A method and system for high-precision length determination of continuously cast steel billets

Abstract

This invention relates to the field of rebar length measurement, and more particularly to a high-precision length measurement method and system for continuously cast rebar billets. The method includes the following steps: installing an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain a raw image stream; preprocessing the raw image stream to obtain an enhanced image stream; determining the end position of the billet based on the enhanced image stream; performing pixel displacement conversion calculation based on the end position of the billet to obtain the real-time length value of the billet; correcting the image measurement motion error of the real-time length value to generate a corrected length value; extracting the target length measurement, and performing real-time rolling tracking of the corrected length value; when the target length measurement is detected to be equal to the standard length measurement, driving the cutting equipment to complete the length measurement. This invention achieves high-precision length measurement and measurement of continuously cast rebar billets, improving the accuracy of length measurement and cutting and increasing production efficiency.

Description

Technical Field

[0001] This invention relates to the field of steel bar length measurement, and in particular to a high-precision length measurement method and system for continuously cast steel bar billets. Background Technology

[0002] With the increasing demands for precision and performance in modern construction and infrastructure projects, the precision control of steel reinforcement, as a core material in building structures, is particularly crucial. Traditional methods for determining the length of continuously cast steel billets rely primarily on manual experience and fixed parameter settings, adjusting the cutting length through on-site operations or periodic inspections. While these methods ensure basic production stability to a certain extent, their predictive accuracy is limited due to dynamic changes in factors such as molten steel flow rate, billet cooling rate, roll friction coefficient, and ambient temperature. This leads to cutting deviations, material waste, reduced production efficiency, and even impacts the quality of subsequent processing stages. Existing methods largely depend on manual monitoring or periodic measurements, lacking real-time and automation capabilities, making it difficult to cope with the complex fluctuations in high-speed continuous casting production. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention proposes a high-precision length-fixing method and system for continuously cast steel billets, thereby resolving at least one of the aforementioned technical issues.

[0004] To achieve the above objectives, the present invention provides a high-precision length-fixing method for continuously cast steel billets, comprising the following steps: Step S1: Install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; Step S2: Perform image preprocessing on the original image stream to obtain an enhanced image stream; determine the end position of the billet based on the enhanced image stream; Step S3: Perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; Step S4: Perform image measurement motion error correction on the real-time length value to generate a corrected length value; Step S5: Extract the target fixed length and perform real-time rolling tracking of the corrected length value. When the target fixed length is detected to be equal to the standard fixed length, drive the cutting equipment to complete the fixed length cutting.

[0005] This specification provides a high-precision length-fixing system for continuously cast steel billets, used to perform the high-precision length-fixing method for continuously cast steel billets as described above, including: The image acquisition module is used to install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; The image preprocessing module is used to preprocess the original image stream to obtain an enhanced image stream; and to determine the end position of the billet based on the enhanced image stream. The calculation module is used to perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; The error correction module is used to correct the image measurement motion error of the real-time length value and generate a corrected length value; The length-fixing module is used to extract the target length-fixing value and perform real-time rolling tracking of the correction length value. When the target length-fixing value is detected to be equal to the standard length-fixing value, the cutting device is driven to complete the length-fixing cut.

[0006] The specific benefits of this invention are as follows: Installing an industrial-grade infrared camera system above the output roller conveyor of the production line and continuously acquiring images allows for real-time acquisition of thermal radiation information from the billet end and surface, covering the entire width and length of the billet and eliminating delays and positional deviations caused by manual measurement. Through grayscale conversion, filtering, brightness compensation, and edge gradient enhancement, thermal noise, roller conveyor background interference, and local brightness fluctuations can be removed, making the billet end more prominent in the image. Utilizing an enhanced image stream to extract the billet contour allows for precise end positioning, significantly reducing positioning errors and ensuring stable and reliable measurement data, providing a solid visual foundation for high-precision length cutting. Pixel displacement conversion based on the end position accurately maps image pixel coordinates to physical length, enabling real-time billet length measurement. By accumulating the pixel increments of each frame and combining them with camera intrinsic and extrinsic parameter calibration coefficients, the length measurement accuracy can be controlled at the millimeter level. This method can continuously monitor the billet length, ensuring that the measured data is highly consistent with the actual physical length, providing a reliable basis for cutting. By correcting motion errors in the real-time length value, inter-frame errors caused by image measurement and roller conveyor speed fluctuations during high-speed operation can be compensated. By comparing the image measurement increment with the theoretical displacement and performing dynamic correction, the corrected length value becomes stable, smooth, and accurate. Extracting the target length and combining it with the cutting equipment's advance action for real-time rolling tracking allows for precise control of the cutting trigger timing, ensuring complete synchronization between the cutting equipment's action and the billet end reaching the standard length. Automatically generated cutting control signals enable millimeter-level control of cutting errors, achieving high-precision length cutting of continuously cast steel billets, reducing manual intervention, and improving product length consistency and production efficiency. Attached Figure Description

[0007] Figure 1 This is a schematic diagram of the steps of a high-precision length-fixing method for continuously cast steel billets according to the present invention; Figure 2 This is a detailed flowchart illustrating the implementation steps of step S1. Figure 3 This is a flowchart illustrating the detailed implementation steps of step S2. Detailed Implementation

[0008] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0009] This application provides a method and system for high-precision length determination of continuously cast steel billets. The execution entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that can be considered general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio-visual management system, an information management system, and a cloud-based data management system.

[0010] Please see Figures 1 to 3 This invention provides a high-precision length-fixing method for continuously cast steel billets, comprising the following steps: Step S1: Install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; Step S2: Perform image preprocessing on the original image stream to obtain an enhanced image stream; determine the end position of the billet based on the enhanced image stream; Step S3: Perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; Step S4: Perform image measurement motion error correction on the real-time length value to generate a corrected length value; Step S5: Extract the target fixed length and perform real-time rolling tracking of the corrected length value. When the target fixed length is detected to be equal to the standard fixed length, drive the cutting equipment to complete the fixed length cutting.

[0011] In a specific embodiment, taking a continuous casting production line for reinforcing bars as an example, the billet width is 200 mm, the thickness is 180 mm, the operating speed is 1.8~2.2 m / s, the target length is 12 m, the image acquisition frame rate is 80 fps, and an industrial-grade infrared camera is used for end image recognition and length calculation. An industrial-grade infrared camera system is installed 1.5 m above the output roller conveyor of the production line. The infrared camera has a resolution of 1280×1024 pixels and a frame rate of 80 fps. First, camera calibration is performed to obtain the intrinsic parameter matrix, distortion coefficient, and extrinsic parameter matrix, establishing a correspondence between the camera pixel coordinate system and the production line physical coordinate system. After calibration, the infrared camera system is synchronized with the continuous casting billet production control system via industrial Ethernet protocol to achieve time synchronization between image acquisition and billet conveying. After synchronization, each frame of image corresponds to the position of the billet at a specific time point, ensuring high-precision alignment of subsequent length measurement and cutting actions.

[0012] The infrared image is converted from the original thermal radiation image to a grayscale image with a pixel value range of 0-255 to reduce color contrast interference and highlight the brightness difference at the edge of the billet. A Gaussian adaptive filter with a kernel size of 5×5 is used, and the filter coefficients are dynamically adjusted according to local brightness changes in the image to smooth thermal noise and roller conveyor background interference. Through continuous frame brightness difference analysis, the high-temperature radiation fluctuations of the billet are identified, the average brightness deviation of each frame is calculated, and brightness correction is performed using compensation gain to make the billet end more prominent in the grayscale image. The Sobel operator is used to perform gradient enhancement on the brightness-compensated image to enhance the edge features of the billet end. Canny edge detection is used to extract the billet contour, eliminate isolated noise points, and fill in regional holes. The overall billet end position is obtained through inter-frame contour tracking.

[0013] Suppose that the end of the billet is detected at pixel X = 950 in the 100th frame, and it is converted into a physical length L_frame = 5.68 m (from the image reference zero point to the end of the billet) through camera intrinsics and calibration matrix.

[0014] In consecutive image frames, the end pixel displacement ΔX_pixel = X_n+1 - X_n is calculated, and the pixel displacement is converted into the actual length increment ΔL_real. The conversion coefficient is obtained from camera calibration, for example, 1.2 mm per pixel. Let: the end displacement ΔX_pixel from frame 100 to frame 101 be 4 pixels; converted to actual length ΔL_real = 4 × 1.2 mm = 4.8 mm; the accumulated real-time length is: L_real(n+1) = L_real(n) + ΔL_real = 5.68 m + 0.0048 m ≈ 5.6848 m.

[0015] The length increment of each frame is continuously accumulated, and the three-frame window is smoothed by moving average to obtain the real-time length value of the billet, eliminating minor jitter between frames.

[0016] The billet speed V = 2.0 m / s is obtained using a roller conveyor speed sensor at a frame rate of f = 80 fps. The theoretical displacement per frame is ΔL_theory = V / f = 2.0 / 80 = 0.025 m. The image measurement length increment is compared with the theoretical displacement, and the motion error coefficient K_error = ΔL_real / ΔL_theory = 0.0048 / 0.025 ≈ 0.192 is calculated. The length increment is corrected using this error coefficient: ΔL_corrected = ΔL_real / K_error ≈ 0.0048 / 0.192 ≈ 0.025 m. This corrected increment is then added to the real-time length value to obtain the corrected length value L_corrected. This method ensures that the image measurement length matches the actual movement of the billet, improving the dimensional accuracy.

[0017] The target length L_target = 12 m is obtained from the work log. Based on the cutting equipment action delay T_cut = 0.5 s and the billet speed V = 2.0 m / s, the lead time L_ahead = V × T_cut = 2.0 × 0.5 = 1.0 m is calculated.

[0018] Adjust the target length to obtain the standard length: L_standard = L_target - L_ahead = 12 - 1 = 11 m.

[0019] The system continuously tracks and corrects the length value in real time. When L_corrected ≥ L_standard, a cutting control signal is generated to trigger the cutting equipment to operate.

[0020] The cutting control signal is sent to the hydraulic cutting machine, which has a response time of 0.3 seconds, and then performs the billet cutting to the specified length. Rolling tracking ensures that the end of the billet accurately reaches the standard length position before the cutting is triggered, and the final billet length error is controlled within ±3 mm, meeting the high-precision length requirements.

[0021] In this embodiment, see Figure 2 The diagram below illustrates the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include: Install an industrial-grade infrared camera system above the output roller conveyor of the production line; Obtain the intrinsic and extrinsic parameters of the infrared camera and perform camera calibration. The calibrated industrial-grade infrared camera system is synchronized with the continuous casting billet production system using industrial communication protocol timing, and continuous image acquisition is performed to obtain the original image stream.

[0022] In this embodiment, on-site structural and process parameter surveys were conducted in the continuous casting billet output roller conveyor area. Based on the roller conveyor width, billet cross-sectional dimensions, and operating speed, the installation geometry of the infrared camera was determined. The camera was positioned approximately 3.5 m to 5.0 m above the roller conveyor centerline, ensuring its optical axis was essentially perpendicular to the roller conveyor plane and covered the entire detection area. An industrial lens with a focal length of 8 mm to 25 mm was selected according to the installation height and field-of-view requirements, achieving a spatial resolution better than 2 mm / pixel. The field of view was controlled between 25° and 45° to balance coverage and measurement accuracy. For continuous casting billet surface temperatures exceeding 800 ℃, a water-cooled or air-cooled protective shell was installed, along with an infrared transmission window. The window material was selected from germanium or zinc selenide, with a thickness controlled within 2 mm to 5 mm to reduce infrared attenuation and imaging distortion. During installation, a laser rangefinder and level were used to correct the camera's attitude, ensuring a tilt angle error of less than ±0.5°. An industrial vibration damping bracket was used to control the vibration amplitude to within 0.1. Within mm, geometric and radiometric calibrations were performed under stable camera operation. First, a heated infrared calibration plate was used as a feature template, with the surface temperature controlled between 60℃ and 100℃ to ensure infrared imaging contrast. At least 12 sets of images from different angles were acquired by changing the calibration plate's orientation. The Zhang Zhengyou calibration method was used to solve for the camera's intrinsic parameters, including focal length, principal point coordinates, and radial and tangential distortion coefficients, ensuring radial distortion was controlled to the order of 10⁻³ and tangential distortion to below 10⁻³. 4 Subsequently, multiple high-temperature stable reference points were set up on the roller conveyor with a spacing of 500 mm to 1000 mm. Their spatial coordinates were obtained using a total station, and the extrinsic parameter matrix was solved using the PnP algorithm in combination with the image pixel coordinates, so that the reprojection error was controlled within 1.5 pixels, corresponding to an actual spatial error of no more than ±2 mm. On this basis, radiometric correction was performed, and the emissivity was set to 0.75 to 0.85 according to the billet material. Background compensation was also performed in combination with the ambient temperature of 25 ℃ to 45 ℃ to reduce the impact of temperature drift on boundary recognition. After calibration, a standard part of known length (such as a 1 m sample) was selected for online verification. The error distribution was calculated through repeated measurements, so that the overall measurement deviation was controlled within ±0.3%, thereby ensuring that the calibration parameters have good stability and repeatability under actual production conditions.

[0023] The calibrated infrared camera is connected to the continuous casting production control system, and data interaction with the PLC is achieved through industrial Ethernet. Profinet or Modbus TCP protocols are used to transmit signals for billet casting speed, roller encoder displacement, and cutting trigger. A PTP precision time synchronization mechanism is introduced to control the image acquisition timestamp error within ±1 ms, ensuring strict alignment between image data and physical position. The sampling frequency is dynamically set according to the billet running speed. When the speed is 1.5 m / s, the acquisition frame rate is set to 50 fps to 100 fps, ensuring that the displacement between adjacent frames does not exceed 30 mm. The image resolution is selected based on processing capability: 640×512 or 1024×768. Real-time buffering and streaming are performed at the acquisition end. A sliding window mechanism stores the continuous image stream, and Gaussian filtering is used for initial noise reduction. The filtering parameter σ is controlled between 1.0 and 1.5 to suppress thermal noise. The system initiates continuous acquisition when the billet head enters the detection area via a PLC trigger mechanism and stops when the tail leaves, thus forming complete time-series image data. The system also synchronously records velocity curves and time information for subsequent length integration calculations, achieving high-precision fusion of multi-source data including image, displacement, and time.

[0024] In this embodiment, see Figure 3 The diagram below illustrates the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include: Perform grayscale conversion on the original image stream to obtain a grayscale image stream; An adaptive filtering process is applied to the grayscale image stream to obtain a filtered image stream; High-temperature radiation brightness fluctuation identification is performed on the filtered image stream to obtain brightness fluctuation information; Adaptive brightness compensation processing is performed based on brightness fluctuation information to generate a brightness-compensated image stream. Edge gradient enhancement is performed on the brightness-compensated image stream to obtain an enhanced image stream; the end position of the billet is determined based on the enhanced image stream.

[0025] In this embodiment, for the pseudo-color or multi-channel data output by the infrared camera, a weighted average method or a single-channel mapping method based on radiation intensity is used to convert each frame of image into an 8-bit or 12-bit grayscale image, where the grayscale value has a monotonic correspondence with the radiation brightness. In specific implementation, the pixel grayscale range is normalized to the range of 0-255 or 0-4095, and linear stretching is performed in combination with the dynamic range of infrared imaging to enhance the contrast between the billet and the background. To avoid saturation in high-temperature areas, an upper limit threshold is set for the grayscale mapping process (e.g., corresponding to a temperature of 900 ℃), and the part exceeding the threshold is truncated. At the same time, the area below the background temperature (e.g., below 50 ℃) is compressed to increase the proportion of effective information. Under experimental conditions, the acquisition resolution is 1024×768 and the frame rate is 80. The raw image stream at FPS, after grayscale conversion, can reduce the data volume by approximately 60% while maintaining the integrity of boundary information. Furthermore, consistency correction is applied to consecutive frames to eliminate grayscale fluctuations caused by automatic gain variations, keeping the average grayscale difference between adjacent frames within ±3%, thus forming a stable grayscale image stream. A median filtering and adaptive Gaussian filtering strategy is combined, dynamically adjusting the filtering intensity based on the pixel variance within a local window (typically 5×5 or 7×7). When the local variance is large, the filtering weight is reduced to protect edge information; when the variance is small, the smoothing effect is enhanced to remove noise. The Gaussian kernel standard deviation σ in the filtering parameters is set between 0.8 and 1.5, and is adaptively adjusted between frames based on the overall noise level of the image. Simultaneously, a time-dimensional filtering mechanism is introduced, performing a time-domain weighted average of 3 to 5 consecutive frames, with the weight decreasing according to the time distance, thereby further reducing the impact of random fluctuations. Under typical experimental conditions, the signal-to-noise ratio of the filtered image can be improved from approximately 28 dB to 35 dB. The filter output is above dB, while the edge gradient loss is controlled within 5%. In addition, by setting an outlier removal mechanism, pixels with gray-level abrupt changes exceeding the local mean ±3σ are replaced, effectively eliminating transient interferences such as sparks and oxide peeling, thus obtaining a smooth and well-structured filtered image stream.

[0026] A spatiotemporal joint analysis window (7×7 spatial window, 5 frames temporal window) is constructed to calculate the temporal mean and variance of each pixel, obtaining a brightness stability index. When the variance exceeds a set threshold (e.g., grayscale fluctuation exceeds ±15), it is identified as a fluctuating region. Simultaneously, a region segmentation method is introduced to divide the image into several sub-regions (e.g., 10–20 strips along the width), and the average grayscale change rate of each region is calculated to identify the overall radiation drift trend. For periodic fluctuations (e.g., caused by equipment vibration or power fluctuations), their frequency characteristics are analyzed using Fast Fourier Transform to identify typical frequency ranges (generally 1 Hz to 10 Hz). The system uses a combination of zonal linear correction and global normalization to calculate compensation coefficients for each zone and adjust its grayscale value to the level of the reference area, which is usually a central region with stable brightness and no boundaries. The compensation function can be expressed as a linear mapping relationship, where the gain coefficient and bias are determined by the difference between the regional mean and the target mean, and the compensation result is restricted to avoid excessive enhancement that leads to noise amplification. At the same time, a time smoothing mechanism is introduced to perform an exponential weighted average of the compensation coefficients, with weight coefficients ranging from 0.6 to 0.8, to ensure the continuity and stability of the compensation process. In the experiment, this method can reduce the overall brightness non-uniformity of the image from the original ±12% to within ±3%, and improve the contrast of the boundary area by about 20%. In addition, for local abnormal bright spots (such as high-temperature spots or sparks), a local suppression strategy is used for amplitude limiting to ensure that they do not affect the overall compensation effect, thereby generating a compensated image stream with uniform brightness distribution and stable contrast.

[0027] Gradient calculation is performed by combining the Sobel and Laplacian operators. First, the first-order gradients are calculated in the horizontal and vertical directions to obtain gradient magnitude images. Then, the second-order Laplacian response is superimposed to enhance edge details. During the gradient calculation process, a threshold range (e.g., gradient values ​​greater than 20-40) is set to suppress weak edges and noise responses. The edge width is refined using a non-maximum suppression method, keeping the edge thickness within 2-3 pixels. Simultaneously, a contrast-limited adaptive histogram equalization method is used to enhance local regions, further expanding the grayscale difference in the boundary regions. Experimental results show that after enhancement, the edge gradient magnitude is increased by approximately 30%-50%, and the boundary localization error is reduced to within ±1.5 pixels. Furthermore, by using a multi-frame superposition enhancement strategy to fuse edge information in consecutive frames, boundary stability can be further improved, resulting in a clear and continuous enhanced image stream.

[0028] In this embodiment, the specific steps for determining the end position of the billet based on the enhanced image stream are as follows: High-temperature background separation is performed on the enhanced image stream to obtain a separated image; The isolated noise point filtering and region hole filling are performed on the separated image to obtain the target region image of the steel billet. Edge detection is performed on the target area image of the steel billet to extract contour boundary information; Based on the contour boundary information, contour localization and tracking are performed to obtain the overall contour line; Identify the direction of billet movement and determine the position of the billet end based on the overall outline.

[0029] In this embodiment, a background modeling method based on statistical modeling is adopted. First, several frames (usually 50 to 100 frames) of images are collected during periods with no billets or very small billet proportions to construct a background grayscale model. The mean and standard deviation of each pixel are calculated to form a background distribution matrix. In real-time processing, the current frame image and the background model are differentially analyzed. When the pixel grayscale difference exceeds a set threshold (e.g., ±20 grayscale levels), it is determined to be a foreground region, i.e., the billet region. To adapt to the characteristics of high-temperature radiation changing over time, an adaptive update mechanism is introduced to dynamically update the background model with a low weight (e.g., 0.05 to 0.1) so that it can track the slow changes in ambient temperature. At the same time, combined with a temperature threshold segmentation method, regions with radiation intensity above a certain level (e.g., corresponding to temperatures above 600 ℃) are enhanced to improve the distinguishability between the billet and the background.

[0030] Connectivity analysis is used to identify and remove areas smaller than a set threshold (e.g., less than 50 pixels) as noise. Median filtering (3×3 or 5×5 window size) is then used to smooth out local anomalies. Subsequently, morphological closing operations (expansion followed by erosion) are employed to fill any voids within the main body of the billet. A circular kernel with a radius of 3–5 pixels is typically selected as the structuring element to ensure effective filling while avoiding excessive boundary expansion. Furthermore, a region growing algorithm is introduced to expand the billet region from a high-confidence seed point, making the region boundaries more continuous. In typical experiments, this processing can improve the integrity of the target region to over 98% while reducing the false noise detection rate to below 2%. The final result is a billet target region image with continuous boundaries and a complete interior.

[0031] The Canny edge detection algorithm is adopted, which includes steps such as Gaussian smoothing, gradient calculation, non-maximum suppression, and dual-threshold concatenation. The standard deviation σ of the Gaussian filter is set to 1.0 to 1.5 to balance noise reduction and detail preservation. The gradient threshold is usually set to a low threshold of 20 to 30 and a high threshold of 40 to 60 to ensure the integrity and accuracy of edge extraction. At the same time, based on prior information, edges close to the width range of the billet are preferentially preserved, while abnormal directions or short edges are suppressed. To improve boundary continuity, an edge concatenation strategy is introduced to merge edge segments with a distance of less than 5 pixels and the same direction.

[0032] Contour tracking algorithms (such as chain code-based contour tracking methods) are used to connect edge points in an orderly manner to form closed or semi-closed contour curves. Subsequently, the contours are smoothed by using curve fitting methods (such as polynomial fitting or spline curve fitting) to eliminate local jaggedness and make the contours smoother and more continuous. In the time dimension, Kalman filtering is introduced to track and predict key contour points, and the contours of adjacent frames are matched and updated to maintain the continuity of contour changes and reduce jitter, with position fluctuations controlled within ±2 mm. At the same time, shape constraints are applied to the contours based on the geometric characteristics of the billet (such as an approximately rectangular cross-section), and abnormal parts that do not conform to geometric rules are eliminated.

[0033] The displacement vector of the centroid of the profile between consecutive frames is calculated and verified in conjunction with the velocity direction information provided by the roller encoder to determine the actual movement direction of the billet (usually unidirectional linear movement). Based on this, the profile is projected according to the movement direction to find the set of outermost boundary points in the direction of the movement front, and the end boundary position is determined by linear fitting or least squares method. To improve stability, the end positions of 5 to 10 consecutive frames are processed by time-weighted averaging to reduce positioning fluctuations to within ±2 mm. At the same time, length constraints and geometric consistency checks are set to avoid misjudgment due to local occlusion or noise.

[0034] In this embodiment, step S3 includes the following steps: The positional change of the enhanced image stream is calculated based on the end position of the steel billet to obtain the pixel change. Establish transformation coefficients between the image pixel coordinate system and the actual physical space coordinate system based on camera intrinsic and extrinsic parameters; Based on the conversion coefficient, the pixel change is converted into a pixel displacement to obtain the actual length increment; The actual length increment is calculated over time and then smoothed to obtain the real-time length value of the billet.

[0035] In this embodiment, coordinate difference is performed on the end pixel positions of consecutive frames to obtain pixel displacement vectors. Horizontal displacement is used to calculate the length change along the roller conveyor direction, while vertical displacement is used to monitor possible jitter or tilt. To improve calculation accuracy, a sub-pixel positioning method is adopted, where the end contour points are corrected at the sub-pixel level through interpolation or gradient centering, improving the end position accuracy from 1 pixel to 0.2–0.3 pixels. The frame rate is set to 80 fps, the billet running speed is 1.5 m / s, and the pixel displacement between adjacent frames is typically between 5 and 10 pixels. Sub-pixel processing can reduce the measurement error to ±0.5 mm. Simultaneously, to prevent occasional noise or local edge detection errors from affecting pixel changes, a time window filter is introduced, and the displacement of 3–5 consecutive frames is weighted and averaged. After the end pixel displacement is calculated, the pixel coordinates need to be mapped to the actual physical space coordinates to achieve length measurement. Using the calibrated camera intrinsic parameters, including focal length, principal point position, and radial and tangential distortion coefficients, combined with the extrinsic parameters rotation matrix and translation vector, a projection relationship from the pixel coordinate system to the physical space coordinate system is established. By using a pinhole camera model combined with distortion correction, pixels on the image plane are projected onto a physical plane parallel to the roller conveyor, yielding the actual physical distance corresponding to each pixel. The conversion coefficient can be calculated based on the camera's installation height, focal length, and pixel resolution. Each pixel displacement is mapped to the actual length, ensuring measurement accuracy along the direction of motion.

[0036] After obtaining the pixel change and conversion coefficient, the pixel displacement of each frame's end is converted into the actual physical length increment. Multiplying the pixel displacement by the conversion coefficient yields the length increment along the billet's movement direction, while camera projection correction eliminates perspective errors. To ensure the accuracy of the length increment, any abnormal frames that may occur during the end movement are discarded or corrected to avoid instantaneous errors affecting the cumulative length. During the length increment calculation, compensation can be made using the roller conveyor speed and inter-frame time interval to ensure the length increment perfectly corresponds to the actual movement of the billet.

[0037] The actual length increment of each frame is accumulated along the time axis to obtain the real-time length of the billet from its entry into the detection area to its current position. During the accumulation process, a sliding window smoothing or exponential weighted average method is used to smooth the length increment to eliminate instantaneous fluctuations caused by end detection jitter and local errors. The sliding window length can be set to 3 to 5 consecutive frames, and the smoothing weight coefficient is controlled between 0.6 and 0.8 to ensure the continuity and stability of the length curve. At the same time, dual-source correction is performed in conjunction with roller speed information to correct the length increment of abnormal frames, ensuring the continuity and accuracy of the accumulated length. Through this method, the real-time length value can be updated frequently, ensuring stable and reliable length measurement and meeting the industrial application requirements for high-precision length measurement of continuously cast steel billets.

[0038] In this embodiment, step S4 includes the following steps: The billet running speed is calculated based on the speed sensor of the output roller conveyor of the production line; The theoretical displacement is calculated based on the billet's running speed to obtain the theoretical displacement per unit time. Based on the theoretical displacement, the motion error of the enhanced image stream is calculated to obtain the error coefficient; Motion compensation correction is performed on the real-time length value based on the error coefficient to generate a corrected length value.

[0039] In this embodiment, as the billet moves along the output roller conveyor of the production line, a speed sensor installed on the roller conveyor continuously collects the billet's motion information. This sensor generates pulse signals by sensing the rotation of the rollers, with each pulse corresponding to a certain angle of roller rotation. The sensor's acquisition frequency is typically above 100 Hz to ensure accurate capture of the motion state of each time segment even when the billet moves at high speed (up to 2.5 meters per second). To eliminate interference signals generated by roller conveyor vibration, friction, or high-speed operation, the acquired pulse sequence is de-jittered, and a low-pass filter is used to smooth the speed curve. The accumulated pulse data within a continuous time window is converted into instantaneous speed information, resulting in a continuous and smooth output billet speed. The speed measurement system is designed with an accuracy of ±0.5%, accurately reflecting the actual operating state of the billet. After obtaining continuous billet speed information, the theoretical displacement of the billet per unit time needs to be calculated based on the image acquisition time interval. The enhanced image acquisition frame rate is typically set to 80 frames per second, corresponding to an acquisition interval of approximately 12.5 milliseconds per frame. By combining the billet's running speed (e.g., 1.5 to 2.5 meters per second), the theoretical displacement distance of the billet along the roller conveyor within each frame time interval can be determined. To ensure that the theoretical displacement is consistent with the actual continuous movement of the billet, a weighted average processing is required for high-speed fluctuations or instantaneous acceleration and deceleration to smooth short-cycle speed changes, while ensuring that the displacement calculation is synchronized with the image frame. The theoretical displacement per unit time provides a reference standard for comparing subsequent image measurement errors, ensuring that the length calculation fully reflects the true movement of the billet along the roller conveyor, and guaranteeing the continuity and reliability of real-time length measurement.

[0040] After obtaining the theoretical displacement for each frame, it is correlated with the end position change obtained from image processing to calculate the image measurement deviation and obtain the error coefficient. Specifically, the pixel displacement of the end in each frame is converted into physical length and compared with the theoretical displacement; the deviation value reflects the actual error in image measurement. To enhance the stability of error calculation, the deviation values ​​of several consecutive frames are smoothed using a sliding window or weighted averaging, ensuring the error coefficient remains continuous and stable under conditions of high-speed billet movement and slight end jitter. The error coefficient is updated in real time, dynamically reflecting the difference between image measurement and the actual movement of the billet.

[0041] After obtaining the error coefficient, motion compensation correction is performed on the real-time length value to adjust the image measurement results to match the actual displacement of the billet. The compensation operation is performed frame-by-frame, combined with sliding time windows or weighted averaging to ensure the length curve remains smooth and continuous during compensation, eliminating the influence of end detection errors or short-term jitter. The correction process simultaneously considers the actual running speed of the billet and the inter-frame time interval, ensuring that the length value in each frame accurately reflects the actual position of the billet on the production line. The final corrected length value exhibits stability and high accuracy, and can be continuously updated under conditions of billet speed of 1.5 to 2.5 meters per second and a frame rate of 80 frames per second, meeting the requirements for high-precision length measurement and online cutting control of continuously cast steel billets. This method ensures that the length measurement data remains reliable and consistent throughout the entire production process.

[0042] In this embodiment, step S5 includes the following steps: Identify the cutting action time of the cutting equipment; calculate the cutting extraction amount based on the cutting action time and the billet running speed to obtain the lead time; Extract the target length based on the work log; The target length is adjusted based on the lead time to obtain the standard length. Based on the standard fixed length, the standard fixed length is tracked in real time. When the target fixed length is detected to be equal to the standard fixed length, a cutting control signal is generated. The cutting control signal drives the cutting equipment to complete the fixed-length cutting.

[0043] In this embodiment, a cutting start signal is acquired through the cutting equipment's sensors or control interface. This signal typically originates from the action feedback port of the hydraulic or motor control system. The status changes of this signal are monitored in real time. When the equipment is detected to transition from a stationary state to an active state, the timestamp of the cutting start is recorded. Similarly, the end time is recorded after the cutting is completed, yielding the duration of the cutting action. To ensure high-precision control, the acquired signal is filtered and de-jittered to eliminate the influence of electrical noise or short-term triggering errors on time determination. The cutting action time accuracy can be controlled at the millisecond level, meeting the requirements for advance calculation under high-speed billet movement. Simultaneously, a cutting equipment response model can be established based on historical action data to analyze the equipment's start-up delay and hydraulic or mechanical inertia characteristics, providing accurate time parameters for real-time cutting control and ensuring synchronization between the cutting action and billet length measurement. After obtaining the cutting action time, the advance required for the cutting operation can be calculated by combining it with the billet's running speed on the roller conveyor. The advance is the actual distance the billet moves, which needs to be reserved before the cutting equipment completes its start-up and execution of the action, ensuring the cutting position is accurate to the set length. To achieve this calculation, the real-time speed of the billet is multiplied by the cutting action time to obtain the length distance that must be triggered in advance before cutting. Speed ​​data is acquired by a roller conveyor speed sensor, typically ranging from 1.5 to 2.5 meters per second. The cutting action time is generally around 500 milliseconds, from which the lead time is calculated, allowing the billet to move within the range of half a meter to one meter. Time and speed corrections are applied to the continuous cutting action, and the billet speed is smoothed using a sliding window averaging process. This ensures stable and reliable lead time calculations, perfectly matching the cutting trigger time with the actual length of the billet and guaranteeing dimensional accuracy.

[0044] To achieve high-precision length control, target length information needs to be obtained from the work log. The work log typically contains the preset length, cutting sequence, and production batch information for each billet. By parsing the log file, the target length of the currently processed billet is extracted and correlated with real-time length measurements. The extraction process includes verifying log integrity, parsing length units, confirming batch and billet number, etc., to ensure that the target length matches the actual production length. The extracted data is internally converted into values ​​that can be directly used for length control and updated synchronously with real-time length measurements, maintaining accurate tracking of the target length throughout the entire cutting cycle and providing an accurate target reference for subsequent lead adjustments.

[0045] By combining the target length with the lead time, the length is corrected in the direction of billet movement, ensuring that the billet end corresponding to the cutting trigger point reaches the standard length position. To ensure the stability of the adjustment results, the lead time and target length are continuously monitored and updated, taking into account billet speed fluctuations and equipment response delays, so that the standard length can dynamically match the real-time status. A continuous rolling comparison is performed between the real-time measured length and the standard length to track the billet end position. When the real-time length approaches or reaches the standard length, a cutting control signal is generated. The rolling tracking process includes real-time data acquisition, smoothing, and error correction, eliminating the impact of short-term fluctuations in length measurement on trigger judgment. The control signal is immediately sent to the cutting equipment interface after generation, ensuring that the cutting action begins before the billet end precisely reaches the standard length position. This rolling tracking mechanism allows for high-frequency updates, and the response speed meets the conditions for high-speed billet movement, ensuring that the cutting operation remains synchronized with the billet length. After the cutting control signal is sent to the cutting equipment, the equipment executes start, cut, and stop actions to achieve standard-length billet cutting. Real-time monitoring of the equipment status and feedback signals confirms that the cutting action was completed as expected. During the cutting process, length measurement information and equipment status are continuously collected to ensure that the position of the cut end precisely matches the standard length. After the entire cutting operation is completed, the length data and cutting events are synchronously recorded in the operation log as a reference for subsequent quality control and production analysis, achieving high-precision length setting and automated closed-loop control of the continuous casting billet.

[0046] In this embodiment, a high-precision length-fixing system for continuously cast steel billets is provided, used to execute the high-precision length-fixing method for continuously cast steel billets as described above, including: The image acquisition module is used to install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; The image preprocessing module is used to preprocess the original image stream to obtain an enhanced image stream; and to determine the end position of the billet based on the enhanced image stream. The calculation module is used to perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; The error correction module is used to correct the image measurement motion error of the real-time length value and generate a corrected length value; The length-fixing module is used to extract the target length-fixing value and perform real-time rolling tracking of the correction length value. When the target length-fixing value is detected to be equal to the standard length-fixing value, the cutting device is driven to complete the length-fixing cut.

[0047] Therefore, the embodiments should be considered as exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of the equivalents of the application are intended to be included within the invention.

[0048] The above description is merely a specific embodiment of the present invention, enabling those skilled in the art to understand or implement it. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein are implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the present invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features of the invention herein.

Claims

1. A high-precision sizing method for reinforcing steel continuous cast billets, characterized in that, Includes the following steps: Step S1: Install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; Step S2: Perform image preprocessing on the original image stream to obtain an enhanced image stream; Determining the end position of the billet based on enhanced image stream; Step S3: Perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; Step S4: Perform image measurement motion error correction on the real-time length value to generate a corrected length value; Step S5: Extract the target fixed length and perform real-time rolling tracking of the corrected length value. When the target fixed length is detected to be equal to the standard fixed length, drive the cutting equipment to complete the fixed length cutting.

2. The high-precision sizing method of a reinforced continuous cast billet according to claim 1, characterized by, The specific steps of step S1 are as follows: Install an industrial-grade infrared camera system above the output roller conveyor of the production line; Obtain the intrinsic and extrinsic parameters of the infrared camera and perform camera calibration. The calibrated industrial-grade infrared camera system is synchronized with the continuous casting billet production system using industrial communication protocol timing, and continuous image acquisition is performed to obtain the original image stream.

3. The high-precision sizing method of a reinforced continuous cast billet according to claim 2, characterized by, The industrial-grade infrared camera system includes a camera unit and a main unit; The camera component includes an industrial-grade high-resolution infrared camera lens, a stainless steel water-cooled camera housing, a double-shielded and double-insulated video cable, and a shielded cable for the camera. The industrial-grade infrared camera system includes a high-definition fixed-size camera, a color LCD monitor, a mouse, a keyboard, and a cabinet.

4. The high-precision sizing method of a reinforced continuous cast billet according to claim 1, characterized by, The specific steps of step S2 are as follows: Perform grayscale conversion on the original image stream to obtain a grayscale image stream; An adaptive filtering process is applied to the grayscale image stream to obtain a filtered image stream; High-temperature radiation brightness fluctuation identification is performed on the filtered image stream to obtain brightness fluctuation information; Adaptive brightness compensation processing is performed based on brightness fluctuation information to generate a brightness-compensated image stream. Edge gradient enhancement is performed on the brightness-compensated image stream to obtain an enhanced image stream; The end position of the billet is determined based on the enhanced image stream.

5. The high-precision sizing method of a reinforced continuous cast billet according to claim 4, characterized by, The specific steps for determining the billet end position based on the enhanced image stream are as follows: High-temperature background separation is performed on the enhanced image stream to obtain a separated image; The isolated noise point filtering and region hole filling are performed on the separated image to obtain the target region image of the steel billet. Edge detection is performed on the target area image of the steel billet to extract contour boundary information; Based on the contour boundary information, contour localization and tracking are performed to obtain the overall contour line; Identify the direction of billet movement and determine the position of the billet end based on the overall outline.

6. The high-precision sizing method of reinforced continuous cast billets according to claim 1, characterized in that, The specific steps of step S3 are as follows: The positional change of the enhanced image stream is calculated based on the end position of the steel billet to obtain the pixel change. Establish transformation coefficients between the image pixel coordinate system and the actual physical space coordinate system based on camera intrinsic and extrinsic parameters; Based on the conversion coefficient, the pixel change is converted into a pixel displacement to obtain the actual length increment; The actual length increment is calculated over time and then smoothed to obtain the real-time length value of the billet.

7. The high-precision sizing method of reinforced continuous cast billets according to claim 1, characterized in that, The specific steps of step S4 are as follows: The billet running speed is calculated based on the speed sensor of the output roller conveyor of the production line; The theoretical displacement is calculated based on the billet's running speed to obtain the theoretical displacement per unit time. Based on the theoretical displacement, the motion error of the enhanced image stream is calculated to obtain the error coefficient; Motion compensation correction is performed on the real-time length value based on the error coefficient to generate a corrected length value.

8. The high-precision sizing method of a reinforced continuous cast billet according to claim 1, characterized by, The specific steps of step S5 are as follows: Identify the cutting action time of the cutting equipment; calculate the cutting extraction amount based on the cutting action time and the billet running speed to obtain the lead time; Extract the target length based on the work log; The target length is adjusted based on the lead time to obtain the standard length. Based on the standard fixed length, the standard fixed length is tracked in real time. When the target fixed length is detected to be equal to the standard fixed length, a cutting control signal is generated. The cutting control signal drives the cutting equipment to complete the fixed-length cutting.

9. A high-precision sizing system for reinforcing billets, characterized in that it comprises: The method for performing high-precision length determination of continuously cast steel billets as described in claim 1 includes: The image acquisition module is used to install an industrial-grade infrared camera system above the output roller conveyor of the production line to continuously acquire images and obtain the raw image stream; The image preprocessing module is used to preprocess the original image stream to obtain an enhanced image stream; and to determine the end position of the billet based on the enhanced image stream. The calculation module is used to perform pixel displacement conversion calculation based on the end position of the steel billet to obtain the real-time length value of the steel billet; The error correction module is used to correct the image measurement motion error of the real-time length value and generate a corrected length value; The length-fixing module is used to extract the target length-fixing value and perform real-time rolling tracking of the correction length value. When the target length-fixing value is detected to be equal to the standard length-fixing value, the cutting device is driven to complete the length-fixing cut.

Images