A machine vision-based tail-end recognition system and method for finished strip steel.

Abstract

This invention discloses a machine vision-based strip tail recognition system and method, belonging to the field of strip rolling technology. The method includes: acquiring and preprocessing an image of the strip tail; inputting the preprocessed image into a preset deep learning model for classification, outputting the strip tail shape type; extracting strip tail feature parameters based on the preprocessed strip tail image; calculating the required roll gap leveling amount for correcting the asymmetric shape of the strip tail according to the strip tail shape type and the strip tail feature parameters; and sending the roll gap leveling amount to the finishing mill control system to achieve real-time control of strip tail deviation. Using the technical solution of this invention, automatic, accurate, and real-time recognition and quantification of the strip tail planar shape can be achieved, and an effective control strategy can be generated accordingly, thereby suppressing tail deviation, improving tail shape, and enhancing rolling stability and product yield.

Description

Technical Field

[0001] This invention relates to the field of strip rolling technology, and in particular to a machine vision-based tail-end recognition system and method for precision rolled strip steel. Background Technology

[0002] Sheet and strip products, as important industrial raw materials, are widely used in many key fields such as automobile manufacturing, electronics, precision instruments, packaging and transportation, playing an important supporting role in industrial development and economic construction.

[0003] Currently, insufficient control over sheet shape quality, especially asymmetric sheet shape, has become a significant bottleneck restricting the improvement of high-end sheet product quality. In hot continuous rolling production, sheet shape control during the finishing rolling stage is particularly critical, directly affecting the dimensional accuracy, mechanical properties, and surface quality of the final product. Although the control technology for symmetrical sheet shapes (such as center ripples and edge ripples) is relatively mature, the online detection and control of asymmetric sheet shapes (such as deviation and camber) still have significant shortcomings.

[0004] In finishing mills, strip misalignment is a common asymmetric shape defect, mainly caused by uneven deformation of the strip in the width direction, misalignment of the roll system, and wedge-shaped incoming material during the rolling process. Especially when rolling thin-gauge products, the misalignment phenomenon is more pronounced after the strip tail breaks free from the mill tension constraint, easily inducing production accidents such as tail-wagging and steel piling between stands. This not only leads to a decrease in product yield and surface damage but can also damage equipment such as rolls and guides, seriously affecting production stability and continuity.

[0005] Currently, there is a lack of online monitoring methods for strip deviation within finishing mills. Most existing systems can only obtain strip centerline offset information at the finishing mill exit using a width gauge, failing to achieve real-time, accurate detection and early warning of deviation between stands. Operators often rely on experience for intervention, resulting in delayed and insufficient precision. Therefore, developing a system capable of real-time and accurate detection of strip deviation between finishing mill stands is of great significance for improving asymmetric shape control, ensuring rolling stability, and enhancing the quality of high-end plate products. Summary of the Invention

[0006] This invention provides a machine vision-based tail-end recognition system and method for precision rolled strip steel, to solve the technical problems of existing technologies that cannot achieve real-time and accurate detection and early warning of deviation between stands; and that operators often rely on experience to intervene, resulting in lagging control and insufficient accuracy.

[0007] To solve the above-mentioned technical problems, the present invention provides the following technical solution: On the one hand, the present invention also provides a machine vision-based method for identifying the tail end of finished strip steel, comprising: Acquire images of the tail section of the strip and perform image preprocessing on the acquired tail section images; The preprocessed strip tail image is input into a preset deep learning model, and the deep learning model is used to classify the preprocessed strip tail image and output the strip tail shape type. Based on the preprocessed strip tail image, feature parameters of the strip tail are extracted; Based on the strip tail shape type and strip tail characteristic parameters, calculate the roll gap leveling amount of the frame required to correct the asymmetric shape of the strip tail; The roll gap leveling amount is sent to the finishing mill control system to achieve real-time control of the strip tail deviation.

[0008] Further, the image preprocessing of the acquired strip tail image includes: The images of the same strip tail taken by multiple cameras are stitched together to form a complete image of the strip tail. A preset filtering algorithm is used to filter the stitched strip tail image to eliminate noise interference; wherein, the preset filtering algorithm is a bilateral filtering algorithm or a guided filtering algorithm. Missing pixels in the filtered strip tail image are filled using bicubic interpolation. For images with missing pixels, a Laplacian pyramid fusion algorithm is used to eliminate brightness differences.

[0009] Furthermore, the deep learning model is the EfficientNetV2 model, which incorporates the CBAM attention mechanism and Droppath regularization.

[0010] Furthermore, the strip tail shape types include rectangular, operating side single blade, drive side single blade, tongue shape, and dovetail shape.

[0011] Furthermore, the characteristic parameters of the strip tail include common parameters and special shape parameters; among them, Common parameters include: tail length, tail area, tail perimeter, the ratio of the length to the width of the smallest bounding rectangle, and the ratio of the perimeter to the area of ​​the inscribed polygon. For strip steel that is determined to be either an operating-side single blade or a transmission-side single blade, its special shape parameters include: single blade vertex coordinates, single blade cutting edge starting position, and single blade length; For strip steel that is determined to be tongue-shaped, its special shape parameters include: tongue vertex coordinates, tongue starting position, tongue length, and tongue arc similarity. For strip steel that is determined to be dovetail-shaped, its special shape parameters include: dovetail concave point coordinates, dovetail vertex coordinates, dovetail slope, dovetail fork angle, and dovetail length.

[0012] Furthermore, based on the preprocessed strip tail image, feature parameters of the strip tail are extracted, including: Based on the preprocessed strip tail image, the feature parameters of the strip tail are extracted using either an adaptive threshold corner detection algorithm based on edge detection or a sub-pixel corner detection algorithm based on grayscale intensity.

[0013] Furthermore, the step of calculating the required roll gap leveling amount for correcting the asymmetric shape of the strip tail based on the strip tail shape type and strip tail characteristic parameters includes: The tail-end characteristic parameters of the strip are quantified as tail-end asymmetry. Among them, for strip steel that is determined to be either an operating-side single-blade or a transmission-side single-blade, = ;in, The length of a single blade; for strip steel judged to be tongue-shaped, ;in, The length of the tongue; For tongue-shaped arc similarity; for strip steel judged to be dovetail-shaped, ;in, The angle of the swallowtail; The length of the swallowtail; It has a swallowtail-shaped forked corner; Based on the tail asymmetry, calculate the roll gap leveling amount for the current frame: ; in, For the first Roll gap leveling amount for each frame; The width of the strip; The incoming material is wedge-shaped; For export thickness; Poor rolling force; The difference in stiffness between the two sides of the rolling mill; and These are the preset model weight coefficients obtained through regression of historical rolling data; The average stiffness of the rolling mill; Based on the position of the tail section, the roll gap leveling amount of the current frame is multiplied by a preset proportional coefficient to obtain the roll gap leveling amount of the subsequent frames affected after the current frame.

[0014] Furthermore, the formula for calculating the proportionality coefficient is as follows: ; in, Indicates by the first i The roll gap leveling amount of the first frame is obtained. i The proportional coefficient required for roller gap leveling of +1 frame; For the firsti rack exit speed; For strip steel from the first The rack is running to the... Time per rack; For the first rack and the first Distance between racks This is the deformation attenuation coefficient.

[0015] Furthermore, the machine vision-based tail-end recognition method for finished strip steel also includes: Establish a mapping relationship between the strip tail shape parameters and rolling process parameters as a verification indicator: ; in, To verify the indicators; These are the weighting coefficients; For the selected rolling process parameters; For bias terms; The number of rolling process parameters involved in the calculation; When the validation metric exceeds the threshold range, adaptive adjustment of model parameters is initiated; the threshold is dynamically adjusted based on historical data, and its calculation formula is as follows: ; in, For the first Dynamic warning thresholds at any given time; Basic threshold; For smoothing coefficients; This is historical error data.

[0016] Furthermore, the machine vision-based tail-end recognition method for finished strip steel also includes: An alert is issued when a preset abnormal situation is detected; The preset abnormal conditions include: a preset number of consecutive images with substandard quality, a classification confidence level lower than a preset confidence threshold, and the actual deviation from the target value exceeding a preset allowable range.

[0017] In another aspect, the present invention also provides an electronic device comprising a processor and a memory; wherein the memory stores at least one instruction, which is loaded and executed by the processor to implement the above-described method.

[0018] In another aspect, the present invention also provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the above method.

[0019] The beneficial effects of the technical solution provided by this invention include at least the following: 1. This invention innovatively extends machine vision inspection from the conventional measurement of strip body deviation to the online recognition and quantification of complex planar shapes at the tail of the strip, filling a technological gap in this field and providing precise input for closed-loop control of tail shape.

[0020] 2. A deep learning model is used to intelligently classify tail shapes, overcoming the problem that traditional image processing algorithms are insufficient in recognizing irregular and blurry tail contours, and significantly improving the accuracy and robustness of classification.

[0021] 3. This invention establishes a complete chain solution from tail shape recognition and feature parameter extraction to roll gap control calculation, realizing predictive feedforward control based on tail shape, which can effectively suppress deviation and tail swing caused by irregular tail shape, and improve rolling stability and yield.

[0022] 4. While ensuring the advantages of high sampling rate and non-contact detection, the system of the present invention has a high degree of modularity, is easy to modify and integrate into existing production lines, and has strong practicality. Attached Figure Description

[0023] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0024] Figure 1 This is a structural diagram of the machine vision-based tail recognition system for precision rolled strip provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the execution flow of the machine vision-based tail recognition method for finished strip steel provided in an embodiment of the present invention; Figure 3 This is a system block diagram of the electronic device provided in the embodiments of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be described in further detail below with reference to the accompanying drawings.

[0026] First, it should be noted that in the embodiments of the present invention, the words "exemplarily," "for example," etc., are used to indicate that they are examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the term "exemplarily" is intended to present the concept in a specific manner. Furthermore, in the embodiments of the present invention, the meaning expressed by "and / or" can be both, or it can be either one or the other.

[0027] First Embodiment

[0028] To address the limitations of existing technologies in achieving real-time, accurate detection and early warning of strip misalignment between stands, and the reliance on operator experience for intervention resulting in delayed and inaccurate control, this embodiment provides a machine vision-based strip tail identification system. This system is used to implement a machine vision-based strip tail identification method, aiming to detect and control the planar shape of the strip tail during the finishing rolling process in real time and accurately, thereby resolving asymmetric strip shape defects such as misalignment and tail-wagging caused by irregular tail shapes.

[0029] like Figure 1 As shown, the system includes: a camera mounting bracket 1, a camera adjustment device, a water cooling device 5, an air sealing device 6, a servo motor 7, two line scan cameras 8 and lenses 9, a photoelectric converter 10, a fiber optic hub 11, an image processing server 12, a KVM extender 13, and a terminal display 14; among which, The camera mounting bracket 1 is fixedly installed between the finishing mill stands or above the finishing mill stand archway to support the line array camera 8, lens 9, camera adjustment device, water cooling device 5, and air sealing device 6. The camera mounting bracket 1 is a frame-type support structure and is fixed to the foundation of the on-site rolling mill equipment or the stand archway by bolts to ensure that the image acquisition equipment has a stable installation position in environments with high temperature, high humidity, vibration, and iron oxide scale splashing.

[0030] The camera adjustment device is fixedly mounted on the camera mounting bracket 1 and is used to adjust the spatial pose of the line scan camera 8 and the lens 9. Specifically, the camera adjustment device includes a front-to-back pose adjustment device 2, a pitch pose adjustment device 3, and a rotation pose adjustment device 4. Each adjustment device is driven by a servo motor 7 and is used to adjust the front-to-back position, pitch angle, and rotation angle of the line scan camera 8 so that the field of view of the line scan camera 8 can cover the running area of ​​the strip tail.

[0031] Two line-scan cameras 8 and lenses 9 are mounted on camera mounting bracket 1 via camera adjustment devices and positioned between two finishing mill stands. The line-scan cameras 8 use the current stand's strip ejection signal or the signal indicating the strip tail has reached the detection area as trigger signals to acquire image information of the strip tail as it disengages from the mill constraint at an angle downwards. The two line-scan cameras 8 respectively acquire local images of different areas of the strip tail, and these images are then stitched together to form a complete image of the strip tail, used for strip tail shape recognition and feature parameter extraction.

[0032] A water-cooling device 5 is installed around the periphery of the line scan camera 8 and lens 9 to cool and protect the image acquisition equipment, preventing the camera from overheating due to the high-temperature rolling environment. An air-sealing device 6 is installed at the front end of the water-cooling device 5 or on the outside of the lens 9 to continuously supply compressed gas to the front end of the lens, reducing contamination of the lens surface by water mist, dust, and iron oxide scale, thus ensuring the clarity of the acquired images. The power cables and signal cables of the line scan camera 8, as well as the pipelines of the water-cooling device 5 and the air-sealing device 6, are all arranged along the camera mounting bracket 1. The power cables, signal cables, water pipes, and air pipes are all wrapped with protective tubing to prevent damage to the lines and pipelines from high temperatures, mechanical impacts, and on-site contamination.

[0033] The image of the strip tail acquired by the line scan camera 8 is first transmitted via gigabit Ethernet to the photoelectric converter 10 installed at the rolling mill site, where the photoelectric converter 10 converts the electrical signal into an optical signal. Subsequently, the image signal is transmitted via optical fiber to the optical fiber hub 11, and then transmitted by the optical fiber hub 11 to the photoelectric converter 10 on the image processing server 12 side, and finally transmitted to the image processing server 12 via gigabit Ethernet.

[0034] Image processing server 12 is used to perform image preprocessing, tail shape classification, feature parameter extraction, and roll gap leveling calculation on the received strip tail image. Specifically, image processing server 12 stitches, filters, completes pixels, and fuses brightness in local images acquired by multiple line scan cameras 8 to obtain a complete strip tail image; then, the preprocessed strip tail image is input into a preset deep learning model to identify the strip tail shape type; then, based on the identification results, the corresponding tail feature parameters are extracted, and combined with rolling process parameters, the roll gap leveling amount used to suppress strip tail deviation or tail swing is calculated.

[0035] The image processing server 12 is also connected to the basic automated control system of the finishing mill to send the calculated roll gap leveling amount to the corresponding finishing mill stand, thereby realizing real-time control of the deviation trend caused by the asymmetrical shape of the strip tail. At the same time, the image processing server 12 is connected to the terminal display 14 through the KVM extender 13. The terminal display 14 is used to display the strip tail shape type, tail characteristic parameters, real-time deviation trend, roll gap leveling suggestion amount, system operating status, and early warning information.

[0036] In this embodiment, the line scan camera 8 can be a line scan CMOS camera. Upon receiving a signal indicating that the current stand has thrown steel or that the tail of the strip has reached the detection area, the line scan camera 8 begins acquiring images and continuously acquires images of the strip tail at a preset acquisition frequency. For each acquired image frame, the line scan camera 8 transmits the image data in real time to the image processing server 12 via Gigabit Ethernet, a photoelectric converter 10, a fiber optic hub 11, and a fiber optic transmission line. After completing image processing and data calculation, the image processing server 12 displays the recognition results and control calculation results on the terminal display 14 via a KVM extender 13, and sends the calculated roll gap leveling amount to the finishing mill basic automation control system.

[0037] The machine vision-based tail-end recognition method for finished strip steel implemented based on the above system is as follows: Figure 2 As shown, it includes: S1, acquire the image of the tail of the strip and perform image preprocessing on the acquired tail image of the strip; Specifically, in this embodiment, when the finishing mill stand detects a strip ejection signal, the image acquisition unit is triggered to start working, continuously acquiring image sequences of the strip tail leaving the roll constraint until the tail completely leaves the detection area. The acquired images are transmitted in real time to the image processing server via a fiber optic network. The image processing server preprocesses the received image sequences, including image denoising, contrast enhancement, and image stitching, to form a clear panoramic image of the strip tail. The specific preprocessing process is as follows: S11, Image stitching process: stitching together partial images of the same strip steel taken by multiple cameras into a complete image; The specific process of image stitching is as follows: S111, Traverse the linear strip image folder named after the coil number; S112, Create an image-free rectangular window with the required image size after stitching, based on the size of the single linear array image being read; S113, fill the pixel values ​​in the rectangular window sequentially according to the image reading order; S114, loop through each folder to stitch together the images of each roll of steel.

[0038] S12, Image filtering: Use bilateral filtering or guided filtering methods to eliminate noise interference; S13, Image registration and data interpolation: Use bicubic interpolation to fill in missing pixels; S14, Image Fusion: The Laplacian pyramid fusion algorithm is used to eliminate brightness differences.

[0039] After the above preprocessing, a complete two-dimensional image of the tail section of the strip is finally obtained.

[0040] S2, input the preprocessed strip tail image into a preset deep learning model, use the deep learning model to classify the preprocessed strip tail image, and output the strip tail shape type; Specifically, in this embodiment, after preprocessing such as image stitching, filtering, pixel completion, and brightness fusion, the preprocessed strip tail image is input into the trained C-EfficientNetV2 classification model. The C-EfficientNetV2 classification model uses the EfficientNetV2 network as its basic feature extraction network, and introduces a Convolutional Block Attention (CBAM) module in the feature extraction stage of the EfficientNetV2 network. Simultaneously, a DropPath regularization strategy is introduced in the network branches with residual connections to enhance the model's ability to recognize key shape features of the strip tail, such as local contours, unilateral extensions, concave bifurcation, and central convexity.

[0041] Furthermore, the construction process of the C-EfficientNetV2 classification model includes the following steps: S21, acquire image samples of the strip tail section and label the image samples according to their categories. The category labeling includes five categories: rectangular tail section, operating side single-blade tail section, transmission side single-blade tail section, dovetail tail section, and tongue-shaped tail section; S22, the strip tail image sample is preprocessed by stitching together the local strip tail images acquired by the line scan camera, and then filtering, registering, interpolating and fusing the stitched image to obtain a complete strip tail image; then the complete strip tail image is scaled proportionally according to a preset size and grayscale normalization is performed to form a standardized image sample that can be input into a deep learning model. S23, using the EfficientNetV2 network as the backbone network, and utilizing its convolutional layers and feature extraction modules to perform multi-scale feature extraction on standardized image samples, to obtain the basic feature map F of the strip tail image; S24, the basic feature map F is input into the CBAM attention module. The CBAM attention module includes a channel attention module and a spatial attention module; wherein, the channel attention module is used to generate channel attention weights based on the contribution of different feature channels to the tail shape classification result, and the spatial attention module is used to generate spatial attention weights based on the positional distribution of tail contour edges, corners, concave points, and convex areas in the image; the channel attention weights and spatial attention weights are applied to the basic feature map F in sequence to obtain the enhanced tail shape feature map F′; S25 introduces a DropPath regularization strategy in the feature extraction branch with residual connections in the EfficientNetV2 network. During model training, some residual branches are randomly dropped according to a preset path drop probability, allowing the model to learn different feature transfer paths in different training batches. During model inference, the random path drop operation is disabled, preserving the complete network structure for strip tail shape classification. This regularization strategy reduces the risk of overfitting to local noise, brightness variations, and small sample differences, and improves the model's generalization ability to strip tail images under different rolling conditions. S26, the enhanced tail shape feature map F′ is input into the global average pooling layer and the fully connected classification layer, and the classification probability corresponding to each tail shape category is output through the Softmax function; the category with the highest classification probability is selected as the shape recognition result of the current strip tail image.

[0042] Furthermore, the C-EfficientNetV2 classification model employs transfer learning for parameter initialization during training. Specifically, the EfficientNetV2 network parameters pre-trained on a general image dataset are first used as the initial parameters of the backbone network. Then, the strip tail image samples are divided into training and test sets. The training set is used to fine-tune the model, and the test set is used to validate the model's classification performance. During training, the strip tail image is used as input, and the corresponding tail shape category label is used as output. The cross-entropy loss function is used to calculate the error between the predicted category and the true category, and the network parameters are updated through backpropagation.

[0043] Furthermore, the strip tail shape types include rectangular, single-blade on the operating side, single-blade on the transmission side, tongue-shaped, and dovetail-shaped. Specifically, a rectangular tail indicates that the strip tail extends relatively evenly on both sides; a single-blade tail on the operating side and a single-blade tail on the transmission side indicate that the strip tail forms a single-sided oblique cut towards the operating side or transmission side, respectively; a tongue-shaped tail indicates that the central area of ​​the strip tail protrudes outwards; and a dovetail-shaped tail indicates that the central part of the strip tail is concave and the two sides extend in a forked manner.

[0044] S3, Based on the preprocessed strip tail image, extract the strip tail feature parameters; Specifically, in this embodiment, tail feature parameters are extracted based on the classification results using an edge contour algorithm; and these parameters are then converted into physical dimensions using camera calibration parameters. For all types, the following common parameters are extracted: Tail length The distance between the starting position and the end position of the strip tail; Tail pixel area The area encompassed by the outer contour of the strip tail edge is calculated using the following formula: ; Tail pixel perimeter The sum of all pixels on the edge contour of the steel strip tail; Minimum circumscribed rectangle length ,Width ; The ratio of the length to the width of the smallest bounding rectangle ; Shape factor, the ratio of the perimeter to the area of ​​the equivalent inscribed polygon. .

[0045] For specific shapes, extract specific parameters: Single-blade shape: The Harris sub-pixel corner detection algorithm is used to locate the vertex of the single blade. and the starting point of the cutting edge Calculate the single-blade pixel length .

[0046] Tongue shape: Locate the apex and starting position of the tongue shape, and calculate the length of the tongue shape. The similarity between the curve and the standard circular arc is calculated by fitting the contour curve. .

[0047] Swallowtail shape: Locate the concave point and two vertices of the swallowtail, and calculate the slope of the swallowtail. Swallowtail fork angle θ and swallowtail length The formula for calculating the fork angle is: .

[0048] Convert pixel coordinates to actual physical dimensions. The length direction is determined by the strip speed. (m / s) and sampling frequency (Hz) conversion: The width direction utilizes a preset pixel-to-millimeter ratio. Conversion: In this embodiment, =12m / s, =1000Hz, =0.1mm / pixel, the measurement of a certain swallowtail-shaped tail. =450 pixels, then the actual swallowtail slope =450×0.1=45mm.

[0049] Among them, the feature parameter extraction adopts an adaptive threshold corner detection algorithm based on edge detection and a sub-pixel corner detection algorithm based on gray intensity; The corner measure formula for the adaptive threshold corner detection algorithm based on edge detection is:

[0050] Sub-pixel corner detection based on grayscale intensity uses the Harris corner response function:

[0051] in, , It is an empirical constant (0.04-0.06).

[0052] Furthermore, edge detection employs a strip tail contour detection method combining an adaptive threshold Canny operator and sub-pixel contour localization. This method, based on the traditional Canny edge detection process, considers the characteristics of the strip tail image, such as uneven brightness, local edge breaks, iron oxide scale interference, water mist interference, and roller conveyor background interference. It improves the threshold selection method, effective detection area, edge connectivity, and edge localization accuracy to enhance the stability and accuracy of strip tail contour extraction.

[0053] Specifically, the strip tail contour detection method based on the combination of adaptive threshold Canny operator and sub-pixel contour localization includes: 1) Image grayscale conversion and filtering: The preprocessed strip tail image is converted into a grayscale image, and Gaussian filtering is used to smooth the grayscale image to suppress the influence of water mist, iron oxide scale, roller background reflection and image transmission noise on the edge detection results. 2) Determination of effective detection area: Based on the width direction position of the strip in the image and the area where the tail is located, the region of interest of the strip tail is determined, and edge detection is performed only in the region of interest to reduce the interference of the roller background, cooling water area and invalid image boundary on edge extraction; 3) Gradient magnitude and gradient direction calculation: The gradient operator is used to calculate the gradient magnitude and gradient direction of each pixel in the region of interest to obtain candidate pixels at the edge of the strip tail. 4) Non-maximum suppression: Non-maximum suppression is performed on candidate edge pixels along the gradient direction to retain the pixels with the largest local gradient magnitude, thus obtaining the refined candidate edges; 5) Adaptive Dual Threshold Determination: Instead of using fixed thresholds, high and low thresholds are adaptively determined based on the gradient amplitude distribution of the current strip tail image. Specifically, the high threshold is determined based on the mean, median, or preset quantile of the gradient amplitude within the region of interest, and the low threshold is determined according to a preset ratio. When the overall image brightness is low or local brightness is uneven, the high and low thresholds are corrected based on grayscale statistics to improve the stability of edge extraction under different rolling conditions. 6) Edge connection and noise removal: Connect candidate edges according to high and low thresholds, and retain weak edge points connected to strong edges; then perform connected component filtering on the edge image to remove isolated noise edges with an area smaller than the preset threshold or a length smaller than the preset threshold, and connect locally broken tail contour edges through morphological closing operation to obtain a continuous outer contour of the strip tail. 7) Subpixel contour localization: Based on the pixel-level tail contour, a bilinear interpolation method is used to perform subpixel-level correction on the edge position. Specifically, based on the gray values ​​of adjacent pixels in the neighborhood of the edge point and their distance weights, the subpixel position of the edge point between adjacent pixels is calculated, thereby improving the positioning accuracy of the strip tail contour, corner points, and inflection points; 8) Tail contour output: Extract the tail edge, corner point, concave point and convex point of the strip steel from the continuous outer contour point set, and calculate the corresponding tail length, tail area, tail perimeter, single blade length, tongue length, dovetail slope, dovetail fork angle and dovetail length and other feature parameters based on the tail shape classification results of the strip steel.

[0054] Through the above processing, the method improves the stability of strip tail edge extraction under different brightness conditions by using adaptive thresholds. On the other hand, it reduces false detections caused by on-site noise by limiting the region of interest, filtering connected components and morphological correction. Furthermore, it improves the extraction accuracy of edge points, corner points and inflection points by sub-pixel contour localization.

[0055] S4. Based on the strip tail shape type and strip tail characteristic parameters, calculate the roll gap leveling amount of the frame required to correct the asymmetric shape of the strip tail. Specifically, in this embodiment, based on the tail type and characteristic parameters, and combined with the rolling process model, a roll gap leveling control model is established to calculate the downstream stand roll gap leveling amount required to correct the asymmetric shape of the tail. The roll gap leveling control model is established based on the principle of constant volume and the rolling mechanism. Taking into account factors such as tail asymmetry, incoming material wedge shape, and mill stiffness difference, it calculates the downstream stand roll gap adjustment required to correct asymmetric deformation. According to the principle of constant volume, the relationship between the inlet slab wedge shape and the outlet slab wedge shape is as follows:

[0056] in, , These represent the thicknesses of the inlet slab on the operating side and the transmission side, respectively. , These represent the lengths of the operating side and drive side of the exit slab, respectively. Asymmetry at the tail end of the strip. With radius of curvature The relationship is:

[0057] Roll gap adjustment The calculation model is as follows:

[0058] in, To eliminate the adjustment required for the incoming material wedge shape, The adjustment required to eliminate material deviation. The adjustment amount required to eliminate the difference in mill stiffness.

[0059] The specific formula for calculating the roll gap adjustment is as follows: Operational adjustment amount:

[0060] in,

[0061] Drive-side adjustment amount:

[0062] in, The plasticity coefficient of the rolling mill. For the rigidity of the rolling mill, The rolling force is poor.

[0063] First, the tail feature parameters are quantified into tail asymmetry L. a As the main input to the control model, it is defined as the offset of the tail geometry center relative to the theoretical centerline of the strip, or determined by specific shape parameters (such as single-blade length). , dovetail slope ) was calculated; For single-blade type: = ; Regarding tongue shape: ; For swallowtail shape: ; Establish roll gap leveling amount The computational model.

[0064] This model is based on the principles of rolling force balance and metal plastic flow, and takes into account tail asymmetry. Incoming material wedge Difference in stiffness between the two sides of the rolling mill Poor rolling force strip width and export thickness Factors such as:

[0065] in, For the first Difference in roller gap leveling amount between the operating side and the drive side of the machine frame (unit: mm). Tail asymmetry; The width of the strip; The incoming material is wedge-shaped; For export thickness; Poor rolling force; The difference in stiffness between the two sides of the rolling mill; and The preset model weight coefficients, which are related to deformation resistance and friction conditions, are obtained by regression analysis using historical rolling data. This represents the average stiffness of the rolling mill.

[0066] Based on the location of the tail section (such as the F1 frame throwing steel), the calculated total leveling amount Dynamically assign weights to affected subsequent racks (such as racks F3 to F7) according to a preset control strategy. The determination is made considering factors such as frame spacing, the principle of equal strip flow rate per second, and deformation and penetration, i.e., the first... The actual leveling command for the frame is The multi-rack collaborative control strategy is as follows: The F3 stand is controlled based on the F1 steel throwing signal, and the deviation and wedge shape of the strip tail at the F3 stand inlet are calculated. The F4 frame and the F3 frame are adjusted together, and the deviation feedforward roll gap adjustment amount of the F4 frame is proportionally distributed to the F5 frame. The F5 frame and the F4 frame are adjusted together, and the deviation feeder roll gap adjustment amount of the F5 frame is proportionally distributed to the F6 frame. The F6 frame and the F5 frame are adjusted together, and the deviation feeder roll gap adjustment of the F6 frame is proportionally distributed to the F7 frame. The F7 frame is adjusted in conjunction with the F4, F5, and F6 frames to calculate the final roll gap adjustment.

[0067] The proportionality coefficient is determined based on the inter-stand distance, strip speed, and deformation characteristics. The specific calculation formula is as follows:

[0068] in, For the first i rack exit speed, For strip steel from rack running to rack time, This refers to the distance between racks. The deformation attenuation coefficient characterizes the degree of attenuation of asymmetric deformation at the tail of the strip during transmission between adjacent stands, and its value ranges from 0 to 1. This deformation attenuation coefficient is determined based on historical rolling data under the same steel grade, specification, or similar rolling procedures. Specifically, it is obtained by statistically analyzing the changes in tail deviation, tail asymmetry, or roll gap leveling response between adjacent stands, using least squares regression, sliding window statistics, or on-site calibration methods. When historical data is insufficient, an initial value can be set based on offline simulation results or on-site experience, and corrected during production based on the actual deviation control effect of subsequent stands.

[0069] Taking the F3 rack as an example, when the F1 rack throws steel and detects the dovetail tail, the input parameters are: =52mm (from) =45mm and other parameters were calculated). =1250mm, =12mm, =0.5mm, =200kN, =50kN / mm, =5000kN / mm. Substitute into the formula to calculate.

[0070] A multi-rack collaborative control strategy is adopted. ΔS3 is used as the primary leveling variable for rack F3 and distributed to subsequent racks according to the feedforward ratio. =0.7 , =0.5 , =0.3 , =0.2 .like A positive value indicates that the operating side roll gap should be increased by | | / 2, Reduced roller gap on the drive side| | / 2.

[0071] S5 sends the roll gap leveling amount to the finishing mill control system to achieve real-time control of strip tail deviation.

[0072] Specifically, the image processing server will calculate the roll gap leveling instructions for each frame ( to The data is transmitted in real time to the finishing mill's basic automation system via industrial Ethernet. Simultaneously, the following information is displayed and updated in real time on the terminal monitor: The classification results of the tail shape of the strip (e.g., dovetail shape); Tail feature parameters (e.g.: =45mm, =85°); Real-time calculation of the roller gap leveling amount for each frame; Real-time deviation trend curve of strip steel.

[0073] Furthermore, the method in this embodiment also includes a real-time data verification mechanism. A mapping relationship is established between the strip tail shape parameters and rolling process parameters:

[0074] in, The validation metric is used to characterize the degree of matching between the current strip tail shape recognition result and the rolling conditions; The number of rolling process parameters involved in the verification calculation; For the normalized first Each process parameter (rolling force, temperature, speed, etc.); For the first The weighting coefficients corresponding to each rolling process parameter; This is a bias term. The rolling process parameters include one or more of the following: rolling force, rolling force difference, strip temperature, stand exit speed, inlet thickness, outlet thickness, incoming material wedge shape, roll gap leveling amount, and strip tension. The weighting coefficients are... and bias terms Determined based on regression analysis of historical rolling data.

[0075] When verifying metrics If the value exceeds the preset threshold range, it indicates that there is a deviation between the current identification result, feature parameter extraction result, or control calculation result and the actual rolling condition. At this time, the adaptive adjustment of model parameters will be initiated.

[0076] In addition, the system in this embodiment also includes fault diagnosis and early warning functions. When a preset number of consecutive images are found to be substandard, the classification confidence level is lower than a preset confidence threshold, or the deviation between the actual deviation and the target deviation exceeds the allowable range, the system issues an early warning.

[0077] Furthermore, the warning threshold is dynamically adjusted based on historical data, and its calculation formula is as follows:

[0078] in, For the first Dynamic warning thresholds at any given time; Basic threshold; For smoothing coefficients; This is historical error data. The basic threshold... The smoothing coefficient is determined based on the allowable deviation of the production line, image quality judgment standards, classification confidence requirements, or historical stable production data; The value range is 0~1, and it is calibrated according to the degree of historical error fluctuation or the effect of on-site control; the historical error data The value is determined based on the deviation between the actual test results and the target value within the preset time window.

[0079] All process data mentioned above, including original images, classification results, feature parameters, control commands, timestamps, and coil numbers, are synchronously stored in the central database for process optimization, model iteration, and production traceability. The system uses a real-time feedback mechanism to correct the C-EfficientNetV2 model parameters online, enabling adaptive recognition of tail shapes for different steel grades.

[0080] This completes all the processes for detecting and controlling the planar shape of the tail section of precision rolled strip steel based on machine vision.

[0081] It should be noted that the core of this invention lies in applying machine vision and deep learning technologies to the online recognition of the shape of the strip tail, and establishing a complete closed-loop solution from shape recognition to roll gap control. During system deployment, a sufficient number and variety of strip tail image samples must be collected in advance to train the deep learning model, and the control model coefficients must be determined by regression based on actual rolling process data from the production line. and During the production process, the system operates fully automatically, requiring no human intervention. It has at least the following advantages: 1. It innovatively realizes online intelligent recognition and quantification of the complex planar shape of the tail end of precision rolled strip steel, filling the technological gap in this field.

[0082] 2. Employing a deep learning model, it possesses powerful recognition capabilities and high accuracy for irregular and blurred tail contours.

[0083] 3. A feedforward control model based on tail shape characteristics was established, which can adjust the roll gap in advance and proactively, effectively suppressing tail deviation and tail swing, and improving rolling stability and yield.

[0084] 4. The system has a high degree of integration and modular design, which makes it easy to upgrade existing production lines and is highly practical.

[0085] Second Embodiment

[0086] This embodiment provides an electronic device, such as... Figure 3As shown, the electronic device includes a processor and a memory; wherein the processor and the memory can be connected via a communication bus; the memory stores at least one instruction, which is loaded and executed by the processor to implement the method of the first embodiment described above. Furthermore, the electronic device may also include a transceiver, the processor and the transceiver can be connected via a communication bus, and the transceiver is used to communicate with other devices.

[0087] Below, in conjunction with Figure 3 A detailed introduction to each component of this electronic device is provided below: The processor is the control center of the electronic device. The electronic device may include multiple processors, each of which can be a single-core processor (single-CPU) or a multi-core processor (multi-CPU). The term "processor" can refer to a single processor or a collective term for multiple processing elements. For example, a processor can be one or more central processing units (CPUs), other general-purpose processors, application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), one or more field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor can perform various functions of the electronic device by running or executing software programs stored in memory and by calling data stored in memory.

[0088] In a specific implementation, as one example, the processor may include one or more CPUs, for example... Figure 3 CPU0 and CPU1 shown are, of course, merely illustrative examples.

[0089] The memory is used to store the software program that executes the solution of the present invention, and the processor controls its execution. For specific implementation methods, please refer to the above method embodiments, which will not be repeated here.

[0090] Optionally, the memory may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory may be integrated with the processor or exist independently, and may be accessed through the interface circuit of the electronic device ( Figure 3 (Not shown in the image) is coupled to the processor; however, this embodiment of the invention does not impose specific limitations on this.

[0091] The transceiver may include a receiver and a transmitter. Figure 3 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function. The transceiver can be integrated with the processor or exist independently, and can be connected through the interface circuit of the electronic device (…). Figure 3 (Not shown in the image) is coupled to the processor, and this embodiment of the invention does not specifically limit this.

[0092] In addition, it should be noted that, Figure 3 The structure of the electronic device shown is not intended to limit the device. Actual devices may include more or fewer components than shown, or combine certain components, or have different component arrangements. Furthermore, the technical effects achieved by this electronic device when performing the method of the first embodiment described above can be referenced to the technical effects described in the first embodiment; therefore, they will not be repeated here.

[0093] Third Embodiment

[0094] This embodiment provides a computer-readable storage medium storing at least one instruction, which is loaded and executed by a processor to implement the method of the first embodiment described above. The computer-readable storage medium may be a ROM, random access memory, CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc. The instruction stored therein can be loaded and executed by a processor in a terminal.

[0095] Furthermore, it should be noted that the present invention can be provided as a method, apparatus, or computer program product. Therefore, embodiments of the present invention can take the form of a completely or partially hardware embodiment, a completely or partially software embodiment, or an embodiment combining software and hardware aspects. Moreover, when implemented in software, embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any usable medium accessible to a computer or a data storage device such as a server or data center containing one or more sets of usable media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive (SSD).

[0096] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0097] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1The functions specified in one or more boxes. These computer program instructions may also be loaded onto a computer or other programmable data processing terminal equipment to cause a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0098] It should also be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element. Furthermore, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, and B alone, where A and B can be singular or plural. Additionally, the character " / " in this text generally indicates an "or" relationship between the preceding and following objects, but it can also indicate an "AND / OR" relationship. Please refer to the context for specific interpretations. "At least one" refers to one or more items, while "more than" refers to two or more items. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of single or multiple items. For example, at least one of a, b, or c can be represented as: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or multiple.

[0099] Furthermore, it is understood that in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0100] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0101] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of functional modules / units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the shown or discussed mutual couplings or direct couplings or communication connections may be through some interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms. Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, i.e., they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs. Additionally, the functional units in the various embodiments of this invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0102] If the method is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0103] Finally, it should be noted that the above description is merely a preferred embodiment of the present invention. It should be pointed out that although preferred embodiments of the present invention have been described, those skilled in the art, once they understand the basic inventive concept of the present invention, can make several improvements and modifications without departing from the principles described herein. These improvements and modifications should also be considered within the scope of protection of the present invention. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

Claims

1. A machine vision-based method for identifying the tail end of finished rolled strip steel, characterized in that, include: Acquire images of the tail section of the strip and perform image preprocessing on the acquired tail section images; The preprocessed strip tail image is input into a preset deep learning model, and the deep learning model is used to classify the preprocessed strip tail image and output the strip tail shape type. Based on the preprocessed strip tail image, feature parameters of the strip tail are extracted; Based on the strip tail shape type and strip tail characteristic parameters, calculate the roll gap leveling amount of the frame required to correct the asymmetric shape of the strip tail; The roll gap leveling amount is sent to the finishing mill control system to achieve real-time control of the strip tail deviation.

2. The machine vision-based tail-end recognition method for finished strip steel as described in claim 1, characterized in that, The image preprocessing of the acquired strip tail image includes: The images of the same strip tail taken by multiple cameras are stitched together to form a complete image of the strip tail. A preset filtering algorithm is used to filter the stitched strip tail image to eliminate noise interference; wherein, the preset filtering algorithm is a bilateral filtering algorithm or a guided filtering algorithm. Missing pixels in the filtered strip tail image are filled using bicubic interpolation. For images with missing pixels, a Laplacian pyramid fusion algorithm is used to eliminate brightness differences.

3. The machine vision-based tail-end recognition method for precision rolled strip steel as described in claim 1, characterized in that, The deep learning model is the EfficientNetV2 model, which incorporates the CBAM attention mechanism and Droppath regularization.

4. The machine vision-based tail-end recognition method for finished strip steel as described in claim 1, characterized in that, The tail shape types of strip steel include rectangular, single blade on the operating side, single blade on the drive side, tongue-shaped, and dovetail-shaped.

5. The machine vision-based tail-end recognition method for finished strip steel as described in claim 4, characterized in that, The characteristic parameters of the strip tail include common parameters and special shape parameters; among them, Common parameters include: tail length, tail area, tail perimeter, the ratio of the length to the width of the smallest bounding rectangle, and the ratio of the perimeter to the area of ​​the inscribed polygon. For strip steel that is determined to be either an operating-side single blade or a transmission-side single blade, its special shape parameters include: single blade vertex coordinates, single blade cutting edge starting position, and single blade length; For strip steel that is determined to be tongue-shaped, its special shape parameters include: tongue vertex coordinates, tongue starting position, tongue length, and tongue arc similarity. For strip steel that is determined to be dovetail-shaped, its special shape parameters include: dovetail concave point coordinates, dovetail vertex coordinates, dovetail slope, dovetail fork angle, and dovetail length.

6. The machine vision-based tail-end recognition method for precision rolled strip steel as described in claim 1, characterized in that, The extraction of strip tail feature parameters based on the preprocessed strip tail image includes: Based on the preprocessed strip tail image, the feature parameters of the strip tail are extracted using either an adaptive threshold corner detection algorithm based on edge detection or a sub-pixel corner detection algorithm based on grayscale intensity.

7. The machine vision-based tail-end recognition method for precision rolled strip steel as described in claim 1, characterized in that, The step of calculating the required roll gap leveling amount for correcting the asymmetric shape of the strip tail based on the strip tail shape type and characteristic parameters includes: The tail-end characteristic parameters of the strip are quantified as tail-end asymmetry. Among them, for strip steel that is determined to be either an operating-side single-blade or a transmission-side single-blade, = ;in, The length of a single blade; for strip steel judged to be tongue-shaped, ;in, The length of the tongue; For tongue-shaped arc similarity; for strip steel judged to be dovetail-shaped, ;in, The angle of the swallowtail; The length of the swallowtail; It has a swallowtail-shaped forked corner; Based on the tail asymmetry, calculate the roll gap leveling amount for the current frame: ； in, For the first Roll gap leveling amount for each frame; The width of the strip; The incoming material is wedge-shaped; For export thickness; Poor rolling force; The difference in stiffness between the two sides of the rolling mill; and These are the preset model weight coefficients obtained through regression of historical rolling data; The average stiffness of the rolling mill; Based on the position of the tail section, the roll gap leveling amount of the current frame is multiplied by a preset proportional coefficient to obtain the roll gap leveling amount of the subsequent frames affected after the current frame.

8. The machine vision-based tail-end recognition method for finished strip steel as described in claim 7, characterized in that, The formula for calculating the proportionality coefficient is: ； in, Indicates by the first i The roll gap leveling amount of the first frame is obtained. i The proportional coefficient required for roller gap leveling of +1 frame; For the first i rack exit speed; For strip steel from the first The rack is running to the... Time per rack; For the first rack and the first Distance between racks This is the deformation attenuation coefficient.

9. The machine vision-based tail-end recognition method for precision rolled strip steel as described in claim 1, characterized in that, The machine vision-based tail-end recognition method for finished strip steel also includes: Establish a mapping relationship between the strip tail shape parameters and rolling process parameters as a verification indicator: ； in, To verify the indicators; These are the weighting coefficients; For the selected rolling process parameters; For bias terms; The number of rolling process parameters involved in the calculation; When the validation metric exceeds the threshold range, adaptive adjustment of model parameters is initiated; the threshold is dynamically adjusted based on historical data, and its calculation formula is as follows: ； in, For the first Dynamic warning thresholds at any given time; Basic threshold; For smoothing coefficients; This is historical error data.

10. The machine vision-based tail-end recognition method for precision rolled strip steel as described in claim 1, characterized in that, The machine vision-based tail-end recognition method for finished strip steel also includes: An alert is issued when a preset abnormal situation is detected; The preset abnormal conditions include: a preset number of consecutive images with substandard quality, a classification confidence level lower than a preset confidence threshold, and the actual deviation from the target value exceeding a preset allowable range.

Images