A method for identifying operation intentions in continuous casting billet-out area

Abstract

The present invention discloses a method for identifying operating intentions in a continuous casting billet-outing area, which belongs to the technical field of operating intention identification and is used in a continuous casting billet-outing area including a roller table and an overhead crane. The method comprises: obtaining the relative displacement of a plurality of clamps of a grab hook based on the collected image of the continuous casting billet-outing area to judge the opening and closing state of the grab hook and obtain the relative position of the grab hook relative to the electronic fence and the safety line; obtaining the residence time and movement trend of the grab hook in the electronic fence based on the relative position change of the grab hook in a plurality of images collected at intervals; and judging the operating intention based on the opening and closing state, residence time and movement trend of the grab hook. The present invention realizes the automatic identification and precise scheduling of operating intentions in the continuous casting billet-outing area through the deep integration of machine vision and intelligent algorithms. It not only solves the pain points of low efficiency and large hidden dangers of traditional manual scheduling, but also has significant advantages in environmental adaptability, response speed, safety and intelligence level.

Description

Technical Field

[0001] The present invention belongs to the technical field of industrial operation intention recognition, and in particular relates to an operation intention recognition method for a continuous casting billet discharge area. Background Art

[0002] In the steelmaking and rolling production process, the continuous casting discharging area is a critical link connecting the continuous casting machine and the hot delivery roller table. Its operating efficiency and safety directly impact the production capacity and product quality of the entire production line. Traditionally, the discharging area's discharging operations rely primarily on manually dispatched overhead cranes (bridge cranes). However, as the modern steel industry continues to strive for higher production efficiency and intelligent technology, this traditional operating model has gradually exposed numerous technical bottlenecks.

[0003] In the existing technology, the scheduling of overhead crane operations relies entirely on manual command. Ground operators need to communicate repeatedly with the overhead crane cab via intercom to confirm the clamping status of the overhead crane grab, the movement trajectory, and the placement of the ingot. Manual information transmission is subject to delays and errors, and the response speed of the overhead crane is limited, making it difficult to meet the high-frequency, short-cycle production needs. In addition, operators need to stay in the high-temperature roller area for a long time, and rely on experience to judge the safe distance between the overhead crane grab and the roller, which is prone to misjudgment due to obstructed vision or environmental interference. Operators need to maintain a high level of attention, frequently monitor the status and issue instructions, which can easily lead to misjudgment due to fatigue, thus affecting operational safety.

[0004] Therefore, some companies have attempted to introduce automated scheduling solutions based on sensors (such as ultrasonic or infrared sensors). However, the continuous casting process often operates in complex conditions characterized by high temperature, high humidity, and dense dust, with roller surface temperatures reaching hundreds of degrees Celsius. Traditional contact sensors (such as lidar and infrared rangefinders) are susceptible to failure in these extreme environments due to high-temperature oxidation, dust adhesion, or mechanical wear, making it impossible to accurately obtain real-time information about the position of the overhead crane hook and the billet.

[0005] In addition, even if basic distance detection can be achieved, it is impossible to intelligently classify the opening and closing status, movement trends and operation intentions of the overhead crane grab hook. It still needs to rely on ground operators to observe with the naked eye or manually transmit through walkie-talkies, and the information lag is significant.

[0006] Therefore, the existing continuous casting billet-out area operations are not stable enough in high-temperature environments, are highly dependent on manual labor, and lack the ability to intelligently identify the overhead crane's operating intentions. There is an urgent need for an operating intention recognition method that can operate stably under complex working conditions and has dynamic monitoring functions. Summary of the Invention

[0007] The present invention provides a method for identifying the operating intention of a continuous casting billet discharge area to solve the problems in the prior art of insufficient stability of existing continuous casting billet discharge area operations under high temperature environments, strong dependence on manual labor, and lack of intelligent recognition capability of overhead crane operating intentions, resulting in low automation, precision and safety of billet on-line and off-line operations.

[0008] The technical solution adopted in the present invention is:

[0009] A method for identifying operation intentions in a continuous casting billet discharging area, the method being used in a continuous casting billet discharging area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook, the method comprising:

[0010] Based on the collected image of the continuous casting billet discharge area, the relative displacement of the plurality of clamps of the grab hook is obtained to determine the opening and closing state of the grab hook, and the relative position of the grab hook relative to the electronic fence and the safety line is obtained;

[0011] Obtaining the residence time and movement trend of the grappling hook within the electronic fence according to the relative position change of the grappling hook in the plurality of images collected at intervals;

[0012] The operation intention is judged according to the opening and closing state, the dwell time and the movement trend of the grab hook, wherein the operation intention includes at least any one of online operation, waiting to be offline, offline operation and transit.

[0013] The method for identifying operation intentions in the continuous casting billet-out zone disclosed in the present invention also has the following additional technical features:

[0014] Online operation judgment, specifically:

[0015] If the grapple is in a clamping state, the electronic fence is triggered from the outside to the inside, and the grapple stays for 3 seconds or more and there is a downward trend, it is determined to be on-line operation;

[0016] During the on-line operation, the overhead crane is controlled by identifying the position of the billet relative to the roller table.

[0017] The overhead crane is controlled by identifying the position of the billet relative to the roller table, specifically:

[0018] The roller conveyor is provided with a position line and a stop line;

[0019] When the billet reaches the in-position line, a speed reduction instruction is sent to the overhead crane;

[0020] When the billet reaches the stop line, an emergency stop command is sent to the overhead crane.

[0021] The casting billet reaches the in-position line and the stop line, specifically:

[0022] When the billet covers the in-place line and the distance by which the edge of the billet exceeds the in-place line is less than or equal to a first over-line determination threshold, it is determined that the billet has reached the in-place line;

[0023] When the billet covers the stop line and the distance by which the edge of the billet exceeds the stop line is less than or equal to a second crossing-line determination threshold, it is determined that the billet reaches the stop line.

[0024] Waiting for offline judgment, specifically:

[0025] If the grab hook is in an unclamped state, and the grab hook stays for more than or equal to 5 seconds and is located on the safety line, it is determined to be waiting to go down the line.

[0026] Offline operation judgment, specifically:

[0027] If the hook is in an unclamped state, and the hook stays for 3 seconds or longer and shows a downward trend, it is determined to be offline operation;

[0028] During the off-line operation, the strand is clamped by a grabber so as to be separated from the roller table.

[0029] Border crossing judgment, specifically:

[0030] If the hook triggers the electronic fence from the outside to the inside, and the hook stays for more than or equal to 3 seconds and there is no downward trend, it is determined to be a crossing;

[0031] During the transit, safety detection is performed by the relative position of the grappling hook relative to the safety line.

[0032] The safety detection is specifically:

[0033] When the distance between the grappling hook and the safety line is greater than or equal to the safety line threshold, it is determined to be safe;

[0034] Otherwise, it is determined to be a collision and an alarm operation is performed.

[0035] Determine the intention of the assignment, specifically:

[0036] Collect videos for a fixed time period greater than 5 seconds, sparsely sample at intervals of seconds to obtain multiple images captured at intervals, and derive the operation intent and confidence based on the model;

[0037] When the confidence level is greater than or equal to 0.9, the operation intention is accurately determined and is used for operation control;

[0038] When the confidence level is greater than or equal to 0.7 and less than 0.9, the operation intention is determined based on the image verification result through image-assisted verification for operation control;

[0039] When the confidence level is less than 0.7, it is determined to be an invalid action.

[0040] The present invention also provides an operation intention recognition system for a continuous casting billet discharge area, which is used in a continuous casting billet discharge area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook. The system includes:

[0041] An image processing and analysis module is used to process the collected images, determine the opening and closing status of the grappling hook, and obtain the relative position of the grappling hook relative to the electronic fence and the safety line to obtain the residence time and movement trend;

[0042] an operation intention recognition module, configured to determine the operation intention based on the opening and closing state, dwell time, and movement trend of the grapple, wherein the operation intention includes at least one of online operation, waiting to be offline, offline operation, and transit;

[0043] The slab in-place detection module is used to control the overhead crane by identifying the position of the slab relative to the roller table;

[0044] The safety detection module is used to perform safety detection based on the relative position of the grappling hook relative to the safety line.

[0045] Due to the adoption of the above technical solution, the beneficial effects achieved by the present invention are as follows:

[0046] 1. This invention uses image-based, non-contact data acquisition to address the vulnerability of traditional contact sensors (such as lidar and infrared rangefinders) to failure in high-temperature (roller surface temperatures reaching hundreds of degrees Celsius), high-humidity, and dusty environments. This prevents data interruptions caused by high-temperature oxidation, dust adhesion, or steam interference. Furthermore, image acquisition requires no physical contact, enabling long-term stable operation and reducing equipment maintenance costs. It also ensures continuous operation in the continuous casting process.

[0047] By processing images, the system can determine the opening and closing status, dwell time, and movement trends of the grabber hook, thereby determining operational intent, such as on-line, off-line, waiting, and transit, achieving millisecond-level response. This not only eliminates the delays and errors associated with manual information transmission but also improves the overhead crane's response speed, meeting high-frequency production demands. Furthermore, the realized operational intent facilitates automated control of subsequent processes.

[0048] Therefore, by determining work intent through image processing, automated dispatching can replace manual scheduling. Ground operators no longer need to communicate frequently via intercom or constantly monitor the status of the overhead crane hook. They only need to intervene when a system anomaly occurs, reducing manual intervention. Automatically determining work intent reduces operator workload and the rate of misjudgments due to fatigue. This allows workers to focus on handling complex anomalies, improving overall work quality.

[0049] In summary, this invention, through the deep integration of machine vision and intelligent algorithms, achieves automated recognition of operational intent and precise scheduling in the continuous casting billet-stripping area. This not only addresses the pain points of traditional manual scheduling, which suffer from low efficiency and significant safety hazards, but also provides key technical support for the intelligent transformation of the steel industry through the deployment of high-temperature resistant equipment and a dynamic monitoring mechanism. It offers significant advantages in environmental adaptability, responsiveness, safety, and intelligence. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0051] Figure 1 It is a flow chart of the method for identifying the operation intention in the continuous casting billet-out zone according to one embodiment of the present invention. DETAILED DESCRIPTION

[0052] In order to more clearly illustrate the overall concept of the present invention, a detailed description is given below in an exemplary manner in conjunction with the accompanying drawings.

[0053] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Therefore, the scope of protection of the present invention is not limited to the specific embodiments disclosed below.

[0054] like Figure 1 As shown, a method for identifying operating intentions in a continuous casting billet-out area is used in a continuous casting billet-out area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook.

[0055] It should be noted that the present invention is applied to the continuous casting billet area. The continuous casting billet area is a key link in the steel production process and is mainly composed of the following core components:

[0056] Roller tables are used to transport the ingots produced by the continuous casting machine to the heating furnace or subsequent processing. The roller table surface temperature can reach hundreds of degrees Celsius, and the process is characterized by high temperature, high humidity, and dense dust. Overhead cranes use hooks to grip the ingots and lift them.

[0057] In addition, electronic fences are installed on both sides of the roller conveyor to demarcate the continuous casting operation area, serving as virtual boundaries. A safety line is set within a 1m height interval above the roller conveyor to dynamically monitor the safe distance between the grab hook and the roller conveyor.

[0058] Specifically, physical coordinates are mapped to the image coordinate system using visual calibration algorithms (such as perspective transformation and homography), defining the virtual boundaries (electronic fences) on both sides of the roller track. Collaborating with the overhead crane operator, a three-dimensional safety reference surface is dynamically fitted within a 1m height range above the roller track to generate the safety line position.

[0059] The method includes S100: obtaining the relative displacement of multiple clamps of the grab hook based on the collected image of the continuous casting billet discharge area to determine the opening and closing state of the grab hook and obtain the relative position of the grab hook relative to the electronic fence and the safety line.

[0060] The core purpose of this step is to replace traditional sensors with non-contact visual monitoring to solve the technical problem of determining the status and position of the overhead crane hook in the high-temperature and dusty environment of the continuous casting billet discharge area. Replacing traditional sensors such as lidar and infrared rangefinders with high-temperature industrial cameras can avoid equipment failure caused by high-temperature oxidation and dust adhesion, and adapt to high-temperature environments.

[0061] In addition, in this step, the relative displacement analysis of the clamp is used to determine the opening and closing status of the grab hook (clamped / unclamped), and the relative position of the grab hook, the electronic fence, and the safety line is calculated, providing basic data for subsequent operation intention identification.

[0062] Specifically, an industrial camera with a resolution of ≥4K and a frame rate of ≥30fps is installed on the top of the roller conveyor to cover the entire operating range of the overhead crane grab hook.

[0063] Furthermore, to determine the clamp open / close state, the clamp outline is extracted through edge detection (such as the Canny algorithm) or object segmentation (such as the YOLO / U-Net model), and the relative displacement between the clamps (such as the clamp opening / closing angle) is calculated. For example, a clamp opening / closing angle of less than 30° is considered closed (clamped), and greater than 60° is considered open (unclamped).

[0064] The relative position of the grapple relative to the electronic fence and the safety line is determined. When the grapple enters the electronic fence area, subsequent operation intention analysis logic is triggered. The distance between the key point at the bottom of the clamp and the safety line is calculated (e.g., >500mm is considered safe to leave), dynamically monitoring the safe distance between the grapple and the roller conveyor.

[0065] It should be noted that this step is to judge the opening and closing state and relative position based on a single-frame video image according to image recognition technology. The image data can also be provided by taking pictures at regular intervals by a camera. The present invention does not impose any restrictions on this.

[0066] This step solves the technical bottleneck of judging the state and position of the grab hook in the high-temperature environment of the continuous casting billet discharge area through the deep integration of machine vision and intelligent algorithms, providing a reliable data basis for subsequent operation intention identification.

[0067] S200: Obtaining the residence time and movement trend of the grappling hook within the electronic fence according to the relative position change of the grappling hook in a plurality of images collected at intervals.

[0068] The core purpose of this step is to analyze the dwell time and movement trend of the grappling hook within the electronic fence through dynamic analysis of multiple images, providing key input for subsequent operation intention classification (such as going online, going offline, and passing through), and avoiding misjudgment caused by single-frame information.

[0069] The model captures video clips using a sliding window, sparsely sampling multiple images for input, and outputs dwell time and motion trends. Specifically, based on these images, an object detection model (such as YOLO / U-Net) extracts key points of the grappling hook (such as the clamp base and the spreader reference point). Its positional changes within the geo-fence are calculated to determine its motion trends. Furthermore, the dwell time of the grappling hook is recorded from the moment it enters the geo-fence.

[0070] It should be noted that the movement trend includes a downward trend. The optical flow analysis of the hook position in consecutive frame images (3D CNN extracts optical flow features) is used to determine whether there is a downward trend (such as the clamp position continuously moving downward).

[0071] Single images are susceptible to environmental interference (such as dust obstruction and light fluctuations). This step improves the accuracy of motion trend determination by analyzing multiple images using continuous optical flow features (extracted by a 3D CNN) and dwell time thresholds. Furthermore, dwell time and downward trends (continuous downward movement) are used to comprehensively determine operational intent, avoiding misjudgments based on single-source information.

[0072] This step solves the technical bottleneck of determining the hook's dwell time and movement trend in the continuous casting billet-out area through the deep integration of machine vision and intelligent algorithms, providing a reliable data basis for subsequent operation intention classification and significantly improving operation efficiency and safety.

[0073] S300: judging the operation intention according to the opening and closing state, the dwelling time and the movement trend of the grab hook, wherein the operation intention includes at least any one of online operation, waiting to be offline, offline operation and transit.

[0074] The core purpose of this step is to achieve accurate classification of the overhead crane operation intentions in the continuous casting billet-cutting area (online, waiting to go offline, offline, in transit) through multi-dimensional data fusion (hook opening and closing status, dwell time, movement trend), thereby replacing traditional manual scheduling and improving operation efficiency and safety.

[0075] Inputs in this step include the hook's open / closed state, specifically, whether it is clamped or unclamped. The clamped state refers to the billet being transported. The dwell time is also included, which is the duration after the hook enters the electronic fence; and the movement trend, specifically whether there is a downward trend (continuous downward movement in the image).

[0076] It is understood that for the on-line operation, the grab hook is in a clamping state, clamping the billet from the outside to the inside into the electronic fence, and there is a downward trend within the electronic fence. At this time, it is necessary to control the grab hook to descend to place the billet on the corresponding position of the roller.

[0077] When waiting to be unloaded, the grappling hook is in the unclamped state and stays within the electronic fence at the safety line for a period of time. During this time, there is no need to control the overhead crane, which can avoid frequent starts and stops of the overhead crane due to misjudgment and improve dispatching efficiency.

[0078] During off-line operations, the grab hook is in an unclamped state, remaining within the electronic fence and showing a downward trend. At this time, the overhead crane needs to be controlled to use the grab hook to grab the ingot from the roller table and place it off-line in the ingot temporary storage area.

[0079] For crossing detection, if the hook enters the electronic fence from the outside and stops but does not descend, it is important to pay attention to the distance between the key point of the hook and the safety line (>500mm is considered safe departure) to issue a safety warning to avoid collisions with the hook.

[0080] Understandably, ground operators frequently communicate with overhead crane operators via intercom, which can lead to information delays and errors. This step, however, analyzes multiple images and calculates the hook's dwell time and movement trends in real time, reducing manual intervention and enabling corresponding overhead crane operations, thereby improving the crane's response speed.

[0081] This step, through the deep integration of machine vision and intelligent algorithms, overcomes the technical bottleneck of task intent classification in the continuous casting billet-out area. Combining the hook's status, dwell time, and movement trends, task intent is categorized into four categories: on-line, waiting to be off-line, off-line, and in transit. This provides a basis for subsequent control instructions. This enables automated overhead crane scheduling, replacing manual control and reducing information transmission delays and errors.

[0082] As a preferred embodiment of the present invention, online operation judgment is specifically as follows:

[0083] If the grapple is in a clamping state, the electronic fence is triggered from the outside to the inside, and the grapple stays for 3 seconds or more and there is a downward trend, it is determined to be on-line operation;

[0084] During the on-line operation, the overhead crane is controlled by identifying the position of the billet relative to the roller table.

[0085] The core purpose of this embodiment is to deeply integrate the opening and closing status, dwell time and movement trend of the grab hook to achieve accurate recognition and automatic control of the overhead crane's online operation intention in the continuous casting billet discharge area.

[0086] Through comprehensive judgment of the hook status, electronic fence triggering, dwell time and movement trend, it replaces traditional manual scheduling, reduces information transmission delays and errors, and realizes automated scheduling.

[0087] It is understandable that when operating on the line, the overhead crane grabs the ingot through the grab hook, triggers the electronic fence from the outside to the inside from the ingot temporary storage area outside the electronic fence, stays inside the electronic fence, and descends to place the ingot on the roller.

[0088] Therefore, the grappling hook in a clamped state and triggering the electronic fence from the outside inward are used as the criteria for online operation. The grappling hook must remain within the electronic fence for ≥ 3 seconds to avoid misjudgment due to misoperation. In addition, the downward trend is combined with further restrictions on online operation, further improving the accuracy of online operation intention judgment.

[0089] This implementation method accurately determines the intention of online operations by determining the state, dwell time, and movement trend of the grab hook during operations in the continuous casting billet-out area, providing a reliable data basis for subsequent automated scheduling and significantly improving operational efficiency and safety.

[0090] As an example of this embodiment, the overhead crane is controlled by identifying the position of the billet relative to the roller table, specifically:

[0091] The roller conveyor is provided with a position line and a stop line;

[0092] When the billet reaches the in-position line, a speed reduction instruction is sent to the overhead crane;

[0093] When the billet reaches the stop line, an emergency stop command is sent to the overhead crane.

[0094] The core purpose of this embodiment is to achieve accurate identification and automatic control of the position of the billet during the online operation of the overhead crane in the continuous casting billet discharge area through the deep integration of machine vision and intelligent algorithms on the basis of online operation judgment.

[0095] It should be noted that the position line and stop line are preset at the end of the roller conveyor, and the physical coordinates are mapped to the image coordinate system based on the visual calibration algorithm.

[0096] The casting billet reaches the in-position line when the front edge of the casting billet reaches the in-position line. The casting billet reaches the stop line when the front edge of the casting billet reaches the stop line.

[0097] When the vision system detects that the front edge of the billet coincides with the in-place line, it sends a speed reduction command to the overhead crane. At this time, the image coordinates are mapped to physical space using a perspective transformation algorithm to calculate the degree of overlap between the billet edge and the in-place line.

[0098] When the billet completely covers the stop line, an emergency stop command is sent to the overhead crane. At this time, the image coordinates are mapped to physical space using a perspective transformation algorithm to calculate the overlap between the billet edge and the stop line.

[0099] This embodiment is a key step in achieving on-line operation scheduling. By identifying the positional relationship between the billet and the roller table's in-position and stop lines, it sends a speed reduction or emergency stop command to the overhead crane, precisely controlling the crane to ensure the billet is precisely aligned with the roller table, avoiding production failures caused by positional deviations. Furthermore, by replacing traditional manual scheduling with automated scheduling, information transmission delays and errors are reduced, the overhead crane's response speed is improved, and operational efficiency is enhanced.

[0100] Specifically, the casting reaches the in-position line and the stop line, specifically:

[0101] When the billet covers the in-place line and the distance by which the edge of the billet exceeds the in-place line is less than or equal to a first over-line determination threshold, it is determined that the billet has reached the in-place line;

[0102] When the billet covers the stop line and the distance by which the edge of the billet exceeds the stop line is less than or equal to a second crossing-line determination threshold, it is determined that the billet reaches the stop line.

[0103] This embodiment identifies the positional relationship between the billet and the roller position line and stop line, and sends a speed reduction or emergency stop command to the overhead crane to ensure that the billet is accurately aligned with the roller, avoids production failures caused by position deviation, and realizes accurate identification and automatic control of the billet position during the overhead crane online operation in the continuous casting billet discharge area.

[0104] The slab reaches the in-position line, specifically when the front edge of the slab overlaps the in-position line (crossing line determination threshold ±15mm). The slab reaches the stop line, specifically when the slab completely covers the stop line (crossing line determination threshold ±10mm).

[0105] The positional relationship between the in-line and stop lines is calculated in real time, and a threshold is set for line determination to avoid production failures caused by position deviations. For example, a sudden crossing of the stop line can lead to misjudgment of position, resulting in a failure to stop the overhead crane in time and causing a processing accident.

[0106] This embodiment further realizes accurate determination of the position of the ingot in the continuous casting billet discharge area through detailed judgment of reaching the in-position line and reaching the stop line, realizes precise control of the overhead crane, and significantly improves operation efficiency and safety.

[0107] As a preferred embodiment of the present invention, waiting for offline judgment is specifically as follows:

[0108] If the grab hook is in an unclamped state, and the grab hook stays for more than or equal to 5 seconds and is located on the safety line, it is determined to be waiting to go down the line.

[0109] The core purpose of this implementation is to achieve accurate identification and automated scheduling of overhead cranes waiting to be unloaded in the continuous casting billet-discharging area through the deep integration of machine vision and intelligent algorithms. By identifying the unclamped state of the grab hook, its dwell time, and the safety line position, it can determine whether the overhead crane is waiting to be unloaded, thus avoiding frequent starts and stops of the overhead crane due to misjudgment and improving scheduling efficiency.

[0110] It is understood that waiting for unloading is a working state between loading and unloading operations. In this state, the billet has been placed on the roller table, and the grabber is not clamped. The grabber remains at the safety line position above the roller table, waiting for the billet to be finished before unloading.

[0111] The grappling hook stays in the safety line area for ≥5 seconds, emphasizing the waiting time of the grappling hook in the safety line area to avoid misjudgment due to the passage of an empty hook.

[0112] After it is determined to be waiting to go offline, a pause command is sent to the overhead crane control system, and the overhead crane enters pause mode until a new dispatch command is received. This prevents the overhead crane from starting due to misjudgment, avoids the tedious operation of controlling the grab hook to move out of the electronic fence area and then move it back into the electronic fence area, and reduces energy consumption and mechanical wear.

[0113] In addition, when it is determined to be waiting to go offline, controlling the grab hook in the safety line area can also avoid accidents caused by collision between the grab hook and the billet.

[0114] This embodiment determines whether the overhead crane is in a waiting-to-go state by identifying the unclamped state, dwell time, and safety line position of the grab hook, providing a reliable data basis for subsequent overhead crane operation control, significantly reducing operation difficulty and improving operation safety.

[0115] As a preferred embodiment of the present invention, offline operation judgment is specifically as follows:

[0116] If the hook is in an unclamped state, and the hook stays for 3 seconds or longer and shows a downward trend, it is determined to be offline operation;

[0117] During the off-line operation, the strand is clamped by a grabber so as to be separated from the roller table.

[0118] The core purpose of this implementation is to achieve precise recognition and automated control of the overhead crane's off-line operation in the continuous casting billet-discharging area through the deep integration of machine vision and intelligent algorithms. Off-line operations are determined by comprehensively determining the hook's status (unclamped), dwell time (≥3 seconds), and movement trend (descending), replacing traditional manual scheduling and reducing information transmission delays and errors.

[0119] It is understood that during the off-line operation, the grab hook is located within the electronic fence, in an unclamped state, and continuously descends to clamp the billet and transport it from the roller conveyor to the billet temporary storage area. Therefore, the off-line operation is determined by the grab hook's unclamped state, a residence time of 3 seconds or more within the electronic fence, and a downward trend (continuous downward movement of the frame image position). The overhead crane performs the clamping action to remove the billet from the roller conveyor.

[0120] This implementation method uses the deep integration of machine vision and intelligent algorithms to determine the offline operation based on the hook status, dwell time and movement trend in the continuous casting billet discharge area, automatically dispatch the overhead crane to perform the offline operation, reduce manual intervention, and improve the response speed of the overhead crane.

[0121] As a preferred embodiment of the present invention, border crossing judgment is specifically as follows:

[0122] If the hook triggers the electronic fence from the outside to the inside, and the hook stays for more than or equal to 3 seconds and there is no downward trend, it is determined to be a crossing;

[0123] During the transit, safety detection is performed by the relative position of the grappling hook relative to the safety line.

[0124] The core purpose of this implementation is to achieve precise identification and safety monitoring of the overhead crane's transit status in the continuous casting billet-casting area through the deep integration of machine vision and intelligent algorithms. By identifying the hook's triggering of the electronic fence, its dwell time, and its movement trend, the system determines whether the overhead crane is in transit, avoiding misjudgment of other operations (such as on-line operations) and improving scheduling efficiency. During transit, the positional relationship between the hook and the safety line is dynamically monitored to ensure that the overhead crane remains in a safe area during transit, reducing the risk of collision.

[0125] It is understandable that transit intent is a distinct operational intent from on-line operations, waiting to go offline, and off-line operations. During a transit intent, the grapple may or may not be in a clamped state. Therefore, the grapple's open or closed state is not used in transit determination. Furthermore, during a transit intent, the grapple typically triggers the electronic fence and moves horizontally within the electronic fence, i.e., it remains within the electronic fence for a period of time without any downward movement, and then triggers the electronic fence again to pass through the electronic fence.

[0126] Therefore, in this embodiment, the conditions for judging the intention to cross the border are that the grappling hook triggers the electronic fence from the outside to the inside, the grappling hook stays in the electronic fence for ≥3 seconds, and there is no downward trend.

[0127] During the border crossing operation, it is necessary to control the height of the grab hook, that is, to perform safety detection through the relative position of the grab hook relative to the safety line. The safety detection is specifically:

[0128] When the distance between the grappling hook and the safety line is greater than or equal to the safety line threshold, it is determined to be safe;

[0129] Otherwise, it is determined to be a collision and an alarm operation is performed.

[0130] This embodiment enables real-time safety monitoring and collision warnings for the overhead crane operating area in the continuous casting billet ejection zone. By determining the distance threshold between the hook and the safety line, accidents caused by close proximity, such as collisions between the overhead crane and the roller conveyor or personnel, can be avoided, accurately determining the safety status. When the distance between the hook and the safety line falls below the preset threshold, an alarm is triggered, ensuring the safety of the operating area and providing dynamic warnings.

[0131] This embodiment uses a visual calibration algorithm to calculate the distance between the key points of the grappling hook (such as the sling reference point) and the safety line. The perspective transformation algorithm maps the image coordinates to the physical space and monitors the distance changes between the grappling hook and the safety line in real time.

[0132] If the distance between the hook and the safety line is greater than or equal to a preset threshold (e.g., ≥500mm), the system is considered safe. If the distance between the hook and the safety line is less than or equal to a preset threshold (e.g., less than 500mm), the system is considered to be in a collision risk state. When a collision risk is identified, an emergency stop command is sent to the overhead crane control system, triggering an audible and visual alarm.

[0133] This embodiment calculates the distance between the key point of the grab hook and the safety line in real time (for example, if it is >500mm, it is determined to be safe to leave), thereby avoiding collision between the overhead crane and the roller table and ensuring operation safety.

[0134] As a preferred embodiment of the present invention, the operation intention is judged as follows:

[0135] Collect videos for a fixed time period greater than 5 seconds, sparsely sample at intervals of seconds to obtain multiple images captured at intervals, and derive the operation intent and confidence based on the model;

[0136] When the confidence level is greater than or equal to 0.9, the operation intention is accurately determined and is used for operation control;

[0137] When the confidence level is greater than or equal to 0.7 and less than 0.9, the operation intention is determined based on the image verification result through image-assisted verification for operation control;

[0138] When the confidence level is less than 0.7, it is determined to be an invalid action.

[0139] This implementation aims to reduce misjudgment rates and ensure the reliability of operational intent through multi-image analysis and confidence-level processing. Furthermore, control strategies are dynamically adjusted based on confidence thresholds (≥0.9, ≥0.7) to avoid erroneous operations caused by low confidence levels.

[0140] Capture video clips of a fixed time period (>5 seconds) to ensure coverage of key overhead crane operation actions (such as hook movement, clamping, and disengagement from the roller conveyor). For example, use a sliding window to capture an 8-second video clip for operation intent recognition.

[0141] Sparse sampling is performed at intervals of seconds (e.g., 1 frame per second) to obtain multiple frames of video images (e.g., 8 frames), reducing data redundancy and retaining key action features.

[0142] In this embodiment, the work intention classification model adopts a two-stream network model (Two-Stream CNN) or 3DCNN, inputs sparsely sampled multi-frame images, and outputs the probability and confidence of four types of work intentions (Score∈[0,1]).

[0143] At high confidence (Score ≥ 0.9), control signals (such as grip, detach, and pause) are directly triggered. At medium confidence (0.7 ≤ Score < 0.9), an image detection model (such as YOLO / U-Net) is used to perform real-time verification of the current frame (detection targets: hook key points, safety line position). At low confidence (Score < 0.7), the action is considered invalid and discarded to avoid misoperation.

[0144] It should be noted that when the Score is ≥0.9, the operation intention is accurately determined, and control instructions (such as going online, going offline, or passing through) can be directly sent to the overhead crane control system. When the Score is 0.7 ≤ Score <0.9, image-assisted verification (such as detecting whether the grabber is holding the ingot or whether the electronic fence has been triggered) is used to determine the final intention based on the verification results, improving decision-making reliability. When the Score is <0.7, the action is determined to be invalid, and the system records the abnormal event and waits for reanalysis in subsequent frames.

[0145] This implementation comprehensively determines the operation intention based on confidence thresholds (≥0.9, ≥0.7) to avoid misjudgments. Furthermore, through confidence level tiering (Score ≥0.9 triggers direct control, while 0.7 ≤ Score <0.9 invokes image-assisted verification), it ensures high accuracy and a low misjudgment rate.

[0146] The present invention further provides an operation intention recognition system for a continuous casting billet discharge area, which is used in a continuous casting billet discharge area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook. The system includes:

[0147] An image processing and analysis module is used to process the collected images, determine the opening and closing status of the grappling hook, and obtain the relative position of the grappling hook relative to the electronic fence and the safety line to obtain the residence time and movement trend;

[0148] an operation intention recognition module, configured to determine the operation intention based on the opening and closing state, dwell time, and movement trend of the grapple, wherein the operation intention includes at least one of online operation, waiting to be offline, offline operation, and transit;

[0149] The slab in-place detection module is used to control the overhead crane by identifying the position of the slab relative to the roller table;

[0150] The safety detection module is used to perform safety detection based on the relative position of the grappling hook relative to the safety line.

[0151] Therefore, any effect of the method for identifying the working intention of the continuous casting billet stripping area can be achieved, which will not be described in detail here.

[0152] Anything not described in the present invention can be achieved by adopting or drawing on existing technologies.

[0153] The various embodiments in this specification are described in a progressive manner, and the same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on the differences from other embodiments.

[0154] The foregoing is merely an embodiment of the present invention and is not intended to limit the present invention. It will be apparent to those skilled in the art that various modifications and variations of the present invention are possible. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention are intended to be included within the scope of the claims of the present invention.

Claims

1. A method for identifying operation intentions in a continuous casting billet stripping area, characterized in that: For a continuous casting billet discharge area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook, the method comprising: Based on the collected image of the continuous casting billet discharge area, the relative displacement of the plurality of clamps of the grab hook is obtained to determine the opening and closing state of the grab hook, and the relative position of the grab hook relative to the electronic fence and the safety line is obtained; Obtaining the residence time and movement trend of the grappling hook within the electronic fence according to the relative position change of the grappling hook in the plurality of images collected at intervals; Determine the operation intention according to the opening and closing state, dwell time and movement trend of the grab hook, wherein the operation intention at least includes on-line operation, waiting for off-line operation, off-line operation and transit; If the grab hook is in a clamping state, the electronic fence is triggered from the outside to the inside, and the grab hook stays for more than or equal to 3 seconds and there is a downward trend, it is determined to be on-line operation. If the hook is in an unclamped state and the hook stays for more than or equal to 5 seconds and is located on the safety line, it is determined to be waiting to go down the line. If the hook is in an unclamped state, and the hook stays for 3 seconds or longer and has a downward trend, it is determined to be offline operation. If the grappling hook triggers the electronic fence from the outside to the inside, and the grappling hook stays for more than or equal to 3 seconds and there is no downward trend, it is determined to be a crossing.

2. The method for identifying operation intentions in a continuous casting billet-out zone according to claim 1, characterized in that: Online operation judgment, specifically: During the on-line operation, the overhead crane is controlled by identifying the position of the billet relative to the roller table.

3. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 2, characterized in that: The overhead crane is controlled by identifying the position of the billet relative to the roller table, specifically: The roller conveyor is provided with a position line and a stop line; When the billet reaches the in-position line, a speed reduction instruction is sent to the overhead crane; When the billet reaches the stop line, an emergency stop command is sent to the overhead crane.

4. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 3, characterized in that: The casting billet reaches the in-position line and the stop line, specifically: When the billet covers the in-place line and the distance by which the edge of the billet exceeds the in-place line is less than or equal to a first over-line determination threshold, it is determined that the billet has reached the in-place line; When the billet covers the stop line and the distance by which the edge of the billet exceeds the stop line is less than or equal to a second crossing-line determination threshold, it is determined that the billet reaches the stop line.

5. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 1, characterized in that: Offline operation judgment, specifically: During the off-line operation, the strand is clamped by a grabber so as to be separated from the roller table.

6. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 1, characterized in that: Border crossing judgment, specifically: During the transit, safety detection is performed by the relative position of the grappling hook relative to the safety line.

7. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 6, characterized in that: The safety detection is specifically: When the distance between the grappling hook and the safety line is greater than or equal to the safety line threshold, it is determined to be safe; Otherwise, it is determined to be a collision and an alarm operation is performed.

8. The method for identifying operation intentions in a continuous casting billet stripping area according to claim 1, characterized in that: Determine the intention of the assignment, specifically: Collect videos for a fixed time period greater than 5 seconds, sparsely sample at intervals of seconds to obtain multiple images captured at intervals, and derive the operation intent and confidence based on the model; When the confidence level is greater than or equal to 0.9, the operation intention is accurately determined and is used for operation control; When the confidence level is greater than or equal to 0.7 and less than 0.9, the operation intention is determined based on the image verification result through image-assisted verification for operation control; When the confidence level is less than 0.7, it is determined to be an invalid action.

9. A system for identifying operation intentions in a continuous casting billet-out area, characterized in that: It is used for the continuous casting billet discharge area including a roller table and an overhead crane, wherein electronic fences are provided on both sides of the roller table, a safety line is provided above the roller table, and the overhead crane includes a grab hook. The system includes: An image processing and analysis module is used to process the collected images, determine the opening and closing status of the grappling hook, and obtain the relative position of the grappling hook relative to the electronic fence and the safety line to obtain the residence time and movement trend; The operation intention recognition module is used to judge the operation intention according to the opening and closing state, dwell time and movement trend of the grab hook, wherein the operation intention includes at least on-line operation, waiting for off-line operation, off-line operation and transit. If the grab hook is in a clamping state, the electronic fence is triggered from the outside to the inside, and the grab hook stays for more than or equal to 3 seconds and there is a downward trend, it is determined to be on-line operation. If the hook is in an unclamped state and the hook stays for more than or equal to 5 seconds and is located on the safety line, it is determined to be waiting to go down the line. If the hook is in an unclamped state, and the hook stays for 3 seconds or longer and has a downward trend, it is determined to be offline operation. If the hook triggers the electronic fence from the outside to the inside, and the hook stays for more than or equal to 3 seconds and there is no downward trend, it is determined to be a crossing; The slab in-place detection module is used to control the overhead crane by identifying the position of the slab relative to the roller table; The safety detection module is used to perform safety detection based on the relative position of the grappling hook relative to the safety line.

Images