Method, device and system for controlling tail length of flying shear head

Abstract

This invention discloses a method, device, and system for controlling the head and tail cutting length of a flying shear, belonging to the field of metallurgical automation control technology. The method includes: acquiring an image sequence from the start of the flying shear to the complete cutting of the rolled piece; performing contour connectivity analysis on the infrared images to screen out candidate key frames where the rolled piece is completely separated by the shear blade; acquiring the candidate key frames and the visible light images corresponding to multiple adjacent frames before and after the candidate key frames as images to be verified; inputting the images to be verified into a trained rolled piece state classification model to obtain a target image at the moment when the rolled piece is completely cut by the shear blade; calculating the actual head and tail cutting length of the rolled piece based on the target image; and correcting the target setting length for the next cutting by the flying shear based on the deviation between the actual head and tail cutting length and the preset target length. This solves the technical problem in related technologies that it is impossible to measure the head and tail cutting length of a flying shear in real time and accurately.

Description

Technical Field

[0001] This invention relates to the field of metallurgical automation control technology, and in particular to a method, device and system for controlling the cutting length of a flying shear head and tail. Background Technology

[0002] Long steel products are an important category of steel products, generally referring to steel products whose length is much greater than their cross-sectional dimensions, mainly including bars, wire rods, and sections. Long products are widely used in construction, infrastructure, and various manufacturing industries. A typical production process involves hot-rolled steel billets from self-heating furnaces or continuous casting through roughing mills, intermediate mills, and subsequent mills on a long product production line to produce qualified finished products. Flying shears are installed between the mills on the long product production line for cutting off the ends of rolled products and handling accidental breakage. After the billets have been rolled through multiple mills, the performance and internal quality of the ends deteriorate. In extreme cases, internal cracks may appear at the ends, posing a safety hazard to downstream rolling processes. Flying shears are needed to remove the problematic ends to facilitate smooth rolling by subsequent mills. The cutting accuracy of the end length is a crucial factor affecting the yield of the production line. Precise measurement of the end length is one of the key factors ensuring high-precision cutting by the flying shears.

[0003] Existing measurement methods include manual and automatic methods. The manual method involves manually removing the cut steel head and tail from the hopper, allowing them to cool, and then measuring their length using a tape measure or other measuring tools. The flying shear operator then adjusts the next cutting length based on the actual cutting length. The automatic method involves automatically calculating the actual head / tail cutting length based on the hot metal detector before the flying shear and the head and tail speeds of the rolled piece.

[0004] Specifically, such as Figure 1 As shown, the distance from the pre-shear hot metal detector (heat detector) HMD_1 to the center line of the flying shear is L1; the distance from HMD_1 to the head / tail of the rolled piece during shearing is L2; ​​the actual head cutting length is L3; the actual tail cutting length is L4; the head rolling speed is V1; and the tail rolling speed is V2. For the above automatic measurement method, refer to... Figure 2 The diagram shown illustrates the calculation of the head / tail length of the flying shear. The head rolling speed V1 = MT_SPD1 * 3.1416 * WD1 / Gear1, and the tail rolling speed V2 = MT_SPD2 * 3.1416 * WD2 / Gear2 / k. MT_SPD1 is the upstream mill motor speed, WD1 is the upstream mill work roll diameter, Gear1 is the upstream mill speed ratio, MT_SPD2 is the downstream mill motor speed, WD2 is the downstream mill work roll diameter, Gear2 is the downstream mill speed ratio, and k is the downstream mill elongation coefficient.

[0005] The actual head cutting length L3 = L2 - L1 and the actual tail cutting length L4 = L1 - L2 are calculated based on the above measurement data. For the head cutting, L2 = V1 * (t2 - t1), where t1 is the moment the head of the rolled piece reaches the heat inspection HMD_1, and t2 is the moment the shear blade closes for shearing. For the tail cutting, L2 = V2 * (t3 - t4), where t3 is the moment the tail of the rolled piece leaves the heat inspection HMD_1, and t4 is the moment the shear blade closes for shearing.

[0006] Based on the existing measurement methods, on the one hand, manual measurement methods suffer from significant detection lag and cannot measure in real time for each cut; on the other hand, automatic measurement methods rely on the accuracy of the head and tail speeds of the rolled piece and the stability of the signal from the thermal metal detector. However, the actual head and tail speeds of the rolled piece are affected by steel temperature and roll wear, resulting in large fluctuations in the measured head and tail shearing lengths compared to the actual shearing lengths. Therefore, the measurement methods in related technologies have the technical problem of being unable to measure the head and tail shearing lengths of the flying shear in real time and accurately. Summary of the Invention

[0007] One of the technical problems that this invention aims to solve is the inability of existing technologies to accurately measure the cutting length of the flying shear head in real time.

[0008] To address the aforementioned technical problems, in a first aspect, embodiments of the present invention provide a method for controlling the head and tail cutting length of a flying shear. The method includes: acquiring an image sequence from the start of the flying shear to the complete cutting of the workpiece, wherein the image sequence includes infrared images and visible light images; performing contour connectivity analysis on the infrared images to filter out candidate keyframes where the workpiece is completely separated by the shear blade; acquiring the candidate keyframes and visible light images corresponding to multiple adjacent frames before and after the candidate keyframes as images to be verified; inputting the images to be verified into a trained workpiece state classification model to obtain a target image at the moment when the workpiece is completely cut by the shear blade; calculating the actual head and tail cutting length of the workpiece based on the target image; and correcting the target setting length for the next cutting by the flying shear based on the deviation between the actual head and tail cutting length and the preset target length.

[0009] Optionally, acquiring the image sequence from the start of the flying shear to the complete cutting of the rolled piece includes: acquiring the trigger signal generated when the position detection device detects the head or tail of the rolled piece and the rolled piece speed; determining the flying shear start distance based on the rolled piece speed and a preset flying shear response time, wherein the flying shear start distance is the distance the rolled piece travels within the preset flying shear response time, and the preset flying shear response time is the duration from the start of the flying shear to the closing of the shear blade; responding to the trigger signal, based on the rolled piece speed, the flying shear start distance, the target set length for this cutting, and the distance between the position detection device and the shear blade closing point... The flying shear start delay is determined by a preset distance between the two parts, including a head-cutting start delay and a tail-cutting start delay. The image acquisition delay is determined based on the preset flying shear response time, the workpiece cutting time, and the image acquisition response time. The workpiece cutting time is the duration from the shear blade contacting the workpiece to the complete cutting of the workpiece, and the image acquisition response time is the response time of the vision system from receiving the image acquisition command to starting image acquisition. After detecting the workpiece, the flying shear is started after the preset start delay, and after the flying shear is started, the vision system is triggered to acquire an image sequence after the preset image acquisition delay.

[0010] Optionally, the step of inputting the image to be verified into the rolling mill state classification model and outputting the target image of the moment when the rolling mill is completely cut by the shear blade includes: inputting the image to be verified into the trained rolling mill state classification model to obtain the rolling mill state classification result corresponding to each frame of the image, wherein the rolling mill state classification result includes a first confidence score of the rolling mill in the uncut state and a second confidence score of the rolling mill in the completely cut state; obtaining a confidence score sequence of the images to be verified arranged in chronological order based on the second confidence score of each frame of the image; identifying abrupt frames in the confidence score sequence where the second confidence score changes from below a first threshold to above a second threshold, wherein the first threshold is less than the second threshold; and determining the image of the abrupt frame as the target image of the moment when the rolling mill is completely cut by the shear blade.

[0011] Optionally, calculating the actual head and tail length of the rolled piece based on the target image includes: obtaining the calibration physical length of the calibration mark set at the shear blade closure point, and the calibration pixel distance of the calibration mark in the visible light image; determining the pixel equivalent based on the calibration physical length and the calibration pixel distance, wherein the pixel equivalent is the actual physical length represented by each pixel in the target image; extracting the rolled piece contour from the infrared image of the target image, mapping the rolled piece contour to the corresponding visible light image through calibration parameters to obtain a fused image; calculating the pixel distance from the farthest end of the head and tail to the shear blade dividing line in the fused image; and calculating the actual head and tail length of the rolled piece based on the pixel distance and the pixel equivalent.

[0012] Optionally, the step of correcting the target length for the next cut by the flying shear based on the deviation between the actual head and tail length and the preset target length includes: determining the current cutting deviation based on the actual head and tail length and the preset target length; weighting and accumulating the current cutting deviation with multiple historical cutting deviations to obtain a cumulative cutting error; and correcting the preset target length based on the cumulative cutting error, and using the corrected length as the target length for the next cut by the flying shear.

[0013] On the other hand, the present invention also provides a control device for the cutting length of the flying shear head and tail, the device comprising: a first acquisition module for acquiring an image sequence from the start of the flying shear to the complete cutting of the workpiece, wherein the image sequence includes infrared images and visible light images; a filtering module for performing contour connectivity analysis on the infrared images to filter out candidate keyframes in which the workpiece is completely separated by the shear blade; a second acquisition module for acquiring the candidate keyframes and visible light images corresponding to multiple adjacent frames before and after the candidate keyframes as images to be verified; an input module for inputting the images to be verified into a trained workpiece state classification model to obtain a target image at the moment when the workpiece is completely cut by the shear blade; a calculation module for calculating the actual cutting length of the workpiece head and tail based on the target image; and a correction module for correcting the target setting length of the next cutting by the flying shear according to the deviation between the actual cutting length and the preset target length.

[0014] On the other hand, the present invention also provides a control system for the cutting length of the head and tail of a flying shear, the system comprising: a calibration plate, removably mounted at the closing reference position of the flying shear blade for vision system shutdown calibration; a position detection device for generating a trigger signal when the head of the rolled piece arrives or the tail leaves; a shearing control module connected to the position detection device for receiving the trigger signal and generating a flying shear start command and an image acquisition command based on the position trigger signal, the target set length for this shearing, and preset parameters, the preset parameters including a flying shear start distance, a preset flying shear response time, a rolled piece cutting time, and an image acquisition response time; and a flying shear drive control module connected to the shearing control module and the flying shear drive motor for receiving... The system receives the flying shear start command sent by the shearing control module, performs closed-loop speed and torque control on the flying shear drive motor, and drives the flying shear to perform the shearing action; the vision system, located on the flying shear operation side and connected to the shearing control module, is used to receive the image acquisition command sent by the shearing control module, and acquire an image sequence of the flying shear shearing area, the image sequence including infrared images and visible light images; the control device described above is connected to the vision system and the shearing control module respectively, and is used to receive the image sequence, calculate the actual head and tail length based on the target image, and generate the target set length for the next shearing based on the deviation between the actual head and tail length and the preset target length, and output it to the shearing control module.

[0015] On the other hand, the present invention also provides an electronic device, including: a processor, and a memory communicatively connected to the processor; the processor executes computer execution instructions stored in the memory to implement the control method as described above.

[0016] On the other hand, the present invention also provides a machine-readable storage medium storing instructions for causing a machine to execute the control method described above.

[0017] On the other hand, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the control method described above.

[0018] The above technical solution employs infrared and visible light cameras positioned on the flying shear's operating side to acquire images of the shearing process in real time. Candidate keyframes are screened through infrared image connectivity analysis, and the visible light images are further validated using a deep learning model to accurately pinpoint the target image where the rolled piece is completely cut. The actual head and tail lengths are then calculated based on calibrated pixel equivalents, and the next shearing length is adjusted using a weighted cumulative error correction. This achieves online real-time measurement and closed-loop control of the flying shear head and tail lengths, effectively avoiding misjudgments caused by environmental interference and improving the accuracy of keyframe selection. Simultaneously, the fusion of infrared and visible light images ensures the accuracy of length measurement; through error accumulation and weighted correction, the actual shearing length quickly converges to the target value. This solves the technical problem of not being able to measure the flying shear head and tail lengths accurately in real time in related technologies, achieving real-time and accurate measurement and control of the flying shear head and tail lengths, significantly improving the shearing accuracy and production line yield.

[0019] Other features and advantages of the embodiments of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0020] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings: Figure 1 This is a schematic diagram of the flying shear process location provided by the existing solution; Figure 2 This is a schematic diagram of the calculation of the head / tail length of the flying shear provided by the existing solution; Figure 3 This is a flowchart illustrating a method for controlling the cutting length of a flying shear head and tail according to an embodiment of the present invention; Figure 4 This is a schematic diagram of camera length measurement calibration provided by an embodiment of the present invention. Figure 5 This is a schematic diagram (top view) of the flying shear head / tail length measurement provided by an embodiment of the present invention. Figure 6 This is a schematic diagram of the flying shear head length detection provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the flying shear tail length detection provided by an embodiment of the present invention; Figure 8 This is a schematic diagram of pixel measurement of the cutting length of the flying shear head and tail provided in an embodiment of the present invention; Figure 9 This invention provides a control device for the cutting length of the head and tail of a flying shear. Figure 10This is a control system diagram for the cutting length of the flying shear head and tail provided in an embodiment of the present invention.

[0021] Figure Labels 1: Upstream rolling mill; 2: Flying shear inlet guide groove; 3: Rolled workpiece; 4: Shear blade; 5: Flying shear housing; 6: Flying shear drive motor; 7: Binocular industrial camera; 8: Downstream rolling mill; 9: Calibration plate; Explanation of key terms Rolled products: Rolled products refer to steel billets that have undergone plastic processing through a rolling mill. They are the core objects of plastic deformation in the steel production process, and their processing directly affects the product's dimensional accuracy, geometry, surface quality, and microstructure.

[0022] HMD: Hot Metal Detector, used to detect the presence of rolled metal.

[0023] HMI: Human-Machine Interface. Detailed Implementation

[0024] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0025] It should be noted that the acquisition, transmission, storage, use, and processing of data in the technical solution of this application all comply with relevant laws and regulations. In the embodiments of this application, certain existing industry solutions such as software, components, and models may be mentioned. These should be considered exemplary, intended only to illustrate the feasibility of implementing the technical solution of this application, and do not imply that the applicant has already used or necessarily used such solutions.

[0026] Method Implementation Examples This invention provides a method for controlling the cutting length of the head and tail of a flying shear. Figure 3 This is a flowchart illustrating a method for controlling the cutting length of a flying shear head and tail according to an embodiment of the present invention. Figure 3 As shown, the method includes steps S101 to S106.

[0027] Step S101: Acquire an image sequence from the start of the flying shear to the complete cutting of the rolled piece, wherein the image sequence includes infrared images and visible light images; Specifically, in order to accurately capture the moment when the rolled piece is completely cut off by the shear blade, it is necessary to accurately trigger the vision system to acquire images within the critical time window based on the flying shear control timing and rolled piece position information.

[0028] In an optional embodiment, an image sequence is acquired from the start of the flying shear until the rolled piece is completely cut off, through the following steps S1011 to S1015: Step S1011: Obtain the trigger signal generated by the position detection device when the head of the rolled piece arrives or the tail of the rolled piece leaves, as well as the speed of the rolled piece; The aforementioned position detection device is installed on the inlet side of the flying shear. The straight-line distance between its installation position and the flying shear blade is a preset fixed value. This position detection device is preferably a hot metal detector. When the head of the rolled piece reaches the position detection device or the tail of the rolled piece leaves the position detection device, the device generates a trigger signal and sends it to the shearing control module. Specifically, if a head-cutting operation is currently being performed, the head-cutting process is triggered when the head of the rolled piece reaches the position detection device; if a tail-cutting operation is currently being performed, the tail-cutting process is triggered when the tail of the rolled piece leaves the position detection device. Simultaneously, the shearing control module acquires the current running speed of the rolled piece, which can be calculated based on the motor speed, work roll diameter, and speed ratio of the upstream mill.

[0029] Step S1012: Based on the rolling speed and the preset flying shear response time, determine the flying shear start distance, wherein the flying shear start distance is the distance the rolling piece moves forward within the preset flying shear response time, and the preset flying shear response time is the duration from the start of the flying shear to the closing of the shear blade; The preset flying shear response time is the duration from the start of the flying shear to the closing of the shear blades, including the control system response time and the mechanical system response time. Specifically, the control system response time is the time from when the shear control module issues the start command to when the flying shear drive control module starts driving the motor; the mechanical system response time is the time from when the motor starts running to when the shear blades are fully closed. The flying shear starting distance is the distance the workpiece travels at its current speed within the preset flying shear response time, which is the product of the workpiece speed and the preset flying shear response time.

[0030] Step S1013: In response to the trigger signal, determine the flying shear start delay based on the rolling speed, flying shear start distance, target set length for this shearing, and preset distance between the position detection device and the shear blade closing point. The flying shear start delay includes head cutting start delay and tail cutting start delay. For the above steps, the target length set for this cut is the instruction length used in this cut. During the first cut, this length is the preset target length (preset initial value) of the system. From the second cut onwards, this length is a value calculated by the control device based on the measured deviation of the previous cut, using the preset target length as a basis. That is, target length set = preset target length + cumulative error, where the cumulative error is the weighted sum of multiple cut deviations (the specific calculation method is described in step S106 below). This value is pre-stored in the cut control module for use in this cut. It should be noted that the preset target length, as a fixed expected value set by the system, does not change with the number of cuts. Each correction is to the target length set for actual execution.

[0031] Specifically, the aforementioned flying shear start-up delay includes the head-cutting start-up delay and the tail-cutting start-up delay, which are calculated as follows: The head-cutting start delay is the time interval from the trigger signal at the head of the rolled piece to the start of the flying shear, and the calculation formula is as follows: t_dly_head=(HMDX+LcutRef_n-D_str) / V, where HMDX is the preset distance between the position detection device and the shear blade closing point, LcutRef_n is the target setting length for this shearing, D_str is the flying shear start distance, and V is the rolling speed; The tail-cutting start delay is the time interval from the trigger signal at the tail of the rolled piece to the start of the flying shear. The calculation formula is: t_dly_tail=(HMDX-LcutRef_n-D_str) / V.

[0032] The above calculations ensure that after the flying shear is started, the closing position of the shear blade is precisely aligned with the head or tail position to be removed on the rolled piece.

[0033] Step S1014: Determine the image acquisition delay based on the preset flying shear response time, the workpiece cutting time, and the image acquisition response time. The workpiece cutting time is the duration from when the shear blade contacts the workpiece to when the workpiece is completely cut off. The image acquisition response time is the response time of the vision system from receiving the image acquisition command to starting image acquisition. Step S1015: After detecting the head or tail of the rolled piece, the flying shear is started after a start delay, and after the flying shear is started, the vision system is triggered to acquire an image sequence after an image acquisition delay.

[0034] For steps S1014-S1015 above, the image acquisition response time is the response time of the vision system from receiving the image command to starting to acquire the first frame of image. Furthermore, the image acquisition delay is determined according to the following formula: t_dly_img=t_str - t_cut - t_ph, where t_str is the preset flying shear response time, t_cut is the workpiece cutting time, and t_ph is the image acquisition response time.

[0035] After detecting the head or tail of the rolled piece, the shearing control module first activates the flying shear after a start delay, and then sends an image acquisition command to the vision system after an image acquisition delay, starting from the moment the flying shear is activated. This delay control ensures that the vision system promptly acquires multiple image sequences from the moment the rolled piece is first sheared to when it is completely severed, thus guaranteeing that the acquired image sequences include keyframes showing the rolled piece being completely cut.

[0036] Specifically, the vision system is located on the flying shear operating side and includes a binocular industrial camera consisting of an infrared camera and a visible light camera. After internal and external parameter calibration and registration, the lenses of the two cameras are aimed at the shear blade closure point of the flying shear housing, with the field of view covering the rolled workpiece, the shear blade, and the flying shear inlet guide groove area. After receiving the image acquisition command, the vision system triggers the infrared camera and the visible light camera to start continuously acquiring image sequences. When the current shearing is completed, the shearing control module sends a photo-taking stop command to the vision system, ending the current image acquisition, thereby obtaining the infrared and visible light images during the current shearing process.

[0037] Step S102: Perform contour connectivity analysis on the infrared image to screen out candidate keyframes in which the rolled piece is completely separated by the shear blade; Specifically, in the acquired image sequence, it is necessary to identify the frame from a large number of images where the rolled piece is completely severed by the shear blade, serving as the time reference for subsequent precise length measurement. This is achieved by performing contour connectivity analysis on the infrared images to filter out candidate keyframes where the rolled piece is completely separated by the shear blade, using the following method: The acquired infrared image sequence was analyzed frame by frame. Because the infrared camera clearly images the high-temperature rolled piece, the piece appears as a bright area in the infrared image, forming a clear contrast with the background environment. This allows for accurate segmentation of the piece's outline from the background. For each frame of the infrared image, image segmentation algorithms can be used to extract the piece's outline region. For example, a threshold segmentation method can be used, setting a segmentation threshold based on the temperature difference between the rolled piece and the background, and then binarizing the infrared image to extract the piece region. Alternatively, an edge detection algorithm can be used to identify the boundary between the rolled piece and the background, thereby obtaining the piece's outline. The image segmentation algorithm used here is not limited.

[0038] Furthermore, after extracting the workpiece contour, connectivity analysis is performed on the contour region. When the shear blade contacts the workpiece but has not yet cut it off, the workpiece contour remains continuous at the shear blade position, belonging to the same connected region. When the workpiece is completely cut off by the shear blade, the originally continuous workpiece contour shows gaps at the shear blade position, forming two independent connected regions. By comparing the changes in the number of connected regions of the workpiece contour in adjacent frames, when the number of connected regions in a certain frame changes from 1 in the previous frame to 2 in the current frame, that frame is marked as a candidate keyframe where the workpiece has been completely separated by the shear blade.

[0039] Step S103: Obtain candidate keyframes and visible light images corresponding to multiple adjacent frames before and after the candidate keyframes, as images to be verified; Specifically, the candidate keyframes initially pinpointed the moment when the rolled piece was completely cut off. However, due to the complex on-site environment, connectivity analysis based solely on infrared images may lead to misjudgments, necessitating secondary verification using visible light images. This verification process involves acquiring the candidate keyframes and the visible light images corresponding to multiple adjacent frames before and after them, as the images to be verified, through the following method: Using the candidate keyframes selected in step S102 as a benchmark, the visible light image corresponding to the infrared image of that frame is obtained. Simultaneously, considering that deep learning models can more accurately determine the cut-off moment by analyzing the confidence changes between consecutive frames when analyzing image sequences, it is also necessary to obtain the visible light images corresponding to multiple adjacent frames before and after the candidate keyframe. For example, taking the candidate keyframe as the center, N frames are taken forward and N frames are taken backward, and the visible light images corresponding to these frames are combined as the image sequence to be verified. The value of N can be set according to the camera frame rate.

[0040] Step S104: Input the image to be verified into the trained rolling mill state classification model to obtain the target image of the rolling mill at the moment when it is completely cut off by the shear blade. Specifically, candidate keyframes selected through infrared images may be misjudged, requiring a second verification of the visible light image sequence using a deep learning model to accurately pinpoint the moment when the rolled piece is completely cut off.

[0041] In an optional embodiment, the image to be verified is input into a rolling mill state classification model, and the target image at the moment when the rolling mill is completely cut off by the shear blade is output. This includes: inputting the image to be verified into a trained rolling mill state classification model to obtain the rolling mill state classification result corresponding to each frame of the image, wherein the rolling mill state classification result includes a first confidence score for the rolling mill in an uncut state and a second confidence score for the rolling mill in a completely cut state; obtaining a confidence score sequence of the images to be verified arranged in chronological order based on the second confidence score of each frame of the image; identifying abrupt frames in the confidence score sequence where the second confidence score changes from below a first threshold to above a second threshold, wherein the first threshold is less than the second threshold; and determining the image of the abrupt frame as the target image at the moment when the rolling mill is completely cut off by the shear blade.

[0042] The process involves inputting the images to be verified into a trained rolling mill state classification model to obtain the rolling mill state classification result for each frame. This model, which can employ deep learning classification, is pre-trained using a large number of labeled flying shear head and tail images, specifically designed to identify "uncut" and "completely cut" states. The rolling mill state classification result includes a first confidence score for the uncut state and a second confidence score for the completely cut state; the sum of these two scores is 1. Further, based on the second confidence score of each frame, a confidence score sequence of the images to be verified, arranged chronologically, is obtained. Within this sequence, frames where the second confidence score changes from below a first threshold to above a second threshold are identified, where the first threshold is less than the second threshold. For example, the first threshold can be set to 0.2, and the second threshold to 0.8. Finally, the image of the frame with the change in confidence score is determined as the target image at the moment the rolling mill is completely cut by the shear blade. This combination of infrared initial screening and secondary verification using a deep learning classification model improves the accuracy of keyframe selection, providing a reliable time reference for subsequent length measurements.

[0043] Step S105: Calculate the actual head and tail lengths of the rolled piece based on the target image; Specifically, after accurately locking the target image of the rolled piece being completely cut off by the shear blade, it is necessary to measure the actual length of the cut-off head and tail based on the target image to provide a basis for subsequent shearing correction.

[0044] In an optional embodiment, the actual head and tail length of the rolled piece is calculated based on the target image, including: obtaining the calibration physical length of the calibration mark set at the shear blade closure point, and the calibration pixel distance of the calibration mark in the visible light image; determining the pixel equivalent based on the calibration physical length and the calibration pixel distance, wherein the pixel equivalent is the actual physical length represented by each pixel in the target image; extracting the rolled piece contour from the infrared image of the target image, and mapping the rolled piece contour to the corresponding visible light image through calibration parameters to obtain a fused image; calculating the pixel distance from the farthest end of the head and tail to the shear blade dividing line in the fused image; and calculating the actual head and tail length of the rolled piece based on the pixel distance and the pixel equivalent.

[0045] In this embodiment, before the system is put into use, a movable calibration sign of length Lc is pre-set at the closure point of the shear blade for pixel calibration of the camera length measurement. Under sufficient lighting conditions, a visible light camera is used to photograph the calibration sign at the shear blade position, and the number of pixels PXc (calibrated pixel distance) along the length direction of the calibration sign is extracted from the camera image. Based on the calibrated physical length Lc and the calibrated pixel distance PXc, the pixel equivalent Kc is determined, which is the actual measured length of two adjacent pixels along the length direction. The calculation formula is Kc = Lc / PXc.

[0046] The profile of the rolled piece is extracted from the infrared image of the target image. Because infrared cameras provide clear imaging of high-temperature rolled pieces, the profile can be accurately segmented from the background, obtaining a precise and interference-free profile. Using pre-calibrated camera intrinsic and extrinsic parameters, the profile extracted from the infrared image is mapped onto the corresponding visible light image at the same time, and then superimposed to obtain a fused image. This fused image possesses both the accurate segmentation capability of the infrared image and the high-resolution characteristics of the visible light image, laying the foundation for precise measurement.

[0047] Furthermore, in the fused image, the extension lines of the upper and lower shear blades are used as the shear blade dividing lines to divide the image into the workpiece body side and the cut-off head and tail side. The farthest contour point of the cut-off head or tail of the workpiece is located, and the vertical distance from this point to the shear blade dividing line is measured, denoted as pixel distance PX3. Based on pixel distance PX3 and pixel equivalent Kc, the actual cut-off head and tail length Lp of the workpiece is calculated using the formula Lp = PX3 × Kc. In this step, the calculated actual cut-off head and tail length can be sent to the shearing control module for subsequent correction of the shearing set length.

[0048] Step S106: Based on the deviation between the actual head and tail cutting length and the preset target length, adjust the target setting length for the next cutting by the flying shear.

[0049] In this embodiment, the preset target length is the expected cutting length set by the system, and the target set length is the actual command length issued to the flying shear to perform the cutting operation. During the first cutting, the target set length is equal to the preset target length, and subsequent adjustments are made based on the deviation.

[0050] Specifically, based on the actual head and tail lengths Lp and the preset target length LcutTgt, the current shearing deviation Lerr is determined, calculated using the formula Lerr = LcutTgt - Lp. This current shearing deviation can be weighted and accumulated with multiple historical shearing deviations to obtain the cumulative shearing error LerrSum. The formula for weighted accumulation is LerrSum = k × ΣLerr_i, where i is the number of shearing operations and k is the weighting coefficient. The weighting coefficient k ranges from 0.5 to 1.0 and is used to adjust the influence of historical deviations on the current correction. The number of historical shearing deviations included in the accumulation can be set according to actual needs; for example, the maximum accumulation number can be set to 10, with rolling updates when this limit is exceeded. Further, based on the cumulative shearing error LerrSum, the preset target length is corrected to obtain the target length LcutRef_(n+1) for the next shearing operation, calculated using the formula LcutRef_(n+1) = LcutTgt + LerrSum. Finally, the corrected setting length LcutRef_(n+1) is sent to the shearing control module. During the next shearing operation, the flying shear will control the shearing according to the target setting length.

[0051] It should be noted that through the above-mentioned deviation accumulation and weighted correction, the actual head and tail cutting lengths can be quickly converged to near the preset target length, eliminating the fluctuations caused by single measurement deviations and improving the cutting accuracy of the flying shear and the yield of the production line.

[0052] Based on the above embodiments and optional embodiments, such as Figure 4 , Figure 5 , Figure 6 , Figure 7 and Figure 8 As shown, the present invention provides an optional embodiment, which is described in detail below. It should be noted that the following steps are described in detail using the head cutting length measurement as an example. Compared with the tail cutting length measurement method, the only difference in the method of measuring the tail cutting length is that the detection object is replaced by the tail of the rolled piece, and the corresponding tail cutting delay calculation is triggered according to the signal of the tail reaching the HMD. The subsequent image processing, length calculation and deviation correction steps are the same as those of the head cutting and will not be described again.

[0053] Before the system is put into use, camera length measurement calibration needs to be completed. For example... Figure 4 As shown, a movable calibration mark 9 with a length of Lc = 200mm is set at the closing point of the shear blade 4. After the binocular industrial camera 7 is installed and fixed, under sufficient lighting conditions, a visible light camera is used to photograph the calibration mark at the shear blade position. The number of pixels PXc = 2000 along the length direction of the calibration mark is extracted from the camera image. Based on the calibration physical length Lc and the calibration pixel distance PXc, the pixel equivalent Kc = Lc / PXc = 200 / 2000 = 0.1mm / pixel is calculated. This pixel equivalent represents the actual measured object length corresponding to two adjacent pixels along the length direction. The following explanation mainly focuses on calculating the cutting head length as an example.

[0054] S1: Acquire an image sequence from the start of the flying shear to the complete cutting of the rolled piece, wherein the image sequence includes infrared images and visible light images.

[0055] Based on the above steps, such as Figure 5 As shown, the vision system includes a binocular industrial camera 7 mounted on the operating side of the flying shear. This binocular industrial camera 7 consists of an infrared camera and a visible light camera, used to acquire image sequences of the flying shear cutting area. The two cameras undergo precise internal and external parameter calibration and registration, with their lenses aligned with the shear blade closure point of the flying shear housing 5, ensuring that the field of view can clearly capture the workpiece 3, the shear blade 4, and the flying shear inlet guide groove 2 area.

[0056] In this step, the image sequence is acquired as follows: First, the trigger signal when the hot metal detector (HMD) detects the rolled piece and the rolled piece speed V=2m / s are acquired. Specifically, during the head-cutting process, the head-cutting process is triggered when the head of the rolled piece reaches the HMD; during the tail-cutting process, the tail-cutting process is triggered when the tail of the rolled piece reaches the HMD. Second, based on the rolled piece speed and the preset flying shear response time, the flying shear start distance D_str=1m is determined, and the target setting length for this cut is LcutRef=0.2m; according to the preset distance HMDX=2m from the HMD to the shear blade closing point, the head-cutting start delay and the tail-cutting start delay are calculated respectively. The head-cutting startup delay is: t_dly_head = (HMDX + LcutRef - D_str) / V = ​​0.6s; The tail-cutting startup delay is: t_dly_tail = (HMDX - LcutRef - D_str) / V = ​​0.4s; Furthermore, based on the preset flying shear response time t_str=0.4s, workpiece cutting time t_cut=0.03s, and image acquisition response time t_ph=0.05s, the image acquisition delay is calculated: t_dly_img=t_str-t_cut-t_ph=0.32s, which means that the camera is triggered to acquire an image sequence based on the calculated image acquisition delay.

[0057] After the shearing control module detects that the head / tail has been cut, it sends a stop image capture command to the control device. The control device then sends a stop image capture command to both cameras and reads the infrared and visible light images captured by the cameras as the image sequence for this shearing process.

[0058] It should be noted that the infrared and visible light images acquired above were preprocessed separately. For the infrared images, temperature pseudo-color mapping and noise suppression were performed to enhance the contrast between the rolled piece and the background. For the visible light images, denoising and edge enhancement were performed to improve image clarity and edge information integrity. This preprocessing improves the robustness of subsequent image segmentation and recognition models.

[0059] S2: Perform contour connectivity analysis on the preprocessed infrared image to screen out candidate keyframes where the rolled piece is completely separated by the shear blade.

[0060] In the above steps, such as Figure 6 and Figure 7 The diagram shows the positional relationship between the rolled piece and the shear blade during head and tail cutting, respectively. Figure 6 This illustrates the state where the section to be removed from the head of the rolled piece is upstream of the shear blade during head cutting. Figure 7The image shows the state during tail cutting: the head of the rolled piece has passed the shear blade and entered the downstream mill, the downstream mill continues to traction the rolled piece, and the tail section to be cut is located upstream of the shear blade. In this step, the infrared image is analyzed frame by frame to extract the rolled piece contour and perform connectivity analysis. Because the infrared camera provides a clear image of the high-temperature rolled piece, the contour can be accurately segmented from the background. When the shear blade contacts the rolled piece but has not yet cut it off, the rolled piece contour remains continuous at the shear blade position, belonging to the same connected region; when the rolled piece is completely cut off by the shear blade, the originally continuous rolled piece contour has gaps at the shear blade position, forming two independent connected regions. By comparing the change in the number of connected regions of the rolled piece contour in adjacent frames, when the number of connected regions in a certain frame changes from 1 in the previous frame to 2 in the current frame, that frame is marked as a candidate keyframe where the rolled piece is completely separated by the shear blade.

[0061] S3: Obtain the candidate keyframe and the visible light images corresponding to its multiple adjacent frames before and after it, as the images to be verified.

[0062] In the above steps, the visible light image corresponding to the infrared image of the selected candidate keyframe is obtained based on the selected candidate keyframe. At the same time, the visible light images corresponding to the N frames before and after the candidate keyframe are obtained, which together serve as the image sequence to be verified. The value of N can be set according to the camera frame rate, for example, N=3.

[0063] S4: Input the image to be verified into the trained rolling mill state classification model to obtain the target image at the moment when the rolling mill is completely cut off by the shear blade.

[0064] In the above steps, the image to be verified is input into a deep learning classification model. This model is pre-trained with a large number of labeled flying shear head and tail images, and is specifically used to identify the two states of "uncut" and "completely cut". The model outputs two confidence scores for each frame of the image; the first confidence score corresponds to the "uncut state" and the second confidence score corresponds to the "completely cut state", and the sum of the two scores is 1.

[0065] Furthermore, based on the second confidence score of each image frame, a chronological confidence score sequence of the images to be verified is obtained. Within this sequence, frames whose second confidence score abruptly changes from below a first threshold (e.g., 0.2) to above a second threshold (e.g., 0.8) are identified and designated as target images. This combination of infrared initial screening and secondary verification using a deep learning classification model improves the accuracy of keyframe selection and provides a reliable time reference for subsequent length measurement.

[0066] It should be noted that the above-mentioned rolling mill status classification model was trained in the following way: a binary classification model based on convolutional neural networks was adopted. Specifically, the model can use a lightweight convolutional neural network structure (such as MobileNet or ResNet-18). The input of the model is a visible light image, and the output is the confidence score of two categories, "not cut" and "completely cut", with the sum of the two scores being 1. During training, visible light images of the flying shear cutting process were collected, and professionals labeled the status (not cut or completely cut) of each frame of the image, forming the training sample set. The images were normalized to 224×224 pixels, the batch size was set to 32, the initial learning rate was set to 0.001, and the model was trained end-to-end using the cross-entropy loss function and gradient descent method for a total of 30 epochs. The model with the highest accuracy on the validation set was selected as the final model. After training, the model was deployed in the control device for real-time inference.

[0067] S5: Calculate the actual cut length of the rolled piece based on the target image.

[0068] Because infrared cameras provide clear images of high-temperature rolled pieces, extracting the rolled piece contour from the infrared image of the target image yields a precise and interference-free contour. Therefore, in this step, as described above... Figure 4 As shown in the provided camera length calibration diagram, a calibration marker 9 is pre-set at the shear blade closure point. The marker length is Lc = 200mm, and the captured image has 2000 pixels (PXc = 2000). The pixel equivalent is calculated as Kc = Lc / PXc = 0.1mm / pixel. Using pre-calibrated camera intrinsic and extrinsic parameters, the workpiece contour extracted from the infrared image is mapped onto the corresponding visible light image at the same time, and the images are superimposed to obtain a fused image. This fused image possesses both the accurate segmentation capability of the infrared image for the workpiece contour and the high-resolution characteristics of the visible light image, laying the foundation for precise measurement.

[0069] like Figure 8 As shown, in the fused image, the extension lines of the upper and lower shear blades are used as the shear blade dividing line. Taking the head cutting as an example, the farthest contour point of the cut-off head of the rolled piece is found, and the vertical distance from this point to the shear blade dividing line is measured and recorded as the pixel distance PX3. Based on the pixel equivalent Kc = 0.1 mm / pixel obtained from the above calibration, the actual head cutting length L3 = PX3 × Kc is calculated. For example, if PX3 = 1500 pixels is measured in the first cut, then L3 = 150 mm. It should be noted that the calculation method of this step is also applicable to calculating the actual tail cutting length L4 = PX4 × Kc, where PX4 is the pixel distance from the farthest end of the tail cutting to the shear blade dividing line.

[0070] Finally, the actual head and tail lengths calculated above are sent to the shearing control module for subsequent shearing setting length correction.

[0071] S6: Based on the deviation between the actual cut length and the preset target length, adjust the target length for the next cut by the flying shear. In this step, the preset target length LcutTgt = 100mm is set. The actual cut length L3 is compared with the preset target length, and the deviation Lerr = LcutTgt - L3 for each cut is calculated. It should be noted that this calculation method is also applicable to the calculation of the tail of the rolled piece, which will not be elaborated here.

[0072] Specifically, after the first cut, the actual cut length L3 = 130mm, then the deviation Lerr_1 = -30mm; After the second cut, the actual cut length L3 = 115mm, then the deviation Lerr_2 = -15mm; After the third cut, the actual cut length L3 = 98 mm, so the deviation Lerr_3 = 2 mm.

[0073] Furthermore, the deviations from each shearing are weighted and accumulated, with a weighting coefficient k = 0.9 and a maximum rolling average of 10 times. The cumulative error calculation formula is LerrSum = k × ΣLerr_i, where i is the number of shearings. The weighting coefficient k here ranges from 0.5 to 1.0, and its value can be reasonably selected according to the actual process conditions.

[0074] Based on the above formula, calculate the cumulative error after each shearing: After the first shearing: LerrSum_1 = 0.9 × (-30) = -27mm; After the second shearing: LerrSum_2 = 0.9 × (-30-15) = 0.9 × (-45) = -40.5mm; After the third shearing: LerrSum_3 = 0.9 × (-30-15+2) = 0.9 × (-43) = -38.7mm.

[0075] The target length for the next shearing is adjusted based on the accumulated error, and the calculation formula is as follows: LcutRef_(n+1) = LcutTgt + LerrSum, where LcutRef_(n+1) is the target length set for the next cut. Therefore, the following calculation is performed: The target length for the second cut is: LcutRef_2 = 100 + (-27) = 83mm; The target length for the third cut is: LcutRef_3 = 100 + (-40.5) = 59.5mm; The target length for the fourth cut is: LcutRef_4 = 100 + (-38.7) = 61.3mm.

[0076] After the above multiple corrections, the actual cutting length can quickly approach the target cutting length of 100mm, thereby effectively improving the cutting accuracy of the flying shear.

[0077] The advantages and technical effects of the present invention are as follows: 1) By adopting a binocular vision scheme that combines infrared and visible light images, and by using a deep learning model to perform double verification on the moment of cutting, it is possible to accurately lock the target image of the rolled piece being completely cut off by the shear blade, effectively avoiding misjudgment caused by on-site environmental factors and improving the accuracy of key frame selection.

[0078] 2) By calibrating the pixel equivalent by calibrating the calibrator and mapping the rolled piece contour extracted from the infrared image to the visible light image for pixel measurement, the advantages of infrared image for accurate segmentation of high-temperature rolled piece contour are utilized, as well as the high resolution of visible light image, thus achieving accurate measurement of the head and tail length.

[0079] 3) By weighting and accumulating the deviations of multiple cuts based on the actual measured length and the preset target length, and then correcting the target length for the next cut, closed-loop control of the cutting length is achieved. This allows the actual head and tail cutting lengths to quickly converge to near the target length, effectively improving the cutting accuracy of the flying shear and the yield of the production line.

[0080] System Implementation Examples This invention also provides a device for controlling the cutting length of the head and tail of a flying shear, such as... Figure 9 As shown, the device includes: a first acquisition module 901, used to acquire an image sequence from the start of the flying shear to the complete cutting of the workpiece, wherein the image sequence includes infrared images and visible light images; a filtering module 902, connected to the first acquisition module 901, used to perform contour connectivity analysis on the infrared images and filter out candidate keyframes in which the workpiece is completely separated by the shear blade; a second acquisition module 903, connected to the filtering module 902, used to acquire the candidate keyframes and the visible light images corresponding to multiple adjacent frames before and after the candidate keyframes as images to be verified; an input module 904, connected to the second acquisition module 903, used to input the images to be verified into a trained workpiece state classification model to obtain a target image at the moment when the workpiece is completely cut by the shear blade; a calculation module 905, connected to the input module 904, used to calculate the actual head and tail lengths of the workpiece based on the target image; and a correction module 906, connected to the calculation module 905, used to correct the target setting length for the next cutting by the flying shear based on the deviation between the actual head and tail lengths and the preset target length.

[0081] This invention also provides a control system for the cutting length of the flying shear head and tail. The system includes: a calibration plate, removably mounted at the closing reference position of the flying shear blade for vision system shutdown calibration; a position detection device for generating a trigger signal when the rolled piece arrives; a shearing control module connected to the position detection device for receiving the trigger signal and generating a flying shear start command and an image acquisition command based on the position trigger signal, the target cutting length for this shearing, and preset parameters, including a flying shear start distance, a preset flying shear response time, a rolled piece cutting time, and an image acquisition response time; and a flying shear drive control module connected to the shearing control module and the flying shear drive motor for... The system receives the flying shear start command from the shearing control module, performs closed-loop speed and torque control on the flying shear drive motor, and drives the flying shear to perform the shearing action. The vision system, located on the flying shear operation side and connected to the shearing control module, receives the image acquisition command from the shearing control module and acquires an image sequence of the flying shear shear area, including infrared and visible light images. The control device, as described above, is connected to both the vision system and the shearing control module. It receives the image sequence, calculates the actual head and tail lengths based on the target image, and generates the target setting length for the next shearing based on the deviation between the actual head and tail lengths and the preset target length, and outputs it to the shearing control module.

[0082] In this embodiment, such as Figure 10 As shown, the system includes: a calibration sign 9, a position detection device, a vision system, a control device, a shearing control module, a flying shear drive control module, a flying shear drive motor 6, and an HMI (Human-Machine Interface) system. The connection relationships and collaborative workflows between the modules are as follows: First, system shutdown calibration is performed. Calibration tag 9 is removably positioned at the closed reference position of shear blade 4. The vision system acquires a visible light image containing calibration tag 9 and sends it to the control device. The control device calculates and stores the pixel equivalent based on the known physical length of calibration tag 9 and its pixel distance in the image, for subsequent pixel-to-physical conversion of the actual length. After calibration, calibration tag 9 is removed from the closed shear blade position, and the system enters normal production mode.

[0083] After entering production mode, the position detection device (preferably a hot metal detector HMD) is installed on the inlet side of the flying shear, with the straight-line distance between its installation position and the shear blade closing point set to a preset fixed value. When the workpiece 3 runs to this position detection device, the device generates a trigger signal and sends it to the shearing control module via hardwire or fieldbus. After receiving the trigger signal, the shearing control module generates two types of instructions based on the target length set for this shearing and preset parameters (including flying shear start distance, preset flying shear response time, workpiece cutting time, and image acquisition response time): a flying shear start instruction and an image acquisition instruction. The flying shear start instruction is sent to the flying shear drive control module via fieldbus, while the image acquisition instruction is sent to the vision system via Ethernet or a dedicated trigger line.

[0084] Upon receiving the flying shear start command, the flying shear drive control module immediately performs closed-loop speed and torque control on the flying shear drive motor 6, driving the flying shear to perform the shearing action. Simultaneously, the flying shear drive control module feeds back the motor status to the shear control module in real time. At the same time, after receiving the image acquisition command, the vision system triggers the binocular industrial camera 7 to begin continuously acquiring image sequences. This binocular industrial camera 7 consists of an infrared camera and a visible light camera, with its lens aimed at the shear blade closure point of the flying shear housing 5, covering the workpiece 3, the shear blade 4, and the flying shear inlet guide groove 2 area. The acquired infrared and visible light images are temporarily stored in the vision system's internal cache.

[0085] After the cropping is complete, the cropping control module sends a stop-image command to the vision system. Once the vision system stops acquiring images, it packages the infrared and visible light image sequences captured during the cropping process and sends them to the control device. Upon receiving the image sequences, the control device performs the following processing sequentially: Contour connectivity analysis is performed on infrared images to screen out candidate keyframes where the workpiece is completely separated by the shear blade; visible light images of the candidate keyframes and multiple adjacent frames before and after them are acquired and input into the workpiece state classification model to identify the target image at the moment when the workpiece is completely cut off; based on the target image and the pixel equivalent stored in the calibration stage, the actual head and tail cutting lengths are calculated; the actual lengths are compared with the preset target lengths to generate the corrected target setting length for the next shearing.

[0086] The control device sends the corrected shearing target length to the shearing control module via the fieldbus. Upon receiving the corrected value, the shearing control module stores it as the target length for the next shearing operation. When the next rolled piece reaches the trigger position, the shearing control module sends a flying shear start command to the flying shear drive control module, completing the shearing process.

[0087] In addition, the HMI human-machine interface system is connected to the shearing control module. Operators can preset parameters such as the shearing target length, weighting coefficient, and rolling average number of times through the HMI, and monitor the actual shearing length, deviation value, and corrected target setting length in real time.

[0088] Through the coordinated operation of the aforementioned calibration sign, vision system, control device, shearing control module, and flying shear transmission control module, the control system achieves online measurement and closed-loop control of the cutting length of the flying shear head and tail, effectively improving the shearing accuracy of the flying shear and the yield of the production line.

[0089] This invention also provides an electronic device, including: a processor and a memory communicatively connected to the processor; the processor executes computer execution instructions stored in the memory to implement the control method as described above.

[0090] This invention also provides a machine-readable storage medium storing instructions for causing a machine to execute the control method described above.

[0091] This invention also provides a computer program product that, when executed on a data processing device, is adapted to perform the initialization steps of the control method described above.

[0092] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0093] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. 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, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0094] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function 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 1 The function specified in one or more boxes.

[0095] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable 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.

[0096] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0097] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.

[0098] Computer-readable media includes both permanent and non-permanent, removable and non-removable media that can store information using any method or technology. Information can be computer-readable instructions, data structures, modules of programs, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transferable medium that can be used to store information accessible by a computing device. As defined herein, computer-readable media does not include transient computer-readable media, such as modulated data signals and carrier waves.

[0099] It should also be noted that the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0100] The above are merely embodiments of this application and are not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method of controlling the tail length of a flying shear head, characterized by, include: Acquire an image sequence from the start of the flying shear to the complete cutting of the rolled piece, wherein the image sequence includes infrared images and visible light images; Contour connectivity analysis was performed on the infrared images to screen out candidate keyframes in which the rolled piece was completely separated by the shear blade; The candidate keyframe and the visible light images corresponding to multiple adjacent frames before and after the candidate keyframe are obtained as images to be verified. The image to be verified is input into the trained rolling mill state classification model to obtain the target image of the rolling mill at the moment when it is completely cut off by the shear blade. Based on the target image, calculate the actual head and tail lengths of the rolled piece; and Based on the deviation between the actual head and tail cutting length and the preset target length, the target setting length for the next cutting by the flying shear is adjusted.

2. The method of claim 1, wherein, The acquisition of the image sequence from the start of the flying shear to the complete cutting of the rolled piece includes: The position detection equipment detects the trigger signal generated when the head of the rolled piece arrives or the tail of the rolled piece leaves, as well as the speed of the rolled piece; Based on the workpiece speed and the preset flying shear response time, the flying shear start distance is determined, wherein the flying shear start distance is the distance the workpiece travels within the preset flying shear response time, and the preset flying shear response time is the time from the start of the flying shear to the closing of the shear blade; In response to the trigger signal, the flying shear start delay is determined based on the rolling speed, the flying shear start distance, the target set length for this shearing, and the preset distance between the position detection device and the shear blade closing point. The flying shear start delay includes a head cutting start delay and a tail cutting start delay. The image acquisition delay is determined based on the preset flying shear response time, the workpiece cutting time, and the image acquisition response time. The workpiece cutting time is the duration from when the shear blade contacts the workpiece to when the workpiece is completely cut off. The image acquisition response time is the response time of the vision system from receiving the image acquisition command to starting image acquisition. After detecting the head or tail of the rolled piece, the flying shear is started after a start delay, and after the flying shear is started, the vision system is triggered to acquire an image sequence after an image acquisition delay.

3. The method of claim 1, wherein, The image to be verified is input into a trained rolling mill state classification model to obtain the target image of the rolling mill at the moment it is completely cut off by the shear blade, including: The image to be verified is input into the trained rolling mill state classification model to obtain the rolling mill state classification result corresponding to each frame of the image. The rolling mill state classification result includes a first confidence score for the rolling mill in the uncut state and a second confidence score for the rolling mill in the completely cut state. Based on the second confidence score of each frame, the confidence score sequence of the images to be verified is obtained in chronological order. In the confidence score sequence, abrupt frames are identified where the second confidence score changes from below a first threshold to above a second threshold, wherein the first threshold is less than the second threshold; The image of the mutation frame is determined as the target image at the moment when the rolled piece is completely cut off by the shear blade.

4. The method according to claim 1, characterized in that, The step of calculating the actual head and tail lengths of the rolled piece based on the target image includes: Obtain the calibration physical length of the calibration mark set at the closure of the shear blade, and the calibration pixel distance of the calibration mark in the visible light image; The pixel equivalent is determined based on the calibrated physical length and the calibrated pixel distance, wherein the pixel equivalent is the actual physical length represented by each pixel in the target image; The outline of the rolled piece is extracted from the infrared image of the target image, and the outline of the rolled piece is mapped to the corresponding visible light image through calibration parameters to obtain a fused image; In the fused image, the pixel distance from the farthest point of the cut-off head and tail to the cutting edge dividing line is calculated; The actual head and tail lengths of the rolled piece are calculated based on the pixel distance and the pixel equivalent.

5. The method according to claim 1, characterized in that, The step of correcting the target cutting length for the next cut based on the deviation between the actual head and tail cutting length and the preset target length includes: The cutting deviation is determined based on the actual head and tail lengths and the preset target length. The current shear deviation is weighted and accumulated with multiple historical shear deviations to obtain the cumulative shear error; and Based on the cumulative shearing error, the preset target length is corrected, and the corrected length is used as the target setting length for the next shear cut.

6. A control device for the cutting length of the head and tail of a flying shear, characterized in that, include: The first acquisition module is used to acquire an image sequence from the start of the flying shear to the complete cutting of the rolled piece, wherein the image sequence includes infrared images and visible light images; The filtering module is used to perform contour connectivity analysis on the infrared image and filter out candidate keyframes in which the rolled piece is completely separated by the shear blade. The second acquisition module is used to acquire the candidate keyframe and the visible light images corresponding to multiple adjacent frames before and after the candidate keyframe, as images to be verified. The input module is used to input the image to be verified into the trained rolling mill state classification model to obtain the target image of the rolling mill at the moment when it is completely cut off by the shear blade. The calculation module is used to calculate the actual head and tail lengths of the rolled piece based on the target image; and The correction module is used to correct the target length of the next cut by the flying shear based on the deviation between the actual head and tail length and the preset target length.

7. A control system for the cutting length of a flying shear head and tail, characterized in that, include: A calibration label is removably mounted at the closure reference position of the flying shear blade for use in vision system shutdown calibration; Position detection equipment is used to generate a trigger signal when the head of the rolled piece arrives or the tail leaves; The shearing control module is connected to the position detection device and is used to receive the trigger signal and generate a flying shear start command and an image acquisition command according to the trigger signal, the target set length for this shearing and preset parameters. The preset parameters include the flying shear start distance, the preset flying shear response time, the workpiece cutting time and the image acquisition response time. The flying shear drive control module is connected to the shear control module and the flying shear drive motor. It is used to receive the flying shear start command sent by the shear control module, perform closed-loop speed and torque control on the flying shear drive motor, and drive the flying shear to perform the shearing action. A vision system, located on the flying shear operation side and connected to the shear control module, is used to receive image acquisition commands sent by the shear control module and acquire image sequences of the flying shear shear cutting area. The image sequences include infrared images and visible light images. The control device as described in claim 6 is connected to the vision system and the cutting control module respectively, and is used to receive the image sequence, calculate the actual head and tail length based on the target image, generate the target setting length for the next cutting based on the deviation between the actual head and tail length and the preset target length, and output it to the cutting control module.

8. An electronic device, characterized in that, include: A processor, and a memory communicatively connected to the processor; The processor executes the computer execution instructions stored in the memory to achieve the purpose of claim 1. The method for controlling the cutting length of the flying shear head and tail as described in any one of the 5.

9. A machine-readable storage medium, characterized in that, The machine-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method for controlling the cutting length of the flying shear head and tail as described in any one of claims 1-5.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the method for controlling the cutting length of the flying shear head and tail according to any one of claims 1-5.

Images