Heating furnace billet visual tracking positioning and loading control method and system

Abstract

The application provides a heating furnace billet visual tracking positioning and loading control method and system, through obtaining heating furnace state data and carrying out correlation analysis and noise reduction processing, obtaining verified state data; obtaining billet tracking data and position recognition results output by an image segmentation calculation model; correcting coarse positioning parameters based on the recognition results, determining a target furnace entry pass combined with the verified state data, and controlling the billet to be conveyed to a billet loading roller; when the billet reaches a camera monitoring area, the head and tail positions are identified by using an image instance segmentation model, the position deviation between the billet center point and the billet loading machine center point is calculated combined with size parameters, a roller speed control curve is generated and is sent to a first level automation system; after the billet is adjusted to the positioning standard by driving the roller, a signal is sent and the control right is transferred; the head and tail positions are monitored in real time by using laser ranging devices on both sides of a billet loading furnace door, and interlocking protection is triggered when the positions exceed a preset range, so that the problem that the final positioning of the billet depends on manual confirmation is solved.

Description

Technical Field

[0001] This application relates to the field of steel rolling production line technology, and in particular to a method and system for visual tracking, positioning and loading control of steel billets in a heating furnace. Background Technology

[0002] In the heating furnace process of steel rolling production lines in the steel industry, steel billets need to be conveyed, positioned, and then loaded into the furnace for heating treatment. The operational stability of this step directly affects the subsequent rolling quality and production efficiency. Due to the complex working conditions of high temperature, high dust, and strong vibration around the heating furnace, the position tracking and loading control during the billet conveying process places stringent requirements on the environmental adaptability of sensing equipment and the robustness of algorithms. The industry urgently needs a control method that can operate stably in extreme environments and achieve high-precision positioning and automated loading of steel billets to adapt to the needs of continuous, high-cycle production.

[0003] Machine vision-based steel loading control systems have become the mainstream technology in this field. Their core architecture includes an image acquisition and processing module, a machine vision target recognition model, a basic automated communication module, and an automatic furnace loading control module. These systems acquire images of steel billets in the conveying area using multiple cameras, extract the position and size features of the billets using image processing algorithms, and then transmit the analysis results to the basic automation system to assist in correcting the billet coordinates and dimensions. This improves the accuracy of calculating and controlling the furnace entry position and is compatible with various mainstream heating furnace types, such as pusher-type and roller hearth-type furnaces.

[0004] However, the aforementioned systems still cannot achieve high-precision closed-loop control throughout the entire billet loading process. The final positioning of the billet still relies on manual confirmation, which not only increases the labor intensity of personnel but also fails to meet the positioning consistency requirements under high-rate-of-production conditions. Furthermore, existing solutions lack a full-link safety monitoring mechanism for the loading process, failing to effectively mitigate safety risks such as billet skewing, interference caused by the billet exceeding its furnace limits, and derailment. Therefore, how to achieve fully automated and precise positioning of the billet conveying process and safe interlocking control of the loading process has become a core technical problem urgently needing to be solved in this field. Summary of the Invention

[0005] This application provides a visual tracking positioning and loading control method and system for steel billets in a heating furnace to solve the problem that the final positioning of steel billets depends on manual confirmation.

[0006] The first aspect of this application provides a method for visual tracking, positioning, and loading control of steel billets in a heating furnace, the method comprising: Acquire heating furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data output by the basic automation layer and process automation layer; The furnace status data is subjected to correlation analysis and noise reduction processing to remove abnormal interference data in order to obtain the verified furnace status data. Simultaneously acquire billet tracking data from the basic automation system and billet position recognition results output by the image segmentation calculation model; Based on the billet position identification result, the billet coarse positioning parameters are corrected, and the target billet entry pass is determined according to the verified heating furnace status data, and the billet is controlled to be transported from the entry verification station to the loading roller conveyor. When the billet arrives at the camera monitoring area corresponding to the target furnace entry channel, the head and tail positions of the billet are identified in real time through the image instance segmentation model. The positional deviation between the center point of the billet and the center point of the charging machine is calculated by combining the size parameters of the billet and the head and tail positions, and a roller speed control curve is generated. The roller conveyor speed control curve is sent to the primary automation system. The drive roller conveyor adjusts the position of the billet until the positioning meets the standard. The condition for meeting the positioning standard is that the positional deviation between the center point of the billet and the center point of the loading machine is less than or equal to a preset threshold, and the axis of the billet is parallel to the feed direction of the loading machine. Send a positioning completion signal to the primary automation system and transfer control of the roller conveyor; During the billet loading process, the head and tail positions of the billet are monitored in real time by laser ranging devices on both sides of the loading furnace door. When the head and tail positions of the billet exceed the preset monitoring range, interlock protection is triggered.

[0007] This application obtains furnace status data and billet tracking data, and combines the position recognition results output by the image segmentation calculation model to correct coarse positioning parameters, controlling the billet to be transported from the furnace entry verification station to the loading roller conveyor. After the billet arrives at the camera monitoring area, the image instance segmentation model is used to identify the head and tail positions and calculate the deviation to generate the roller conveyor speed control curve, driving the roller conveyor to adjust the billet position until the positioning is qualified, and then handing over control authority to the first-level automation system. During the loading process, the head and tail positions are monitored in real time by a laser ranging device, and interlock protection is triggered when the position exceeds the preset range, which helps to improve the billet positioning accuracy and the safety of the loading process, and reduces the reliance on manual confirmation.

[0008] Optionally, the image segmentation calculation model is built based on deep learning and adopts an architecture consisting of an encoder and a decoder. The encoder obtains the edge texture information and contour morphology information of the billet through multi-level feature extraction operations. The decoder restores the spatial details of the image through upsampling operations and skip connection operations, and outputs a pixel-level segmentation mask with the same size as the input image. The segmentation mask is used to calibrate the complete contour range of a single billet.

[0009] The image segmentation calculation model employs a deep learning architecture consisting of an encoder and a decoder. The encoder extracts edge texture and contour information of the billet through multi-level feature extraction, while the decoder restores the spatial details of the image through upsampling and skip connection operations. The output is a pixel-level segmentation mask with the same size as the input image to define the complete contour range of a single billet, which helps to improve the accuracy of billet target extraction and adaptability to complex working conditions.

[0010] Optionally, in the step of performing correlation analysis and noise reduction processing on the heating furnace status data, the correlation analysis is used to match the billet loading data, roller conveyor operation status data, steel charging machine operating status data and furnace environment status data at the same time stamp, and the noise reduction processing is used to filter out abnormal data points caused by high temperature, dust or vibration conditions.

[0011] Correlation analysis and noise reduction processing help improve the consistency of furnace status data over time, reduce the interference of abnormal data points on subsequent control logic, and improve the accuracy and reliability of data input.

[0012] Optionally, the step of correcting the billet coarse positioning parameters based on the billet position identification result includes: The billet position data obtained by the basic automation system through velocity integration mode and grating correction mode is compared with the billet contour position data output by the image segmentation calculation model. Adjust the rough positioning parameters of the billet based on the comparison results.

[0013] The above steps compare the billet position data obtained by the basic automated system through velocity integration mode and grating correction mode with the billet contour position data output by the image segmentation calculation model, and adjust the billet coarse positioning parameters according to the comparison results. This helps to improve the consistency of position data during coarse positioning, reduce the deviation that may be introduced by a single tracking mode, and improve the accuracy of billet coarse positioning.

[0014] Optionally, the step of generating the roller conveyor speed control curve includes: The target rotational speed of the roller conveyor is determined based on the magnitude of the positional deviation. The adjustment frequency of the roller conveyor speed is determined based on the changing trend of the position deviation; The target rotational speed is combined with the adjustment frequency to generate the roller conveyor speed control curve.

[0015] The above steps determine the target speed of the roller conveyor based on the magnitude of the position deviation, determine the adjustment frequency of the roller conveyor speed based on the changing trend of the position deviation, and then combine the target speed and the adjustment frequency to generate the roller conveyor speed control curve. This helps to match the roller conveyor speed control with the actual position deviation of the billet, improve the dynamic response characteristics in the positioning process, and enhance the accuracy of billet position adjustment.

[0016] Optionally, after triggering the interlock protection, the following steps are also included: Control the steel loading machine to stop. The control roller conveyor stops operating, and an abnormal alarm message is marked on the status display screen. The abnormal alarm message includes the billet number, the current head and tail position data, and the direction of exceeding the limit.

[0017] After the interlock protection is triggered, this application helps to stop the operation of the loading machine and the roller conveyor by controlling the loading machine to stop and reducing the safety risks under abnormal conditions. At the same time, by marking abnormal alarm information including billet number, current head and tail position data and over-limit direction on the status display screen, it helps to provide operators with the basis for abnormal location and improve the efficiency of fault diagnosis.

[0018] Optionally, the method further includes: The status display screen shows the model calculation status, steel loading area working condition data, and parameter calculation results in real time. The status display screen is divided into a real-time tracking status area and a video display area. The real-time tracking status area is used to display the real-time tracking progress, the coarse positioning results of each area of ​​the furnace charging roller conveyor, the online billet number, dimensional parameters and real-time position. The video display area is used to display the real-time video stream captured by the camera.

[0019] The above method displays the model calculation status, steel loading area working condition data, and parameter calculation results in real time on the status display screen. The screen is divided into a real-time tracking status area and a video display area. The real-time tracking status area displays the real-time tracking progress, the coarse positioning results of each area of ​​the steel loading roller conveyor, the online billet number, size parameters, and real-time position. The video display area displays the real-time video stream captured by the camera. This helps to present various status information in the billet conveying and positioning process in separate areas, improving the efficiency and intuitiveness of operators' monitoring of the on-site working conditions.

[0020] Optionally, the coverage area of ​​the camera monitoring area corresponds to the bearing surface of the steel loading roller conveyor, and the installation position of the camera satisfies the following conditions: the shooting axis is perpendicular to the bearing surface of the steel loading roller conveyor, and the shooting field of view covers the maximum movement range of a single steel billet on the steel loading roller conveyor.

[0021] By aligning the coverage area of ​​the camera monitoring area with the bearing surface of the loading roller conveyor, and making the shooting axis perpendicular to the bearing surface of the loading roller conveyor, while ensuring that the shooting field of view covers the maximum movement range of a single steel billet on the loading roller conveyor, it helps to improve the geometric accuracy of image acquisition, ensuring that the steel billet is always within the effective monitoring field of view throughout the entire movement process, thereby improving the continuity of visual tracking and the reliability of positioning data.

[0022] Optionally, the time taken for the complete control process from acquiring furnace status data to triggering interlock protection or completing the steel loading action is less than or equal to 100 milliseconds.

[0023] By setting the time taken for the complete control process to less than or equal to 100 milliseconds, it helps to shorten the overall response time from acquiring furnace status data to triggering interlock protection or completing steel loading actions, improve the real-time performance and execution efficiency of the control process, make the connection between visual positioning and equipment actions more compact, and enhance the system's ability to quickly adapt to dynamic working conditions.

[0024] The second aspect of this application provides a visual tracking, positioning, and loading control system for steel billets in a heating furnace, applied to the visual tracking, positioning, and loading control method for steel billets in a heating furnace provided in the first aspect. The system includes: The heating furnace status module is used to acquire heating furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data output by the basic automation layer and process automation layer; and performs correlation analysis and noise reduction processing on the heating furnace status data to remove abnormal interference data in order to obtain verified heating furnace status data. An image segmentation calculation model is used to acquire camera video streams in real time. Based on a convolutional neural network with an encoder-decoder architecture, the model performs preprocessing and pixel-level segmentation on the video stream images and outputs the billet head and tail position recognition results and size parameters. The steel loading information matching and coarse positioning module is used to synchronously acquire the billet tracking data of the basic automation system and the billet position recognition result output by the image segmentation calculation module, correct the billet coarse positioning parameters according to the billet position recognition result, and determine the target furnace entry number of the billet by combining the verified furnace status data output by the furnace status module, and control the billet to be automatically transported from the furnace entry verification station to the steel loading roller conveyor. The precise positioning control module for the steel loading area is used to, after the billet arrives at the camera monitoring area corresponding to the target furnace entry channel, call the image instance segmentation model to track and identify the head and tail positions of the billet in real time, calculate the positional deviation between the center point of the billet and the center point of the steel loading machine by combining the size parameters of the billet and the head and tail coordinates, dynamically generate the roller conveyor speed control curve and send it to the primary automation system, drive the roller conveyor to adjust the position of the billet until the positioning is qualified, and send a positioning completion signal to the primary automation system after the positioning is qualified and transfer the roller conveyor control authority; the judgment condition for the positioning qualification is: the positional deviation between the center point of the billet and the center point of the steel loading machine is less than or equal to a preset threshold, and the axis of the billet is parallel to the feed direction of the steel loading machine; The process control module is installed to monitor the position of the billet head and tail in real time through laser ranging devices set on both sides of the billet loading furnace door during the billet loading process. When the position of the billet head and tail exceeds the preset monitoring range, the interlock protection mechanism is triggered. The status display module is used to display the model calculation status, steel loading area working condition data and parameter calculation results in real time, and supports online setting and dynamic adjustment of system parameters.

[0025] This application establishes a heating furnace status module, an image segmentation calculation model, a steel loading information matching and coarse positioning module, a steel loading area precise positioning control module, a loading process control module, and a status display module. Each module performs functions such as status data reading and verification, video stream pixel-level segmentation, coarse positioning parameter correction and pass determination, position deviation calculation and roller speed control curve generation and distribution, laser ranging monitoring and interlock protection triggering, status information display, and parameter adjustment. This helps to achieve systematic control of the billet from coarse positioning to precise positioning and then to loading, improves the coordination and control accuracy between each link, and reduces the need for manual intervention.

[0026] As can be seen from the above technical solutions, this application provides a method and system for visual tracking, positioning, and loading control of steel billets in a heating furnace. This involves acquiring heating furnace status data, including steel billet loading data output from the basic automation layer and the process automation layer, roller conveyor operation status data, steel charging machine operating data, and furnace environment status data; performing correlation analysis and noise reduction on the heating furnace status data to remove abnormal interference data, thus obtaining verified heating furnace status data; simultaneously acquiring steel billet tracking data from the basic automation system and steel billet position recognition results output from the image segmentation calculation model; correcting the coarse positioning parameters of the steel billet based on the steel billet position recognition results; and determining the target entry pass of the steel billet according to the verified heating furnace status data, thereby controlling the steel billet from... The billet is conveyed from the furnace entry verification station to the loading roller conveyor. When the billet arrives at the camera monitoring area corresponding to the target furnace entry lane, the head and tail positions of the billet are identified in real time using an image instance segmentation model. The positional deviation between the center point of the billet and the center point of the loading machine is calculated based on the billet's size parameters and its head and tail positions, generating a roller conveyor speed control curve. This curve is then sent to the primary automation system. The roller conveyor is driven to adjust the billet's position until it reaches the target position. A positioning completion signal is sent to the primary automation system, and roller conveyor control is transferred. During the billet loading process, the head and tail positions of the billet are monitored in real time by laser ranging devices on both sides of the furnace door. When the head and tail positions of the billet exceed the preset monitoring range, interlocking protection is triggered. This solves the problem of the billet's final positioning relying on manual confirmation. Attached Figure Description

[0027] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A structural block diagram of the visual tracking, positioning, and loading control system for steel billets in a heating furnace provided in this application embodiment; Figure 2 A flowchart illustrating the visual tracking, positioning, and loading control method for steel billets in a heating furnace, as provided in this application embodiment. Detailed Implementation

[0029] The embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described below do not represent all embodiments consistent with this application. They are merely examples of systems and methods consistent with some aspects of this application as detailed in the claims.

[0030] To address the issue of reliance on manual confirmation for the final positioning of steel billets, some embodiments of this application provide a visual tracking positioning and loading control method for steel billets in heating furnaces. This method can be applied to various mainstream heating furnace types, such as pusher-type heating furnaces and roller hearth heating furnaces. By deeply coupling machine vision technology with the basic automation system and process automation system of the production line, fully automatic closed-loop control of the entire process of steel billet from conveying, coarse positioning, precise positioning to loading is achieved.

[0031] It should be noted that the execution order of the steps in the method described in this application is not strictly limited. In practical applications, some steps can be executed in parallel or their order can be adjusted if the process conditions are met, as long as the technical purpose of this application can be achieved.

[0032] The visual tracking, positioning, and loading control method for steel billets in this heating furnace includes: S100: Acquire furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating data, and furnace environment status data output from the basic automation layer and process automation layer.

[0033] It should be understood that the furnace status data specifically includes billet loading data (e.g., billet number, billet size parameters, target furnace entry pass, etc.) output from the basic automation layer and process automation layer, roller conveyor operation status data (e.g., running speed, direction of rotation, and current load status of the furnace entry roller and charging roller), charging machine operating status data (e.g., charging machine forward / backward position, charging machine center point coordinates, charging machine operation status, etc.), and furnace environment status data (e.g., furnace door opening / closing status, furnace temperature, furnace pressure, etc.). This data is collected in real-time from the primary basic automation system (i.e., the basic automation layer) and the secondary process automation system (i.e., the process automation layer) via industrial Ethernet or fieldbus.

[0034] S200: Perform correlation analysis and noise reduction on the heating furnace status data to remove abnormal interference data and obtain verified heating furnace status data.

[0035] It should be understood that correlation analysis is used to match and align various types of data under the same timestamp to ensure the consistency of each set of data in the time dimension; noise reduction is used to filter out abnormal data points caused by harsh working conditions such as high temperature radiation, dust cover or equipment vibration (such as instantaneous messages exceeding the normal range or abrupt sensor readings), and obtain the verified heating furnace status data after removing abnormal interference data.

[0036] S300: Synchronously acquire billet tracking data from the basic automation system and billet position recognition results output by the image segmentation calculation model.

[0037] It should be understood that the billet tracking data includes at least the billet displacement data calculated using the velocity integral mode and the billet position correction data obtained using the raster correction mode. The billet position recognition result includes at least the position information of the billet contour in the image coordinate system and the coordinates of the billet's head and tail.

[0038] S400: Based on the billet position identification results, correct the billet coarse positioning parameters, and determine the target entry pass of the billet according to the verified heating furnace status data, and control the billet to be transported from the entry verification station to the loading roller conveyor.

[0039] Specifically, the actual position of the billet obtained by visual recognition is compared with the calculated position of the billet obtained by the basic automation system, and the position parameters in the coarse positioning process are adjusted according to the deviation.

[0040] S500: When the billet arrives at the camera monitoring area corresponding to the target furnace entry channel, the head and tail positions of the billet are identified in real time through the image instance segmentation model. The positional deviation between the center point of the billet and the center point of the charging machine is calculated by combining the billet size parameters and the head and tail positions, and the roller speed control curve is generated.

[0041] It should be understood that the dimensional parameters of the billet, such as the length and width of the billet, and the positional deviation between the center point of the billet and the center point of the loading machine include the positional deviation along the conveying direction of the roller conveyor and the lateral deviation perpendicular to the conveying direction. When the deviation is large, the roller conveyor is controlled to run at a higher speed to quickly reduce the deviation. When the deviation gradually decreases, the roller conveyor is controlled to run at a lower speed to achieve smooth and precise adjustment.

[0042] S600: Sends the roller conveyor speed control curve to the first-level automation system.

[0043] S700: Drive the roller conveyor to adjust the position of the billet until the positioning meets the standard.

[0044] It should be understood that the primary automation system drives the roller conveyor to adjust the position of the billet according to the control curve until the positioning meets the standard. The criteria for determining whether the positioning meets the standard can be, for example, that the positional deviation between the center point of the billet and the center point of the loading machine is less than or equal to a preset threshold (e.g., ±5mm), and that the axis of the billet is parallel to the feed direction of the loading machine (i.e., the billet is not skewed).

[0045] The criteria for determining whether the positioning meets the standard are: the positional deviation between the center point of the billet and the center point of the loading machine is less than or equal to the preset threshold, and the axis of the billet is parallel to the feed direction of the loading machine.

[0046] It should be understood that the criteria for accurate positioning include two dimensions: First, the positional deviation between the billet's center point and the charging machine's center point is less than or equal to a preset threshold. This preset threshold can be set according to actual process requirements; for example, it can be set to ±3mm for scenarios requiring high precision positioning, and ±5mm or ±10mm for general scenarios. This condition ensures that the billet's positional accuracy in the conveying direction meets the requirements for the charging machine's gripping or pushing. Second, the billet's axis is parallel to the charging machine's feed direction. This condition ensures that the billet's posture is correct when entering the furnace, without any skew. Otherwise, even if the billet's center point is accurate, a skewed billet may still interfere with the structures on both sides of the furnace door when loaded into the furnace, or cause the billet to be misplaced in the furnace, affecting the uniformity of heating. The billet's axis can be determined by using the billet contour output by the image instance segmentation model and calculating the billet's major axis direction using the minimum bounding rectangle or principal component analysis method. The charging machine's feed direction is a known fixed direction (along the roller conveyor direction). When both of the above conditions are met simultaneously, the billet's positioning is deemed satisfactory.

[0047] It should be noted that the specific value of the preset threshold can be adjusted according to factors such as the furnace door width, steel charging machine structural parameters, and billet size of different heating furnaces. Those skilled in the art can set it reasonably according to actual working conditions after reading this application.

[0048] S800: Sends a positioning completion signal to the primary automation system and transfers control of the roller conveyor.

[0049] It should be understood that after the positioning is completed, a positioning completion signal is sent to the primary automation system and the control authority of the roller conveyor is transferred. The primary automation system then takes over the control of the roller conveyor and triggers subsequent automatic steel loading actions.

[0050] S900: During the billet loading process, the head and tail positions of the billet are monitored in real time by laser ranging devices on both sides of the loading furnace door. When the head and tail positions of the billet exceed the preset monitoring range, interlock protection is triggered.

[0051] It should be understood that the laser ranging equipment continuously emits laser beams and receives reflected signals, and calculates the distance between the head and tail of the billet and the structures on both sides of the furnace door in real time. When the head and tail positions of the billet exceed the preset monitoring range, interlock protection is triggered (such as controlling the steel loading machine to stop operating, controlling the roller conveyor to stop running, etc.), thereby preventing safety accidents such as interference between the billet and the structures on both sides of the furnace door or the billet falling off the conveyor.

[0052] It should be noted that in the method steps of this application, the two steps of "acquiring heating furnace status data" and "synchronously acquiring billet tracking data from the basic automation system and billet position recognition results output by the image segmentation calculation model" can be executed in parallel during actual operation. That is, various types of data are continuously collected and processed during system operation, rather than being strictly executed sequentially. In addition, the time taken for the complete control process from acquiring heating furnace status data to triggering interlock protection or completing the steel loading action is less than or equal to 100 milliseconds. This ensures that the connection between visual positioning and equipment action is sufficiently tight and can meet the real-time requirements under high-cycle production.

[0053] In some embodiments, the image segmentation computation model is built based on deep learning and adopts an architecture consisting of an encoder and a decoder. The encoder obtains the edge texture information and contour morphology information of the steel billet through multi-level feature extraction operations; The decoder restores the spatial details of the image through upsampling and skip connection operations, and outputs a pixel-level segmentation mask with the same size as the input image. The segmentation mask is used to define the complete contour range of a single steel billet.

[0054] It should be understood that the image segmentation computation model adopts an architecture consisting of an encoder and a decoder. Its core function is to perform pixel-level semantic segmentation on steel billet images captured by industrial cameras. The encoder downsamples layer by layer through a multi-layer convolutional neural network, extracting image features at different scales at each layer: shallow convolutional kernels capture the edge texture information of the steel billet (such as the iron oxide scale texture on the surface of the steel billet, the straight line features of the edges, etc.), middle convolutional kernels extract the local contour morphology information of the steel billet (such as the corner points of the steel billet, the direction of the sides, etc.), and deep convolutional kernels further abstract the semantic structure information of the steel billet (such as the overall orientation, posture, and other high-level semantics of the steel billet). The encoder captures full-dimensional information from edge texture, contour morphology to semantic structure simultaneously through multi-level feature extraction operations. The decoder gradually restores the spatial resolution of the feature map through upsampling operations (such as transposed convolution or interpolation upsampling), and fuses the high-resolution shallow features output by each layer of the encoder with the upsampled features of the corresponding layer of the decoder through skip connections. This restores the spatial details of the image while preserving high-level semantic information, and finally outputs a pixel-level segmentation mask with the same size as the input image. Each pixel in the segmentation mask is classified as either "steel billet" or "background," thereby accurately defining the complete outline of a single steel billet in the image and effectively adapting to the need for precise extraction of steel billet targets under complex working conditions such as high temperature, dust, and vibration.

[0055] It should be noted that this application does not impose any particular restrictions on the specific network structure of the deep learning model used (such as the number of layers in the encoder and decoder, the size of the convolution kernel, the type of activation function, etc.), as long as it can achieve the above-mentioned pixel-level segmentation function. Those skilled in the art can choose a suitable network structure according to the actual working conditions and computing resources.

[0056] In some embodiments, in the step of performing correlation analysis and noise reduction on the heating furnace status data, the correlation analysis is used to match the billet loading data, roller conveyor operation status data, steel charging machine operating status data and furnace environment status data at the same timestamp, and the noise reduction is used to filter out abnormal data points caused by high temperature, dust or vibration conditions.

[0057] It should be understood that correlation analysis and noise reduction of furnace status data are crucial preliminary steps to ensure the accuracy of subsequent control logic. Correlation analysis specifically refers to the time alignment and matching of billet loading data (e.g., the billet number, length, and width currently being processed), roller conveyor operating status data (e.g., the current speed and direction of each roller section), charging machine operating status data (e.g., the current position and forward / backward stroke of the charging machine), and furnace environment status data (e.g., furnace door opening and furnace temperature) collected at the same time stamp. This ensures consistency of each set of data in the time dimension, avoiding errors in control logic judgment due to data time misalignment. For example, if the roller conveyor speed data and billet position data at a certain time stamp do not correspond precisely in time, it may lead to deviations in billet position calculation. Noise reduction processing is specifically used to filter out abnormal data points caused by high temperature, dust, or vibration conditions. In high-temperature environments, sensors may output instantaneous saturation values; dust may obscure sensors, causing signal loss or abrupt changes; and equipment vibration may cause fluctuations in sensor readings. If these abnormal data points are not filtered, they will severely affect the subsequent billet positioning accuracy and control effect. Noise reduction processing can employ methods such as median filtering, moving average filtering, or statistical outlier removal to eliminate outliers that exceed the normal range or are significantly inconsistent with the trend of preceding and following data.

[0058] It should be noted that correlation analysis and noise reduction can be performed after data acquisition, or they can be completed synchronously during the data inflow using real-time stream processing. The specific implementation method can be flexibly selected according to the system's computing resources and real-time requirements.

[0059] In some embodiments, the step of correcting the billet coarse positioning parameters based on the billet position identification result includes: The billet position data obtained by the basic automation system through velocity integration mode and grating correction mode is compared with the billet contour position data output by the image segmentation calculation model. Adjust the billet rough positioning parameters based on the comparison results.

[0060] It should be understood that during the billet conveying process, the basic automation system obtains coarse positioning data of the billet through two methods: speed integration mode (i.e., calculating the theoretical displacement of the billet by integrating the roller speed and running time) and grating correction mode (i.e., correcting the cumulative error of integration using the grating signal when the billet passes a fixed grating detection point). However, due to factors such as roller slippage, relative sliding between the billet and the roller, and installation errors of the grating detection points, both modes may introduce varying degrees of positioning deviation. Therefore, this application compares the billet position data obtained by the basic automation system through the above two modes with the billet contour position data output by the image segmentation calculation model.

[0061] Specifically, the billet contour position identified in the image coordinate system is transformed to the world coordinate system and the difference is calculated with the billet position tracked by the basic automation system in the same coordinate system. Based on the comparison results (such as the direction and amount of positional deviation between the two), the billet coarse positioning parameters are adjusted (such as offset correction of the initial position in the velocity integral mode, or compensation of the correction amount in the grating correction mode), thereby improving the consistency of position data in the coarse positioning process, reducing the system deviation that may be introduced by a single tracking mode, and improving the accuracy of billet coarse positioning.

[0062] It should be noted that the above comparison and correction can be carried out continuously during the conveying process of each steel billet, or it can be intermittently corrected according to a preset cycle (e.g., once every 100ms). The specific correction frequency can be determined according to the actual working conditions and accuracy requirements.

[0063] In some embodiments, the step of generating the roller conveyor speed control curve includes: The target rotational speed of the roller conveyor is determined based on the magnitude of the positional deviation. The adjustment frequency of the roller conveyor speed is determined based on the changing trend of the position deviation; The target speed and the adjustment frequency are combined to generate the roller conveyor speed control curve.

[0064] It should be understood that the specific steps for generating the roller conveyor speed control curve are as follows: First, determine the target speed of the roller conveyor based on the magnitude of the positional deviation between the billet center point and the loading machine center point. When the deviation is large (e.g., greater than 50mm), set a higher target speed (e.g., 80%–100% of the roller conveyor's rated speed) to quickly reduce the deviation; when the deviation is moderate (e.g., 20mm–50mm), set a moderate target speed (e.g., 40%–60% of the rated speed); when the deviation is small (e.g., less than 20mm), set a lower target speed (e.g., 10%–30% of the rated speed) to achieve smooth fine-tuning. Second, determine the adjustment frequency of the roller conveyor speed based on the trend of the positional deviation. If the deviation is decreasing rapidly (i.e., the billet is rapidly approaching the target position), reduce the adjustment frequency to avoid overshoot; if the deviation changes slowly or remains basically unchanged (i.e., the billet is approaching the target position but has not yet reached the target), increase the adjustment frequency to achieve fine adjustment. Finally, the target rotational speed and adjustment frequency determined above are combined to generate a roller conveyor speed control curve. This control curve can be a piecewise linear speed-time curve or a smoother S-shaped acceleration / deceleration curve, depending on the combined requirements of stability and response speed under actual working conditions. After the generated roller conveyor speed control curve is sent to the primary automation system, the primary automation system drives the roller conveyor frequency converter to adjust the roller conveyor motor speed according to the curve, thereby making the billet position adjustment process both fast and smooth.

[0065] It should be noted that the specific values ​​of the target speed and adjustment frequency mentioned above can be calibrated and optimized on-site according to the equipment characteristics of different production lines (such as the power of the roller conveyor motor, the response speed of the frequency converter, the weight of the billet, etc.), and this application does not impose any special limitations on this.

[0066] In some embodiments, after triggering the interlocking protection, the method further includes: Control the steel loading machine to stop. The control roller conveyor stops operating, and abnormal alarm information is marked on the status display screen. The abnormal alarm information includes the billet number, the current head and tail position data, and the direction of exceeding the limit.

[0067] Specifically, a stop command is sent to the charging machine's drive system, immediately halting its forward and backward movements to prevent equipment damage or safety accidents caused by continued charging operations when the billet position is abnormal. A stop command is also sent to the primary automation system, stopping the charging and furnace-entry rollers to prevent the billet from continuing to move under abnormal conditions, which could worsen skewness or cause it to derail. Simultaneously, anomaly alarm information is displayed on the status screen. This alarm includes the billet number where the anomaly occurred (facilitating quick operator identification of the specific billet), the current head and tail positions (i.e., the distances between the billet's head and tail ends and the structures on both sides of the furnace door, as monitored in real-time by the laser rangefinder), and the direction of the out-of-limit movement (e.g., "left side of head out of limit" or "right side of tail out of limit"). This provides operators with clear information for anomaly location, improving troubleshooting efficiency. It should be noted that after the interlock protection is triggered, the system can further record all relevant data at the time of the anomaly (including roller speed, charging machine position, billet image, etc.), forming a fault log for subsequent analysis. After troubleshooting, operators can release the interlock and resume production by manually resetting or confirming with the system.

[0068] In some embodiments, the method further includes: The status display screen shows the model calculation status, steel loading area working condition data, and parameter calculation results in real time. The status display screen is divided into a real-time tracking status area and a video display area. The real-time tracking status area is used to display the real-time tracking progress, the coarse positioning results of each area of ​​the furnace charging roller conveyor, the online billet number, dimensional parameters and real-time position. The video display area is used to display the real-time video stream captured by the camera.

[0069] It should be understood that the real-time tracking status area is used to display the following information: real-time tracking progress (e.g., the billet number currently being tracked and the percentage of tracking completion), coarse positioning results of each area of ​​the furnace charging roller conveyor (e.g., the specific position of the billet in each section of the furnace charging roller conveyor, the grating detection points it has passed, etc.), online billet number (the number of the billet currently being conveyed and positioned), dimensional parameters (the length, width, etc. of the billet), and real-time position (the real-time coordinates of the billet in the current coordinate system). The video display area is used to display the real-time video stream captured by the industrial camera, allowing operators to intuitively observe the actual status and position of the billet through the video. The above-mentioned partitioned display method enables the clear and organized presentation of various status information during the billet conveying and positioning process, improving the efficiency and intuitiveness of operators' monitoring of the on-site working conditions.

[0070] It should be noted that the status display screen can further include online parameter setting and dynamic adjustment functions. Operators can directly modify system operating parameters (such as positioning deviation threshold, roller speed limit, etc.) on the screen. After receiving the modified parameters, the system will automatically update the corresponding values ​​in the control logic without stopping or restarting the system.

[0071] In some embodiments, the coverage area of ​​the camera monitoring area corresponds to the bearing surface of the loading roller conveyor, and the installation position of the camera satisfies the following conditions: the shooting axis is perpendicular to the bearing surface of the loading roller conveyor, and the shooting field of view covers the maximum movement range of a single steel billet on the loading roller conveyor.

[0072] It should be understood that the coverage area of ​​the camera monitoring area corresponds to the bearing surface of the loading roller conveyor. That is, the camera's field of view should cover the entire area on the loading roller conveyor where the billet may stop and move, ensuring that the billet remains within the effective monitoring field of view throughout the entire process from entering the camera monitoring area to completing its positioning. The camera's installation position meets the following conditions: the shooting axis is perpendicular to the bearing surface of the loading roller conveyor. This vertical installation method minimizes perspective distortion, ensuring an accurate mapping between the geometric dimensions and position of the billet in the image and its actual physical dimensions and position, avoiding measurement errors introduced by tilted shooting. Simultaneously, the shooting field of view covers the maximum movement range of a single billet on the loading roller conveyor. That is, throughout the entire process from entering the monitoring area to completing its positioning, regardless of the billet's position within that area, its overall outline should appear completely within the shooting field of view, ensuring that the image instance segmentation model can completely extract the billet's outline information.

[0073] It should be noted that the specific installation height of the camera and the focal length of the lens should be calculated and selected based on the actual width of the steel loading roller conveyor and the maximum movement range of the steel billet to ensure that the field of view coverage matches the above requirements.

[0074] In some embodiments, the time taken for the complete control flow from acquiring furnace status data to triggering interlock protection or completing the steel loading action is less than or equal to 100 milliseconds.

[0075] It should be understood that a time metric of less than or equal to 100 milliseconds encompasses the entire time frame of data acquisition (obtaining furnace status data, billet tracking data, and image data), data processing (correlation analysis, noise reduction, image segmentation calculation, and position deviation calculation), control decision-making (generating the roller conveyor speed control curve), and command issuance (sending the control curve to the primary automation system). By optimizing the inference speed of the image segmentation calculation model (e.g., using a lightweight network structure and GPU-accelerated inference), simplifying the data preprocessing process, and employing high-speed industrial Ethernet for data communication, the entire closed-loop control process is ensured to be completed within 100 milliseconds. This response time ensures a sufficiently tight connection between visual positioning and equipment actions, enabling the system to respond quickly to dynamic changes in the billet position and meeting the real-time requirements of high-cycle steel rolling production lines.

[0076] It should be noted that 100 milliseconds is an upper limit. In well-optimized systems, the time taken for the complete control flow can be further reduced to 50 milliseconds or even less. The specific time depends on factors such as hardware configuration, network latency, and model complexity.

[0077] See Figure 2 When the billet enters the loading area, the system first receives the billet data to be loaded. On the furnace entry roller conveyor and the billet loading roller conveyor, the system confirms the roller conveyor position based on the target entry roller of the billet after information verification, and initiates the billet coarse tracking process to drive the billet to be automatically transported from the furnace entry verification station to the billet loading roller conveyor. After the billet arrives at the camera monitoring area of ​​the corresponding entry roller conveyor, the precise positioning control module is automatically activated. The image instance segmentation model tracks the head and tail positions of the billet in real time, calculates the position deviation based on the size parameters, and generates roller conveyor control commands to achieve high-precision positioning of the billet. After positioning is completed, the loading process control module is activated to track the entire billet loading process and ensure that the loading process is safe and reliable.

[0078] See Figure 1 This application also provides a visual tracking, positioning, and loading control system for steel billets in a heating furnace, applied to the visual tracking, positioning, and loading control method for steel billets in a heating furnace provided in the above embodiments. The system includes: The heating furnace status module is used to acquire heating furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data output from the basic automation layer and process automation layer. The module also performs correlation analysis and noise reduction on the heating furnace status data to remove abnormal interference data and obtain verified heating furnace status data.

[0079] It should be understood that the furnace status module is used to read in real time the billet loading data (e.g., billet number, billet size parameters such as length and width, target furnace entry pass, etc.), roller conveyor running status data (e.g., running speed, direction of rotation, current load status, roller conveyor motor current, etc.) output from the basic automation layer (Level 1 system) and the process automation layer (Level 2 system), the charging machine operating status data (e.g., charging machine forward and backward position, charging machine center point coordinates, charging machine action status, charging machine travel limit signal, etc.), and the furnace environment status data (e.g., furnace door opening and closing status, furnace temperature, furnace pressure, furnace atmosphere, etc.). The above data is acquired in real time from the Level 1 and Level 2 automation systems via industrial Ethernet (e.g., PROFINET, EtherCAT, or Modbus TCP protocol) or fieldbus. This module performs correlation analysis on the read data, matching and aligning various types of data under the same timestamp according to the time dimension to ensure the consistency of each group of data in the time dimension. It also performs noise reduction processing, using methods such as median filtering, moving average filtering, or statistical outlier removal to filter out abnormal data points caused by harsh operating conditions such as high temperature radiation, dust cover, or equipment vibration (e.g., instantaneous messages exceeding the normal range or abrupt sensor readings), thereby eliminating abnormal interference data and outputting verified heating furnace status data, providing accurate and reliable data support for the subsequent control logic and parameter calculation of the system.

[0080] An image segmentation calculation model is used to acquire camera video streams in real time. Based on an encoder-decoder architecture, a convolutional neural network is used to preprocess and segment the video stream images at the pixel level, and output the billet head and tail position recognition results and size parameters.

[0081] It should be understood that the image segmentation computation model is used to acquire video streams from industrial cameras in real time. A convolutional neural network based on an encoder-decoder architecture preprocesses the video stream images (e.g., image scaling, normalization, denoising) and performs pixel-level segmentation. The model extracts features through multiple layers via the encoder: shallow convolutional kernels capture the edge texture information of the billet, mid-layer convolutional kernels extract the local contour morphology information of the billet, and deep convolutional kernels abstract the semantic structure information of the billet, simultaneously capturing the full-dimensional information of the billet from edge texture and contour morphology to semantic structure. The decoder gradually restores the spatial resolution of the feature map through upsampling operations (e.g., transposed convolution or interpolation upsampling), and fuses the high-resolution shallow features output from each layer of the encoder with the upsampled features of the corresponding layer of the decoder using skip connections. The final output is a pixel-level segmentation mask with the same size as the input image. This segmentation mask accurately defines the complete contour range of a single billet in the image, and then outputs the billet head and tail position recognition results (including the coordinates of the billet head and tail in the image coordinate system) and size parameters (the actual physical size of the billet after camera calibration and pixel value transformation).

[0082] The steel loading information matching and coarse positioning module is used to synchronously acquire the billet tracking data of the basic automation system and the billet position recognition results output by the image segmentation calculation module. Based on the billet position recognition results, the coarse positioning parameters of the billet are corrected, and the target furnace entry number of the billet is determined by combining the verified furnace status data output by the furnace status module. The module then controls the automatic transport of the billet from the furnace entry verification station to the steel loading roller conveyor.

[0083] It should be understood that the billet information matching and coarse positioning module is used to synchronously acquire billet tracking data from the basic automation system (including billet displacement data calculated through velocity integration mode and billet position correction data obtained through grating correction mode) and billet position recognition results output by the image segmentation calculation model (including billet contour position, head and tail coordinates, etc.). This module compares the billet position data obtained by the basic automation system through velocity integration mode and grating correction mode with the billet contour position data output by the image segmentation calculation model. Based on the comparison results (e.g., the direction and amount of positional deviation between the two), it corrects the billet coarse positioning parameters (e.g., offset correction of the initial position in velocity integration mode, or compensation of the correction amount in grating correction mode), thereby improving the consistency of position data during coarse positioning and reducing system deviations that may be introduced by a single tracking mode. At the same time, this module, combined with the verified furnace status data output by the furnace status module, determines the target entry pass for the billet and controls the automatic transport of the billet from the entry verification station to the billet loading roller conveyor.

[0084] Specifically, the system sends roller conveyor start command and track selection command to the primary automation system, driving the furnace feed roller conveyor and steel loading roller conveyor to run at a preset speed, automatically transporting the steel billet from the furnace feed verification station through each section of the roller conveyor to the steel loading roller conveyor area corresponding to the target furnace feed track.

[0085] The precise positioning control module for the steel loading area is used to track and identify the head and tail positions of the billet in real time after the billet arrives at the camera monitoring area corresponding to the target furnace entry channel. It calls the image instance segmentation model to calculate the positional deviation between the center point of the billet and the center point of the steel loading machine by combining the billet's size parameters and head and tail coordinates. It dynamically generates the roller conveyor speed control curve and sends it to the primary automation system to drive the roller conveyor to adjust the billet position until the positioning is qualified. After the positioning is qualified, it sends a positioning completion signal to the primary automation system and transfers the roller conveyor control authority. The judgment condition for the positioning qualification is: the positional deviation between the center point of the billet and the center point of the steel loading machine is less than or equal to the preset threshold, and the billet axis is parallel to the feeding direction of the steel loading machine.

[0086] It should be understood that the precise positioning control module in the steel loading area calculates the coordinates of the billet's center point by combining the billet's dimensional parameters (such as its length and width) with the head and tail coordinates. The billet's center point can be determined by combining the midpoint of the head and tail coordinates with the midpoint of the billet's width direction. This allows for the calculation of the positional deviation between the billet's center point and the loading machine's center point. This positional deviation includes the positional deviation along the roller conveyor direction (i.e., the distance the billet moves forward or backward in the conveying direction) and the lateral deviation perpendicular to the conveying direction (i.e., the distance the billet deviates from the roller conveyor's centerline). Based on this positional deviation, a roller conveyor speed control curve is dynamically generated. When the deviation is large, the roller conveyor is controlled to operate at a higher speed to quickly reduce the deviation. As the deviation gradually decreases, the roller conveyor is controlled to operate at a lower speed to achieve smooth and precise adjustment. This control curve is then sent to the primary automation system to drive the roller conveyor to adjust the billet's position until the positioning is accurate. The positioning compliance criteria are met simultaneously in two dimensions: First, the positional deviation between the billet's center point and the loading machine's center point is less than or equal to a preset threshold (e.g., ±5mm, which can be adjusted according to process requirements); second, the billet's axis is parallel to the loading machine's feed direction (i.e., the billet is not skewed; the angle between the billet's long axis and the loading machine's feed direction is calculated using image segmentation results, and this angle should be less than or equal to a preset angle threshold, e.g., ±1°). When both conditions are met simultaneously, the positioning is considered compliant. After compliance, the module sends a positioning completion signal to the primary automation system and transfers control of the roller conveyor. The primary automation system then takes over roller conveyor control and triggers subsequent automatic loading actions.

[0087] The process control module is installed to monitor the position of the billet head and tail in real time through laser ranging devices set on both sides of the billet loading furnace door during the billet loading process. When the position of the billet head and tail exceeds the preset monitoring range, the interlock protection mechanism is triggered.

[0088] It should be understood that the loading process control module is used to monitor the position of the billet head and tail in real time during the billet loading process using laser ranging devices installed on both sides of the loading furnace door. The laser ranging devices continuously emit laser beams and receive reflected signals, calculating the distance between the billet head and tail and the structures on both sides of the furnace door in real time. When the position of the billet head and tail exceeds the preset monitoring range (for example, the distance between the billet head and the structure on the left side of the furnace door is less than the safety threshold, or the distance between the billet tail and the structure on the right side of the furnace door is less than the safety threshold), the module immediately triggers the interlock protection mechanism: it sends a stop command to the drive system of the loading machine to immediately stop the loading machine's forward and backward movement, sends a roller stop command to the primary automation system to stop the loading rollers and the furnace entry rollers from rotating, and simultaneously marks abnormal alarm information on the status display screen (including the billet number where the abnormality occurred, the current head and tail position data, and the direction of exceeding the limit), thereby preventing safety accidents such as interference between the billet and the structures on both sides of the furnace door or the billet falling off the track. It should be noted that the monitoring and protection functions of this module are active throughout the entire process of billet loading, from the moment the billet enters the furnace door until it is fully loaded into the furnace.

[0089] The status display module is used to display the model calculation status, steel loading area working condition data and parameter calculation results in real time, and supports online setting and dynamic adjustment of system parameters.

[0090] It should be understood that the status display module is used to display the model calculation status in real time (such as the inference progress of the image segmentation model, the current frame rate, processing time, etc.), the working condition data of the steel loading area (such as the current speed of each section of the roller conveyor, the position of the steel loading machine, the furnace door opening, etc.), and the parameter calculation results (such as the coordinates of the billet center point, the position deviation value, the positioning compliance status, etc.). It also supports the online setting and dynamic adjustment of system parameters. Operators can directly modify the system operating parameters (such as the positioning deviation threshold, the upper limit of the roller conveyor speed, the acceleration and deceleration time, etc.) on the status display screen. After receiving the modified parameters, the system automatically updates the corresponding values ​​in the control logic without stopping or restarting the system. The status display module is divided into two core areas: a real-time tracking status area and a video display area. The real-time tracking status area displays the real-time tracking progress (e.g., the billet number currently being tracked and the percentage of tracking completed), the coarse positioning results of each area of ​​the furnace charging roller conveyor (e.g., the specific position of the billet in each section of the furnace charging roller conveyor, the grating detection points that have been passed, etc.), the online billet number (the number of the billet currently being transported and positioned), dimensional parameters (the length, width, etc. of the billet), and the real-time position (the real-time coordinates of the billet in the current coordinate system). The video display area displays the real-time video stream captured by the industrial camera, allowing operators to intuitively observe the actual status and position of the billet through the video.

[0091] like Figure 1As shown, the heating furnace status module, status display module, steel loading information matching and coarse positioning module, steel loading area precise positioning control module, and loading process control module can communicate with the main process scheduling module. The main process scheduling module is used to receive the steel loading production plan issued by the upstream system, coordinate the orderly execution of each functional module according to the steel loading operation process, sort out the execution status of the operation nodes and synchronize the current steel loading operation progress to the downstream process, realize the orderly linkage of each module in the entire steel loading process, ensure the smooth connection of each link of billet tracking, positioning, and loading, and avoid instruction conflicts or action misalignment between different modules.

[0092] The data flow and control flow relationships between the various functional modules of the system are as follows: The heating furnace status module interacts with both the basic automation layer and the process automation layer, reading various status data in real time. After correlation analysis and noise reduction, it outputs verified status data to the steel loading information matching and coarse positioning module and the steel loading area precise positioning control module. The image segmentation calculation model receives video stream input from the industrial camera, performs pixel-level segmentation processing, and outputs billet position recognition results to the steel loading information matching and coarse positioning module and the steel loading area precise positioning control module. The steel loading information matching and coarse positioning module corrects the coarse positioning parameters based on the visual recognition results, determines the target furnace entry pass based on the verified status data, and sends roller control commands to the primary automation system to achieve automatic conveying of the billet from the furnace entry verification station to the steel loading roller. The steel loading area precise positioning control module is activated after the billet enters the camera monitoring area, calls the image instance segmentation model to track the billet position in real time, generates a roller rotation speed control curve, and sends it to the primary automation system to drive the roller to adjust the billet position. After the positioning is achieved, it sends a positioning completion signal. The loading process control module monitors the head and tail positions of the billet in real time using a laser rangefinder, triggering interlock protection in case of abnormalities. The status display module connects to the above modules, displaying various operational information in real time and supporting online parameter setting. In the diagram, solid arrows represent control flow, and dashed arrows represent data flow, illustrating the deep coupling between the vision system and the automation system.

[0093] It should be noted that, Figure 1 The connection relationships between the modules shown are merely exemplary. In actual deployment, data interaction between modules can be achieved through various means such as shared memory, message queues, or industrial communication protocols. This application does not impose any particular limitations on this.

[0094] It should be understood that the above are merely preferred embodiments of this application and are not intended to limit 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 protection of this application.

[0095] It should be noted that the specific working process of each module in the above system embodiments can be referred to the description of the corresponding steps in the method embodiments, and will not be repeated here. In the system provided by this application, each module can be implemented in software form on an industrial computer or edge computing device, or in hardware form (such as FPGA, ASIC and other dedicated chips), or in a combination of software and hardware. Preferably, the image segmentation calculation model and the image instance segmentation model are deployed on an industrial edge computing device equipped with a GPU or NPU using a deep learning inference engine (such as TensorRT, OpenVINO or ONNX Runtime) to meet the real-time requirement of completing the entire control process within 100 milliseconds.

[0096] The coverage area of ​​the "camera monitoring area" in this application corresponds to the bearing surface of the loading roller conveyor. The camera's installation position ensures that the shooting axis is perpendicular to the bearing surface of the loading roller conveyor and that the shooting field of view covers the maximum movement range of a single steel billet on the loading roller conveyor. To adapt to the high-temperature environment around the heating furnace, the industrial camera is preferably equipped with an air-cooled or water-cooled protective sleeve, a high-temperature resistant lens, and a dustproof viewing window to ensure long-term stable operation under extreme conditions. The camera's exposure and white balance parameters can be adaptively adjusted according to the on-site lighting conditions, or multiple sets of parameters can be preset and switched according to day / night and seasonal changes to ensure that clear images of steel billets can be acquired under different lighting conditions.

[0097] The "laser ranging device" in this application can be a laser displacement sensor or a lidar, with a preferred measurement accuracy of ±1mm and a preferred measurement frequency of not less than 50Hz, to ensure that the dynamic changes in the head and tail positions can be captured in real time during the billet loading process. The installation location of the laser ranging device should avoid the direct flame radiation area when the furnace door is open, or be equipped with appropriate heat insulation protection devices to ensure the long service life and measurement reliability of the device.

[0098] The specific values ​​in the above embodiments of this application (such as the ±5mm position deviation threshold, the 100ms response time, and the 50Hz laser measurement frequency) are illustrative and not intended to limit the scope of protection of this application. Those skilled in the art, after reading this specification, can appropriately adjust and optimize these specific parameters according to the equipment characteristics, process requirements, and control precision needs of different production lines. As long as these adjustments do not depart from the technical concept of this application, they should be considered part of the technical solution disclosed in this application.

[0099] As can be seen from the above technical solutions, the embodiments of this application provide a method and system for visual tracking, positioning, and loading control of steel billets in a heating furnace. This involves acquiring heating furnace status data, including steel billet loading data output from the basic automation layer and the process automation layer, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data; performing correlation analysis and noise reduction on the heating furnace status data to remove abnormal interference data and obtain verified heating furnace status data; simultaneously acquiring steel billet tracking data from the basic automation system and steel billet position recognition results output from the image segmentation calculation model; correcting the coarse positioning parameters of the steel billet based on the steel billet position recognition results; and determining the target loading point of the steel billet based on the verified heating furnace status data. The system controls the transport of steel billets from the furnace entry verification station to the loading roller conveyor. When the billet reaches the camera monitoring area corresponding to the target furnace entry lane, the head and tail positions of the billet are identified in real time using an image instance segmentation model. The positional deviation between the billet's center point and the loading machine's center point is calculated based on the billet's dimensional parameters and head and tail positions, generating a roller conveyor speed control curve. This curve is then sent to the primary automation system. The roller conveyor adjusts the billet's position until it reaches the target location. A positioning completion signal is sent to the primary automation system, and control of the roller conveyor is transferred. During billet loading, the head and tail positions of the billet are monitored in real time by laser ranging devices on both sides of the furnace door. Interlocking protection is triggered when the head and tail positions exceed the preset monitoring range. This solves the problem of relying on manual confirmation for the final billet positioning.

Claims

1. A method for visual tracking, positioning, and loading control of steel billets in a heating furnace, characterized in that, The method includes: Acquire heating furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data output by the basic automation layer and process automation layer; The furnace status data is subjected to correlation analysis and noise reduction processing to remove abnormal interference data in order to obtain the verified furnace status data. Simultaneously acquire billet tracking data from the basic automation system and billet position recognition results output by the image segmentation calculation model; Based on the billet position identification result, the billet coarse positioning parameters are corrected, and the target billet entry pass is determined according to the verified heating furnace status data, and the billet is controlled to be transported from the entry verification station to the loading roller conveyor. When the billet arrives at the camera monitoring area corresponding to the target furnace entry channel, the head and tail positions of the billet are identified in real time through the image instance segmentation model. The positional deviation between the center point of the billet and the center point of the charging machine is calculated by combining the size parameters of the billet and the head and tail positions, and a roller speed control curve is generated. The roller conveyor speed control curve is sent to the primary automation system. The drive roller conveyor adjusts the position of the billet until the positioning meets the standard. The condition for meeting the positioning standard is that the positional deviation between the center point of the billet and the center point of the loading machine is less than or equal to a preset threshold, and the axis of the billet is parallel to the feed direction of the loading machine. Send a positioning completion signal to the primary automation system and transfer control of the roller conveyor; During the billet loading process, the head and tail positions of the billet are monitored in real time by laser ranging devices on both sides of the loading furnace door. When the head and tail positions of the billet exceed the preset monitoring range, interlock protection is triggered.

2. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The image segmentation calculation model is built based on deep learning and adopts an architecture consisting of an encoder and a decoder. The encoder obtains the edge texture information and contour morphology information of the steel billet through multi-level feature extraction operations; The decoder restores the spatial details of the image through upsampling and skip connection operations, and outputs a pixel-level segmentation mask with the same size as the input image. The segmentation mask is used to define the complete contour range of a single steel billet.

3. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, In the step of performing correlation analysis and noise reduction on the heating furnace status data, the correlation analysis is used to match the billet loading data, roller conveyor operation status data, steel charging machine operating status data and furnace environment status data at the same time stamp, and the noise reduction is used to filter out abnormal data points caused by high temperature, dust or vibration conditions.

4. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The step of correcting the billet coarse positioning parameters based on the billet position identification result includes: The billet position data obtained by the basic automation system through velocity integration mode and grating correction mode is compared with the billet contour position data output by the image segmentation calculation model. Adjust the rough positioning parameters of the billet based on the comparison results.

5. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The step of generating the roller conveyor speed control curve includes: The target rotational speed of the roller conveyor is determined based on the magnitude of the positional deviation. The adjustment frequency of the roller conveyor speed is determined based on the changing trend of the position deviation; The target rotational speed is combined with the adjustment frequency to generate the roller conveyor speed control curve.

6. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, After the triggering of the interlocking protection, it also includes: Control the steel loading machine to stop. The control roller conveyor stops operating, and an abnormal alarm message is marked on the status display screen. The abnormal alarm message includes the billet number, the current head and tail position data, and the direction of exceeding the limit.

7. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The method further includes: The status display screen shows the model calculation status, steel loading area working condition data, and parameter calculation results in real time. The status display screen is divided into a real-time tracking status area and a video display area. The real-time tracking status area is used to display the real-time tracking progress, the coarse positioning results of each area of ​​the furnace charging roller conveyor, the online billet number, dimensional parameters and real-time position. The video display area is used to display the real-time video stream captured by the camera.

8. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The coverage area of ​​the camera monitoring area corresponds to the bearing surface of the steel loading roller conveyor. The installation position of the camera satisfies the following conditions: the shooting axis is perpendicular to the bearing surface of the steel loading roller conveyor, and the shooting field of view covers the maximum movement range of a single steel billet on the steel loading roller conveyor.

9. The method for visual tracking, positioning, and loading control of steel billets in a heating furnace according to claim 1, characterized in that, The time taken for the complete control process of the method, from acquiring furnace status data to triggering interlock protection or completing steel loading, is less than or equal to 100 milliseconds.

10. A visual tracking, positioning, and loading control system for steel billets in a heating furnace, characterized in that, The method for visual tracking, positioning, and loading control of steel billets in a heating furnace, applicable to any one of claims 1-9, comprises: The heating furnace status module is used to acquire heating furnace status data, which includes billet loading data, roller conveyor operation status data, steel charging machine operating status data, and furnace environment status data output by the basic automation layer and process automation layer; and performs correlation analysis and noise reduction processing on the heating furnace status data to remove abnormal interference data in order to obtain verified heating furnace status data. An image segmentation calculation model is used to acquire camera video streams in real time. Based on a convolutional neural network with an encoder-decoder architecture, the model performs preprocessing and pixel-level segmentation on the video stream images and outputs the billet head and tail position recognition results and size parameters. The steel loading information matching and coarse positioning module is used to synchronously acquire the billet tracking data of the basic automation system and the billet position recognition result output by the image segmentation calculation module, correct the billet coarse positioning parameters according to the billet position recognition result, and determine the target furnace entry number of the billet by combining the verified furnace status data output by the furnace status module, and control the billet to be automatically transported from the furnace entry verification station to the steel loading roller conveyor. The precise positioning control module for the steel loading area is used to, after the billet arrives at the camera monitoring area corresponding to the target furnace entry channel, call the image instance segmentation model to track and identify the head and tail positions of the billet in real time, calculate the positional deviation between the center point of the billet and the center point of the steel loading machine by combining the size parameters of the billet and the head and tail coordinates, dynamically generate the roller conveyor speed control curve and send it to the primary automation system, drive the roller conveyor to adjust the position of the billet until the positioning is qualified, and send a positioning completion signal to the primary automation system after the positioning is qualified and transfer the roller conveyor control authority; the judgment condition for the positioning qualification is: the positional deviation between the center point of the billet and the center point of the steel loading machine is less than or equal to a preset threshold, and the axis of the billet is parallel to the feed direction of the steel loading machine; The process control module is installed to monitor the position of the billet head and tail in real time through laser ranging devices set on both sides of the billet loading furnace door during the billet loading process. When the position of the billet head and tail exceeds the preset monitoring range, the interlock protection mechanism is triggered. The status display module is used to display the model calculation status, steel loading area working condition data and parameter calculation results in real time, and supports online setting and dynamic adjustment of system parameters.

Images