessory surface and extract edge contour information through the RGB three-color separation edge detection algorithm. The output end is connected to the input end of the data set construction and coordinate parameter integration unit. The data set construction and coordinate parameter integration unit is used to construct an initial three-dimensional point cloud data set including the three-dimensional spatial coordinate parameters of the edge contour feature points based on the edge contour information. The output end is connected to the input end of the data space coordinate transformation and reference coordinate system matching unit. The data spatial coordinate transformation and reference coordinate system matching unit is used to perform spatial coordinate transformation on the initial three-dimensional point cloud data set and match it with the reference coordinate system of cable accessory process design. The output end is connected to the input end of the data area division and sub-point cloud dataset generation unit. The data region division and sub-point cloud dataset generation unit is used to perform region division on the matched three-dimensional point cloud data in combination with the preset geometric parameters of the cable accessory process and obtain the sub-point cloud dataset corresponding to each process partition. The output end is connected to the input end of the distribution feature calculation and process parameter association unit. The distribution feature calculation and process parameter association unit is used to calculate the spatial topology parameters and surface curvature distribution characteristics of each sub-point cloud dataset and associate them with the material shrinkage rate and insulation layer thickness parameters of the cable accessory process. The output end is connected to the input end of the cable accessory process standard parameter comparison and analysis unit. When calculating the spatial topology parameters of the sub-point cloud dataset, a spatial neighborhood relationship model is constructed, expressed as: in, T p For point p Spatial topological correlation degree; N p For point p The number of neighboring points; N ( p ) is a point p The set of neighborhood points; X p , Y p , Z p ), ( X q , Y q , Z q ) are points p and its neighboring points q 3D coordinates; d std Standard distance between adjacent feature points specified for cable accessory manufacturing process; λ The elastic deformation coefficient parameter of the cable accessory material; when calculating the surface curvature distribution characteristics, the allowable deviation parameter of the curvature radius of the cable accessory manufacturing process is also considered. The curvature values ​​of each sub-point cloud dataset are divided into intervals, and the proportion of point clouds in different curvature intervals is counted. When calculating the surface curvature distribution characteristics of the sub-point cloud dataset, a correlation model between curvature and process parameters is established, expressed as: in, K p For point p The average curvature of the region; A p For point p The area where it is located; R q For point q The radius of curvature; η This is the temperature influence coefficient, which is related to the coefficient of thermal expansion of the cable accessory materials; Δ T This is the difference between the actual measured temperature and the standard temperature. T 0 represents the standard temperature parameter; This is the pressure influence coefficient, which is related to the installation pressure parameters of the cable accessories; Δ P This is the difference between the actual installation pressure and the standard installation pressure. P 0 represents the standard installation pressure parameter; The cable accessory process standard parameter comparison and analysis unit receives the spatial topology parameters and surface curvature distribution characteristics output by the distribution characteristic calculation and process parameter correlation unit. It then compares these parameters with the cable accessory process standard parameters and outputs the test results. Specifically, when comparing the parameters of each sub-point cloud dataset with the standard parameters, the insulation resistance parameter of the cable accessory process is used. ρ The contrast threshold for the surface curvature distribution characteristics is adjusted.

Images