Automatic hoisting and transferring system for hot-rolled steel coils and control method thereof

Abstract

This invention discloses an automatic hoisting and transfer system for hot-rolled steel coils and its control method. The invention relates to the field of metallurgical technology, specifically a field panoramic construction module used to acquire video of the hoisting operation area, perform correction, registration, and stitching processing, and construct a panoramic image benchmark. Through the collaborative work of the field panoramic construction module, scene element recognition module, and steel coil trajectory construction and status analysis module, this invention continuously perceives and quantifies the spatial position, movement trajectory, and interaction with production interference factors of the hot-rolled steel coil during hoisting and transfer. When the steel coil is affected by interference factors such as personnel, vehicles, or occupied work areas in the transfer path, the system identifies the interference impact through speed fluctuations and spatial intersections, forming an interference impact distribution map. This system possesses visualized and quantifiable status perception capabilities, reducing the efficiency decline caused by manual judgment and scheduling.

Description

Technical Field

[0001] This invention relates to the field of metallurgical technology, specifically to an automatic hoisting and transfer system for hot-rolled steel coils and its control method. Background Technology

[0002] Hot-rolled steel coils are an important intermediate product in the steel production process. They are typically formed from continuously cast billets through processes such as heating, rough rolling, and finish rolling. They are characterized by their large weight, concentrated volume, high surface temperature, and high surface quality requirements, and are widely used in metallurgy, machinery manufacturing, construction, and energy equipment. After rolling, hot-rolled steel coils require a series of transfer operations within the production workshop, including hoisting, transshipment, temporary storage, and connection to subsequent processes. This transfer process usually relies on overhead cranes, overhead hoists, or automated hoisting equipment in the complex plant environment. The hoisting and transfer of hot-rolled steel coils is mainly under semi-automatic or manually assisted control. The production site often involves multiple work units operating in parallel, personnel and vehicles moving together, and frequent changes in temporary storage materials, making the hoisting path susceptible to interference. The current hot-rolled steel coil hoisting and circulation mainly relies on fixed path planning or manual experience scheduling, lacking systematic perception and quantitative analysis of the actual circulation trajectory of the steel coils, real-time passage efficiency, and the degree of influence of interference factors; To address the aforementioned technical shortcomings, a solution is proposed. Summary of the Invention

[0003] To address the shortcomings of existing technologies, this invention provides an automatic hoisting and transfer system for hot-rolled steel coils and its control method. The system continuously tracks the transfer behavior of hot-rolled steel coils and comprehensively evaluates the impact of on-site interference factors on path passage efficiency, thereby optimizing the hoisting path and improving efficiency.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an automatic hoisting and transfer system for hot-rolled steel coils and its control method, comprising: The on-site panoramic construction module is used to acquire video of the hoisting operation area, perform correction, registration and stitching processing, and construct a panoramic image baseline. The scene element recognition module calls the object recognition model to parse the original video footage, identify the selection box, perform spatial calibration and projection processing, and generate a scene element map (MAP). The steel coil trajectory construction and state analysis module performs cross-frame association on the candidate bounding boxes (BOX) of the hot-rolled steel coil target (COL) in the scene element map (MAP), generates a trajectory sequence (Trk), analyzes the influence relationship between the steel coil trajectory and other scene elements, and obtains an interference influence distribution map. The path evaluation and optimization module evaluates the steel coil trajectory based on the interference impact distribution map, constructs a path topology map, and performs path optimization analysis to generate optimized modified paths and optimized display maps.

[0005] A camera device is deployed in the hoisting operation area, and the on-site panoramic construction module is used to acquire raw video footage. Based on the intrinsic parameters of the camera device, radial distortion correction and tangential distortion correction are performed on the original video image to obtain the corrected video image; The overlapping areas between adjacent corrected video frames are extracted, and multi-view registration is performed using feature points in the overlapping areas. The images are then stitched together using a weighted fusion method to generate a panoramic image.

[0006] The scene element recognition module calls the object recognition model to parse the original video footage. The object recognition model includes: data preprocessing submodule PRE, feature backbone submodule BAK, feature fusion submodule FPN, candidate generation submodule RPN, detection and classification submodule DET, instance segmentation submodule SEG, path semantic parsing submodule PAT, and temporal correction submodule TSC. Among them, the data preprocessing submodule PRE performs scale unification, brightness equalization, and motion blur suppression on the original video footage to obtain standardized frame data FRM; The Feature Backbone Submodule (BAK) performs multi-layer convolutional coding on the standardized frame data (FRM) to extract low-level edge contour features, mid-level texture morphology features, and high-level semantic structure features, resulting in a multi-layer feature map (FTR). The Feature Fusion Submodule (FPN) performs top-down and laterally connected multi-scale fusion on the multi-layer feature map (FTR) to generate a multi-scale fused feature map (PFT). The candidate generation submodule RPN generates candidate regions PRO based on the multi-scale fusion feature PFT, and performs boundary regression on the candidate regions PRO to obtain candidate boxes BOX; The detection and classification submodule DET performs category discrimination on the region corresponding to the candidate box BOX and outputs the feature category TAG. The feature category TAG includes: hot-rolled steel coil target COL, production interference factor INT, and operation occupied area OCC. The feature category TAG is marked on the candidate box BOX. The instance segmentation submodule SEG performs pixel-level mask prediction on the candidate bounding box (BOX) and outputs the feature position (POS) and feature area (ROI) of the candidate bounding box (BOX). A fixed marker is selected as a spatial reference point, and a spatial calibration mapping relationship between image coordinates and on-site physical coordinates is established based on the spatial reference point. The candidate bounding box (BOX), feature location (POS), and feature area (ROI) are projected onto the panoramic image based on the spatial calibration mapping relationship to generate a scene feature map (MAP).

[0007] The steel coil trajectory construction and state analysis module analyzes the influence relationship between the steel coil trajectory and other scene elements, including: Based on the scene element map MAP, cross-frame association processing is performed on the spatial position of the candidate bounding boxes (BOX) of the hot-rolled steel coil target COL in the continuously corrected video frame. A trajectory identifier Tid is assigned to the candidate bounding boxes of the hot-rolled steel coil target COL. Based on the association rules of position continuity, candidate bounding boxes with a position distance of less than a preset association threshold and the same motion direction in adjacent frames are matched to generate the trajectory sequence Trk of the steel coil target. Based on the spatial coordinate difference and time interval between adjacent time periods in the trajectory sequence Trk, the turnover speed Spd of the steel coil during the time period is calculated using the following formula: Where Dis is the actual displacement distance between adjacent trajectory points in the scene feature map MAP of the hot-rolled steel coil target COL, and Tim is the time difference between two frames at the acquisition position of the candidate box BOX. The position points of the same steel coil target in consecutive video frames are associated in chronological order to generate a steel coil trajectory line. The multiple flow velocities Spd within the same steel coil trajectory line are weighted and summed to obtain the flow velocity characterization value of the steel coil trajectory line.

[0008] The steel coil trajectory construction and state analysis module calculates the start time and end time of the candidate box (BOX) to obtain the occupancy time of the candidate box (BOX). The trajectory line of the steel coil is spatially superimposed with the region of interest (ROI) of the candidate boxes (BOX) of other feature categories (TAGs), and the intersection ratio between the trajectory line and the ROI is calculated to obtain the trajectory intersection degree value. Before and after the candidate box BOX of the hot-rolled steel coil target COL enters the ROI of the candidate box BOX of the other element category TAG, the flow rate characterization value Spd1 of the candidate box BOX of the hot-rolled steel coil target COL when entering the ROI of the candidate box BOX of the other element category TAG, and the flow rate characterization value Spd2 when leaving the ROI of the candidate box BOX of the other element category TAG, are calculated. The difference between the flow rate characterization value Spd1 and the flow rate characterization value Spd2 is used as the disturbance speed change.

[0009] The steel coil trajectory construction and state analysis module normalizes the occupancy time, trajectory intersection value, and interference speed change to obtain normalized occupancy time rD, normalized trajectory intersection value rC, and normalized interference speed change rS, and calculates the influence score Sco. Wherein, the weighting coefficients wD, wC, and wS are the weighting coefficients for the normalized occupancy time rD, the normalized trajectory intersection value rC, and the normalized disturbance velocity change rS, respectively, and p is the nonlinear coupling index; The influence score Sco, along with the candidate boxes BOX and steel coil trajectory lines corresponding to the other element categories TAG, are marked in the panoramic image to generate an interference influence distribution map.

[0010] The path evaluation and optimization module performs joint analysis of production speed paths and interference scores to generate optimized and modified paths, including: For the steel coil trajectory line shown in the interference impact distribution map, calculate the comprehensive evaluation value Sp of the path: Where aur is the flow velocity characterization value of the steel coil trajectory line, cur is the influence score Sco, and α and β are weighting coefficients; A path topology graph is constructed using the steel coil trajectory line described in the interference impact distribution map. The steel coil trajectory line is used as an edge in the path topology graph, and the comprehensive path evaluation value Sp is used as the path passage cost. In the aforementioned path topology graph, based on the initial task objective and the target task objective as path constraints, a path search and optimization algorithm is invoked to solve for the path combination with the minimum travel cost; When the interference score of a certain path segment is higher than the preset interference threshold and its corresponding flow speed is lower than the preset speed threshold, the probability of that path segment being selected as the optimal path is reduced; The obtained path segments are sequentially stitched together to generate an optimized and modified path; the optimized and modified path is then mapped onto the panoramic map according to the spatial calibration mapping relationship to generate an optimized display map.

[0011] An automatic hoisting and transfer control method for hot-rolled steel coils includes: S1. Collect original video footage of the hot-rolled steel coil hoisting operation area, perform distortion correction, multi-view registration, and image stitching on the original video footage, and construct a panoramic image benchmark; S2. Based on the original video footage, the object recognition model is invoked to identify the steel coil target, production interference factors, and work-occupied area, generating a scene element map (MAP). S3. Perform cross-frame association processing on the steel coil target in the scene element map to obtain the continuous trajectory sequence of the steel coil. Based on the spatial overlap, temporal occupation, and state change relationship between the hot-rolled steel coil and the candidate boxes of other element categories (TAGs), generate an interference impact distribution map. S4. Based on the interference impact distribution map, conduct a comprehensive evaluation of the on-site path, construct a path topology map, perform path search and optimization processing, and generate optimized modified paths and optimized display maps.

[0012] This invention provides an automatic hoisting and transfer system for hot-rolled steel coils and its control method. Compared with the prior art, it has the following advantages: This invention utilizes the collaborative work of a field panoramic construction module, a scene element recognition module, and a steel coil trajectory construction and status analysis module to continuously perceive and quantify the spatial position, movement trajectory, and interaction with production interference factors of hot-rolled steel coils during hoisting and transfer. When the steel coil is affected by interference factors such as personnel, vehicles, or work areas occupied in the transfer path, the interference impact is identified through speed fluctuations and spatial intersections, and an interference impact distribution map is generated. It has the ability to visualize and quantify status perception, reducing the efficiency decline caused by manual judgment and scheduling. This invention maps the efficiency of steel coil passage to the degree of on-site interference as the cost of path passage, and constructs path topology relationships and performs path optimization analysis to identify potential bottleneck paths with high interference and low efficiency and generate optimized modified paths. This enables hoisting scheduling to prioritize low-interference and high-efficiency passage routes, reduce the stagnation of steel coils in congested areas, and improve the overall hoisting flow efficiency and production cycle stability. Attached Figure Description

[0013] Figure 1 This is a schematic diagram of an automatic hoisting and transfer control method for hot-rolled steel coils according to the present invention; Figure 2 This is a schematic diagram of the system structure framework of the present invention; Figure 3 This is a schematic diagram of the structural framework of the item recognition model of the present invention. Detailed Implementation

[0014] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0015] Please see Figure 1 This application provides an automatic hoisting and transfer system for hot-rolled steel coils and its control method, including: The on-site panoramic construction module is used to acquire video of the hoisting operation area, perform correction, registration and stitching processing, and construct a panoramic image baseline. The scene element recognition module calls the object recognition model to parse the original video footage, identify the selection box, perform spatial calibration and projection processing, and generate a scene element map (MAP). The steel coil trajectory construction and state analysis module performs cross-frame association on the candidate bounding boxes (BOX) of the hot-rolled steel coil target (COL) in the scene element map (MAP), generates a trajectory sequence (Trk), analyzes the influence relationship between the steel coil trajectory and other scene elements, and obtains an interference influence distribution map. The path evaluation and optimization module evaluates the steel coil trajectory based on the interference impact distribution map, constructs a path topology map, and performs path optimization analysis to generate optimized modified paths and optimized display maps.

[0016] Cameras were deployed in the hoisting operation area, and a panoramic construction module was used to capture raw video footage. Based on the intrinsic parameters of the camera device, radial distortion correction and tangential distortion correction are performed on the original video image to obtain the corrected video image; The overlapping areas between adjacent corrected video frames are extracted, and multi-view registration is performed using feature points in the overlapping areas. The images are then stitched together using a weighted fusion method to generate a panoramic image.

[0017] The scene element recognition module calls the object recognition model to parse the original video footage. The object recognition model includes: data preprocessing submodule PRE, feature backbone submodule BAK, feature fusion submodule FPN, candidate generation submodule RPN, detection and classification submodule DET, instance segmentation submodule SEG, path semantic parsing submodule PAT, and temporal correction submodule TSC. Among them, the data preprocessing submodule PRE performs scale unification, brightness equalization, and motion blur suppression on the original video footage to obtain standardized frame data FRM; The feature backbone submodule BAK performs multi-layer convolutional coding on the standardized frame data FRM to extract low-level edge contour features, mid-level texture morphology features, and high-level semantic structure features, resulting in a multi-layer feature map FTR. The Feature Fusion Submodule (FPN) performs top-down and lateral connection multi-scale fusion on the multi-layer feature map FTR to generate the multi-scale fused feature PFT. The candidate generation submodule RPN generates candidate regions PRO based on multi-scale fusion feature PFT, and performs boundary regression on the candidate regions PRO to obtain candidate boxes BOX; The detection and classification submodule DET performs category discrimination on the region corresponding to the candidate box (BOX) and outputs feature category TAGs. The feature category TAGs include: hot-rolled steel coil target COL, production interference factors INT, and operation occupied area OCC. The feature category TAGs are then marked on the candidate box (BOX). The instance segmentation submodule SEG performs pixel-level mask prediction on the candidate bounding box (BOX) and outputs the feature location (POS) and feature area (ROI) of the candidate bounding box (BOX). Fixed landmarks are selected as spatial reference points, and a spatial calibration mapping relationship between image coordinates and on-site physical coordinates is established based on these spatial reference points. The candidate bounding box (BOX), feature location (POS), and feature area (ROI) are projected onto the panoramic image based on spatial calibration mapping to generate a scene feature map (MAP).

[0018] The steel coil trajectory construction and state analysis module analyzes the influence relationship between the steel coil trajectory and other scene elements, including: Based on the scene element map (MAP), cross-frame association processing is performed on the spatial position of the candidate bounding boxes (BOX) of the hot-rolled steel coil target (COL) in the continuously corrected video frame. A trajectory identifier (Tid) is assigned to the candidate bounding boxes (BOX) of the hot-rolled steel coil target (COL). Based on the association rules of position continuity, candidate bounding boxes (BOX) in adjacent frames with a position distance less than a preset association threshold and the same motion direction are matched to generate the trajectory sequence Trk of the steel coil target. Based on the spatial coordinate difference and time interval between adjacent time periods in the trajectory sequence Trk, the turnover speed Spd of the steel coil during the time period is calculated using the following formula: Where Dis is the actual displacement distance between adjacent trajectory points in the scene feature map MAP of the hot-rolled steel coil target COL, and Tim is the time difference between two frames at the acquisition position of the candidate box BOX. The position points of the same steel coil target in consecutive video frames are associated in chronological order to generate a steel coil trajectory line. The multiple flow velocities Spd within the same steel coil trajectory line are weighted and summed to obtain the flow velocity characterization value of the steel coil trajectory line.

[0019] The steel coil trajectory construction and state analysis module calculates the start time and end time of the candidate box (BOX) to obtain the occupancy time of the candidate box. Spatially overlay the steel coil trajectory line with the feature range ROIs of the candidate boxes of other feature categories (TAGs), calculate the intersection ratio between the steel coil trajectory line and the feature range ROIs, and obtain the trajectory intersection degree value; Before and after the candidate box BOX of the hot-rolled steel coil target COL enters the ROI of the candidate box BOX of other element category TAG, calculate the flow rate characterization value Spd1 of the candidate box BOX of the hot-rolled steel coil target COL when it enters the ROI of the candidate box BOX of other element category TAG, and the flow rate characterization value Spd2 when it leaves the ROI of the candidate box BOX of other element category TAG. The difference between the flow rate characterization value Spd1 and the flow rate characterization value Spd2 is used as the disturbance speed change.

[0020] The steel coil trajectory construction and state analysis module normalizes the occupancy time, trajectory intersection value, and disturbance velocity change to obtain the normalized occupancy time rD, normalized trajectory intersection value rC, and normalized disturbance velocity change rS, and calculates the influence score Sco. Wherein, the weighting coefficients wD, wC, and wS are the weighting coefficients for the normalized occupancy time rD, the normalized trajectory intersection value rC, and the normalized disturbance velocity change rS, respectively, and p is the nonlinear coupling index; Mark the candidate boxes (BOX) and steel coil trajectory lines that affect the score Sco and other corresponding element categories (TAG) in the panoramic image to generate an interference impact distribution map.

[0021] In one embodiment of the present invention, in view of the problems of complex working environment, many interference factors and dynamic changes in hoisting path during the hoisting and transfer of hot-rolled steel coils, multiple sets of camera devices are deployed above or at fixed structural positions in the steel coil storage area, main hoisting channel, intersection node and densely occupied area of ​​the hoisting operation area, according to a preset coverage overlap ratio. The on-site panoramic construction module uniformly accesses the original video images collected by each camera device, and performs radial distortion correction, tangential distortion correction and viewpoint unification processing on the original video images based on the internal parameters and installation posture parameters of the camera devices, so as to eliminate the influence of lens distortion and installation angle differences on spatial measurement accuracy.

[0022] Furthermore, the on-site panoramic construction module extracts the overlapping areas between adjacent camera images, determines the spatial correspondence between images through feature point matching and multi-view registration, and uses a weighted fusion method to complete the seamless stitching of multiple images, thereby constructing a panoramic image benchmark that corresponds one-to-one with the physical space of the hoisting operation area, so that subsequent recognition and analysis are carried out under a unified spatial coordinate system.

[0023] Based on the panoramic image benchmark, the scene element recognition module calls the object recognition model to analyze the continuous video frames frame by frame. Through multi-layer feature extraction and classification, it identifies different categories of scene elements such as hot-rolled steel coil targets, personnel, vehicles, and work areas, and generates corresponding candidate boxes and spatial ranges for each element. Subsequently, it establishes a spatial calibration mapping relationship between image coordinates and on-site physical coordinates by combining pre-selected fixed markers, and projects the position information of each candidate box onto the panoramic image benchmark to generate a structured scene element map.

[0024] The steel coil trajectory construction and state analysis module takes the scene element map as input and performs cross-frame association processing on the candidate boxes of hot-rolled steel coil targets in continuous video frames. It assigns a unique trajectory identifier to the same steel coil target through position continuity and motion direction consistency rules, and forms a continuous trajectory sequence of steel coils. At the same time, based on the spatial displacement and time interval of adjacent position points in the trajectory sequence in the panoramic image reference, it calculates the actual turnover speed of the steel coil in different time periods and different path segments, and maps the turnover speed to the corresponding steel coil trajectory segment.

[0025] Furthermore, the steel coil trajectory construction and status analysis module overlays and analyzes the spatial range of production interference factors such as personnel, vehicles, and work-occupied areas in the scene element map, automatically identifying the spatial intersection and temporal overlap relationships between the steel coil trajectory and the interference factors. When the steel coil trajectory enters or crosses the spatial range of the interference factor, the system records the changes in the coil's turnover speed before entering, during crossing, and after leaving the interference area, and calculates the corresponding interference impact score by combining the duration of the interference factor's occupation, the size of the occupied area, and the degree of trajectory intersection.

[0026] The interference impact score is further bound to the spatial location of the steel coil trajectory segment and interference elements in the panoramic image baseline, and displayed in a visual overlay to form an interference impact distribution map that reflects the degree of impact of different areas on the steel coil hoisting efficiency. Through this interference impact distribution map, the key areas with high interference and high risk in the hoisting path can be presented intuitively.

[0027] Beneficial effects: This invention, through the collaborative work of the on-site panoramic construction module, scene element recognition module, and steel coil trajectory construction and status analysis module, continuously perceives and quantifies the spatial position, movement trajectory, and interaction with production interference factors of hot-rolled steel coils during hoisting and circulation. When the steel coil is affected by interference factors such as personnel, vehicles, or work areas occupied in the circulation path, the interference impact is identified through speed fluctuations and spatial intersections, and an interference impact distribution map is formed. It has the ability to visualize and quantify status perception, reducing the efficiency decline caused by manual judgment and scheduling.

[0028] The path evaluation and optimization module performs joint analysis of production speed paths and interference scores to generate optimized and modified paths, including: For the steel coil trajectory line in the interference impact distribution map, calculate the comprehensive evaluation value Sp of the path: Where aur is the flow rate characteristic value of the steel coil trajectory line, cur is the influence score Sco, and α and β are weighting coefficients; A path topology graph is constructed using the steel coil trajectory lines in the interference impact distribution map, with the steel coil trajectory lines serving as edges in the path topology graph, and the comprehensive path evaluation value Sp serving as the path travel cost; In the path topology graph, based on the initial task objective and the target task objective as path constraints, the path search and optimization algorithm is invoked to find the path combination with the minimum travel cost; When the interference score of a certain path segment is higher than the preset interference threshold and its corresponding flow speed is lower than the preset speed threshold, the probability of that path segment being selected as the optimal path is reduced; The obtained path segments are sequentially stitched together to generate an optimized and modified path; the optimized and modified path is then mapped to a panoramic map according to the spatial calibration mapping relationship to generate an optimized display map.

[0029] In a preferred embodiment of the present invention, in order to address the problem of reduced overall hoisting efficiency of hot-rolled steel coils due to unreasonable path selection and concentrated local interference during hoisting and transfer, the path evaluation and optimization module performs a unified quantitative evaluation and optimization process on the steel coil transfer path after obtaining the interference impact distribution map output by the steel coil trajectory construction and state analysis module.

[0030] The path evaluation and optimization module uses the steel coil trajectory line in the interference impact distribution map as the basic analysis object. It performs joint modeling processing on the flow speed characterization value and impact score corresponding to each steel coil trajectory line. Through the preset weighted fusion rules, it uniformly maps the actual passage efficiency of the steel coil on the path segment and the interference intensity of the corresponding area into a path comprehensive evaluation value Sp, making different path segments comparable and transforming the traditional path advantage and disadvantage problem that relies on experience judgment into a calculable path cost problem.

[0031] Furthermore, the path evaluation and optimization module, based on the spatial topological relationships in the panoramic image benchmark, abstracts the steel coil trajectory line into an edge structure in the path topology graph, and the start and end positions of the trajectory line into a node structure in the path topology graph. It uses the comprehensive path evaluation value Sp as the passage cost of the corresponding edge, thereby constructing a path topology graph model that reflects the actual flow status on site. The path topology graph is also constrained by the physical boundaries of the site, the equipment layout relationship, and the hoisting operation rules to ensure that the optimization analysis results conform to the actual production conditions.

[0032] Furthermore, the path evaluation and optimization module, based on the initial and target operation objectives of the current hoisting operation, calls the path search and optimization algorithm to minimize the cost of the path topology map. By comparing the passage costs of different path combinations, it selects the path scheme with the optimal passage efficiency under the comprehensive consideration of flow speed and interference intensity. When a path segment is detected to have a high interference score in the interference impact distribution map and the corresponding flow speed is consistently lower than the normal level, the path evaluation and optimization module automatically reduces the priority of selecting this path segment as the optimal path during the path search process, thereby guiding the optimization algorithm to avoid potential congestion bottleneck areas.

[0033] Furthermore, the path evaluation and optimization module stitches together the optimal path segment combination obtained from the solution according to the actual flow order to generate an optimized modified path for guiding hoisting operations. Based on the spatial calibration mapping relationship, the optimized modified path is superimposed and mapped onto the panoramic image reference to form an optimized display map. The optimized display map can provide a clear and executable path decision basis for the hoisting scheduling system or manual dispatchers, so that hoisting operations can be carried out along low-interference and high-efficiency paths while meeting on-site operational constraints, thereby reducing the waiting time of steel coils in high-interference areas.

[0034] Beneficial effects: This invention maps the efficiency of steel coil passage to the degree of on-site interference as the path passage cost, and identifies potential bottleneck paths with high interference and low efficiency by constructing path topology and performing path optimization analysis, and generates optimized and modified paths; thus enabling hoisting scheduling to prioritize low interference and high efficiency passage routes, reduce the stagnation of steel coils in congested areas, and improve the overall hoisting flow efficiency and production cycle stability.

[0035] An automatic hoisting and transfer control method for hot-rolled steel coils includes: S1. Collect original video footage of the hot-rolled steel coil hoisting operation area, perform distortion correction, multi-view registration, and image stitching on the original video footage, and construct a panoramic image benchmark; S2. Based on the original video footage, the object recognition model is invoked to identify the steel coil target, production interference factors, and work-occupied area, generating a scene element map (MAP). S3. Perform cross-frame association processing on the steel coil target in the scene element map to obtain the continuous trajectory sequence of the steel coil. Based on the spatial overlap, temporal occupation, and state change relationship between the hot-rolled steel coil and the candidate boxes of other element categories (TAGs), generate an interference impact distribution map. S4. Based on the interference impact distribution map, conduct a comprehensive evaluation of the on-site path, construct a path topology map, perform path search and optimization processing, and generate optimized modified paths and optimized display maps.

[0036] The above embodiments can be implemented, in whole or in part, by software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. A computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. Computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via a wired or wireless network. A computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0037] Some of the data in the above formulas are numerical calculations with dimensions removed, and the contents not described in detail in this specification are all prior art known to those skilled in the art.

[0038] The above embodiments are only used to illustrate the technical methods of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical methods of the present invention without departing from the spirit and scope of the technical methods of the present invention.

Claims

1. An automatic hoisting and transfer system for hot-rolled steel coils, characterized in that... ,include: The on-site panoramic construction module is used to acquire video of the hoisting operation area, perform correction, registration and stitching processing, and construct a panoramic image baseline. The scene element recognition module calls the object recognition model to parse the original video footage, identify the selection box, perform spatial calibration and projection processing, and generate a scene element map (MAP). The steel coil trajectory construction and state analysis module performs cross-frame association on the candidate bounding boxes (BOX) of the hot-rolled steel coil target (COL) in the scene element map (MAP), generates a trajectory sequence (Trk), analyzes the influence relationship between the steel coil trajectory and other scene elements, and obtains an interference influence distribution map. The path evaluation and optimization module evaluates the steel coil trajectory based on the interference impact distribution map, constructs a path topology map, and performs path optimization analysis to generate optimized modified paths and optimized display maps.

2. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 1, characterized in that... , A camera device is deployed in the hoisting operation area, and the on-site panoramic construction module is used to acquire raw video footage. Based on the intrinsic parameters of the camera device, radial distortion correction and tangential distortion correction are performed on the original video image to obtain the corrected video image; The overlapping areas between adjacent corrected video frames are extracted, and multi-view registration is performed using feature points in the overlapping areas. The images are then stitched together using a weighted fusion method to generate a panoramic image.

3. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 1, characterized in that... The scene element recognition module calls the object recognition model to parse the original video footage. The object recognition model includes: a data preprocessing submodule PRE, a feature backbone submodule BAK, a feature fusion submodule FPN, a candidate generation submodule RPN, a detection and classification submodule DET, an instance segmentation submodule SEG, a path semantic parsing submodule PAT, and a temporal correction submodule TSC. Among them, the data preprocessing submodule PRE performs scale unification, brightness equalization, and motion blur suppression on the original video footage to obtain standardized frame data FRM; The Feature Backbone Submodule (BAK) performs multi-layer convolutional coding on the standardized frame data (FRM) to extract low-level edge contour features, mid-level texture morphology features, and high-level semantic structure features, resulting in a multi-layer feature map (FTR). The Feature Fusion Submodule (FPN) performs top-down and laterally connected multi-scale fusion on the multi-layer feature map (FTR) to generate a multi-scale fused feature map (PFT). The candidate generation submodule RPN generates candidate regions PRO based on the multi-scale fusion feature PFT, and performs boundary regression on the candidate regions PRO to obtain candidate boxes BOX; The detection and classification submodule DET performs category discrimination on the region corresponding to the candidate box BOX and outputs the feature category TAG. The feature category TAG includes: hot-rolled steel coil target COL, production interference factor INT, and operation occupied area OCC. The feature category TAG is marked on the candidate box BOX. The instance segmentation submodule SEG performs pixel-level mask prediction on the candidate bounding box (BOX) and outputs the feature position (POS) and feature area (ROI) of the candidate bounding box (BOX). A fixed marker is selected as a spatial reference point, and a spatial calibration mapping relationship between image coordinates and on-site physical coordinates is established based on the spatial reference point. The candidate bounding box (BOX), feature location (POS), and feature area (ROI) are projected onto the panoramic image based on the spatial calibration mapping relationship to generate a scene feature map (MAP).

4. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 1, characterized in that... The steel coil trajectory construction and state analysis module analyzes the influence relationship between the steel coil trajectory and other scene elements, including: Based on the scene element map MAP, cross-frame association processing is performed on the spatial position of the candidate bounding boxes (BOX) of the hot-rolled steel coil target COL in the continuously corrected video frame. A trajectory identifier Tid is assigned to the candidate bounding boxes of the hot-rolled steel coil target COL. Based on the association rules of position continuity, candidate bounding boxes with a position distance of less than a preset association threshold and the same motion direction in adjacent frames are matched to generate the trajectory sequence Trk of the steel coil target. Based on the spatial coordinate difference and time interval between adjacent time periods in the trajectory sequence Trk, the turnover speed Spd of the steel coil during the time period is calculated using the following formula: Where Dis is the actual displacement distance between adjacent trajectory points in the scene feature map MAP of the hot-rolled steel coil target COL, and Tim is the time difference between two frames at the acquisition position of the candidate box BOX. The position points of the same steel coil target in consecutive video frames are associated in chronological order to generate a steel coil trajectory line. The multiple flow velocities Spd within the same steel coil trajectory line are weighted and summed to obtain the flow velocity characterization value of the steel coil trajectory line.

5. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 1, characterized in that... The steel coil trajectory construction and state analysis module calculates the start time and end time of the candidate box (BOX) to obtain the occupancy time of the candidate box (BOX). The trajectory line of the steel coil is spatially superimposed with the region of interest (ROI) of the candidate boxes (BOX) of other feature categories (TAGs), and the intersection ratio between the trajectory line and the ROI is calculated to obtain the trajectory intersection degree value. Before and after the candidate box BOX of the hot-rolled steel coil target COL enters the ROI of the candidate box BOX of the other element category TAG, the flow rate characterization value Spd1 of the candidate box BOX of the hot-rolled steel coil target COL when entering the ROI of the candidate box BOX of the other element category TAG, and the flow rate characterization value Spd2 when leaving the ROI of the candidate box BOX of the other element category TAG, are calculated. The difference between the flow rate characterization value Spd1 and the flow rate characterization value Spd2 is used as the disturbance speed change.

6. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 5, characterized in that... The steel coil trajectory construction and state analysis module normalizes the occupancy time, trajectory intersection value, and interference speed change to obtain normalized occupancy time rD, normalized trajectory intersection value rC, and normalized interference speed change rS, and calculates the influence score Sco. Wherein, the weighting coefficients wD, wC, and wS are the weighting coefficients for the normalized occupancy time rD, the normalized trajectory intersection value rC, and the normalized disturbance velocity change rS, respectively, and p is the nonlinear coupling index; The influence score Sco, along with the candidate boxes BOX and steel coil trajectory lines corresponding to the other element categories TAG, are marked in the panoramic image to generate an interference influence distribution map.

7. The automatic hoisting and transfer system for hot-rolled steel coils according to claim 1, characterized in that... The path evaluation and optimization module performs joint analysis of production speed path and interference score and generates optimized modified paths, including: For the steel coil trajectory line shown in the interference impact distribution map, calculate the comprehensive evaluation value Sp of the path: Where aur is the flow velocity characterization value of the steel coil trajectory line, cur is the influence score Sco, and α and β are weighting coefficients; A path topology graph is constructed using the steel coil trajectory line described in the interference impact distribution map. The steel coil trajectory line is used as an edge in the path topology graph, and the comprehensive path evaluation value Sp is used as the path passage cost. In the aforementioned path topology graph, based on the initial task objective and the target task objective as path constraints, a path search and optimization algorithm is invoked to solve for the path combination with the minimum travel cost; When the interference score of a certain path segment is higher than the preset interference threshold and its corresponding flow speed is lower than the preset speed threshold, the probability of that path segment being selected as the optimal path is reduced; The obtained path segments are sequentially stitched together to generate an optimized and modified path; the optimized and modified path is then mapped onto the panoramic map according to the spatial calibration mapping relationship to generate an optimized display map.

8. A method for automatic hoisting and transfer control of hot-rolled steel coils, characterized in that... ,include: S1. Collect original video footage of the hot-rolled steel coil hoisting operation area, perform distortion correction, multi-view registration, and image stitching on the original video footage, and construct a panoramic image benchmark; S2. Based on the original video footage, the object recognition model is invoked to identify the steel coil target, production interference factors, and work-occupied area, generating a scene element map (MAP). S3. Perform cross-frame association processing on the steel coil target in the scene element map to obtain the continuous trajectory sequence of the steel coil. Based on the spatial overlap, temporal occupation, and state change relationship between the hot-rolled steel coil and the candidate boxes of other element categories (TAGs), generate an interference impact distribution map. S4. Based on the interference impact distribution map, conduct a comprehensive evaluation of the on-site path, construct a path topology map, perform path search and optimization processing, and generate optimized modified paths and optimized display maps.

Images