A visual measurement method for punch position deviation of an electric cross arm

Abstract

This invention belongs to the field of industrial vision technology, specifically relating to a visual measurement method for the position deviation of punched holes in power crossarms. The method includes: generating a grayscale image of the punched hole; obtaining a set of point points on the punched hole edge; obtaining the edge curvature variation rate; generating a set of effective arc point points; calculating the actual center of the punched hole; obtaining the position deviation of the punched hole; outputting the crossarm defect type; generating sorting control instructions; and outputting crossarm sorting data. This invention solves the technical problem of existing technologies that cannot distinguish between the smooth, true arc of the punched hole edge and the abrupt change in physical burrs caused by stamping, resulting in interference terms in the extracted boundary data, leading to inaccurate geometric center fitting, distorted position deviation measurement, and inability to guide the automatic rejection of defective products on the production line.

Description

Technical Field

[0001] This invention relates to the field of industrial vision technology. More specifically, this invention relates to a visual measurement method for the positional deviation of punched holes in an electric crossarm. Background Technology

[0002] With the continuous advancement of power infrastructure construction, the demand for crossarm components of power transmission towers, used to support conductors and insulators in transmission lines, is increasing daily. Power crossarms are typically manufactured from angle steel or channel steel through machining processes such as cutting and punching. During the assembly line processing of crossarms, the positional accuracy of the punched holes directly determines the efficiency and structural safety of subsequent tower assembly. Therefore, quickly and accurately measuring the punching position deviation is a crucial step in controlling product quality and reducing the defect rate.

[0003] In related technologies, for example, Chinese patent document CN209559076U discloses a special ruler for measuring the dimensions of crossarms of power transmission towers. This Chinese patent document uses a specially designed mechanical scale ruler in conjunction with manual measurement to obtain dimensional data such as the hole spacing of the crossarm. This processing method can achieve basic physical dimension extraction and preliminary processing accuracy assessment during offline sampling inspection or small-batch production.

[0004] However, analysis of the performance of existing technologies under actual high-speed production line electric crossarm stamping quality inspection conditions reveals that this method, which relies on traditional physical contact testing or conventional basic machine vision for rough fitting, has a flaw in contact-based geometric center alignment calculation that ignores the interference of edge burr deformation. First, the physical micromorphology of the hole edges in stamped parts has spatial irregularities: at the moment of stamping and cutting, the steel plate is easily subjected to shear force, resulting in burr protrusions or collapsed corners at the hole edges, exhibiting an uneven, rough cross-section. Existing technologies, when extracting the hole contour, often uniformly extract all edge pixels and perform global circle equation fitting. The algorithm cannot distinguish between smooth, real arcs and abnormal physical burrs, causing the model to assign higher weights to burr points when calculating the geometric center, thus deviating the center coordinates. Second, due to the complex lighting conditions in industrial environments and the surface reflection of metal materials, traditional inspection methods fail to effectively filter out false edge effects caused by local highlights. When the algorithm evaluates the punching coordinates, it fails to eliminate the false edge contours caused by irregular deformation and reflection, resulting in erroneous measurements of the center position. This measurement model fails to reflect the physical reality of rough edges on holes in industrial stamped parts, resulting in the system being unable to accurately calculate the actual positional deviation of punched holes in power crossarms, and lacking a sorting process mechanism for fully automated rejection on high-speed dynamic production lines. Summary of the Invention

[0005] To address the technical problem that existing technologies cannot distinguish between the smooth, true arc of the punched edge and the abrupt physical burrs caused by stamping during calculation, resulting in interference terms in the extracted boundary data, inaccurate geometric center fitting, distorted position deviation measurement, and inability to guide the automatic rejection of defective products on the production line, this invention provides a visual measurement method for the position deviation of punched holes in power crossarms, comprising:

[0006] S1: Generate a punched grayscale image by executing an industrial camera image acquisition and target area cropping method; S2: Based on the pixel gradient of the punched grayscale image, execute a quadratic curve interpolation algorithm to obtain the punched edge point set; obtain the edge curvature variation rate according to the local coordinate tangential distribution characteristics of the punched edge point set; generate an effective arc point set by performing adaptive threshold condition filtering on the edge curvature variation rate; S3: Based on the effective arc point set, perform the least squares circular algebra equation partial derivative operation to calculate the actual punched center; analyze the standard process drawing to obtain the theoretical punched position; based on the actual punched center and the pre-obtained theoretical punched position, perform vector space difference calculation to obtain the punched position deviation; based on the punched position deviation, perform multi-level tolerance interval numerical comparison classification to output the crossarm defect type; S4: Based on the crossarm defect type, establish a timing tracking queue by collecting external transmission pulse data and perform message mapping to generate sorting control instructions; execute sorting actions based on the sorting control instructions and output crossarm sorting data.

[0007] This invention addresses the shortcomings of existing methods for detecting punched holes in power crossarms. These methods rely on manual caliper checks or rough visual fitting, failing to distinguish between true arcs and stamping burr deformations. This leads to center-of-circuit fitting errors, distorted positional measurements, and a lack of automated sorting throughout the entire process in high-speed production lines. Sub-pixel-level edge extraction improves contour positioning accuracy, overcoming hardware resolution limitations. Adaptive curvature rate filtering effectively eliminates interference from stamping deformations such as burrs and collapsed corners, purifying the fitted samples from the data source. Least-squares-based pure arc fitting ensures accurate center-of-circuit positioning. Multi-level tolerance grading standardizes quality assessment, avoiding subjective errors from manual judgment. Combined with synchronous sorting control along the production line sequence, it achieves full automation from detection to rejection. This method improves the measurement accuracy and detection efficiency of punched hole positional deviations, meeting the full-volume quality inspection requirements of high-speed production lines.

[0008] Preferably, generating a punched grayscale image includes: An industrial camera was used to perform continuous exposure acquisition on the physical crossarm of the power line on the production line to obtain a two-dimensional pixel matrix containing holes and the surrounding metal texture as the initial crossarm image. A connected component labeling algorithm based on global grayscale threshold was used to scan the initial crossarm image, identify and remove background clutter features, and extract the target holes and their boundary neighborhoods as punched grayscale images.

[0009] Preferably, obtaining the set of punched edge points includes: The first-order partial derivatives of the grayscale values ​​of each pixel are calculated by applying the difference operator along the horizontal and vertical directions of the punched grayscale image, and the image gradient matrix representing the edge transition intensity is synthesized by combining the horizontal and vertical gradient components. Local gradient extrema are located in the image gradient matrix, and a one-dimensional parabolic equation is established along the gradient direction where the extrema are located to perform quadratic curve interpolation calculation, so as to solve for the edge extrema coordinates with floating-point accuracy and generate the punched edge point set.

[0010] Preferably, obtaining the edge curvature variation rate includes: Based on the punching edge point set constituting the punching contour coordinate sequence, a contour sliding window for feature extraction is constructed point by point; within each contour sliding window, the tangential direction change between the lines connecting adjacent two-dimensional coordinate points is calculated using the forward difference rule of calculus to obtain the contour angle difference, which is used as the edge curvature rate at the center point of the window.

[0011] Preferably, generating a valid set of arc points includes: The global arithmetic mean of all edge curvature variations is calculated to generate an edge abrupt change threshold. The edge curvature variation of each window center point is then compared with this threshold. If the edge curvature variation of the current window center point is greater than the threshold, the point is determined to be a physical burr or corner collapse defect, and its edge calculation weight in subsequent solutions is forcibly cleared to discard the data stream. If the edge curvature variation of the current window center point is less than or equal to the threshold, the point is determined to be a smooth real arc segment, the corresponding point is assigned full weight, and its floating-point coordinate data is recorded, recombined, and used to extract and generate a set of valid arc points.

[0012] This invention generates a mutation threshold by calculating the global arithmetic mean of the edge curvature variation rate and establishes an adaptive numerical comparison mechanism, which can accurately determine physical burr defects and smooth real arc segments. By forcibly clearing the weight of the center point of the abnormal window and discarding it, the purity of the final extracted effective arc point set is ensured, and the weight pollution caused by the subsequent deformation interference terms in the solution of the geometric analytical equation is avoided.

[0013] Preferably, calculating the actual punching center includes: Substitute the coordinate data of the effective arc point set into the algebraic circle equation of the two-dimensional plane to establish a system of fitting system equations containing the center coordinates and radius as unknown parameters; use the least squares circle fitting algorithm to construct the residual sum of squares function of all effective data points in the effective arc point set to the boundary of the fitting circle, and take the first-order partial derivatives with respect to the center coordinates and set them equal to zero to establish a system of linear equations, and solve the actual punching center by matrix inversion.

[0014] This invention establishes a system of fitting equations containing the coordinates of the center and the radius as unknown parameters. It uses the least squares algorithm to construct the residual sum of squares function and finds the minimum value of its partial derivative. Substituting the result into the matrix inverse, it abandons the traditional centroid calculation method that is easily affected by incomplete contours. It directly restores the geometric center of the physical hole theory by solving the pure algebraic inner product.

[0015] Preferably, obtaining the punching position deviation includes: The standard process drawing file is parsed from the underlying data communication interface to extract the theoretical punching position, which represents the correct two-dimensional spatial coordinates of the theoretical punching. Based on the laws of plane analytical geometry, the absolute difference between the center of the actual punching circle and the theoretical punching position in the two-dimensional plane coordinate system in the horizontal and vertical axes is calculated and recorded as the absolute difference between the center of the horizontal and vertical axes. The Euclidean distance between the center of the line is solved using the Euclidean formula. The calculated absolute difference between the center of the horizontal and vertical axes and the Euclidean distance between the center of the line are merged, formatted and encapsulated, and output as the punching position deviation for error verification.

[0016] This invention extracts the correct two-dimensional spatial coordinates from standard process drawings and calculates the absolute difference between the actual punching center and the theoretical position in the horizontal and vertical axes, as well as the Euclidean distance between the center of the straight line, based on the laws of plane analytical geometry. This realizes the transformation of isolated measurement coordinates into relative dynamic differences that reflect the mechanical process execution error, avoids screen reference drift, and establishes a reliable reference system for production line tolerance verification.

[0017] Preferably, the output crossarm defect types include: Extract the nominal dimensional parameters of the theoretical punching position to generate an acceptable upper limit and a rework upper limit, with the acceptable upper limit being less than the rework upper limit; take the straight-line distance between the two centers in the punching position deviation as the comprehensive distance parameter; when the comprehensive distance parameter is less than or equal to the acceptable upper limit, the corresponding defect type is acceptable product; when the comprehensive distance parameter is greater than the acceptable upper limit but less than or equal to the rework upper limit, the corresponding defect type is product requiring repair; when the comprehensive distance parameter is greater than the rework upper limit, the corresponding defect type is scrap product.

[0018] This invention utilizes nominal dimensional parameters and process tolerances to establish multi-level tolerance ranges, and compares and maps the punching position deviation with the qualified upper limit and rework upper limit. It can automatically assess the severity of the error and output the identification of qualified products, products that need rework, or scrap products. This not only eliminates the dimensional uncertainty of manual judgment, but also provides accurate data guidance for rework correction processes based on the main offset direction.

[0019] Preferably, generating sorting control instructions includes: The system acquires the encoder pulse equivalent parameters of the production line, collects the transmission pulse data of the production line conveyor belt through sensors, and extracts mechanical delay parameters from the encoder pulse equivalent parameters to establish a time-series tracking queue that maps the visual inspection time to the physical rejection station displacement distance. The system converts the classification status identifier of the crossarm defect type into a low-level communication message that can be recognized by the underlying programmable logic controller and corresponds to the control register address. At the moment when the action execution is performed by matching the mechanical delay parameters during the countdown of the time-series tracking queue, the low-level communication message is sent down through the entire process via the industrial bus to generate sorting control instructions that trigger external hardware actions.

[0020] Preferably, the output cross-carrier sorting data includes: The underlying programmable logic controller parses the received sorting control instructions and drives the underlying pneumatic push rod mechanism on the assembly line through the electromagnetic reversing valve to perform physical rejection or release operations on the current crossarm being tested, either pushing it off the track or allowing it to pass. Simultaneously, it collects the execution feedback signal of the magnetic switch of the cylinder at the end of the stroke of the underlying pneumatic push rod mechanism, associates the execution feedback signal with the crossarm defect type corresponding to this quality inspection process, packages it into a data entity with a timestamp, generates crossarm sorting data, and stores it in the underlying database.

[0021] The beneficial effects of this invention are as follows: it improves the automation and intelligence level of punching quality inspection in the power fittings processing industry, providing a non-contact, high-precision online visual inspection solution for high-speed stamping production lines. Compared with traditional manual sampling inspection, this invention can achieve full-volume online inspection of products, improving inspection efficiency and reducing the labor intensity and human error of quality inspectors; accurate deviation measurement and grading judgment ensure the consistency of product quality before leaving the factory, reducing hole interference problems in subsequent tower assembly; the automatic sorting process and data traceability mechanism realize digital management of production quality, providing data support for stamping process optimization and quality traceability, and helping to promote the intelligent upgrading of the power crossarm processing industry. Attached Figure Description

[0022] Figure 1 A flowchart illustrating a visual measurement method for the positional deviation of punched holes in an electric crossarm according to the present invention is shown schematically. Figure 2 The schematic diagram illustrates the principle of measuring the position deviation of the punching hole in the power crossarm in this invention. Detailed Implementation

[0023] This invention discloses a visual measurement method for the positional deviation of punched holes in a power crossarm, referring to... Figure 1 This includes steps S1-S4: S1: Generate a grayscale image of the punched hole.

[0024] It should be noted that when acquiring punching images on a high-speed punching production line for power crossarms, the system faces technical challenges such as interference from ambient light variations and metal surface reflections, as well as background clutter from the conveyor belt. Considering the low efficiency of traditional physical measuring tools and their susceptibility to distortion caused by manual operation and subjective fatigue, this traditional processing paradigm, when facing large-scale continuous production, is prone to physical interference and rework of numerous hole-position components in subsequent tower assembly processes due to its lag. This invention employs a non-contact continuous exposure acquisition scheme using an industrial camera, combined with a connected component marking target area extraction algorithm, to perform automated image acquisition, eliminating physical contact components and effectively isolating background clutter from the production line. This transforms the punching state of the physical crossarm into a high signal-to-noise ratio two-dimensional digital matrix for processing by the backend visual algorithm. In the global data chain, this invention serves as the data entry point for the visual measurement system. The punched grayscale image it extracts directly provides a highly focused grayscale brightness calculation entity for subsequent first-order partial derivative calculations along the gradient direction. The acquisition of this basic image matrix and the region locking action lay the data reference benchmark for defining the physical contour boundary of the punch and for subsequent high-precision sub-pixel edge coordinate calculation.

[0025] Specifically, by executing an industrial camera image acquisition and target area detection method, a punching grayscale image is generated, including: An industrial camera was used to perform continuous exposure acquisition on the physical crossarm of the power line on the production line to obtain a two-dimensional pixel matrix containing holes and the surrounding metal texture as the initial crossarm image. A connected component labeling algorithm based on global grayscale threshold was used to scan the initial crossarm image, identify and remove background clutter features, and extract the target holes and their boundary neighborhoods as punched grayscale images.

[0026] Furthermore, the grayscale image of the punched hole characterizes the two-dimensional brightness distribution of the punched holes and surrounding metal texture on the crossarm surface. Extracting this image aims to remove background clutter from the production line and pinpoint the core measurement area. The region localization algorithm employs a connected component labeling model based on a grayscale threshold during the actual sinking process to ensure automated extraction of the target punched hole contour. This invention avoids the computational waste associated with global searches in extracting the target detection region, allowing the data obtained to form the source input for all subsequent edge extraction and deviation calculations.

[0027] At this point, the initial crossarm image and the punched hole grayscale image were obtained.

[0028] S2: Generate a valid set of circular arc points.

[0029] It should be noted that, given the unavoidable micro-roughness of the metal shear surface produced when industrial punch presses cut thick steel plates, conventional pixel-level edge detection struggles to obtain high-precision, smooth contours that reflect the true geometric shape. Existing technologies generally employ integer-level pixel gradient operators combined with fixed thresholds to perform coarse edge extraction when locating hole edges. This approach is limited by the resolution bottleneck of the camera's physical sensor and easily introduces stepped jagged errors in oblique and curvature change regions, causing geometric distortion in the early stages of feature extraction and leading to deviations in subsequent analytical fitting results of the circle center. To address this issue of insufficient low-level positioning accuracy, this invention introduces gradient calculation based on first-order partial derivatives and quadratic curve interpolation algorithms. This enables the analysis system to overcome the limitations of the hardware discrete pixel grid, sinking the edge extreme value positioning accuracy down to the pixel level and finding the true physical transition points in continuous grayscale transition bands. This technique effectively ensures the accuracy of extracting the spatial coordinates of the punch hole contour point array.

[0030] Specifically, based on the pixel gradient of the punched grayscale image, a quadratic curve interpolation algorithm is performed to obtain the punched edge point set, including: The first-order partial derivatives of the grayscale values ​​of each pixel are calculated by applying the difference operator along the horizontal and vertical directions of the punched grayscale image, and the image gradient matrix representing the edge transition intensity is synthesized by combining the horizontal and vertical gradient components. Local gradient extrema are located in the image gradient matrix, and a one-dimensional parabolic equation is established along the gradient direction where the extrema are located to perform quadratic curve interpolation calculation, so as to solve for the edge extrema coordinates with floating-point accuracy and generate the punched edge point set.

[0031] It should be noted that the quadratic curve interpolation calculation includes: at the local maximum gradient magnitude point determined along the gradient direction, taking this extreme point and its two adjacent sampling points along the gradient direction, for a total of three points, requiring that the gradient magnitude of the intermediate point is not less than the gradient magnitude of its two adjacent points. Using the gradient magnitudes of these three points as input, a one-dimensional parabola with a defined opening direction is constructed along the gradient direction, passing through these three sampling points. Since the true edge position corresponds to the continuous peak of the gradient magnitude, the position of the parabola's axis of symmetry is used as the sub-pixel level edge position estimate. Then, the offset of this axis of symmetry relative to the intermediate integer point is superimposed onto the integer coordinates of the intermediate point along the unit direction of the gradient direction to obtain the floating-point level edge coordinates. The punched edge point set is a set of multiple floating-point coordinates that break through the physical pixel grid limitations, representing the smooth position of the punched edge at the microscopic scale. In the specific sinking execution, the quadratic curve interpolation algorithm uses an existing mathematical model that fits a parabola along the gradient direction and solves for the axis of symmetry to accurately calculate the sub-pixel coordinates. To clarify the algorithm logic, the system calculates and generates floating-point coordinates based on the gradual transition law of pixel grayscale when constructing the point set. Mathematical calculation replaces the rough integer rounding to eliminate jagged edges from the bottom layer of the algorithm.

[0032] Thus, a set of punched edge points representing a high-precision profile was obtained.

[0033] It should be noted that when evaluating the geometric morphology of the edges of crossbeam holes manufactured by heavy stamping processes, it is important to consider that the steel inevitably produces physical deformation interferences such as burrs, protrusions, or pits at the edges under shear force. Existing visual measurement algorithms mostly lack screening mechanisms and treat all extracted edge contour window center points as valid data equally. This traditional logic fails to establish a screening mechanism for local micro-geometric mutations, causing the algorithm to be often biased by burr points with errors when fitting the center, resulting in center drift and inaccurate calculation of hole position coordinate deviation. To solve this deformation interference problem, this invention proposes a curvature variation evaluation method based on curve differential features. By analyzing the tangential direction change between adjacent points within the contour sliding window of the punched edge point set, the physical burr morphology at the micro level is transformed into a quantitative digital representation of curve smoothness. This method digitally marks irregular geometric mutations at the initial contour feature level to expose various edge defects caused by stamping die aging or steel tearing.

[0034] Preferably, the edge curvature variation rate is obtained based on the local coordinate tangential distribution characteristics of the punched edge point set, including: Based on the punching edge point set constituting the punching contour coordinate sequence, a contour sliding window for feature extraction is constructed point by point; within each contour sliding window, the tangential direction change between the lines connecting adjacent two-dimensional coordinate points is calculated using the forward difference rule of calculus to obtain the contour angle difference, which is used as the edge curvature rate at the center point of the window.

[0035] Furthermore, the window size of the contour sliding window is set to a range of 5 to 15 points, preferably 10 points. If the value is less than 5 points, it will be too sensitive to small image noise, causing normal arcs to be misjudged as burrs and failing to retain enough samples in subsequent fitting. If the value is greater than 15 points, the smooth transition will cause the algorithm to excessively ignore real small collapses and produce false negatives. Therefore, the preferred value of 10 can achieve a numerical balance between identifying physical anomalies and preserving normal arc features. The punching contour coordinate sequence refers to the ordered coordinate point sequence obtained by concatenating and sorting the disordered set of punching edge points according to their geometric adjacency relationship on the punching closed contour. Starting from any edge point in the set of edge points, the process involves selecting a starting point and, based on the connectivity and adjacency relationships between these points, sequentially finding and connecting the next adjacent edge point along the same circumferential direction of the punch profile, either clockwise or counterclockwise, until returning to the starting point. This organizes the discrete set of edge points into an ordered coordinate sequence along the profile. In this sequence, the preceding and following points of any given coordinate point are the adjacent two-dimensional coordinate points mentioned when constructing the sliding window and calculating the connection between adjacent points. The edge curvature variation characterizes the degree of curvature change of the punch profile within a small neighborhood and directly reflects the presence of physical burrs or notches at that location. The specific calculation steps for the edge curvature variation are as follows: First, along the extracted punch profile coordinate sequence, select a small sliding window containing multiple adjacent points. Calculate the angle of the line connecting the first and last points within the window. Calculate the actual total length of the line connecting all adjacent points within the window, i.e., the profile arc length. Divide the angle change of the first and last points of the window by the profile arc length within the window; the result is the edge curvature variation at the center point of the window.

[0036] Thus, the edge curvature variation rate for determining the edge smoothness state was obtained.

[0037] It should be noted that this invention employs a smoothness classification and elimination method. By calculating the mean curvature variation rate, an edge abrupt change threshold is generated, and conditional branch judgment is performed on the edge curvature variation rate. This establishes a hierarchical mechanism of zero-weight elimination and full-trust retention. This data cleaning mechanism ensures the purity of the contour data participating in subsequent simultaneous calculations, thereby avoiding weight pollution caused by deformation interference terms in solving geometric analytical equations. From the perspective of data flow, this invention relies on the pre-generated edge curvature variation rate to perform synchronous truncation and feature recombination operations on the corresponding punching edge point set. The final output effective arc point set realizes the transformation from the original contour containing physical impurities to a set of pure metal arc segments, thus providing high-quality sample coordinates for the subsequent least-squares algebraic circle fitting operation.

[0038] Preferably, an effective set of arc points is generated by performing adaptive threshold filtering on the edge curvature variation rate, including: Calculate the global arithmetic mean of all edge curvature variations to generate an edge abrupt change threshold, and then compare the edge curvature variation at the center point of each window with the numerical value of this edge abrupt change threshold. If the edge curvature variation rate of the current window center point is greater than the edge abrupt change threshold, the point is determined to be a physical burr or corner collapse defect, and its edge calculation weight in subsequent solutions is forcibly cleared to perform data stream discarding; if the edge curvature variation rate of the current window center point is less than or equal to the edge abrupt change threshold, the point is determined to be a smooth real arc segment, the corresponding point is given full weight, and its floating-point coordinate data is recorded for recombination and combination to extract and generate a valid arc point set.

[0039] Furthermore, the effective arc point set is a collection of pure geometric arc coordinates after eliminating interference from physical deformations such as burrs and gaps. The smoothness classification and elimination algorithm is implemented using basic numerical comparison and conditional branching logic during the actual execution. This invention employs edge smoothness classification and elimination logic, the mathematical essence of which is to forcibly clear the calculation weight of the abnormal window center point before algebraic fitting operations to avoid numerical interference from abrupt geometric features in the solution equations, thus ensuring the effective purity of the retained arc coordinate information.

[0040] Thus, an effective set of arc points was obtained that can eliminate deformation interference terms and ensure the purity of the data used in the fitting.

[0041] S3: Output crossarm defect type.

[0042] It should be noted that when crossing the boundary between discrete edge coordinates and the macroscopic mechanical hole location, the system needs to solve the analytical calculation problem of converting the residual effective circular arc to the standard geometric center with high precision. Traditional measurement models often use the midpoint of a rectangular bounding box or the arithmetic mean centroid calculation method to perform a rough hole position deduction. Although this method can output results quickly, its computational link is sensitive to edge symmetry and lacks rigorous quadratic algebraic equation support. This results in automated quality inspection systems being unable to track micron-level hole center deviations and facing low measurement reliability in rigorous assembly line inspection environments. To improve the anti-interference capability of hole center positioning, this invention introduces an algebraic circle equation solving algorithm based on least squares residual analysis. The system adopts analytical logic of constructing the minimum value of partial derivatives and solving the matrix, replacing the centroid calculation method which is easily affected by contour incompleteness. This eliminates the geometric offset caused by the asymmetry of the cut and endows the measurement results with algebraic interpretation capabilities.

[0043] Specifically, based on the effective set of circular arc points, the partial derivative operation of the least squares circular algebra equation is performed to calculate the actual punching center, including: Substitute the coordinate data of the effective arc point set into the algebraic circle equation of the two-dimensional plane to establish a system of fitting system equations containing the center coordinates and radius as unknown parameters; use the least squares circle fitting algorithm to construct the residual sum of squares function of all effective data points in the effective arc point set to the boundary of the fitting circle, and take the first-order partial derivatives with respect to the center coordinates and set them equal to zero to establish a system of linear equations, and solve the actual punching center by matrix inversion.

[0044] Furthermore, the matrix inversion operation is specifically represented as follows: Based on the least squares circle fitting, the algebraic equation of the circle is used. All effective arc sampling points are substituted one by one to construct the sum of squares of the algebraic residuals from each point to the fitted circle as the objective function. To minimize this sum of squares, the first-order partial derivatives of each undetermined coefficient in the circle equation are calculated and set to zero, resulting in a set of linear equations about these undetermined coefficients. This set of linear equations is written in matrix form where the product of the coefficient matrix and the vector of undetermined coefficients equals a constant vector. Then, the matrix inversion operation is performed on the coefficient matrix to solve for each undetermined coefficient. These coefficients are then used to reconstruct the center coordinates and radius. The actual punching center represents the two-dimensional coordinates of the absolute geometric center of the punched hole, theoretically perfect after eliminating interference from surrounding burrs and corner deformation. In practical implementation, the least squares circle fitting algorithm constructs the sum of squares of the geometric distance errors from the point to the center and uses matrix operations to solve for the unknown parameters of the center's horizontal and vertical coordinates and the radius. To improve the accuracy of the center positioning, the system does not use the global data averaging method that includes all edge defects. Instead, it performs a direct inner product calculation between the pure coordinate set and the algebraic equation of the circle to reversely reconstruct the actual punching center. This provides a reliable comparison benchmark for subsequent hole position deviation calculation.

[0045] Thus, the actual center of the punch, representing the absolute geometric center of the theory, was obtained.

[0046] It should be noted that when evaluating whether the quality of the crossarm stamping position meets the factory acceptance standards, simply obtaining the coordinates of the isolated hole center lacks practical engineering comparison significance and cannot guide the entire production and sorting process. This invention reshapes the measurement and evaluation system using relative coordinate transformation and algebraic difference calculation mechanisms. The system avoids using absolute screen coordinates that deviate from drawing specifications, extracts the theoretical punching position including computer-aided design dimensions, and performs relative vector geometric difference derivation in combination with the actual punching center. This method converts isolated measurement coordinates into relative dynamic differences reflecting mechanical process execution errors. This invention utilizes the actual punching center to retrieve the theoretical punching position pre-analyzed and imported into the system to calculate the two-dimensional deviation distance, obtaining the punching position deviation, characterizing the distance difference between the actual processed position and the design drawing requirements, thereby providing driving data for subsequent product tolerance grading verification and control signal output.

[0047] Preferably, standard process drawings are analyzed to obtain the theoretical punching position. Based on the actual punching center and the pre-obtained theoretical punching position, vector space difference calculation is performed to obtain the punching position deviation, including: The standard process drawing file is parsed from the underlying data communication interface to extract the theoretical punching position, which represents the correct two-dimensional spatial coordinates of the theoretical punching. Based on the laws of plane analytical geometry, the absolute difference between the center of the actual punching circle and the theoretical punching position in the two-dimensional plane coordinate system in the horizontal and vertical axes is calculated and recorded as the absolute difference between the center of the horizontal and vertical axes. The Euclidean distance between the center of the line is solved using the Euclidean formula. The calculated absolute difference between the center of the horizontal and vertical axes and the Euclidean distance between the center of the line are merged, formatted and encapsulated, and output as the punching position deviation for error verification.

[0048] Furthermore, the punching position deviation is precisely calculated to determine the positional deviation between the current punching position of the actual crossarm and the theoretical requirement of the standard engineering drawings. Relative coordinate transformation and algebraic difference calculations are employed, and during the actual sinking operation, the two-point distance formula and vector difference operations from plane analytical geometry are used. The process of determining the offset avoids manual visual estimation and relies solely on objective numerical distance calculation mechanisms. The theoretical punching position is obtained by analyzing standard drawing files or importing calibration parameters from standard samples to establish a theoretical reference anchor point, avoiding benchmark drift caused by manual screen marking. Under this objective comparison logic, the system establishes a reliable reference system, thus achieving stable judgment between identifying punching offset tendencies and ensuring measurement consistency even under minor vibration conditions on the production line.

[0049] Thus, the punching position deviation, which characterizes the degree of actual processing misalignment, was obtained.

[0050] It should be noted that, as the digital classification and judgment step in the visual measurement method, this invention uses the punching position deviation obtained in the preceding steps as the basis for judgment, and maps the specific crossarm defect type backward based on the multi-level tolerance boundaries derived from the design benchmark. Specifically, the system constructs a multi-level floating tolerance range by multiplying the nominal dimensional parameters by the ratio of the process allowable tolerance, based on the theoretical design benchmark, and establishes a mapping logic from position deviation to quality grade label; then, by using numerical comparison and conditional branch interval mapping, it realizes the automated assessment of error severity, thereby establishing a quantitative conversion relationship from physical displacement deviation to classification label, so that each output defect judgment result has traceable engineering quality orientation.

[0051] Preferably, based on the punching position deviation, a multi-level tolerance range numerical comparison and classification is performed to output the crossarm defect type, including: The nominal dimensional parameters of the theoretical punching position are extracted and algebraically multiplied with the ratios of two different pre-stored allowable tolerances for stamping processes in the industrial system to generate an acceptable upper limit and a rework upper limit, with the acceptable upper limit being less than the rework upper limit. The straight-line distance between the two centers of the absolute difference between the centers of the horizontal and vertical axes in the punching position deviation and the Euclidean distance between the straight center is taken as the comprehensive distance parameter. When the comprehensive distance parameter is less than or equal to the acceptable upper limit, it is determined to be within the acceptable tolerance zone, and the corresponding defect type is a qualified product, which is released for circulation. When the comprehensive distance parameter is greater than the acceptable upper limit but less than or equal to the rework upper limit, it is determined to be within the correctable error range, and the corresponding defect type is a product requiring rework. Furthermore, based on the relative magnitude and the positive or negative sign of the absolute difference between the centers of the horizontal and vertical axes in the punching position deviation, the main direction of the punching offset is determined, providing directional guidance for the rework correction process. When the comprehensive distance parameter is greater than the rework upper limit, it is determined to be within the position error range of the overflow tolerance upper limit, and the corresponding defect type is a scrap product, which is rejected.

[0052] Furthermore, the crossarm defect type is a digital classification label characterizing whether the current power crossarm product is a qualified product, a product requiring rework, or a scrap product. The tolerance boundary judgment algorithm is implemented using numerical logic comparison and multi-branch conditional statement mapping during execution. The quantitative conversion relationship between position deviation and product qualification status in this invention avoids the uncertainty of judgment scale caused by relying on manual visual inspection in the past, and establishes the objective interpretability of online full-quantity inspection on the production line. It should be noted that the comprehensive distance parameter is the straight-line distance between the two centers obtained by synthesizing the two axial offset components of the lateral and longitudinal directions. It is a single scalar reflecting the overall offset amplitude and is used to grade the error amplitude. The axial absolute difference between the lateral and longitudinal directions is used to determine the offset direction. Both are taken from the same punch position deviation package and are fully used in subsequent steps. There are no intermediate quantities that are calculated but idle. The position tolerance corresponding to the upper limit of qualification is 0.5mm, and the position tolerance corresponding to the upper limit of rework is 1mm. The two thresholds are determined according to the industry precision standards for power fitting stamping and can be adjusted according to actual process requirements. The comprehensive distance parameter is the physical distance between the actual center and the theoretical center of the circle, in mm, which is calculated from the pixel coordinates using the camera calibration coefficient.

[0053] Thus, the types of crossarm defects that guide the sorting process were obtained.

[0054] S4: Output crossarm sorting data.

[0055] It should be noted that, to address the technical challenge of ensuring the accurate and timely conversion of digital results into mechanical rejection actions in the sorting mechanism, this invention introduces a timing-synchronous communication algorithm with pulse compensation as a conversion module. A position encoding and tracking mechanism is used to execute a queue delay-triggered operation based on displacement alignment, anchoring the classification results into a physical reference system. Communication mapping logic is then used to perform encapsulation conversion, establishing a safe boundary for the software-to-mechanical action conversion. In the output stage, this invention extracts the crossarm defect type generated previously and combines it with an externally transmitted pulse signal. Through the delay fusion of these two signals in the queue, a sorting control command containing the physical rejection mission is generated. This command connects the entire control process, from feature perception to pneumatic push rod intervention, and establishes a control architecture that drives the physical terminal based on calculation results.

[0056] Specifically, based on the crossarm defect type, a timing tracking queue is established by collecting external transmission pulse data, and message mapping is performed to generate sorting control instructions, including: The system acquires the encoder pulse equivalent parameters of the production line, collects the transmission pulse data of the production line conveyor belt through sensors, and extracts mechanical delay parameters from the encoder pulse equivalent parameters to establish a time-series tracking queue that maps the visual inspection time to the physical rejection station displacement distance. The system converts the classification status identifier of the crossarm defect type into a low-level communication message that can be recognized by the underlying programmable logic controller and corresponds to the control register address. At the moment when the action execution is performed by matching the mechanical delay parameters during the countdown of the time-series tracking queue, the low-level communication message is sent down through the entire process via the industrial bus to generate sorting control instructions that trigger external hardware actions.

[0057] Furthermore, the sorting control command represents the final communication entity across the entire vision measurement system to the physical manufacturing terminal, used to drive the pneumatic push rod or rejection baffle. The timing synchronization communication algorithm, during actual execution, employs underlying communication protocol message conversion and pulse counting delay triggering logic. The execution of the timing pulse alignment operation establishes the direct kinematic process between the visual calculation time difference and the physical product movement space difference. The system outputting this command signifies a shift from the reliance on manual handling of defective products to a fully automated, digitally driven process, from edge smoothness rejection perception to precise interception of defective crossarms in the physical workshop. Ultimately, this completes the automated pre-sorting work for the stamping quality of power crossarms.

[0058] At this point, the sorting control instructions for transmitting visual classification results to the underlying mechanical rejection actions have been obtained. It should be noted that after the sorting control instructions are sent out from the main line, the system faces the challenge of accurately translating the instantaneous virtual electrical signals into physical interception and rejection actions, and collecting hardware action feedback to form a quality traceability log. If the end-point control network lacks a physical action feedback collection mechanism and data storage method, the production line is prone to issues such as push rod jamming or insufficient air pressure, leading to missed rejection of already determined non-conforming parts, resulting in a rigid disconnect in the entire quality inspection process. This invention addresses this by combining the execution of physical action instructions at the control layer with the collection of feedback signals from end-point sensors, thus feeding digital visual decisions back into the mechanical production cycle to achieve archiving of feedback throughout the entire quality inspection process.

[0059] Preferably, the sorting action is executed based on the sorting control command, and the crossarm sorting data is output, including: The underlying programmable logic controller parses the received sorting control instructions and drives the underlying pneumatic push rod mechanism on the assembly line through the electromagnetic reversing valve to perform physical rejection or release operations on the current crossarm being tested, either pushing it off the track or allowing it to pass. Simultaneously, it collects the execution feedback signal of the magnetic switch of the cylinder at the end of the stroke of the underlying pneumatic push rod mechanism, associates the execution feedback signal with the crossarm defect type corresponding to this quality inspection process, packages it into a data entity with a timestamp, generates crossarm sorting data, and stores it in the underlying database.

[0060] Furthermore, in the overall data fusion chain, this invention serves as the physical execution and status archiving node of the industrial measurement method. It receives the instruction protocol packets generated in the preceding sequence and directly drives the pneumatic mechanism to output physical blocking actions. Subsequently, it captures hardware status feedback, thereby generating crossarm sorting data records containing time attributes. The solidified storage of this data effectively replaces manual counting and provides system log credentials for the quality traceability of all crossarm production data.

[0061] It should be noted that, Figure 2This diagram illustrates the principle of measuring the position deviation of punched holes in a power crossarm. The diagram shows a two-dimensional coordinate system formed by the horizontal and vertical directions. The dots on the circular outline represent the effective arc point set, indicating the smooth, true arc segments retained after curvature variation filtering. The crosses represent burr / collapse points, indicating abnormal edge interference points generated by the stamping process. The square points with crosses inside represent the actual punched hole center, obtained by least-squares circle fitting calculation from the effective arc point set. The square point in the upper right corner represents the theoretical punched hole position, the ideal hole position reference set by the standard process. The horizontal line segment between two points represents the absolute difference between the centers along the horizontal axis, the vertical line segment represents the absolute difference between the centers along the vertical axis, and the diagonal line segment represents the Euclidean distance between the two centers. These three together constitute the quantitative index of the punched hole position deviation. This diagram visually presents the core logic of this invention: using adaptive threshold filtering to remove abnormal points such as burrs and collapses while retaining the effective arc point set. It embodies the measurement principle of fitting the actual punched hole center based on a pure arc segment and then calculating the position deviation by vector space difference with the theoretical punched hole position.

[0062] This completes the visual measurement method for the positional deviation of punched holes in power crossarms.

[0063] While various embodiments of the invention have been shown and described in this specification, it will be apparent to those skilled in the art that such embodiments are provided by way of example only. Many modifications, alterations, and alternatives will occur to those skilled in the art without departing from the spirit and essence of the invention.

Claims

1. A visual measurement method for the positional deviation of punched holes in an electric crossarm, characterized in that, include: S1: Generate a punched grayscale image by executing an industrial camera image acquisition and target area cropping method; S2: Based on the pixel gradient of the punched grayscale image, perform a quadratic curve interpolation algorithm to obtain the punched edge point set; according to the local coordinate tangential distribution characteristics of the punched edge point set, obtain the edge curvature variation rate; by performing adaptive threshold condition filtering on the edge curvature variation rate, generate an effective arc point set; S3: Based on the effective set of circular arc points, perform the least squares circular algebra equation partial derivative operation to calculate the actual punching center; analyze the standard process drawing to obtain the theoretical punching position; based on the actual punching center and the pre-obtained theoretical punching position, perform vector space difference calculation to obtain the punching position deviation; based on the punching position deviation, perform multi-level tolerance interval numerical comparison and classification to output the crossarm defect type. S4: Based on the crossarm defect type, a timing tracking queue is established by collecting external transmission pulse data, and message mapping is performed to generate sorting control instructions; sorting actions are executed based on the sorting control instructions, and crossarm sorting data is output.

2. The visual measurement method for the punching position deviation of a power crossarm according to claim 1, characterized in that, The generation of the punched grayscale image includes: An industrial camera was used to perform continuous exposure acquisition on the physical crossarm of the power line on the production line to obtain a two-dimensional pixel matrix containing holes and the surrounding metal texture as the initial crossarm image. A connected component labeling algorithm based on global grayscale threshold was used to scan the initial crossarm image, identify and remove background clutter features, and extract the target holes and their boundary neighborhoods as punched grayscale images.

3. The visual measurement method for the punching position deviation of an electric crossarm according to claim 1, characterized in that, The process of obtaining the set of punched edge points includes: The first-order partial derivatives of the grayscale values ​​of each pixel are calculated by applying the difference operator along the horizontal and vertical directions of the punched grayscale image, and the image gradient matrix representing the edge transition intensity is synthesized by combining the horizontal and vertical gradient components. Local gradient extrema are located in the image gradient matrix, and a one-dimensional parabolic equation is established along the gradient direction where the extrema are located to perform quadratic curve interpolation calculation, so as to solve for the edge extrema coordinates with floating-point accuracy and generate the punched edge point set.

4. The visual measurement method for the punching position deviation of a power crossarm according to claim 1, characterized in that, The process of obtaining the edge curvature variation rate includes: Based on the punching edge point set constituting the punching contour coordinate sequence, a contour sliding window for feature extraction is constructed point by point; within each contour sliding window, the tangential direction change between the lines connecting adjacent two-dimensional coordinate points is calculated using the forward difference rule of calculus to obtain the contour angle difference, which is used as the edge curvature rate at the center point of the window.

5. The visual measurement method for the punching position deviation of an electric crossarm according to claim 1, characterized in that, The generation of the effective arc point set includes: The global arithmetic mean of all edge curvature variations is calculated to generate an edge abrupt change threshold. The edge curvature variation of each window center point is then compared with this threshold. If the edge curvature variation of the current window center point is greater than the threshold, the point is determined to be a physical burr or corner collapse defect, and its edge calculation weight in subsequent solutions is forcibly cleared to discard the data stream. If the edge curvature variation of the current window center point is less than or equal to the threshold, the point is determined to be a smooth real arc segment, the corresponding point is assigned full weight, and its floating-point coordinate data is recorded, recombined, and used to extract and generate a set of valid arc points.

6. The visual measurement method for the punching position deviation of a power crossarm according to claim 1, characterized in that, The calculation of the actual punching center includes: Substitute the coordinate data of the effective arc point set into the algebraic circle equation of the two-dimensional plane to establish a system of fitting system equations containing the center coordinates and radius as unknown parameters; use the least squares circle fitting algorithm to construct the residual sum of squares function of all effective data points in the effective arc point set to the boundary of the fitting circle, and take the first-order partial derivatives with respect to the center coordinates and set them equal to zero to establish a system of linear equations, and solve the actual punching center by matrix inversion.

7. The visual measurement method for the punching position deviation of a power crossarm according to claim 1, characterized in that, The method of obtaining the punching position deviation includes: The standard process drawing file is parsed from the underlying data communication interface to extract the theoretical punching position, which represents the correct two-dimensional spatial coordinates of the theoretical punching. Based on the laws of plane analytical geometry, the absolute difference between the center of the actual punching circle and the theoretical punching position in the two-dimensional plane coordinate system in the horizontal and vertical axes is calculated and recorded as the absolute difference between the center of the horizontal and vertical axes. The Euclidean distance between the center of the line is solved using the Euclidean formula. The calculated absolute difference between the center of the horizontal and vertical axes and the Euclidean distance between the center of the line are merged, formatted and encapsulated, and output as the punching position deviation for error verification.

8. The visual measurement method for the positional deviation of punched holes in an electric crossarm according to claim 1, characterized in that, The types of defects in the output crossarm include: Extract the nominal dimensional parameters of the theoretical punching position to generate an acceptable upper limit and a rework upper limit, with the acceptable upper limit being less than the rework upper limit; take the straight-line distance between the two centers in the punching position deviation as the comprehensive distance parameter; when the comprehensive distance parameter is less than or equal to the acceptable upper limit, the corresponding defect type is acceptable product; when the comprehensive distance parameter is greater than the acceptable upper limit but less than or equal to the rework upper limit, the corresponding defect type is product requiring repair; when the comprehensive distance parameter is greater than the rework upper limit, the corresponding defect type is scrap product.

9. The visual measurement method for the punching position deviation of an electric crossarm according to claim 1, characterized in that, The generation of sorting control instructions includes: The system acquires the encoder pulse equivalent parameters of the production line, collects the transmission pulse data of the production line conveyor belt through sensors, and extracts mechanical delay parameters from the encoder pulse equivalent parameters to establish a time-series tracking queue that maps the visual inspection time to the physical rejection station displacement distance. The system converts the classification status identifier of the crossarm defect type into a low-level communication message that can be recognized by the underlying programmable logic controller and corresponds to the control register address. At the moment when the action execution is performed by matching the mechanical delay parameters during the countdown of the time-series tracking queue, the low-level communication message is sent down through the entire process via the industrial bus to generate sorting control instructions that trigger external hardware actions.

10. A visual measurement method for the positional deviation of punched holes in an electric crossarm according to claim 1, characterized in that, The output crossarm sorting data includes: The underlying programmable logic controller parses the received sorting control instructions and drives the underlying pneumatic push rod mechanism on the assembly line through the electromagnetic reversing valve to perform physical rejection or release operations on the current crossarm being tested, either pushing it off the track or allowing it to pass. Simultaneously, it collects the execution feedback signal of the magnetic switch of the cylinder at the end of the stroke of the underlying pneumatic push rod mechanism, associates the execution feedback signal with the crossarm defect type corresponding to this quality inspection process, packages it into a data entity with a timestamp, generates crossarm sorting data, and stores it in the underlying database.

Images