Intelligent hoisting system and method for self-decision transport container

Abstract

The invention discloses a self-decision-making transportation container intelligent hoisting system and method, relates to the technical field of radioactive substance transportation equipment, and solves the technical problem that an existing hoisting tool is difficult to adapt to precise operation requirements under complex working conditions due to lack of intelligent sensing and autonomous decision-making capabilities. According to the technical scheme, the method is characterized in that the position of the lifting lug is recognized and positioned through an improved YOLOv8 network and an improved edge detection algorithm, high-precision positioning of the lifting lug is achieved, and the positioning efficiency is improved; the lazy arm is provided with a laser distance measuring sensor so as to feed back distance data in real time and control errors in a closed-loop mode; the power-off instant locking and safety real-time monitoring functions are achieved, and the safety is remarkably improved. The lifting operation in the nuclear industry can be promoted to be transformed to be intelligent and unmanned, and a standardized solution is provided for transportation of similar radioactive substances; the technology can be expanded to heavy load scenes such as ports and wind power, and industrial equipment is helped to be intelligently upgraded.

Description

Technical Field

[0001] This application relates to the field of radioactive material transport equipment technology, and in particular to an intelligent hoisting system and method for self-decision transport containers. Background Technology

[0002] As industrial automation evolves towards intelligence and unmanned operation, the problems of traditional UF6 transport container hoisting equipment, such as reliance on manual experience, low positioning accuracy, and insufficient safety redundancy, are becoming increasingly prominent. This is especially true in high-risk fields like nuclear power, where the hoisting of cylindrical containers places even higher demands on equipment reliability, adaptability, and safety.

[0003] Currently, intelligent lifting devices are used in actual transportation activities. For example, Chinese patent application CN 119750365 A discloses an intelligent lifting device and its lifting method for handling cable reels, which relates to the field of cable production equipment technology. The intelligent lifting device can lift cable reels of different outer diameters, with strong workpiece adaptability; it can also lift four small reels at a time, resulting in high work efficiency; the rigid guide telescopic column of the intelligent lifting device ensures the positioning accuracy of the lifting device, making the handling of reels automatic, efficient, and safe, thereby improving the automation level of the cable industry.

[0004] However, existing lifting devices generally lack intelligent sensing and autonomous decision-making capabilities, making it difficult to adapt to the precise operation requirements under complex working conditions. Summary of the Invention

[0005] This application provides a self-decision-making intelligent lifting system and method for transport containers. Its technical purpose is to enable intelligent lifting equipment for transport containers to have intelligent perception and autonomous decision-making capabilities, so as to adapt to the precise operation requirements under complex working conditions.

[0006] The above-mentioned technical objective of this application is achieved through the following technical solution: A self-decision-making intelligent lifting system for transport containers includes a mechanical structure, a vision system, an electronic control system, and a data fusion module. The mechanical structure includes two booms, each equipped with a safety locking device and a laser rangefinder. The vision system includes a global camera, an industrial camera, and an image processing module. After the container is placed in the lifting area, the global camera captures images of the container. The image processing module determines the relative position of the container with respect to the two booms based on the container images, and then feeds the relative position back to the electrical control system. The electrical control system controls the two booms to move as a whole based on the relative position so that the center of the two booms is directly opposite the center of the container. An industrial camera captures images of the lifting lugs of the container. The image processing module identifies and locates the position of the lifting lugs based on the images, and then feeds the position of the lifting lugs back to the electrical control system. The electronic control system controls the boom to move downwards based on the position of the lugs. After the laser rangefinder detects that the two booms have reached the preset height, it sends the preset height to the data fusion module. The data fusion module then synchronizes the preset height to the electronic control system, and the electronic control system controls the two booms to stop moving downwards. The electrical control system controls the two booms to move from the middle to both ends according to the boom-load coupling dynamic model until they hook onto the lifting lugs. The lifting lugs are then locked by the safety locking device, and then the two booms are raised to lift the container to the designated position. The industrial camera captures images of the container's lifting lugs, and the image processing module identifies and locates the lifting lug positions based on these images, including: Industrial cameras were used to capture images of the container's lifting lugs under different lighting, angle, and obstruction conditions. The image processing module identifies the location of the sling in the image using a trained improved YOLOv8 network, and then locates the sling position using an improved edge detection algorithm. The backbone of the improved YOLOv8 network includes a MobileNetV3 lightweight module, and the neck network includes a multi-scale feature fusion module and a CBAM module. The CBAM module combines channel attention and spatial attention mechanisms. The total loss of the improved YOLOv8 network includes regression loss, classification loss, and keypoint loss, expressed as: Total Loss = Regression Loss + Classification Loss + Keypoint Loss. The regression loss is EIoU Loss, the classification loss is Focal Loss, and the keypoint loss is L1 loss or Wing Loss.

[0007] Preferably, the mechanical structure further includes a harmonic reducer, a servo motor, a linear guide, a ball screw, a slewing support bearing, and a synchronous pulley mechanism. Both booms move left and right via the linear guide and servo motor, move up and down via the ball screw and harmonic reducer, and rotate 360° continuously via the slewing support bearing. The distance between the two booms is adjusted by the synchronous pulley mechanism driven by the servo motor.

[0008] Preferably, the safety locking device is composed of an electromagnetic-hydraulic composite braking system, which includes a hydraulic brake, a plunger pump, an accumulator, a supercapacitor, a hydraulic cylinder, brake pads, and mechanical chucks. The hydraulic brake is connected to the plunger pump, the plunger pump is connected to the accumulator and the hydraulic cylinder, the accumulator is connected to the supercapacitor, the supercapacitor is connected to the hydraulic brake, the hydraulic cylinder is connected to the brake pads, and the brake pads are connected to the mechanical chucks.

[0009] Preferably, the global camera is equipped with a wide-angle lens, and the industrial camera is equipped with a telecentric lens.

[0010] Preferably, the electronic control system includes a servo driver and an embedded controller. The laser rangefinder and vision system input the collected data to the embedded controller, which then sends the data to the servo driver. The servo driver drives the mechanical structure and safety locking device according to the collected data.

[0011] Preferably, the boom-load coupled dynamic model is described by the Lagrange-Euler equations, expressed as: ; in, These represent the position vector, velocity vector, and acceleration vector of the boom, respectively. This represents the control torque vector of the boom. Indicates the external disturbance torque. This represents the inertia matrix.

[0012] Preferably, the training process of the improved YOLOv8 network includes: Industrial cameras were used to capture images of the container's lifting lugs under different lighting, angle, and obstruction conditions. Use the annotation tool to annotate the lugs in the lug image and generate a YOLO format label file; A synthetic dataset of stilts was generated using a Generative Adversarial Network (GAN). The YOLO format label file and synthetic dataset are input into the improved YOLOv8 network for training until the improved YOLOv8 network converges, resulting in the trained improved YOLOv8 network.

[0013] Preferably, the step of locating the position of the lug by improving the edge detection algorithm includes: The pixel-level edges of the identified lugs are extracted based on the Canny operator, and the lugs are coarsely located based on the pixel-level edges. The sub-pixel edge extraction operator based on local gray area effect fitting extracts the sub-pixel level edges of the lug after coarse positioning, and accurately positions the lug based on the sub-pixel level edges. Among them, based on the equation of a straight line y=ax+b The obtained local gray area is represented as: ; ; in, M , L , R Both represent the area enclosed by the subpixel edge; S M , S L , SR All represent local grayscale areas; A , B This represents the grayscale values ​​on both sides of the edge; h The selected window width size; Based on curve equations The obtained local gray area is represented as: ; ; The subpixel-level edge of the lug is obtained by local grayscale area.

[0014] Preferably, the hoisting system further includes a software system, which is connected to both the data fusion module and the electrical control system. The data fusion module is connected to the vision system and receives visual data from the vision system and positional data from the laser rangefinder. The software system manually issues control commands to the electrical control system based on the visual data and positional data provided by the data fusion module. The electrical control system controls the mechanical structure and safety locking device based on the control commands.

[0015] A self-decision-making intelligent lifting method for transport containers, the lifting method being used in a lifting system, the lifting method comprising: After the container is placed in the lifting area, the container image is acquired. Based on the container image, the relative position of the container with respect to the two booms is determined. Then, based on the relative position, the two booms are moved as a whole so that the center of the two booms is directly opposite the center of the container. The image of the container's lifting lugs is acquired, and the position of the lifting lugs is identified and located based on the image. The boom is controlled to move downward according to the position of the lifting lugs. At the same time, it is detected whether the two booms have reached the preset height. If they have, the two booms are controlled to stop moving downward. Otherwise, the booms are controlled to move downward until they reach the preset height. Based on the boom-load coupling dynamics model, the two booms are controlled to move from the middle to both ends until they hook onto the lifting lugs. Then the lifting lugs are locked, and the two booms are raised to lift the container to the designated position.

[0016] The above technical solutions can achieve at least some of the following technical effects: The self-decision-making intelligent lifting system and method for transport containers described in this application, through improved YOLOv8 network and improved edge detection algorithm, identifies and locates the lifting lug position, achieving high-precision positioning of the lifting lug and improving positioning efficiency; the boom is equipped with a laser rangefinder sensor to provide real-time feedback of spacing data and close-loop control of errors; it has instantaneous locking upon power failure and real-time safety monitoring functions, significantly enhancing safety. It can promote the transformation of nuclear industry lifting operations towards intelligence and unmanned operation, providing standardized solutions for the transportation of similar radioactive materials; the technology can be extended to heavy-duty scenarios such as ports and wind power, contributing to the intelligent upgrading of industrial equipment. Attached Figure Description

[0017] Figure 1 This is a framework diagram of the self-decision transport container intelligent hoisting system in the embodiments of this application; Figure 2 This is a frame diagram of the mechanical structure in the embodiments of this application; Figure 3 This is a schematic diagram of the mechanical structure in the embodiments of this application; In the diagram: 1-Boom; 2-Harmonic reducer; 3-Servo motor; 4-Linear guide; 5-Ball screw; 6-Slewing bearing; 7-Pulley mechanism; 8-Laser rangefinder; 9-Global camera; 10-Industrial camera; 11-Image processing module; 12-Wide-angle lens; 13-Telecentric lens; 14-Servo driver; 15-Embedded controller; 16-Hydraulic brake; 17-Plunger pump; 18-Accumulator; 19-Supercapacitor; 20-Hydraulic cylinder; 21-Brake pad; 22-Mechanical gripper; 23-Data fusion module; 24-Software system. Detailed Implementation

[0018] The technical solution of this application will be described in detail below with reference to the accompanying drawings.

[0019] like Figure 1 As shown, the self-decision transport container intelligent hoisting system described in this application includes a mechanical structure, a vision system, an electrical control system, and a data fusion module 23. The mechanical structure includes two four-degree-of-freedom hinged linkage booms 1, each equipped with a safety locking device and a laser rangefinder sensor 8. The vision system includes a high-resolution global camera 9, an industrial camera 10, and an image processing module 11.

[0020] Preferably, such as Figure 2As shown, the mechanical structure also includes a harmonic reducer 2, a servo motor 3, a linear guide rail 4, a ball screw 5, a slewing support bearing 6, and a synchronous pulley mechanism 7. Both booms 1 move left and right (i.e., in the X / Y axis direction) via the linear guide rail 4 and the servo motor 3, adjusting the horizontal position of the booms. They move up and down (in the Z axis direction) via the ball screw 5 and the harmonic reducer 2. Both booms 1 have slewing support bearings 6 at their ends, allowing for 360° continuous rotation. The distance between the two booms 1 is adjusted by the servo motor 3 driving the synchronous pulley mechanism 7, with a laser rangefinder sensor 8 providing real-time feedback of the distance data and closed-loop error control.

[0021] Preferably, the global camera 9 is equipped with a wide-angle lens 12 to cover the panoramic monitoring of the hoisting area, and the industrial camera 10 is equipped with a telecentric lens 13 for capturing local details of the hoisting lug.

[0022] Preferably, the electronic control system includes a servo driver 14 and an embedded controller 15, equipped with a processor, supports EtherCAT bus communication, uses a multi-axis cooperative control algorithm to control the movement of the boom, and combines fuzzy PID to compensate for load disturbances.

[0023] The laser rangefinder 8 and the vision system input the collected data to the embedded controller 15. The embedded controller 15 sends the data to the servo driver 14. The servo driver 14 drives the mechanical structure and safety locking device according to the collected data.

[0024] Preferably, the safety locking device is composed of an electromagnetic-hydraulic composite braking system, which includes a hydraulic brake 16, a piston pump 17, an accumulator 18, a supercapacitor 19, a hydraulic cylinder 20, brake pads 21, and mechanical chucks 22.

[0025] Specifically, when the system needs to be locked (e.g., after the spreader grabs a container), under the braking of the hydraulic brake 16, the plunger pump 17 pumps hydraulic oil from the oil tank (not listed) into the accumulator 18 and the hydraulic cylinder 20. The hydraulic cylinder 20 pushes the brake pad 21 to contact the lifting lug or the container, generating a locking force.

[0026] Accumulator 18 is used to store hydraulic energy to charge supercapacitor 19. In the event of a power outage or other emergency, accumulator 18 can also release the stored hydraulic oil to maintain braking pressure for a period of time.

[0027] Supercapacitor 19 serves as an emergency power source, providing power to piston pump 17 or solenoid valve (not listed) when the main power supply to the system fails, ensuring that the hydraulic system can operate normally for a period of time (e.g., 15 seconds) to complete safety locking or emergency braking.

[0028] Mechanical gripper 22 serves as a redundant backup. In the event of a failure of the hydraulic braking system (e.g., hydraulic leakage or insufficient pressure), it is driven by spring or gravity to automatically clamp the lifting lug or container, preventing it from falling.

[0029] In the event of a power outage, the supercapacitor 19 immediately supplies power to the hydraulic system's control circuit and piston pump 17 via the hydraulic brake 16, enabling the hydraulic cylinder 20 to actuate the brake pads 21 for braking. Simultaneously, the accumulator 18 provides additional hydraulic oil to ensure sufficient hydraulic pressure during braking. If the hydraulic system completely fails, the brake pads 21 push the mechanical chuck 22, which then engages, achieving physical locking through spring preload, independent of the electronic control system, ensuring rigid fixation within half a second after a power outage. This safety locking device ensures safe locking under both normal and emergency conditions.

[0030] The working principle of the above-mentioned self-decision transport container intelligent hoisting system is as follows: (1) After the container is placed in the hoisting area, the container image is acquired by the global camera. The image processing module determines the relative position of the container with respect to the two booms based on the container image, and then feeds back the relative position to the electrical control system. The electrical control system controls the two booms to move as a whole based on the relative position so that the center of the two booms is facing the center of the container.

[0031] Specifically, the global camera coordinate system and the world coordinate system of the boom are calibrated in advance. When the container image occupies the center position of the image, the center of the boom is directly opposite the center position of the container.

[0032] (2) The industrial camera acquires images of the lifting lugs of the container. The image processing module identifies and locates the lifting lug position based on the lifting lug image, and then feeds back the lifting lug position to the electrical control system.

[0033] Specifically, the industrial camera acquires images of the container's lifting lugs, and the image processing module identifies and locates the lifting lug positions based on the images, including: 100: Use industrial cameras to collect images of the container's lifting lugs under different lighting, angle, and obstruction conditions.

[0034] 101: The image processing module identifies the position of the lugs in the lug image through the trained improved YOLOv8 network, and then locates the position of the lugs through the edge detection algorithm.

[0035] The improved YOLOv8 network backbone includes a lightweight MobileNetV3 module, which reduces the number of parameters by 40% and improves inference speed to 30ms / frame. The YOLOv8 backbone consists of multiple stages, each containing multiple C2f modules. Each C2f module is replaced with a lightweight MobileNetV3 module, adjusted according to parameters such as the number of input / output channels and stride. The lightweight MobileNetV3 module reduces computation by using depthwise separable convolutions and attention mechanisms. In the YOLOv8 network, replacing some modules in the backbone with lightweight MobileNetV3 modules reduces model parameters and computational cost while maintaining high feature extraction capabilities.

[0036] YOLOv8 networks typically include FPN and PANet structures for multi-scale feature fusion. This application adds a CBAM module after each feature fusion layer to enhance the propagation of important features. CBAM (Convolutional Block Attention Module) is an attention module that combines channel attention and spatial attention, helping the model better focus on important features and enhancing its feature fusion capabilities.

[0037] To address the imbalance between positive and negative samples (the ear of the object usually occupies only a small portion of the image area), Focal Loss (cross-entropy loss) is introduced for classification loss; EIoU (Enhanced IoU, EIoU = IoU - (overlap loss + center distance loss + aspect ratio loss)) is used for regression loss to better optimize bounding box regression, while taking into account center distance, overlap area and aspect ratio.

[0038] Furthermore, since ear detection requires high-precision localization, this application improves the YOLOv8 network loss by adding a loss function for sub-pixel-level center point localization, making it more sensitive to small errors and suitable for keypoint localization. Assuming the model needs to output: bounding box (bbox): 4 values ​​(x, y, w, h), class: 1 value (ear), keypoint: 2 values ​​(offset of the ear center point relative to the bounding box), this application designs a multi-task loss function: Total loss = regression loss (EIoU) + classification loss (Focal Loss) + keypoint loss (L1 loss or WingLoss).

[0039] Preferably, the training process of the improved YOLOv8 network includes: 200: Use an industrial camera to collect images of the container's lifting lugs under different lighting, angle, and occlusion conditions.

[0040] 201: Use a labeling tool (such as LabelImg) to label the lugs in the lug image and generate a YOLO format label file.

[0041] 202: A synthetic dataset of stilts generated using a Generative Adversarial Network (GAN).

[0042] The training data of the improved YOLOv8 network was augmented to enhance the feature extraction capability in occluded scenes. The synthetic dataset contains 100,000 images of hanging ears, covering working conditions such as fogging, reflection, and occlusion (simulating 30%-70% occlusion rate). A GAN generative adversarial network was used to generate virtual data under extreme lighting conditions.

[0043] 203: Input the YOLO format label file and synthetic dataset into the improved YOLOv8 network for training until the improved YOLOv8 network converges, and obtain the trained improved YOLOv8 network.

[0044] Preferably, the step of locating the position of the lug using an edge detection algorithm includes: 300: The pixel-level edges of the identified lugs are extracted based on the Canny operator, and the lugs are coarsely located based on the pixel-level edges; 301: The sub-pixel edge extraction operator based on local gray-scale area effect fitting extracts the sub-pixel level edges of the lug after coarse positioning, and accurately positions the lug based on the sub-pixel level edges.

[0045] Specifically, when the edge is a straight line, the local gray area is calculated using the equation of the straight line. When the edge is a curve, the local gray area is calculated using the equation of the curve.

[0046] Based on linear equations y=ax+b The obtained local gray area is represented as: ; ; in, M , L , R Both represent the area enclosed by the subpixel edge; S M , S L , S R All represent local grayscale areas; A , B This represents the grayscale values ​​on both sides of the edge; h The selected window width size.

[0047] Based on curve equations The obtained local gray area is represented as: ; ; 302: Obtain the subpixel-level edge of the lug by local grayscale area.

[0048] (3) After the lifting lug position is identified and located, the electric control system controls the boom to move downward according to the lifting lug position. After the laser rangefinder detects that the two booms have reached the preset height, it sends the preset height to the data fusion module. The data fusion module then synchronizes the preset height to the electric control system, and the electric control system controls the two booms to stop moving downward.

[0049] (4) The electrical control system controls the two booms to move from the middle to both ends according to the boom-load coupling dynamic model until they hook the lifting lugs. The lifting lugs are locked by the safety locking device, and then the two booms automatically rise to lift the container to the designated position or move it through the software system 24.

[0050] Preferably, the boom-load coupled dynamic model is described by the Lagrange-Euler equations, expressed as: ; in, These represent the position vector, velocity vector, and acceleration vector of the boom, respectively. This represents the control torque vector of the boom. Indicates the external disturbance torque. This represents the inertia matrix.

[0051] Preferably, the hoisting system further includes a software system, which is connected to both the data fusion module and the electrical control system. The data fusion module is connected to the vision system and receives visual data from the vision system and positional data from the laser rangefinder. The software system manually issues control commands to the electrical control system based on the visual data and positional data provided by the data fusion module. The electrical control system controls the mechanical structure and safety locking device based on the control commands.

[0052] The self-decision-making intelligent hoisting method for transport containers described in this application includes: 10: After the container is placed in the lifting area, the container image is acquired. Based on the container image, the relative position of the container with respect to the two booms is determined. Then, based on the relative position, the two booms are moved as a whole so that the center of the two booms is aligned with the center of the container. 11: Collect images of the container's lifting lugs and identify and locate the lifting lug positions based on the images; 12: Control the boom to move downward according to the position of the lifting lugs, and at the same time check whether the two booms have reached the preset height. If they have, control the two booms to stop moving downward; otherwise, continue to control the booms to move downward until the booms reach the preset height. 13: Based on the boom-load coupling dynamics model, control the two booms to move from the middle to both ends until they hook onto the lifting lugs. Then lock the lifting lugs and raise the two booms to lift the container to the designated position.

[0053] The above are exemplary embodiments of this application, and the scope of protection of this application is defined by the claims and their equivalents.

Claims

1. A self-decision-making intelligent lifting system for transport containers, comprising a mechanical structure, a vision system, an electrical control system, and a data fusion module, wherein the mechanical structure includes two booms, each equipped with a safety locking device and a laser rangefinder sensor; the vision system includes a global camera, an industrial camera, and an image processing module, characterized in that, include: After the container is placed in the lifting area, the global camera captures images of the container. The image processing module determines the relative position of the container with respect to the two booms based on the container images, and then feeds the relative position back to the electrical control system. The electrical control system controls the two booms to move as a whole based on the relative position so that the center of the two booms is directly opposite the center of the container. An industrial camera captures images of the lifting lugs of the container. The image processing module identifies and locates the position of the lifting lugs based on the images, and then feeds the position of the lifting lugs back to the electrical control system. The electronic control system controls the boom to move downwards based on the position of the lugs. After the laser rangefinder detects that the two booms have reached the preset height, it sends the preset height to the data fusion module. The data fusion module then synchronizes the preset height to the electronic control system, and the electronic control system controls the two booms to stop moving downwards. The electrical control system controls the two booms to move from the middle to both ends according to the boom-load coupling dynamic model until they hook onto the lifting lugs. The lifting lugs are then locked by the safety locking device, and then the two booms are raised to lift the container to the designated position. The industrial camera captures images of the container's lifting lugs, and the image processing module identifies and locates the lifting lug positions based on these images, including: Industrial cameras were used to capture images of the container's lifting lugs under different lighting, angle, and obstruction conditions. The image processing module identifies the location of the sling in the image using a trained improved YOLOv8 network, and then locates the sling position using an improved edge detection algorithm. The backbone of the improved YOLOv8 network includes a MobileNetV3 lightweight module, and the neck network includes a multi-scale feature fusion module and a CBAM module. The CBAM module combines channel attention and spatial attention mechanisms. The total loss of the improved YOLOv8 network includes regression loss, classification loss, and keypoint loss, expressed as: Total Loss = Regression Loss + Classification Loss + Keypoint Loss. The regression loss is EIoU Loss, the classification loss is Focal Loss, and the keypoint loss is L1 loss or Wing Loss.

2. The self-decision-making intelligent hoisting system for transport containers as described in claim 1, characterized in that, The mechanical structure also includes a harmonic reducer, a servo motor, a linear guide, a ball screw, a slewing support bearing, and a synchronous belt pulley mechanism. Both booms move left and right via the linear guide and servo motor, move up and down via the ball screw and harmonic reducer, and rotate 360° continuously via the slewing support bearing. The distance between the two booms is adjusted by the synchronous belt pulley mechanism driven by the servo motor.

3. The self-decision-making intelligent hoisting system for transport containers as described in claim 2, characterized in that, The safety locking device is composed of an electromagnetic-hydraulic composite braking system, which includes a hydraulic brake, a piston pump, an accumulator, a supercapacitor, a hydraulic cylinder, brake pads, and mechanical chucks. The hydraulic brake is connected to the piston pump, the piston pump is connected to the accumulator and the hydraulic cylinder, the accumulator is connected to the supercapacitor, the supercapacitor is connected to the hydraulic brake, the hydraulic cylinder is connected to the brake pads, and the brake pads are connected to the mechanical chucks.

4. The self-decision-making intelligent hoisting system for transport containers as described in claim 3, characterized in that, The global camera is equipped with a wide-angle lens, and the industrial camera is equipped with a telecentric lens.

5. The self-decision-making intelligent hoisting system for transport containers as described in claim 4, characterized in that, The electronic control system includes a servo driver and an embedded controller. The laser rangefinder and vision system input the collected data to the embedded controller, which then sends the data to the servo driver. The servo driver drives the mechanical structure and safety locking device based on the collected data.

6. The self-decision-making intelligent hoisting system for transport containers as described in claim 5, characterized in that, The boom-load coupled dynamic model is described by the Lagrange-Euler equations, expressed as: ； in, These represent the position vector, velocity vector, and acceleration vector of the boom, respectively. This represents the control torque vector of the boom. Indicates the external disturbance torque. This represents the inertia matrix.

7. The self-decision-making intelligent hoisting system for transport containers as described in claim 6, characterized in that, The training process of the improved YOLOv8 network includes: Industrial cameras were used to capture images of the container's lifting lugs under different lighting, angle, and obstruction conditions. Use the annotation tool to annotate the lugs in the lug image and generate a YOLO format label file; A synthetic dataset of stilts was generated using a Generative Adversarial Network (GAN). The YOLO format label file and synthetic dataset are input into the improved YOLOv8 network for training until the improved YOLOv8 network converges, resulting in the trained improved YOLOv8 network.

8. The intelligent lifting system for self-decision-making transport containers as described in claim 7, characterized in that, The method of locating the lug position by improving the edge detection algorithm includes: The pixel-level edges of the identified lugs are extracted based on the Canny operator, and the lugs are coarsely located based on the pixel-level edges. The sub-pixel edge extraction operator based on local gray area effect fitting extracts the sub-pixel level edges of the lug after coarse positioning, and accurately positions the lug based on the sub-pixel level edges. Among them, based on the equation of a straight line y=ax+b The obtained local gray area is represented as: ； ； in, M , L , R Both represent the area enclosed by the subpixel edge; S M , S L , S R All represent local grayscale areas; A , B This represents the grayscale values ​​on both sides of the edge; h The selected window width size; Based on curve equations The obtained local gray area is represented as: ； ； The subpixel-level edge of the lug is obtained by local grayscale area.

9. The self-decision-making intelligent lifting system for transport containers as described in claim 8, characterized in that, The hoisting system also includes a software system, which is connected to both the data fusion module and the electrical control system. The data fusion module is connected to the vision system and receives visual data from the vision system and positional data from the laser rangefinder. The software system manually sends control commands to the electrical control system based on the visual data and positional data provided by the data fusion module. The electrical control system then controls the mechanical structure and safety locking device based on these control commands.

10. A self-decision-making intelligent lifting method for transport containers, wherein the lifting method is used in the lifting system described in any one of claims 1-9, characterized in that, The hoisting method includes: After the container is placed in the lifting area, the container image is acquired. Based on the container image, the relative position of the container with respect to the two booms is determined. Then, based on the relative position, the two booms are moved as a whole so that the center of the two booms is directly opposite the center of the container. The image of the container's lifting lugs is acquired, and the position of the lifting lugs is identified and located based on the image. The boom is controlled to move downward according to the position of the lifting lugs. At the same time, it is detected whether the two booms have reached the preset height. If they have, the two booms are controlled to stop moving downward. Otherwise, the booms are controlled to move downward until they reach the preset height. Based on the boom-load coupling dynamics model, the two booms are controlled to move from the middle to both ends until they hook onto the lifting lugs. Then the lifting lugs are locked, and the two booms are raised to lift the container to the designated position.

Images