A method for post-processing of a lashing wire end by a lashing robot and a lashing robot system
By using image acquisition, recognition, and tapping processing through a binding robot system, the problem of exposed binding wire ends has been solved, achieving automation and quality reliability in the binding process, and improving the safety and durability of building structures.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- HUNAN DILUER TECHNOLOGY CO LTD
- Filing Date
- 2026-03-02
- Publication Date
- 2026-06-02
AI Technical Summary
Existing binding robots cannot effectively handle the cut ends of the binding wires, resulting in exposed ends that threaten the safety and durability of building structures, and they cannot automate the binding process.
By using a binding robot system, which employs a robotic arm, vision module, and tapping execution module, image acquisition, end recognition, tapping decision-making, and planning are performed to achieve automated processing of the binding wire ends.
It has achieved automated processing of the wire binding ends, improved construction quality and efficiency, eliminated potential engineering quality hazards, and supported the full automation of the binding process.
Smart Images
Figure CN122129128A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of intelligent manufacturing technology, and in particular to a method and system for post-processing the ends of binding wires using a binding robot. Background Technology
[0002] According to the relevant provisions on rebar tying in the "Code for Acceptance of Construction Quality of Concrete Structures" (GB 50204), the ends of the cut tie wires after tying must be bent and pressed into the rebar cage, and it is absolutely forbidden for them to extend beyond the concrete cover. However, existing tying robots only perform the "tightening" and "cutting" actions, lacking effective control over the state of the cut ends. These exposed ends can lead to: ① It forms a corrosion pathway deep into the concrete: accelerating the corrosion of internal steel bars, seriously threatening the safety and durability of the building structure; ② Reduce the thickness of the protective layer: This reduces the fire resistance and structural strength of the components.
[0003] Currently, these exposed ends are inspected manually by workers who use hammers or special tools to check and tap each exposed end along the rebar cage. However, this requires manual intervention and cannot automate the entire binding process. Summary of the Invention
[0004] To address the technical problems in the prior art, this invention provides a method and a binding robot system for post-processing the ends of binding wires using a binding robot, thereby facilitating the automation of the entire binding process.
[0005] To achieve the above-mentioned objectives, the present invention provides the following technical solution: A method for post-processing the ends of binding wires using a binding robot, the binding robot comprising a robotic arm and a striking execution module, the striking execution module being fixed to the robotic arm and equipped with a striking head for striking the ends; the method includes: Image acquisition of the binding area; Identify and locate the coordinates of the end point; Tapping decision: Determine whether the end needs to be tapped. If not, end; if yes, proceed to the next step. Planning: Plan the pose of the robotic arm and the motion parameters of the striking head during the striking action; Execution of the tapping action: The robotic arm moves to the tapping position according to the planned pose, and the tapping execution module performs the tapping action according to the planned motion parameters.
[0006] Preferably, after performing the tapping step, the method further includes: Quality verification: Determine whether the end meets the standard after tapping. If yes, the process ends; otherwise, return to the tapping decision step.
[0007] Preferably, the step of identifying and locating the end coordinates includes: using an improved Faster R-CNN wire end recognition model based on an attention-based mechanism to segment and identify the wire end from the rebar background, and calculating the precise position and vector direction of the wire end in three-dimensional space; the improved Faster R-CNN wire end recognition model is as follows: using ResNet as the basic backbone network for Faster R-CNN feature extraction, and integrating the SENet module into some convolutional layers of ResNet; the integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling operation, then the two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map to achieve feature recalibration and optimization.
[0008] Preferably, the striking decision step includes: comparing the recognition result with the built-in specification library to determine whether the end is too long or has an unfavorable orientation, which may pose a risk of intruding into the concrete, and deciding whether to strike; in the planning step, the pose of the robotic arm is planned by enabling the striking head to strike the root of the end from the most effective direction, and the motion parameters of the striking head are calculated based on the size of the end and the angle to bend.
[0009] A tying robot system, comprising: robotic arm; The vision module is used to acquire images of the binding area; The striking execution module is equipped with a striking head for striking the end, which is fixed on the robotic arm so that it can be moved by the robotic arm; The central processing module connects the vision module, the robotic arm, and the striking execution module. It is used to identify and locate the coordinates of the end head based on the image acquired by the vision module, and then make a striking decision. If a striking is required, it plans the pose of the robotic arm and the motion parameters of the striking head when striking, and sends the planning results to the robotic arm and the striking execution module. The robotic arm moves to the striking position according to the planned pose, and the striking execution module performs the striking action according to the planned motion parameters.
[0010] Preferably, the central processing module is further configured to perform quality verification after tapping, determine whether the end meets the standard after tapping, and if not, make another tapping decision.
[0011] Preferably, the identification and localization of the wire end coordinates is as follows: using an improved Faster R-CNN wire end recognition model based on an attention-based mechanism, the wire end is segmented and identified from the rebar background, and the precise position and vector direction of the wire end in three-dimensional space are calculated; the improved Faster R-CNN wire end recognition model is as follows: using ResNet as the basic backbone network for Faster R-CNN feature extraction, and integrating the SENet module into some convolutional layers of ResNet; the integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling operation, then the two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map to achieve feature recalibration and optimization.
[0012] Preferably, the striking decision is as follows: the identification result is compared with the built-in specification library to determine whether there is a risk of the end penetrating the concrete, and then a decision is made on whether to strike; the planning of the robot arm's pose and striking energy during striking is as follows: the robot arm's pose is planned by enabling the striking head to strike the root of the end from the most effective direction, and the motion parameters of the striking head are calculated based on the size of the end and the angle to bend.
[0013] Preferably, the striking execution module further includes a miniature high-pressure pneumatic system and a cylinder-piston mechanism. The miniature high-pressure pneumatic system serves as a power source, providing pneumatic energy. The cylinder-piston mechanism serves as a motion conversion mechanism, converting the pneumatic energy into the required linear impact, so that the compressed air output by the miniature high-pressure pneumatic system enters the rear chamber of the cylinder, pushing the piston to accelerate along the cylinder. The striking head is installed at the front end of the piston rod.
[0014] Preferably, the striking execution module further includes a servo motor and a cam-follower mechanism, wherein the servo motor serves as a drive source; the cam-follower mechanism serves as a motion conversion mechanism, converting the rotational motion of the servo motor into the required linear impact; and the striking head is located at the front end of the follower.
[0015] Compared with existing technologies, the method and system for post-processing of tie wire ends using a tying robot provided by this invention, by setting a tapping execution module on the robotic arm, performs steps such as image acquisition, identification and positioning of end coordinates, tapping decision, planning, and execution of tapping in the tying area. This enables the tying robot to automatically and "intelligently" check and remove each unqualified tie wire end after completing the tying, thereby facilitating the realization of full-process, high-quality automated tying and significantly improving construction efficiency and quality. Attached Figure Description
[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 This is a flowchart illustrating a method for post-processing the ends of binding wires using a binding robot, according to an embodiment of the present invention. Figure 2 for Figure 1 A schematic diagram of the Faster R-CNN network structure in the method shown; Figure 3 for Figure 1 The diagram shows the ResNet50 network configuration and residual structure in the method shown, where (a) is a network configuration diagram and (b) is a residual structure diagram. Figure 4 for Figure 1 A schematic diagram of the SENet network structure in the method shown; Figure 5 for Figure 1 A schematic diagram of the network structure of the improved Faster R-CNN wire end recognition model in the method shown; Figure 6 This is a schematic diagram of the structure of a tying robot system provided in one embodiment of the present invention; Figure 7 for Figure 6 The diagram shows the structure of the striking execution module in the tying robot system. Figure 8 for Figure 6 The diagram shows another implementation of the striking execution module in the tying robot system.
[0018] In the diagram, 1 is the robotic arm; 2 is the vision module; 3 is the striking execution module; 4 is the tying gun; 31 is the first channel steel connector; 32 is the air inlet; 33 is the cylinder-piston mechanism; 34 is the first striking head; 51 is the second channel steel connector; 52 is the servo motor; 53 is the cam-follower mechanism; 54 is the striking rod; and 55 is the second striking head. Detailed Implementation
[0019] To enable those skilled in the art to better understand the technical solutions in this application, the technical solutions in the embodiments of this application will be clearly and completely described below. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0020] It should be noted that when a component is referred to as being "fixed to" or "set on" another component, it can be directly on or indirectly set on the other component; when a component is referred to as being "connected to" another component, it can be directly connected to or indirectly connected to the other component.
[0021] It should be noted that the structures, proportions, sizes, etc., shown in the accompanying drawings of this specification are only for the purpose of assisting those skilled in the art in understanding and reading the content disclosed in the specification, and are not intended to limit the conditions under which this application can be implemented. Therefore, they have no substantial technical significance. Any modifications to the structure, changes in the proportions, or adjustments to the size should still fall within the scope of the technical content disclosed in this application, provided that they do not affect the effects and purposes that this application can produce.
[0022] like Figures 1 to 5 As shown, this embodiment of the invention provides a method for post-processing the ends of binding wires using a binding robot, for processing the ends left after the binding wires have been cut. The binding robot includes a robotic arm and a striking execution module, the striking execution module being fixed to the robotic arm and equipped with a striking head for striking the ends.
[0023] The method includes: Image acquisition of the binding area; Identify and locate the coordinates of the end point; Tapping decision: Determine whether the end needs to be tapped. If not, end; if yes, proceed to the next step. Planning: Plan the pose (position and attitude) of the robotic arm and the motion parameters of the striking head during the striking process; Execution of the tapping action: The robotic arm moves to the tapping position according to the planned pose, and the tapping execution module performs the tapping action according to the planned motion parameters.
[0024] In this embodiment, the step of identifying and locating the end coordinates includes: using an improved Faster R-CNN wire end recognition model based on an attention-based mechanism to segment and identify the wire end from the rebar background, and calculating the precise position and vector direction of the wire end in three-dimensional space; the improved Faster R-CNN wire end recognition model is as follows: using ResNet as the basic backbone network for Faster R-CNN feature extraction, and integrating the SENet module into some convolutional layers of ResNet; the integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling operation, then the two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map to achieve feature recalibration and optimization.
[0025] The principle of the improved Faster R-CNN splinter detection model that incorporates an attention mechanism includes: The residual network ResNet50 (or other ResNet networks) is used as the basic backbone network for feature extraction in Faster R-CNN, and a Squeeze-and-Excitation Network (SENet) is integrated to enhance the model's attention mechanism, such as... Figure 2 As shown.
[0026] ResNet addresses the degradation problem in deep learning network training by introducing a residual learning framework, enabling the training of deeper networks to achieve better-performing models. Figure 3 (a) shows the specific network configuration of ResNet50, while Figure 3 (b) provides a detailed description of typical residual structures.
[0027] Unlike traditional methods that directly learn the original mapping relationship, ResNet trains network layers to learn the residual function F(x)=H(x). The output of this function is then added back to the input x, forming F(x) + x. This design simplifies the optimization objective of the network layers, making the optimization process more efficient.
[0028] The primary purpose of introducing residual structures into ResNet is to address the vanishing and exploding gradient problems that arise as network depth increases. While the overall trend in network layers is towards increasing depth, simply stacking more layers—the traditional stacking structure—leads to gradient problems that hinder network convergence. ResNet's residual learning approach offers a more fundamental solution. ResNet addresses the degradation problem of deep neural networks through the idea of residual learning: by adding a few identity mapping layers to a shallow network whose accuracy has plateaued, it achieves the goal of increasing network depth without increasing error.
[0029] However, SENet (SENet's network structure is as follows) Figure 4 (As shown) SENet focuses on the relationships between feature channels. It optimizes network functionality by explicitly modeling the interdependencies between these channels. Specifically, during the feature learning process, SENet recalibrates feature channels by evaluating the contribution of each channel to the overall features, and then enhances features beneficial to the current task and suppresses features of lower importance based on these contribution values.
[0030] This patent integrates the SENet module into ResNet50. The SENet module is designed to preprocess the input of each convolutional layer and apply the preprocessing to the output feature map after the convolution operation. This strategy introduces a channel-level attention mechanism to dynamically adjust the importance of input feature channels and optimize the feature processing flow of the entire model.
[0031] The SENet module first extracts global information from the feature maps of each channel through global average pooling. Then, two fully connected layers perform feature compression and expansion, respectively, and generate weights for each channel using the ReLU activation function. These weights are used to weight each channel of the original feature maps, achieving feature recalibration and optimization, and enhancing the network's efficiency in capturing key information.
[0032] Figure 5 This embodiment demonstrates the improved Faster R-CNN model architecture, which integrates ResNet50 and SENet modules. As shown in the figure, the SENet module is strategically integrated into the first three bottleneck convolutional layers of ResNet50. Channel-level attention mechanisms are introduced at these critical stages, greatly optimizing the model's ability to process and recognize ligation point wire features.
[0033] In this embodiment, the knocking decision step includes: comparing the identification result with the built-in specification library to determine whether the end is too long or has an unfavorable orientation, which may pose a risk of intruding into the concrete, and then deciding whether to knock it.
[0034] In this embodiment, in the planning step, the pose of the robotic arm is planned by making the striking head strike the root of the end from the most effective direction (usually obliquely pointing into the inside of the steel frame). The motion parameters of the striking head are calculated based on the size of the end and the angle to bend, and the required striking energy (if the striking execution module is pneumatic) or servo motor parameters (if the striking execution module is electric) are further obtained.
[0035] In this embodiment, after the tapping step, the process further includes: quality verification: determining whether the tapping of the end meets the standard. If yes, the process ends; otherwise, it returns to the tapping decision step. This re-inspection step ensures the final quality, forming a complete closed loop of "perception-decision-execution-verification," thus guaranteeing the reliability of the system.
[0036] like Figure 6 As shown, this embodiment also provides a tying robot system, including: Robotic arm 1; Vision module 2 is used to acquire images of the binding area (using a high-resolution industrial camera fixed to the end of robotic arm 1 to take pictures and identify nodes before and after binding). The striking execution module 3 is equipped with a striking head for striking the end, which is fixed on the robotic arm 1 so that it can be moved by the robotic arm 1. The central processing module connects to vision module 2, robotic arm 1, and striking execution module 3. It is used to identify and locate the end coordinates based on the image acquired by vision module 2, and then make a striking decision. If striking is required, it plans the pose of robotic arm 1 and the motion parameters of the striking head during striking, and sends the planning results to robotic arm 1 and striking execution module 3. Robotic arm 1 moves to the striking position according to the planned pose, and striking execution module 3 performs the striking action according to the planned motion parameters.
[0037] In this embodiment, the central processing module is also used to perform quality verification after tapping, to determine whether the end meets the standard after tapping, and if not, to make a tapping decision.
[0038] The identification and localization of the wire tie end coordinates involves using an improved Faster R-CNN wire tie end recognition model based on an attention-based mechanism to segment and identify the wire tie end from the rebar background, and calculating the precise position and vector direction of the wire tie end in three-dimensional space. The improved Faster R-CNN wire tie end recognition model uses a ResNet residual network as the basic backbone network for Faster R-CNN feature extraction, and integrates the SENet module into some convolutional layers of ResNet. The integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling. Then, two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map, achieving feature recalibration and optimization.
[0039] The knocking decision is as follows: the identification result is compared with the built-in specification library to determine whether there is a risk of the end penetrating the concrete, and then decide whether to knock it.
[0040] The pose and striking energy of the robotic arm 1 during the striking are planned as follows: the pose of the robotic arm 1 is planned by making the striking head strike the root of the end in the most effective direction, and the motion parameters of the striking head are calculated based on the size of the end and the angle to bend.
[0041] like Figure 7 As shown, in this embodiment, the striking execution module 3 is pneumatic and is firmly fixed to the side of the body of the binding gun 4 by a rigid metal mounting bracket (first channel steel connector 31), ensuring that its working area is completely isolated from the wire twisting / cutting mechanism of the binding gun 4 in space and does not interfere with each other. The striking execution module 3 includes a miniature high-pressure pneumatic system, a cylinder-piston mechanism 33, a first linear bearing, and a first striking head 34.
[0042] The miniature high-pressure pneumatic system serves as a power source, providing pneumatic energy. It features rapid response and precise pressure control, capable of receiving electrical signal commands from the controller and accurately adjusting the pressure and flow rate of the output gas, thereby providing controllable input for the striking force and stroke.
[0043] The cylinder-piston mechanism 33, acting as a motion conversion mechanism, converts pneumatic energy into the required linear impact. Compressed air output from the miniature high-pressure pneumatic system enters the rear chamber of the cylinder through the inlet 32, pushing the piston to accelerate along the cylinder barrel. A high-speed solenoid valve is installed between the miniature high-pressure pneumatic system and the rear chamber of the cylinder. During a strike, compressed air enters the rear chamber through the high-speed solenoid valve. When the solenoid valve reverses, the force of the return spring or the reverse airflow drives the piston to retract rapidly, preparing for the next strike. The cylinder stroke is optimized to achieve high acceleration and impact force at the moment of impact.
[0044] The first linear bearing, serving as a guide and support structure, connects to the piston rod, restricting the piston's movement to only the axial direction, thus ensuring the precision and reliability of the striking motion. In this embodiment, the piston rod is housed within a high-precision linear bearing integrated into a modular alloy housing, providing robust support for the entire piston assembly. This strictly limits its movement to the axial direction, preventing any lateral swaying or deflection, ensuring consistent positioning accuracy for each strike, and significantly improving the mechanism's lifespan.
[0045] The first striking head 34 is mounted at the front end of the piston rod. This first striking head 34 features a circular, flat, block-like design and is precision-ground from tungsten carbide, exhibiting extremely high hardness and wear resistance. This unique design aims to create optimized surface contact with the cylindrical binding wire, thereby more effectively converting the impact force into a torque that causes the binding wire to plastically bend upon impact.
[0046] like Figure 8 As shown, the striking actuator module 3 can also be electrically powered, designed as a highly integrated and compact electromechanical actuator. Its core function is to convert the electrical control signal into a short-stroke, high-speed, and force-controllable precise linear impact motion. The striking actuator module 3 is securely fixed to the side or rear of the binding gun 4 via a rigid metal mounting bracket (second channel steel connector 51), ensuring that its working area is completely spatially isolated from the wire twisting / cutting mechanism of the binding gun 4, preventing interference. The striking actuator module 3 includes a servo motor 52, a cam-follower mechanism 53, a second linear bearing or bushing, and a second striking head 55.
[0047] The servo motor 52 serves as the drive source. It can be a miniature high-response servo motor 52 with high torque density and precise position control capability. It can receive pulse commands from the controller and accurately output specific rotation angles, speeds and accelerations, thereby providing controllable input for the striking force and stroke.
[0048] The cam-follower mechanism 53 acts as a motion conversion mechanism, converting the rotational motion of the servo motor 52 into the required linear impact. The output shaft of the servo motor 52 is rigidly connected to a custom-profiled cam or eccentric wheel. A high-strength striking rod 54 acts as a follower, its tail always pressed against the profile of the cam by a preloaded spring. When the cam is driven by the motor to rotate to the lift phase, its profile pushes the striking rod 54 to overcome the spring pressure and accelerate its extension along its axis; when the cam enters the return phase, the return spring's force drives the striking rod 54 to retract rapidly, preparing for the next strike. The cam profile is optimized to achieve high acceleration and impact force of the striking head instantaneously.
[0049] A linear bearing or bushing serves as a guide and support structure, connecting the striking rod 54 and restricting its movement to the axial direction only, ensuring the precision and reliability of the striking motion. The striking rod 54 is housed within a high-precision second linear bearing or hardened steel bushing. This second linear bearing or hardened steel bushing is press-fitted into a modular aluminum alloy housing, providing robust support for the entire striking rod 54, strictly limiting its movement to the axial direction only, eliminating any lateral sway or deflection, ensuring repeatable positioning accuracy for each strike, and significantly improving the mechanism's lifespan.
[0050] The second striking head 55, located at the front end of the striking rod 54, is made of high-strength hardened tool steel through precision grinding, possessing extremely high hardness and wear resistance. Its head is not a sharp design, but rather machined into a slightly curved surface or a small-angle wedge-shaped plane. This unique design aims to form optimized surface or line contact with the cylindrical binding wire, thereby more effectively converting the impact force into a torque that causes the binding wire to plastically bend upon impact, rather than cutting or slipping it.
[0051] Compared with existing technologies, the method and system for post-processing the ends of the binding wire using a binding robot in this embodiment have the following advantages: ① Quality reliability: The treatment of wire binding ends has been changed from "probability of passing" to "guaranteed passing", which fundamentally eliminates an important hidden danger to engineering quality and improves the service life of building structures; ② Fully automated process: It has achieved complete unmanned operation of the binding work, which is a key step towards "lights-out construction site"; ③ High adaptability: The vision module can adapt to various complex nodes and random end orientations, and the intelligent algorithm can make optimal decisions; ④ High efficiency: The entire post-processing is completed within seconds, matching the binding rhythm, and does not significantly affect the overall work efficiency; ⑤ Easy to modify: The tapping execution module can be integrated into the existing tying robot as an additional function, making it easy to promote.
[0052] The above description is merely an embodiment of the present invention. It should be noted that those skilled in the art can make improvements without departing from the inventive concept of the present invention, but these improvements all fall within the protection scope of the present invention.
Claims
1. A method for post-processing the ends of binding wires using a binding robot, characterized in that, The binding robot includes a robotic arm and a striking execution module, the striking execution module being fixed to the robotic arm and equipped with a striking head for striking the end; the method includes: Image acquisition of the binding area; Identify and locate the coordinates of the end point; Tapping decision: Determine whether the end needs to be tapped. If not, end; if yes, proceed to the next step. Planning: Plan the pose of the robotic arm and the motion parameters of the striking head during the striking action; Execution of the tapping action: The robotic arm moves to the tapping position according to the planned pose, and the tapping execution module performs the tapping action according to the planned motion parameters.
2. The method for post-processing the end of the binding wire using a binding robot according to claim 1, characterized in that, After performing the tapping step, the following is also included: Quality verification: Determine whether the end meets the standard after tapping. If yes, the process ends; otherwise, return to the tapping decision step.
3. The method for post-processing the end of the binding wire using a binding robot according to claim 1, characterized in that, The step of identifying and locating the end coordinates includes: using an improved Faster R-CNN wire end recognition model based on an attention-based mechanism to segment and identify the wire end from the rebar background, and calculating the precise position and vector direction of the wire end in three-dimensional space; the improved Faster R-CNN wire end recognition model is as follows: using ResNet as the basic backbone network for Faster R-CNN feature extraction, and integrating the SENet module into some convolutional layers of ResNet; the integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling operation, then the two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map to achieve feature recalibration and optimization.
4. The method for post-processing the end of the binding wire using a binding robot according to claim 1, characterized in that, The striking decision-making step includes: comparing the recognition result with the built-in specification library to determine whether the end is too long or has an unfavorable orientation, which may pose a risk of intruding into the concrete, and deciding whether to strike it; in the planning step, the pose of the robotic arm is planned by enabling the striking head to strike the root of the end from the most effective direction, and the motion parameters of the striking head are calculated based on the size of the end and the angle to bend.
5. A tying robot system, characterized in that, include: robotic arm; The vision module is used to acquire images of the binding area; The striking execution module is equipped with a striking head for striking the end, which is fixed on the robotic arm so that it can be moved by the robotic arm; The central processing module connects the vision module, the robotic arm, and the striking execution module. It is used to identify and locate the coordinates of the end head based on the image acquired by the vision module, and then make a striking decision. If a striking is required, it plans the pose of the robotic arm and the motion parameters of the striking head when striking, and sends the planning results to the robotic arm and the striking execution module. The robotic arm moves to the striking position according to the planned pose, and the striking execution module performs the striking action according to the planned motion parameters.
6. The tying robot system according to claim 5, characterized in that, The central processing module is also used to perform quality verification after tapping, to determine whether the end meets the standard after tapping, and if not, to make another tapping decision.
7. The tying robot system according to claim 5, characterized in that, The identification and localization of the wire tie end coordinates is achieved by using an improved Faster R-CNN wire tie end recognition model based on an attention-based mechanism to segment and identify the wire tie end from the rebar background, and to calculate the precise position and vector direction of the wire tie end in three-dimensional space. The improved Faster R-CNN wire tie end recognition model uses ResNet as the basic backbone network for Faster R-CNN feature extraction, and integrates the SENet module into some convolutional layers of ResNet. The integration method is as follows: the SENet module first extracts global information from the feature map of each channel through global average pooling. Then, two fully connected layers perform feature compression and expansion respectively, and generate weights for each channel through the ReLU activation function. These weights are used to weight each channel of the original feature map to achieve feature recalibration and optimization.
8. The tying robot system according to claim 5, characterized in that, The striking decision is as follows: the recognition result is compared with the built-in specification library to determine whether there is a risk of the end penetrating the concrete, and then it is determined whether to strike. The planning of the robot arm's pose and striking energy during the striking is as follows: the robot arm's pose is planned by making the striking head hit the root of the end from the most effective direction, and the motion parameters of the striking head are calculated based on the size of the end and the angle to bend.
9. The tying robot system according to claim 5, characterized in that, The striking execution module also includes a miniature high-pressure pneumatic system and a cylinder-piston mechanism. The miniature high-pressure pneumatic system serves as a power source, providing pneumatic energy. The cylinder-piston mechanism serves as a motion conversion mechanism, converting the pneumatic energy into the required linear impact, allowing the compressed air output by the miniature high-pressure pneumatic system to enter the rear chamber of the cylinder, pushing the piston to accelerate along the cylinder. The striking head is mounted on the front end of the piston rod.
10. The tying robot system according to claim 5, characterized in that, The striking execution module also includes a servo motor and a cam-follower mechanism. The servo motor serves as the drive source; the cam-follower mechanism serves as a motion conversion mechanism, converting the rotational motion of the servo motor into the required linear impact; and the striking head is located at the front end of the follower.