A method and system for controlling the size of copper flat conductors based on machine vision

Abstract

This application relates to the technical field of copper flat conductor size control, and discloses a method and system for copper flat conductor size control based on machine vision. The method includes: acquiring continuous frame images of the copper flat conductor; identifying and extracting the edge lines of the copper flat conductor based on each frame image; identifying jitter frames in the continuous frame images; setting a non-uniform time window for each jitter frame and performing fusion compensation on the edge lines of the jitter frames; extracting the size parameters of the copper flat conductor based on the edge lines in each frame image and calculating the size residual of the copper flat conductor; identifying the trend drift of the copper flat conductor size based on the size residual of the copper flat conductor in the continuous frame images; extracting the size error data of the copper flat conductor and inputting it into a PID controller to suppress the trend drift of the copper flat conductor size. This application can solve the problems of inaccurate edge line identification and misjudgment of size offset trends caused by conductor edge jitter, achieving more precise size control.

Description

Technical Field

[0001] This application relates to the technical field of copper flat conductor size control, specifically to a method and system for copper flat conductor size control based on machine vision. Background Technology

[0002] Copper flat conductors are key components in fields such as power transmission and electronic equipment manufacturing. Non-contact inspection and control technology based on machine vision is gradually becoming the mainstream solution for dimensional control in copper flat conductor production. Although machine vision technology has achieved some application in the dimensional inspection and control of copper flat conductors, some technical bottlenecks still exist in actual production environments, restricting further improvements in dimensional control accuracy.

[0003] During high-speed transport, copper flat conductors are prone to slight jitter or edge warping due to mechanical vibration and uneven feeding, resulting in blurred, offset, or discontinuous conductor edges in the acquired continuous frame images. Traditional static edge detection algorithms rely solely on grayscale changes in a single frame to extract edges, failing to distinguish between true edges and pseudo-edges caused by jitter, easily leading to edge extraction errors. During long-term continuous processing, copper flat conductors are affected by the cumulative effects of roll wear, material temperature changes, and metal fatigue, causing slow, trend-based drift in dimensional parameters, such as a gradual increase in thickness over rolling time. Existing technologies often control dimensional residuals within a single frame or short time window, lacking the ability to track and predict residual trends over long periods. Misjudging trend drift as random noise leads to delayed response in the control system; excessive sensitivity to short-term fluctuations triggers frequent adjustments to processing parameters, exacerbating dimensional instability. Current dimensional control systems primarily analyze residuals at the level of single numerical comparisons, failing to delve into the time-series characteristics of residuals and failing to distinguish between trend-based growth and random fluctuations in residuals. The limitations of this analysis lead to a lack of precise basis for subsequent control parameter adjustments, which can easily result in over-adjustment or under-adjustment.

[0004] For example, Chinese patent CN114322801B discloses a method for measuring the diameter of shaft-type parts based on parallel binocular machine vision. The steps include: S1, adjusting the camera; S2, capturing an image of a dot calibration plate, and acquiring the image for image enhancement and noise reduction to obtain a new image; S3, based on the new image, acquiring the pixel-level distance D1 of the camera and its angle θ1 relative to the horizontal coordinate, and the pixel-level distance D2 of the lower camera and its angle θ2 relative to the horizontal coordinate; S4, adjusting the field of view of the captured image to ensure that angles θ1 and θ2 are within the allowable range of measurement error; S5, acquiring the actual dimension Dreal corresponding to the lower edge of f1 and the upper edge of f2, where Dreal is the sum of the distance from C1 to the lower edge of the image and the distance from C2 to the upper edge of the image, and Dreal is the diameter of the shaft-type part. This technical solution is simple to operate, highly versatile, and has good real-time performance and accuracy, improving the efficiency of shaft diameter measurement for shaft-type parts.

[0005] Chinese patent application CN108844476A discloses a machine vision-based method for detecting aperture size tolerances, comprising the following steps: placing the object to be detected on a feeding conveyor belt; a photoelectric sensor detecting the object; a robotic arm automatically gripping the object; transferring the object to the vision area; the robotic arm positioning itself at a first detection position; data being collected for the first time using a vision acquisition device and transmitted to the analysis data section for detection and analysis; the robotic arm positioning itself at a second detection position; data being collected for the second time using a vision acquisition device and transmitted to the analysis data section for detection and analysis; the robotic arm positioning itself at a third detection position; data being collected for the third time using a vision acquisition device and transmitted to the analysis data section for detection and analysis; data analysis of the object's accuracy category; data feedback to the robotic arm controller; the robotic arm transferring the object to a classification area; the robotic arm releasing the object; the unloading conveyor belt carrying the object out; and the detection ending. This technical solution offers high detection efficiency, stable operation, low detection cost, ease of use, and accurate detection.

[0006] All of the above technical solutions suffer from the problems mentioned in the background: they lack the ability to track and predict the long-term trend of workpiece dimensional error changes, which easily leads to measurement errors and control delays.

[0007] The information disclosed in this background section is intended only to enhance the understanding of the overall background of this application and should not be construed as an admission or in any way implying that the information constitutes prior art known to those skilled in the art. Summary of the Invention

[0008] The technical problem to be solved by this application is to overcome the defects of the prior art and provide a method and system for adjusting the size of copper flat conductors based on machine vision, so as to solve the problems of inaccurate edge line recognition and misjudgment of size offset trend caused by edge jitter of copper flat conductors, and achieve more accurate size control.

[0009] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0010] On the one hand, this application provides a method for adjusting the size of copper flat conductors based on machine vision, including the following steps:

[0011] Acquire consecutive frame images of the copper flat conductor; identify and extract the edge lines of the copper flat conductor based on each frame image;

[0012] Identify jitter frames in consecutive frames of images based on the edge lines of the copper flat conductor in each frame of the image.

[0013] A non-uniform time window is set for each jittery frame, and edge lines of the jittery frames are fused and compensated.

[0014] The size parameters of the copper flat conductor are extracted based on the edge lines in each frame of the image, and the size residual of the copper flat conductor is calculated.

[0015] Based on the dimensional residuals of copper flat conductors in consecutive frame images, the trend drift of copper flat conductor dimensions is identified;

[0016] Extract the dimensional error data of the copper flat conductor and input it into the PID controller to suppress the trend of dimensional drift of the copper flat conductor.

[0017] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the step of identifying jitter frames in continuous frame images specifically includes:

[0018] Set sampling points on the edge lines in each frame of the image; for any frame of the image, record the vertical and horizontal coordinates of each sampling point in the image.

[0019] Calculate the difference in vertical coordinates between any two adjacent sampling points, and use it as the jitter value of the edge line;

[0020] Arrange the jitter values ​​of the edge lines into a jitter value sequence according to the horizontal coordinates of the corresponding sampling points from smallest to largest;

[0021] The jitter amplitude and jitter frequency of the edge line are calculated based on the jitter value sequence;

[0022] If the jitter amplitude of the edge line is greater than the preset jitter amplitude threshold, or the jitter frequency is greater than the preset jitter frequency threshold, then the image containing the edge line is a jitter frame.

[0023] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the method for calculating the jitter frequency is as follows: record the positive and negative signs of each jitter value in the jitter value sequence; if any two adjacent jitter values ​​in the jitter value sequence have opposite positive and negative signs, it is recorded as one jitter count; record the total number of jitter counts in the jitter value sequence, divide it by the number of jitter values ​​in the jitter value sequence, and obtain the jitter frequency;

[0024] The method for identifying jittery frames in continuous frame images further includes: calculating the mean of the vertical coordinates of all sampling points on each edge line as the height value of the corresponding edge line; performing outlier detection on the height values ​​of all edge lines; and identifying the image containing the edge line corresponding to any outlier as a jittery frame.

[0025] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the method for setting a non-uniform time window for any jittery frame is as follows:

[0026] Extract the N consecutive frames before the jittered frame and the N consecutive frames after the jittered frame, which together with the jittered frame form a non-uniform time window; N is a positive integer;

[0027] For any frame of an image within a non-uniform time window, the jitter amplitude and jitter frequency of the edge lines are normalized and dimensionless, and then weighted and summed to obtain the jitter factor of the corresponding frame image.

[0028] If the jitter factor of any frame image in the non-uniform time window is greater than the preset jitter factor threshold, the corresponding frame image is removed from the non-uniform time window.

[0029] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the method for fusion compensation of the edge lines of any jittery frame is as follows:

[0030] Calculate the acquisition time difference between each frame of the image and the jitter frame in the non-uniform time window;

[0031] The fusion weight of each frame image is assigned based on the acquisition time difference between each frame image and the jitter frame and the jitter factor, wherein the acquisition time difference and the jitter factor are both negatively correlated with the fusion weight.

[0032] Based on the fusion weights, the edge lines of each frame in the non-uniform time window are weighted and fused to obtain the fused edge lines of the jittery frame.

[0033] Local fitting is performed on the fused edge line to obtain the edge line of the jitter frame after fusion compensation.

[0034] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the continuous frame images of the copper flat conductor are continuous frame images containing the copper flat conductor region collected during the transportation process after the copper flat conductor is processed and formed; each frame image corresponds to one copper flat conductor.

[0035] The process of identifying and extracting the edge lines of the copper flat conductor specifically includes: identifying the copper flat conductor region from the image, performing edge detection on the copper flat conductor region to obtain the outline of the copper flat conductor; and extracting the edge lines of the copper flat conductor based on the outline of the copper flat conductor.

[0036] The size parameter of the copper flat conductor is either its width or its thickness; the calculation of the size residual of the copper flat conductor specifically includes: subtracting the preset reference size parameter from the size parameter of the copper flat conductor in each frame image to obtain the size residual of the copper flat conductor in each frame image;

[0037] The dimensional error data of the copper flat conductor includes the dimensional residuals of the copper flat conductor in each frame of the image acquired from the current moment.

[0038] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the method identifies the trend drift of the copper flat conductor size based on the dimensional residuals of the copper flat conductor in consecutive frame images, specifically including:

[0039] The dimensional residuals of the copper flat conductor in each frame of the image are organized into a time series of dimensional residuals according to the order of acquisition of the corresponding frame images from early to late.

[0040] Based on the time series of the size residuals, a forward time window and a reverse time window for the size residuals are constructed. Specifically, this includes: extracting the size residuals of the copper flat conductor from the most recently acquired M consecutive frames of images from the size residual sequence, and forming the reverse time window of the size residual sequence; M is a positive integer; inputting the time series of the size residuals into a trained time series prediction model to obtain the predicted values ​​of the copper flat conductor size residuals in the next M consecutive frames of images, and forming the forward time window of the size residuals.

[0041] The evolution trend of the dimensional residuals is extracted from the forward time window and the reverse time window, respectively;

[0042] Based on the evolution trend of dimensional residuals in the forward and reverse time windows, the trend drift of copper flat conductor dimensions is identified.

[0043] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the evolution trend of the size residuals for any time window in the forward and reverse time windows includes the residual slope and the residual mean; for any time window, the method for extracting the residual slope is as follows: curve fitting is performed on the continuous size residuals in the forward or reverse time window, and the average slope of the fitted curve is calculated as the residual slope of the corresponding time window;

[0044] For any time window, the method for extracting the mean residual is as follows: take the absolute value of each size residual in the forward or reverse time window and then calculate the average value to obtain the mean residual of the corresponding time window;

[0045] For the positive time window, the evolution trend of the size residuals also includes the residual volatility, which is extracted as follows: calculate the standard deviation of each size residual in the positive time window and divide it by the mean of the residuals in the positive time window to obtain the residual volatility of the positive time window.

[0046] As a preferred embodiment of the machine vision-based copper flat conductor size control method described in this application, the identification of the trend drift of the copper flat conductor size specifically includes:

[0047] If the residual slopes of both the forward and reverse time windows are positive or both are negative, and the difference between the residual slopes of the forward and reverse time windows is less than a preset slope difference threshold, then the first criterion for trend drift is met.

[0048] The ratio of the mean residual of the forward time window to the mean residual of the reverse time window is calculated as the amplitude growth rate of the size residual; if the amplitude growth rate is greater than a preset residual growth threshold and the residual volatility of the forward time window is less than a preset residual volatility threshold, then the second criterion of trend drift is satisfied.

[0049] If both the first and second criteria for trend drift are met, then the dimensions of the copper flat conductor exhibit a trend drift.

[0050] Secondly, this application provides a machine vision-based copper flat conductor size control system, including an image acquisition module, an image recognition module, an image processing module, an error recognition module, and a size control module; wherein:

[0051] The image acquisition module is used to acquire continuous frame images of the copper flat conductor;

[0052] The image recognition module identifies and extracts the edge lines of the copper flat conductor based on each frame of the image;

[0053] The image processing module is used to identify jittery frames in continuous frame images and to perform edge fusion compensation on the jittery frames by setting a non-uniform time window;

[0054] The error identification module calculates the dimensional residual of the copper flat conductor in each frame of the image based on the edge line, and identifies the trend of copper flat conductor size drift.

[0055] The size control module is equipped with a PID controller to suppress the tendency of copper flat conductor size drift.

[0056] Compared with the prior art, the beneficial effects achieved by this application are as follows:

[0057] This application addresses the problem of edge warping and slight jitter in copper flat conductors during high-speed transmission, which can lead to blurred or discontinuous image edges. By identifying jitter frames in continuous image frames and applying a non-uniform time window to compensate for edge line fusion, dynamic defects can be effectively handled, resulting in smooth and reliable edge information. Based on the optimized edge lines, the width, thickness, and other dimensional parameters of the copper flat conductor can be directly and accurately extracted, reducing the impact of inaccurate edge information on dimensional parameter extraction.

[0058] This application can effectively identify the trend of copper flat conductor size drift by calculating the size residual and deeply analyzing the time series characteristics of the size residual in continuous frame images, avoiding misjudging it as noise disturbance and causing hysteresis adjustment, thus making size control more timely and accurate. Attached Figure Description

[0059] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein:

[0060] Figure 1 A flowchart of a machine vision-based method for adjusting the size of a copper flat conductor provided in this application;

[0061] Figure 2 A schematic diagram of a machine vision-based copper flat conductor size control system provided for this application;

[0062] Figure 3 This is a schematic diagram of the scene for acquiring images of a copper flat conductor provided in this application.

[0063] The meanings of the main reference numerals in the figure are as follows: 1. Conveying platform; 2. Copper flat conductor; 3. Image acquisition module; 4. Image of copper flat conductor. Detailed Implementation

[0064] The technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the embodiments and specific features in the embodiments are detailed descriptions of the technical solution of this application, rather than limitations thereof. In the absence of conflict, the embodiments and technical features in the embodiments can be combined with each other.

[0065] Example 1

[0066] This embodiment introduces a method for controlling the size of copper flat conductors based on machine vision, referring to... Figure 1 The method includes the following steps:

[0067] Acquire consecutive frame images of the copper flat conductor; identify and extract the edge lines of the copper flat conductor based on each frame image;

[0068] The process of identifying and extracting the edge lines of the copper flat conductor specifically includes: identifying the copper flat conductor region from the image, performing edge detection on the copper flat conductor region to obtain the outline of the copper flat conductor; and extracting the edge lines of the copper flat conductor based on the outline of the copper flat conductor.

[0069] In this embodiment, the upper edge line of the outline of the copper flat conductor is preferably extracted. (Refer to...) Figure 3 Images 4 of the copper flat conductor are all cross-sectional images of the copper flat conductor. The copper flat conductor region is approximately rectangular, and its outline includes four sides: top, bottom, left, and right. The top side is extracted as the edge line. The continuous frame images are continuous frame images containing the copper flat conductor region acquired by the image acquisition module 3 during the transportation process on the conveying platform 1 after the copper flat conductor 2 has been processed and formed; each frame image corresponds to one copper flat conductor.

[0070] Identifying jitter frames in consecutive images based on the edge lines of the copper flat conductor in each frame; specifically including:

[0071] Set sampling points on the edge lines of each frame image;

[0072] For any frame of an image, record the vertical and horizontal coordinates of each sampling point in the image;

[0073] Optionally, the pixel rows and pixel columns of each frame image are numbered, with the bottommost pixel row of the image numbered as row 1, and the numbering increasing sequentially from bottom to top; the vertical coordinate of any sampling point on any edge line is the number of the pixel row where the sampling point is located; the leftmost pixel column of the image is numbered as column 1, and the numbering increasing sequentially from left to right; the horizontal coordinate of any sampling point on any edge line is the number of the pixel column where the sampling point is located.

[0074] Calculate the difference in vertical coordinates between any two adjacent sampling points as the jitter value of the edge line; optionally, when calculating the jitter value, subtract the vertical coordinate of the sampling point with the smaller horizontal coordinate from the vertical coordinate of the sampling point with the larger horizontal coordinate.

[0075] Arrange the jitter values ​​of the edge lines into a jitter value sequence according to the horizontal coordinates of the corresponding sampling points from smallest to largest;

[0076] The jitter amplitude and jitter frequency of the edge line are calculated based on the jitter value sequence. The method for calculating the jitter frequency is as follows: record the positive or negative sign of each jitter value in the jitter value sequence; if any two adjacent jitter values ​​in the jitter value sequence have opposite signs, it is recorded as one jitter count; record the total number of jitter counts in the jitter value sequence, divide it by the number of jitter values ​​in the jitter value sequence, and obtain the jitter frequency; optionally, in this embodiment, the standard deviation of all jitter values ​​in the jitter value sequence is calculated as the jitter amplitude of the corresponding edge line.

[0077] If the jitter amplitude of the edge line is greater than the preset jitter amplitude threshold, or the jitter frequency is greater than the preset jitter frequency threshold, then the image containing the edge line is a jitter frame.

[0078] The method for identifying jittery frames in consecutive frame images further includes: calculating the mean of the vertical coordinates of all sampling points on each edge line as the height value of the corresponding edge line; performing outlier detection on the height values ​​of all edge lines; and defining the image containing the edge line corresponding to any outlier as a jittery frame. Optionally, this embodiment uses Z-score (standard score method) to perform outlier detection on the height values ​​of all edge lines.

[0079] A non-uniform time window is set for each jittery frame, and edge lines of the jittery frames are fused and compensated.

[0080] The method for setting a non-uniform time window for any jittery frame is as follows:

[0081] Extract the N consecutive frames before the jittered frame and the N consecutive frames after the jittered frame, which together with the jittered frame form a non-uniform time window; N is a positive integer;

[0082] For any frame of an image within a non-uniform time window, the jitter amplitude and jitter frequency of the edge lines are normalized and dimensionless, and then weighted and summed to obtain the jitter factor of the corresponding frame image.

[0083] If the jitter factor of any frame image in the non-uniform time window is greater than the preset jitter factor threshold, the corresponding frame image is removed from the non-uniform time window.

[0084] The method for performing edge fusion compensation on any jittery frame is as follows:

[0085] Calculate the acquisition time difference between each frame of the image and the jitter frame in the non-uniform time window;

[0086] The fusion weight of each frame is assigned based on the acquisition time difference between each frame and the jitter frame, and the jitter factor. Both the acquisition time difference and the jitter factor are negatively correlated with the fusion weight. Optionally, when calculating the jitter factor, the sum of the weights of the jitter amplitude and jitter frequency is 1. In this embodiment, the preferred formula for assigning the fusion weight to any frame within a non-uniform time window is as follows:

[0087] w = (1-D)·exp(-α·t);

[0088] Where w is the fusion weight, D is the jitter factor of the corresponding frame image, α is the adjustment coefficient, and its value is set by those skilled in the art based on actual needs; t is the acquisition time difference between the corresponding frame image and the jitter frame. Based on the above formula, the larger the jitter factor and the larger the acquisition time difference with the jitter frame, the lower the fusion weight of the image, and the smaller the impact on the edge line fusion compensation of the jitter frame, thus preventing the fusion compensation from introducing too much interference.

[0089] Based on the fusion weights, the edge lines of each frame in the non-uniform time window are weighted and fused to obtain the fused edge lines of the jittery frame.

[0090] Optionally, the weighted fusion of the edge lines of each frame image is performed as follows: the fusion weights of each frame image are normalized so that the sum of the fusion weights of all frames in the non-uniform time window is 1; each pixel on the edge line of each frame image is numbered in ascending order of horizontal coordinates; each pixel on the edge line of the jittered frame is marked as a target pixel, and all pixels with the same number on the edge lines in the non-uniform time window are extracted as reference pixels of the target pixel; the vertical coordinates of all reference pixels are weighted and summed to obtain the fusion value of the vertical coordinates of the target pixel, where the weight of any reference pixel is the fusion weight of the corresponding frame image; the fusion value of the vertical coordinates of each pixel on the edge line of the jittered frame is used to replace the original value of the vertical coordinates, and each pixel is connected to obtain the fused edge line of the jittered frame.

[0091] The fused edge lines are locally fitted to obtain the edge lines for jitter frame fusion compensation. Optionally, this embodiment uses sliding window smoothing or cubic spline interpolation to achieve local fitting of the fused edge lines, suppressing edge jaggedness and short-term fluctuations, resulting in clear and smooth final edge lines.

[0092] This application identifies jittery frames and uses a non-uniform time window to fuse and compensate the edge lines of these frames. Even when dynamic defects such as jitter and warping exist in the copper flat conductor image, smooth and reliable edge information can still be obtained. Copper flat conductors often experience edge warping and slight jitter during high-speed transport, leading to blurred or discontinuous image edges. Traditional static edge detection algorithms are prone to errors due to this. In contrast, this application can ultimately output stable, low-bias boundary information for size determination, improving the accuracy and stability of subsequent size detection and control.

[0093] The size parameters of the copper flat conductor are extracted based on the edge lines in each frame of the image, and the size residual of the copper flat conductor is calculated.

[0094] The dimensional parameter of the copper flat conductor is either its width or its thickness. By performing fusion compensation on the edge lines of the jittered frame, a clear and flat edge line is obtained, and the width or thickness of the copper flat conductor can be directly extracted. For example, in this embodiment, the upper edge line of the copper flat conductor is extracted, the length of the upper edge line is the width of the copper flat conductor, and the average height difference between the upper edge line and the transmission plane is the thickness of the copper flat conductor.

[0095] The calculation of the dimensional residual of the copper flat conductor specifically includes: subtracting the preset reference dimensional parameter from the dimensional parameter of the copper flat conductor in each frame of the image to obtain the dimensional residual of the copper flat conductor in each frame of the image.

[0096] Based on the dimensional residuals of copper flat conductors in consecutive frame images, the trend drift of copper flat conductor dimensions is identified; specifically including:

[0097] The dimensional residuals of the copper flat conductor in each frame of the image are organized into a time series of dimensional residuals according to the order of acquisition of the corresponding frame images from early to late.

[0098] Based on the time series of the size residuals, a forward time window and a reverse time window of the size residuals are constructed; specifically, this includes: extracting the size residuals of the copper flat conductors from the most recently acquired M consecutive frames of images from the size residual sequence, and forming the reverse time window of the size residual sequence; M is a positive integer;

[0099] The time series of the size residuals is input into a trained time series prediction model, which calculates and outputs predicted values ​​of the copper flat conductor size residuals in M ​​consecutive future images; the predicted values ​​of the copper flat conductor size residuals in the M consecutive future images form a positive time window of the size residuals. Optionally, this embodiment configures a recurrent neural network model as the time series prediction model.

[0100] The evolution trend of the dimensional residuals is extracted from the forward time window and the reverse time window, respectively;

[0101] For any time window in the forward or reverse time window, the evolution trend of the dimensional residuals includes the residual slope and the residual mean; for any time window, the method for extracting the residual slope is as follows: perform curve fitting on the continuous dimensional residuals in the forward or reverse time window, and calculate the average slope of the fitted curve as the residual slope of the corresponding time window.

[0102] For any time window, the method for extracting the mean residual is as follows: take the absolute value of each size residual in the forward or reverse time window and then calculate the average value to obtain the mean residual of the corresponding time window;

[0103] For the positive time window, the evolution trend of the size residuals also includes the residual volatility, which is extracted as follows: calculate the standard deviation of each size residual in the positive time window and divide it by the mean of the residuals in the positive time window to obtain the residual volatility of the positive time window.

[0104] Based on the evolution trend of dimensional residuals in the forward and reverse time windows, the trend drift of copper flat conductor dimensions is identified; specifically including:

[0105] If the residual slopes of both the forward and reverse time windows are positive or both are negative, and the difference between the residual slopes of the forward and reverse time windows is less than a preset slope difference threshold, then the first criterion for trend drift is met; otherwise, there is no trend drift. When the residual slopes of the forward and reverse time windows have the same sign and their values ​​are relatively close, it indicates that the directions of the residual slopes of the two time windows are highly consistent, and the trend of dimensional residual changes can continue from the reverse time window to the forward time window. If the directions of change of the residual slopes of the two windows are opposite or the slope difference is large, then there are irregular disturbances such as short-term fluctuations in the dimensional residuals.

[0106] The ratio of the mean residual value of the forward time window to the mean residual value of the reverse time window is calculated as the amplitude growth rate of the dimensional residual. If the amplitude growth rate is greater than a preset residual growth threshold, and the residual volatility of the forward time window is less than a preset residual volatility threshold, then the second criterion for trend drift is met; otherwise, there is no trend drift. The amplitude growth rate indicates whether the dimensional residual is amplified in the forward time window. A large amplitude growth rate indicates that the dimensional residual is expanding and will not fall back on its own. The residual volatility indicates the degree of instability of the future dimensional residual growth trend. A small residual volatility indicates that the growth of the dimensional residual is relatively stable, i.e., there is trend drift. Conversely, if the amplitude growth rate is small, the dimensional residual does not show a significant amplification trend; if the residual volatility is large, it is a repetitive oscillating dimensional deviation, not trend drift.

[0107] If both the first and second criteria for trend drift are met, then the dimensions of the copper flat conductor exhibit a trend drift.

[0108] During long-term continuous processing, the dimensional parameters of copper flat conductors, such as width and thickness, exhibit a slow, trend-like drift. Existing technologies easily misinterpret this as noise disturbances, leading to hysteresis in adjustment. This embodiment, by continuously tracking the dimensional residuals within reverse and forward time windows, can effectively distinguish between trend-like changes and occasional disturbances in the dimensional residuals, thereby improving the accuracy of dimensional adjustment of copper flat conductors.

[0109] Extract the dimensional error data of the copper flat conductor and input it into the PID controller to suppress the trend of dimensional drift of the copper flat conductor.

[0110] The dimensional error data of the copper flat conductor includes the dimensional residuals of the copper flat conductor in each frame of image acquired from the current moment.

[0111] The method of suppressing the trend drift of copper flat conductors using a PID controller specifically includes: calculating the control quantity for each time step based on the continuous dimensional residuals of the copper flat conductors; and adjusting the processing parameters of the copper flat conductors based on the control quantity for each time step.

[0112] During continuous processing such as rolling, extrusion, or straightening, the dimensional accuracy of copper flat conductors is affected by various processing parameters. For example, the roll gap directly controls the compression ratio of thickness and width; the feeding speed changes the rheological state of the metal and the cooling time, affecting the deformation of the copper flat conductor. A PID controller can output control quantities for any processing parameter. For instance, to adjust the roll gap, the PID controller outputs a control quantity that can be adjusted by an electric adjusting screw or a hydraulic system; to adjust the feeding speed, the PID controller outputs a control quantity that can be adjusted by the speed of a speed-regulating motor.

[0113] Example 2

[0114] This embodiment is the second embodiment of this application; it is based on the same inventive concept as Embodiment 1, and refers to... Figure 2 This embodiment introduces a machine vision-based copper flat conductor size control system, including an image acquisition module, an image recognition module, an image processing module, an error recognition module, and a size control module; wherein:

[0115] The image acquisition module is used to acquire continuous frame images of the copper flat conductor; the image acquisition module is equipped with a camera, which can continuously acquire images during the transportation process of the copper flat conductor.

[0116] The image recognition module identifies and extracts the edge lines of the copper flat conductor based on each frame of the image; the image recognition module is equipped with image recognition algorithm and edge detection algorithm, which can identify the copper flat conductor area in the image and extract the edge lines of the copper flat conductor.

[0117] The image processing module is used to identify jittery frames in continuous image frames and to perform edge line fusion compensation for jittery frames by setting a non-uniform time window. The image processing module calculates the jitter amplitude and jitter frequency of the edge lines in each image frame to identify jittery frames, and performs weighted fusion of the edge lines of each image frame in the non-uniform time window to perform edge line fusion compensation for jittery frames.

[0118] The error identification module calculates the dimensional residual of the copper flat conductor in each frame of the image based on the edge line and identifies the trend of the copper flat conductor size drift. The error identification module establishes a positive time window and a negative time window for the size error, and sets a first criterion and a second criterion for the trend drift based on the evolution trend of the dimensional residual in the two time windows.

[0119] The size control module is equipped with a PID controller to suppress the trend-based drift of the copper flat conductor size. The input of the PID controller is the dimensional residual of the copper flat conductor in multiple consecutive frames of images, and the output is the control quantity for each time step. The copper flat conductor production line adjusts the processing parameters of the copper flat conductor based on the control quantity for each time step.

[0120] The specific functions of each module described above are as described in the relevant content of the machine vision-based copper flat conductor size control method in Example 1, and will not be repeated here.

[0121] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0122] The embodiments of this application have been described above with reference to the accompanying drawings. However, this application is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of this application without departing from the spirit and scope of protection of this application, and these forms are all within the protection scope of this application.

Claims

1. A method for controlling the size of copper flat conductors based on machine vision, characterized in that: Includes the following steps: Acquire consecutive frame images of the copper flat conductor; identify and extract the edge lines of the copper flat conductor based on each frame image; Identify jitter frames in consecutive frames of images based on the edge lines of the copper flat conductor in each frame of the image. A non-uniform time window is set for each jittery frame, and edge lines of the jittery frames are fused and compensated. The size parameters of the copper flat conductor are extracted based on the edge lines in each frame of the image, and the size residual of the copper flat conductor is calculated. Based on the dimensional residuals of copper flat conductors in consecutive frame images, the trend drift of copper flat conductor dimensions is identified; Extract the dimensional error data of the copper flat conductor and input it into the PID controller to suppress the trend of dimensional drift of the copper flat conductor.

2. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 1, characterized in that: The identification of jitter frames in consecutive frame images specifically includes: Set sampling points on the edge lines in each frame of the image; for any frame of the image, record the vertical and horizontal coordinates of each sampling point in the image. Calculate the difference in vertical coordinates between any two adjacent sampling points, and use it as the jitter value of the edge line; Arrange the jitter values ​​of the edge lines into a jitter value sequence according to the horizontal coordinates of the corresponding sampling points from smallest to largest; The jitter amplitude and jitter frequency of the edge line are calculated based on the jitter value sequence; If the jitter amplitude of the edge line is greater than the preset jitter amplitude threshold, or the jitter frequency is greater than the preset jitter frequency threshold, then the image containing the edge line is a jitter frame.

3. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 2, characterized in that: The method for calculating the jitter frequency is as follows: record the positive or negative sign of each jitter value in the jitter value sequence; if any two adjacent jitter values ​​in the jitter value sequence have opposite signs, then it is recorded as one jitter count; Record the total number of jitters in the jitter value sequence, divide it by the number of jitter values ​​in the jitter value sequence, and obtain the jitter frequency; The method for identifying jittery frames in continuous frame images further includes: calculating the mean of the vertical coordinates of all sampling points on each edge line as the height value of the corresponding edge line; performing outlier detection on the height values ​​of all edge lines; and identifying the image containing the edge line corresponding to any outlier as a jittery frame.

4. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 3, characterized in that: The method for setting a non-uniform time window for any jittery frame is as follows: Extract the N consecutive frames before the jittered frame and the N consecutive frames after the jittered frame, which together with the jittered frame form a non-uniform time window; N is a positive integer; For any frame of an image within a non-uniform time window, the jitter amplitude and jitter frequency of the edge lines are normalized and dimensionless, and then weighted and summed to obtain the jitter factor of the corresponding frame image. If the jitter factor of any frame image in the non-uniform time window is greater than the preset jitter factor threshold, the corresponding frame image is removed from the non-uniform time window.

5. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 4, characterized in that: The method for performing edge fusion compensation on any jittery frame is as follows: Calculate the acquisition time difference between each frame of the image and the jitter frame in the non-uniform time window; The fusion weight of each frame image is assigned based on the acquisition time difference between each frame image and the jitter frame and the jitter factor, wherein the acquisition time difference and the jitter factor are both negatively correlated with the fusion weight. Based on the fusion weights, the edge lines of each frame in the non-uniform time window are weighted and fused to obtain the fused edge lines of the jittery frame. Local fitting is performed on the fused edge line to obtain the edge line of the jitter frame after fusion compensation.

6. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 5, characterized in that: The continuous frame images of the copper flat conductor are continuous frame images of the copper flat conductor region collected during the transportation process after the copper flat conductor is processed and formed; each frame image corresponds to one copper flat conductor. The process of identifying and extracting the edge lines of the copper flat conductor specifically includes: identifying the copper flat conductor region from the image, performing edge detection on the copper flat conductor region to obtain the outline of the copper flat conductor; and extracting the edge lines of the copper flat conductor based on the outline of the copper flat conductor. The size parameter of the copper flat conductor is either its width or its thickness; the calculation of the size residual of the copper flat conductor specifically includes: subtracting the preset reference size parameter from the size parameter of the copper flat conductor in each frame image to obtain the size residual of the copper flat conductor in each frame image; The dimensional error data of the copper flat conductor includes the dimensional residuals of the copper flat conductor in each frame of the image acquired from the current moment.

7. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 6, characterized in that: Based on the dimensional residuals of copper flat conductors in consecutive frame images, the trend drift of copper flat conductor dimensions is identified, specifically including: The dimensional residuals of the copper flat conductor in each frame of the image are organized into a time series of dimensional residuals according to the order of acquisition of the corresponding frame images from early to late. Based on the time series of the size residuals, a forward time window and a reverse time window for the size residuals are constructed. Specifically, this includes: extracting the size residuals of the copper flat conductor from the most recently acquired M consecutive frames of images from the size residual sequence, and forming the reverse time window of the size residual sequence; M is a positive integer; inputting the time series of the size residuals into a trained time series prediction model to obtain the predicted values ​​of the copper flat conductor size residuals in the next M consecutive frames of images, and forming the forward time window of the size residuals. The evolution trend of the dimensional residuals is extracted from the forward time window and the reverse time window, respectively; Based on the evolution trend of dimensional residuals in the forward and reverse time windows, the trend drift of copper flat conductor dimensions is identified.

8. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 7, characterized in that: For any time window in the forward or reverse time window, the evolution trend of the dimensional residuals includes the residual slope and the residual mean; for any time window, the method for extracting the residual slope is as follows: perform curve fitting on the continuous dimensional residuals in the forward or reverse time window, and calculate the average slope of the fitted curve as the residual slope of the corresponding time window. For any time window, the method for extracting the mean residual is as follows: take the absolute value of each size residual in the forward or reverse time window and then calculate the average value to obtain the mean residual of the corresponding time window; For the positive time window, the evolution trend of the size residuals also includes the residual volatility, which is extracted as follows: calculate the standard deviation of each size residual in the positive time window and divide it by the mean of the residuals in the positive time window to obtain the residual volatility of the positive time window.

9. The method for adjusting the size of copper flat conductors based on machine vision as described in claim 8, characterized in that: The identification of the trend drift in the size of the copper flat conductor includes: If the residual slopes of both the forward and reverse time windows are positive or both are negative, and the difference between the residual slopes of the forward and reverse time windows is less than a preset slope difference threshold, then the first criterion for trend drift is met. The ratio of the mean residual of the forward time window to the mean residual of the reverse time window is calculated as the amplitude growth rate of the size residual; if the amplitude growth rate is greater than a preset residual growth threshold and the residual volatility of the forward time window is less than a preset residual volatility threshold, then the second criterion of trend drift is satisfied. If both the first and second criteria for trend drift are met, then the dimensions of the copper flat conductor exhibit a trend drift.

10. A machine vision-based copper flat conductor size control system, used to implement the machine vision-based copper flat conductor size control method as described in any one of claims 1-9, characterized in that: It includes an image acquisition module, an image recognition module, an image processing module, an error recognition module, and a size control module; among which: The image acquisition module is used to acquire continuous frame images of the copper flat conductor; The image recognition module identifies and extracts the edge lines of the copper flat conductor based on each frame of the image; The image processing module is used to identify jittery frames in continuous frame images and to perform edge fusion compensation on the jittery frames by setting a non-uniform time window; The error identification module calculates the dimensional residual of the copper flat conductor in each frame of the image based on the edge line, and identifies the trend of copper flat conductor size drift. The size control module is equipped with a PID controller to suppress the tendency of copper flat conductor size drift.

Images