A reinforcing bar binding method, device, equipment and storage medium

Abstract

The application discloses a steel bar binding method, device, equipment and storage medium, and relates to the field of robots, and is applied to a steel bar binding robot and comprises the following steps: acquiring original point cloud data of a target binding area by using a target camera carried by the steel bar binding robot, and preprocessing the original point cloud data to obtain corresponding processed point cloud data; completing the processed point cloud data based on a preset cylindrical geometric constraint by using a target point cloud completion network based on deep learning to obtain corresponding completed point cloud data of the target binding area; aligning the completed point cloud data with a preset steel bar point cloud template to obtain corresponding aligned point cloud data of the target binding area, and planning a motion track of the steel bar binding robot based on the aligned point cloud data to obtain a target motion track, so that the steel bar binding robot is controlled to bind the steel bars to be bound in the target binding area based on the target motion track. The application improves the steel bar binding precision of the steel bar binding robot.

Description

Technical Field

[0001] This invention relates to the field of robotics, and in particular to a method, apparatus, equipment, and storage medium for tying reinforcing bars. Background Technology

[0002] Currently, traditional solutions for attitude estimation in automated rebar tying can be divided into three categories: The first category uses fixed fixtures and preset paths for mechanical positioning, which relies on high-precision assembly and rigid support, and has extremely poor adaptability to on-site deviations and workpiece deformation; the second category is based on two-dimensional vision to identify the position of intersections, which lacks three-dimensional attitude information and cannot cope with complex spatial arrangements; the third category uses laser scanning or traditional point cloud registration, which can obtain three-dimensional information but is costly, and the matching accuracy drops significantly when dealing with situations such as rebar surfaces without texture or severe self-occlusion, resulting in insufficient robustness and difficulty in meeting the comprehensive requirements of efficiency, cost and reliability on construction sites. Summary of the Invention

[0003] In view of this, the purpose of this invention is to provide a method, apparatus, device, and storage medium for tying reinforcing bars, which can improve the accuracy of reinforcing bar tying. The specific solution is as follows:

[0004] In a first aspect, this application discloses a rebar tying method applied to a rebar tying robot, comprising:

[0005] The original point cloud data of the target binding area is acquired using the target camera mounted on the rebar binding robot, and the original point cloud data is preprocessed to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be bound in the target binding area.

[0006] The target point cloud completion network is used to complete the processed point cloud data based on preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning.

[0007] The completed point cloud data is aligned with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area. Based on the aligned point cloud data, the motion trajectory of the rebar binding robot is planned to obtain the target motion trajectory, so as to control the rebar binding robot to bind the rebar to be bound in the target binding area based on the target motion trajectory.

[0008] Optionally, the preprocessing of the original point cloud data to obtain the corresponding processed point cloud data includes:

[0009] The original point cloud data is downsampled using a preset voxel grid filtering method to obtain the corresponding filtered point cloud data.

[0010] Outlier point cloud data in the filtered point cloud data are identified and removed to obtain the corresponding processed point cloud data.

[0011] Optionally, before using the target point cloud completion network to complete the processed point cloud data based on preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target binding region, the method further includes:

[0012] The target point cloud completion network is obtained by training the initial point cloud completion network based on the encoder-decoder architecture using the target loss function.

[0013] The target loss function includes the target chamfer distance loss function, the cylindrical surface normal consistency loss function, and the curvature consistency loss function.

[0014] Optionally, the encoder of the target point cloud completion network is used to determine the feature representation of the input point cloud data using a multilayer perceptron structure, and the decoder of the target point cloud completion network is used to reconstruct the complete point cloud data corresponding to the input point cloud data based on the feature representation using deconvolution and fully connected layers.

[0015] Optionally, before aligning the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area, the method further includes:

[0016] Obtain a preset rebar point cloud template and save the preset rebar point cloud template locally on the rebar binding robot.

[0017] Optionally, aligning the completed point cloud data with a preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area includes:

[0018] Based on the completed point cloud data and the region type of the target binding area, the target point cloud template is determined from all the preset rebar point cloud templates stored locally by the rebar binding robot.

[0019] Based on a preset principal component analysis method, the completed point cloud data and the target point cloud template are initially aligned to obtain the corresponding preliminary aligned point cloud.

[0020] Using the target iterative nearest point algorithm and target displacement correction strategy, the completed point cloud data and the target point cloud template are registered based on the preliminary aligned point cloud to obtain the aligned point cloud data corresponding to the target binding area.

[0021] Optionally, the step of planning the motion trajectory of the rebar tying robot based on the aligned point cloud data to obtain the target motion trajectory includes:

[0022] The first binding posture of the rebar binding robot is determined, and the second binding posture of the rebar binding robot is determined based on the aligned point cloud data.

[0023] Based on the first binding posture and the second binding posture, the motion trajectory of the target robotic arm of the rebar binding robot is determined, and based on the aligned point cloud data and the robotic arm motion trajectory, the motion trajectory of the rebar binding robot is planned to obtain the target motion trajectory.

[0024] Secondly, this application discloses a rebar tying device for use in a rebar tying robot, comprising:

[0025] The point cloud data acquisition module is used to acquire the original point cloud data of the target binding area using the target camera mounted on the rebar binding robot, and to preprocess the original point cloud data to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be bound in the target binding area.

[0026] The point cloud data completion module is used to complete the processed point cloud data based on preset cylindrical geometric constraints using a target point cloud completion network to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning.

[0027] The rebar tying module is used to align the completed point cloud data with a preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target tying area, and to plan the motion trajectory of the rebar tying robot based on the aligned point cloud data to obtain the target motion trajectory, so as to control the rebar tying robot to tie the rebar to be tied in the target tying area based on the target motion trajectory.

[0028] Thirdly, this application discloses an electronic device, including:

[0029] Memory, used to store computer programs;

[0030] A processor is used to execute the computer program to implement the aforementioned rebar tying method.

[0031] Fourthly, this application discloses a computer-readable storage medium for storing a computer program, wherein the computer program, when executed by a processor, implements the aforementioned rebar tying method.

[0032] In this application, when the rebar tying robot is tying rebars, it uses a target camera mounted on the robot to acquire the original point cloud data of the target tying area, and preprocesses the original point cloud data to obtain corresponding processed point cloud data. The point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be tied in the target tying area. The processed point cloud data is completed using a target point cloud completion network based on preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target tying area. The target point cloud completion network is a point cloud completion network based on deep learning. The completed point cloud data is aligned with a preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target tying area. The motion trajectory of the rebar tying robot is planned based on the aligned point cloud data to obtain a target motion trajectory, so as to control the rebar tying robot to tie the rebars to be tied in the target tying area based on the target motion trajectory. As can be seen, after acquiring the original point cloud data of the target binding area, the rebar tying robot in this application can perform a series of preprocessing operations to obtain the corresponding processed point cloud data, providing high-quality input data for subsequent point cloud completion and registration. Then, using a target point cloud completion network designed for the cylindrical features of the rebar, the processed point cloud data is completed based on the cylindrical geometric constraints, thereby ensuring that the completed point cloud maintains the true geometric shape. Finally, the completed point cloud data is aligned with a preset rebar point cloud template to achieve adaptive registration of the completed point cloud data. Based on the aligned point cloud data, the motion trajectory of the rebar tying robot is planned to obtain the target motion trajectory, which controls the rebar tying robot to bind the rebars to be bound in the target binding area. This effectively solves the problem of point cloud missing caused by rebar self-occlusion and improves registration accuracy and robustness. Attached Figure Description

[0033] To more clearly illustrate the technical solutions in the embodiments of the present invention 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 embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0034] Figure 1 This is a flowchart of a rebar tying method disclosed in this application;

[0035] Figure 2 This is a schematic diagram of a rebar tying device disclosed in this application;

[0036] Figure 3 This is a structural diagram of an electronic device disclosed in this application. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0038] Current traditional solutions to the attitude estimation problem in automated rebar tying mainly fall into three categories: The first category uses fixed clamps and preset paths for mechanical positioning, relying on high-precision assembly and rigid support, which has extremely poor adaptability to on-site deviations and workpiece deformation; the second category is based on two-dimensional vision to identify the position of intersections, lacking three-dimensional attitude information and unable to cope with complex spatial arrangements; the third category uses laser scanning or traditional point cloud registration, which can obtain three-dimensional information but is costly, and the matching accuracy drops significantly when dealing with situations such as rebar surfaces without texture or severe self-occlusion, resulting in insufficient robustness and difficulty in meeting the comprehensive requirements of efficiency, cost, and reliability on construction sites. To solve the above technical problems, this application discloses a rebar tying method that can improve the accuracy of rebar tying.

[0039] See Figure 1 As shown, this invention discloses a rebar tying method applied to a rebar tying robot, comprising:

[0040] Step S11: Use the target camera mounted on the rebar tying robot to acquire the original point cloud data of the target tying area, and preprocess the original point cloud data to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be tied in the target tying area.

[0041] In this embodiment, the target camera (such as an industrial 3D camera) mounted on the rebar tying robot collects the original point cloud data of the rebar intersection area, and preprocesses the original point cloud data to obtain corresponding processed point cloud data. This point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be tied in the target tying area. The preprocessing of the original point cloud data to obtain the corresponding processed point cloud data includes: downsampling the original point cloud data using a preset voxel mesh filtering method to obtain corresponding filtered point cloud data; identifying outlier point cloud data in the filtered point cloud data and removing the outlier point cloud data to obtain the corresponding processed point cloud data.

[0042] In one specific implementation, the collected raw point cloud data is assumed to be an unordered set of points:

[0043] ;

[0044] in, Represents the original point cloud data set. Indicates the first The three-dimensional coordinate vector of a point , This represents the total number of points in the original point cloud dataset.

[0045] First, a coordinate system transformation is performed to change the point cloud from the camera coordinate system. Transform to world coordinate system Points in the camera coordinate system Convert to points in world coordinate system :

[0046] ;

[0047] In homogeneous coordinates, it is represented as .in, This represents the first [number] after transformation to the world coordinate system. Coordinates of a point, This represents a 4×4 homogeneous transformation matrix from the camera coordinate system to the world coordinate system, including rotation and translation. Represents the first in the camera coordinate system The coordinates of the points.

[0048] Then, a voxel grid filtering method is used for downsampling, defining the side length of the voxel cube as... Divide the space into a regular grid:

[0049] ;

[0050] For each voxel If it contains a set of points Then use the center of mass of that voxel. Replace all points:

[0051] ;

[0052] in, This represents the point cloud set after voxel filtering. The parameter representing the side length of a voxel cube typically ranges from 0.5 to 2.0 mm. Indicates the first Individual factors, Indicates the first The centroid coordinates of points in an individual element. Represents a set The number of midpoints.

[0053] Outlier removal can be performed using methods based on statistical distributions. For point clouds... For each point in the array, calculate its distance to... Average distance between the nearest neighbors:

[0054] set up It is a point of The average distance between the nearest neighbors is:

[0055] ;

[0056] in, Point The set of nearest neighbors, This represents the number of nearest neighbors, typically ranging from 30 to 50. Point The average distance to its nearest neighbor. The L2 norm (Euclidean distance) of a vector is used to represent the vector.

[0057] Next, calculate the average distance and standard deviation of the entire point cloud, assuming... There is If there are 10 points, then the average distance is 100. for:

[0058] ;

[0059] Standard deviation for:

[0060] ;

[0061] in:

[0062] This represents the number of points after voxel filtering. Not greater than ; This represents the mean of the average distances between all points in the point cloud; The standard deviation of the average distance between all points in the point cloud;

[0063] Finally, points that meet the following conditions are retained to obtain the final preprocessed point cloud. :

[0064] ;

[0065] in, This represents the standard deviation multiplier threshold, an adjustable parameter used to control the strictness of outlier removal, typically ranging from 1.0 to 2.0. Through the above preprocessing steps, the original point cloud... Converted into clean, downsampled point clouds This provides high-quality input data for subsequent point cloud completion and registration.

[0066] Step S12: Using the target point cloud completion network, the processed point cloud data is completed based on the preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning.

[0067] In this embodiment, a point cloud completion network based on deep learning is designed to address the point cloud missing problem caused by rebar self-occlusion. As a target point cloud completion network, it is specifically designed to handle the cylindrical geometric features of reinforcing bars. Specifically, the target point cloud completion network can be obtained by training an initial point cloud completion network based on an encoder-decoder architecture using target loss functions, including the target chamfer distance loss function, the cylindrical surface normal consistency loss function, and the curvature consistency loss function. In one specific implementation, the encoder of the target point cloud completion network is used to determine the feature representation of the input point cloud data using a multilayer perceptron structure, and the decoder is used to reconstruct the complete point cloud data corresponding to the input point cloud data based on the feature representation using deconvolution and fully connected layers. By incorporating cylindrical geometric constraints, the point cloud completion accuracy can be improved from multiple dimensions.

[0068] In one specific implementation, let the preprocessed incomplete point cloud be... The complete point cloud output by the network is ,but:

[0069] ;

[0070] Wherein, represents the complete point cloud set predicted by the network. This represents the set of learnable parameters of the network.

[0071] The encoder part of the network adopts a multilayer perceptron structure, and the feature representation of the point cloud is calculated as follows:

[0072] ;

[0073] in It is the encoded feature matrix, and D is the feature dimension of the feature matrix.

[0074] The decoder section uses deconvolution and fully connected layers to reconstruct the complete point cloud:

[0075] ;

[0076] Training is supervised using a multi-level loss function. First, an improved chamfer distance loss is used to measure the difference between the predicted point cloud and the true complete point cloud:

[0077] ;

[0078] in, This represents the predicted point cloud set. , Represents a true and complete set of point clouds. This indicates the improved chamfer distance loss.

[0079] Introducing a uniformity loss for the normal direction of the cylindrical surface ensures the geometric continuity of the completed surface:

[0080] ;

[0081] in, Indicates the first point in the predicted point cloud The normal vector of each point This represents the normal vector of the corresponding point in the real point cloud. This indicates the number of corresponding point pairs.

[0082] Introducing curvature uniformity loss to maintain the surface's smoothness:

[0083] ;

[0084] in, This represents the curvature value of the j-th point in the predicted point cloud. This represents the curvature value of the corresponding point in the real point cloud. Indicates the number of points.

[0085] The total loss function can be expressed as the weighted sum of the losses of each item, that is:

[0086] ;

[0087] in, , , The weighting coefficients for each loss satisfy the following conditions: .

[0088] Step S13: Align the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area, and plan the motion trajectory of the rebar binding robot based on the aligned point cloud data to obtain the target motion trajectory, so as to control the rebar binding robot to bind the rebar to be bound in the target binding area based on the target motion trajectory.

[0089] In this embodiment, before aligning the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area, the method further includes: obtaining the preset rebar point cloud template and saving the preset rebar point cloud template locally on the rebar binding robot. Aligning the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area specifically includes: determining the target point cloud template from all preset rebar point cloud templates saved locally on the rebar binding robot based on the completed point cloud data and the region type of the target binding area; performing preliminary alignment of the completed point cloud data and the target point cloud template based on a preset principal component analysis method to obtain the corresponding preliminary aligned point cloud; and registering the completed point cloud data and the target point cloud template based on the preliminary aligned point cloud using the target iterative nearest point algorithm and the target displacement correction strategy to obtain the aligned point cloud data corresponding to the target binding area.

[0090] In one specific implementation, the completed point cloud is... Point cloud of pre-constructed steel reinforcement formwork Alignment is performed using a hierarchical registration strategy.

[0091] Phase 1: Initial alignment based on principal component analysis.

[0092] Computational point cloud Covariance matrix:

[0093] ;

[0094] in It is a point cloud The center of mass.

[0095] Perform eigenvalue decomposition on the covariance matrix:

[0096] ;

[0097] in, , , This represents three eigenvectors. This represents the corresponding eigenvalue.

[0098] Initial rotation matrix Composed of eigenvectors, represented as The initial translation vector is represented as ,in It is a template point cloud The center of mass.

[0099] Phase Two: Optimize Alignment.

[0100] An improved iterative nearest-point algorithm is used for fine registration, with the objective function being:

[0101] ;

[0102] in, These are the corresponding point pairs that have been established. , , It is the weight coefficient of the i-th corresponding point, calculated based on the point-to-point distance.

[0103] In this embodiment, an adaptive displacement correction strategy can also be used to avoid local optima. For example, when continuous... The error decrease in the next iteration is less than the threshold. At that time, a displacement disturbance is applied:

[0104] ;

[0105] in, The unit vector representing the main axial direction of the reinforcing bar. This represents the disturbance step size, typically taken as 0.5-2.0 times the diameter of the reinforcing bar. Indicates the gradient direction of the loss function. This represents the minimum error reduction threshold. After the perturbation, the optimization restarts until the convergence condition is met or the maximum number of iterations is reached. By embedding a displacement perturbation mechanism into traditional ICP, this embodiment can adaptively correct point cloud data along the reinforcing bar axis, effectively avoiding local optima and improving the registration success rate. This two-stage architecture processing flow of "completing first and then registering" effectively solves the problem of missing point clouds caused by reinforcing bar self-occlusion, improving registration accuracy and robustness.

[0106] In this embodiment, considering the rotational symmetry of the rebar intersections and the existence of multiple equivalent binding postures, the motion trajectory of the rebar binding robot is planned based on the aligned point cloud data to obtain the target motion trajectory. Specifically, this may include: determining the current first binding posture of the rebar binding robot, and determining the corresponding second binding posture of the rebar binding robot based on the aligned point cloud data; determining the robotic arm motion trajectory corresponding to the target robotic arm of the rebar binding robot based on the first and second binding postures, and planning the motion trajectory of the rebar binding robot based on the aligned point cloud data and the robotic arm motion trajectory to obtain the target motion trajectory.

[0107] In one specific implementation, four candidate poses are defined. Each candidate pose It is a 4×4 homogeneous transformation matrix, represented as:

[0108] ;

[0109] in, Indicate candidate pose The rotation matrix, Indicate candidate pose The translation vector.

[0110] Next, the current posture of the robotic arm's end effector is calculated. Relative rotation angle with respect to each candidate pose:

[0111] ;

[0112] in, The rotation matrix represents the current posture of the robotic arm's end effector. Represents the trace operation of a matrix (the sum of the diagonal elements). This represents the minimum rotation angle required to rotate from the current pose to the candidate pose k, and the optimal binding pose is selected. ,satisfy:

[0113] ;

[0114] .

[0115] Finally, the optimal binding posture will be determined. Transform to the robot's base coordinate system using coordinate transformation:

[0116] ;

[0117] in, This represents the transformation matrix from the tool coordinate system to the robot base coordinate system. This represents the transformation matrix from the camera coordinate system to the tool coordinate system. This indicates the target pose of the robotic arm's end effector in the base coordinate system.

[0118] For a six-axis robotic arm, the joint angles are solved using inverse kinematics. Let the rotation matrix of the target pose be... The translation vector is Solve for the joint angle vector Calculation of the first joint angle:

[0119] ;

[0120] in , , Indicates the target location The component d_3 represents the length of the third link of the robotic arm, and the other joint angles are solved recursively through geometric relationships. The formula for calculating the second joint angle is:

[0121] ;

[0122] in , The length of the first link. This is the offset of the first joint.

[0123] The joint angles from the third to the sixth joint were obtained through rotation matrix decomposition:

[0124] ;

[0125] ;

[0126] ;

[0127] ;

[0128] in Representing the rotation matrix The Line number Column elements.

[0129] Motion control can employ an adaptive PID algorithm, with the control law as follows:

[0130] ;

[0131] in This represents the joint angle error vector. This represents the desired joint angle vector. This represents the actual joint angle vector. The proportional, integral, and differential gain matrices can be adaptively adjusted according to the motion state. Finally, the robotic arm is driven by robot control commands to complete the binding action, achieving automated rebar binding and thus realizing high-precision, high-efficiency, and robust 6D attitude estimation of the rebar intersection area. This embodiment, based on millimeter-level positioning accuracy and high binding success rate, shortens the single-point binding time, significantly improving efficiency compared to traditional manual binding. Simultaneously, the solution exhibits good adaptability, supporting automatic identification and binding of various rebar diameters and horizontal, vertical, and inclined arrangement methods. Furthermore, by optimizing the motion path and real-time processing capabilities, construction costs are significantly reduced, wear on the robotic arm joints is minimized, and the construction safety environment is fundamentally improved, providing a reliable technical solution for automated rebar binding.

[0132] As can be seen, after acquiring the original point cloud data of the target binding area, the rebar tying robot in this application can perform a series of preprocessing operations to obtain the corresponding processed point cloud data, providing high-quality input data for subsequent point cloud completion and registration. Then, using a target point cloud completion network designed for the cylindrical features of the rebar, the processed point cloud data is completed based on the cylindrical geometric constraints, thereby ensuring that the completed point cloud maintains the true geometric shape. Finally, the completed point cloud data is aligned with a preset rebar point cloud template to achieve adaptive registration of the completed point cloud data. Based on the aligned point cloud data, the motion trajectory of the rebar tying robot is planned to obtain the target motion trajectory, which controls the rebar tying robot to bind the rebars to be bound in the target binding area. This effectively solves the problem of point cloud missing caused by rebar self-occlusion and improves registration accuracy and robustness.

[0133] See Figure 2 As shown, this application discloses a rebar tying device, applied to a rebar tying robot, comprising:

[0134] The point cloud data acquisition module 11 is used to acquire the original point cloud data of the target binding area using the target camera mounted on the rebar binding robot, and to preprocess the original point cloud data to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be bound in the target binding area.

[0135] The point cloud data completion module 12 is used to complete the processed point cloud data based on preset cylindrical geometric constraints using a target point cloud completion network to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning.

[0136] The rebar tying module 13 is used to align the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target tying area, and to plan the motion trajectory of the rebar tying robot based on the aligned point cloud data to obtain the target motion trajectory, so as to control the rebar tying robot to tie the rebar to be tied in the target tying area based on the target motion trajectory.

[0137] As can be seen, after acquiring the original point cloud data of the target binding area, the rebar tying robot in this application can perform a series of preprocessing operations to obtain the corresponding processed point cloud data, providing high-quality input data for subsequent point cloud completion and registration. Then, using a target point cloud completion network designed for the cylindrical features of the rebar, the processed point cloud data is completed based on the cylindrical geometric constraints, thereby ensuring that the completed point cloud maintains the true geometric shape. Finally, the completed point cloud data is aligned with a preset rebar point cloud template to achieve adaptive registration of the completed point cloud data. Based on the aligned point cloud data, the motion trajectory of the rebar tying robot is planned to obtain the target motion trajectory, which controls the rebar tying robot to bind the rebars to be bound in the target binding area. This effectively solves the problem of point cloud missing caused by rebar self-occlusion and improves registration accuracy and robustness.

[0138] In one specific embodiment, the point cloud data acquisition module 11 may include:

[0139] A voxel filtering unit is used to downsample the original point cloud data using a preset voxel grid filtering method to obtain the corresponding filtered point cloud data.

[0140] An outlier data removal unit is used to identify outlier point cloud data in the filtered point cloud data and remove the outlier point cloud data from the filtered point cloud data to obtain the corresponding processed point cloud data.

[0141] In one specific embodiment, the device may further include:

[0142] The network training module is used to train the initial point cloud completion network based on the encoder-decoder architecture using the target loss function to obtain the target point cloud completion network.

[0143] The target loss function includes the target chamfer distance loss function, the cylindrical surface normal consistency loss function, and the curvature consistency loss function.

[0144] In one specific embodiment, the device may further include:

[0145] The template acquisition module is used to acquire a preset rebar point cloud template and save the preset rebar point cloud template locally on the rebar binding robot.

[0146] In one specific embodiment, the rebar tying module 13 may include:

[0147] The template determination unit is used to determine the target point cloud template from all the preset rebar point cloud templates stored locally by the rebar tying robot based on the completed point cloud data and the region type of the target tying area.

[0148] The first point cloud alignment unit is used to perform preliminary alignment of the completed point cloud data and the target point cloud template based on a preset principal component analysis method to obtain a corresponding preliminary aligned point cloud.

[0149] The second point cloud alignment unit is used to register the completed point cloud data with the target point cloud template based on the preliminary aligned point cloud using the target iterative nearest point algorithm and the target displacement correction strategy to obtain the aligned point cloud data corresponding to the target binding area.

[0150] In one specific embodiment, the rebar tying module 13 may include:

[0151] The binding posture determination unit is used to determine the current first binding posture of the rebar binding robot and determine the corresponding second binding posture of the rebar binding robot based on the aligned point cloud data.

[0152] The motion trajectory planning unit is used to determine the motion trajectory of the target robotic arm of the rebar tying robot based on the first tying posture and the second tying posture, and to plan the motion trajectory of the rebar tying robot based on the aligned point cloud data and the robotic arm motion trajectory to obtain the target motion trajectory.

[0153] Furthermore, embodiments of this application also disclose an electronic device, Figure 3 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content of the diagram should not be construed as limiting the scope of this application.

[0154] Figure 3 This is a schematic diagram of the structure of an electronic device 20 provided in an embodiment of this application. Specifically, the electronic device 20 may include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 stores a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the rebar tying method disclosed in any of the foregoing embodiments. Alternatively, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0155] In this embodiment, the power supply 23 is used to provide operating voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and external devices, and the communication protocol it follows can be any communication protocol applicable to the technical solution of this application, and is not specifically limited here; the input / output interface 25 is used to acquire external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs, and is not specifically limited here.

[0156] In addition, the memory 22, as a carrier for resource storage, can be a read-only memory, random access memory, disk, or optical disk, etc. The resources stored thereon can include an operating system 221, computer programs 222, etc., and the storage method can be temporary storage or permanent storage.

[0157] The operating system 221 is used to manage and control the various hardware devices on the electronic device 20 and the computer program 222, which may be Windows Server, Netware, Unix, Linux, etc. In addition to including a computer program capable of performing the rebar tying method disclosed in any of the foregoing embodiments by the electronic device 20, the computer program 222 may further include computer programs capable of performing other specific tasks.

[0158] Furthermore, this application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, it implements the aforementioned rebar tying method. Specific steps of this method can be found in the corresponding content disclosed in the foregoing embodiments, and will not be repeated here.

[0159] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section.

[0160] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0161] The steps of the methods or algorithms described in conjunction with the embodiments disclosed herein can be implemented directly by hardware, a software module executed by a processor, or a combination of both. The software module can be located in random access memory (RAM), main memory, read-only memory (ROM), electrically programmable ROM, electrically erasable programmable ROM, registers, hard disk, removable disk, CD-ROM, or any other form of storage medium known in the art.

[0162] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0163] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for tying reinforcing bars, characterized in that, Applications to rebar tying robots include: The original point cloud data of the target binding area is acquired using the target camera mounted on the rebar binding robot, and the original point cloud data is preprocessed to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be bound in the target binding area. The target point cloud completion network is used to complete the processed point cloud data based on preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning. The completed point cloud data is aligned with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area. The motion trajectory of the rebar binding robot is planned based on the aligned point cloud data to obtain the target motion trajectory, so as to control the rebar binding robot to bind the rebar to be bound in the target binding area based on the target motion trajectory. Before using the target point cloud completion network to complete the processed point cloud data based on preset cylindrical geometric constraints to obtain the completed point cloud data corresponding to the target binding region, the method further includes: The target point cloud completion network is obtained by training the initial point cloud completion network based on the encoder-decoder architecture using the target loss function. The target loss function includes the target chamfer distance loss function, the cylindrical surface normal consistency loss function, and the curvature consistency loss function; Before aligning the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area, the method further includes: Obtain a preset rebar point cloud template and save the preset rebar point cloud template locally on the rebar binding robot; The step of aligning the completed point cloud data with the preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target binding area includes: Based on the completed point cloud data and the region type of the target binding area, the target point cloud template is determined from all the preset rebar point cloud templates stored locally by the rebar binding robot. Based on a preset principal component analysis method, the completed point cloud data and the target point cloud template are initially aligned to obtain the corresponding preliminary aligned point cloud. Using the target iterative nearest point algorithm and target displacement correction strategy, the completed point cloud data and the target point cloud template are registered based on the preliminary aligned point cloud to obtain the aligned point cloud data corresponding to the target binding area.

2. The rebar tying method according to claim 1, characterized in that, The preprocessing of the original point cloud data to obtain the corresponding processed point cloud data includes: The original point cloud data is downsampled using a preset voxel grid filtering method to obtain the corresponding filtered point cloud data. Outlier point cloud data in the filtered point cloud data are identified and removed to obtain the corresponding processed point cloud data.

3. The rebar tying method according to claim 1, characterized in that, The encoder of the target point cloud completion network is used to determine the feature representation of the input point cloud data using a multilayer perceptron structure, and the decoder of the target point cloud completion network is used to reconstruct the complete point cloud data corresponding to the input point cloud data based on the feature representation using deconvolution and fully connected layers.

4. The rebar tying method according to claim 1, characterized in that, The step of planning the motion trajectory of the rebar tying robot based on the aligned point cloud data to obtain the target motion trajectory includes: The first binding posture of the rebar binding robot is determined, and the second binding posture of the rebar binding robot is determined based on the aligned point cloud data. Based on the first binding posture and the second binding posture, the motion trajectory of the target robotic arm of the rebar binding robot is determined, and based on the aligned point cloud data and the robotic arm motion trajectory, the motion trajectory of the rebar binding robot is planned to obtain the target motion trajectory.

5. A rebar tying device, characterized in that, Applications to rebar tying robots include: The point cloud data acquisition module is used to acquire the original point cloud data of the target binding area using the target camera mounted on the rebar binding robot, and to preprocess the original point cloud data to obtain the corresponding processed point cloud data; the point cloud data is the point cloud data corresponding to the rebar intersection points of the rebars to be bound in the target binding area. The point cloud data completion module is used to complete the processed point cloud data based on preset cylindrical geometric constraints using a target point cloud completion network to obtain the completed point cloud data corresponding to the target binding region; the target point cloud completion network is a point cloud completion network based on deep learning. The rebar tying module is used to align the completed point cloud data with a preset rebar point cloud template to obtain the aligned point cloud data corresponding to the target tying area, and to plan the motion trajectory of the rebar tying robot based on the aligned point cloud data to obtain the target motion trajectory, so as to control the rebar tying robot to tie the rebar to be tied in the target tying area based on the target motion trajectory. The device further includes: The network training module is used to train the initial point cloud completion network based on the encoder-decoder architecture using the target loss function to obtain the target point cloud completion network. The target loss function includes the target chamfer distance loss function, the cylindrical surface normal consistency loss function, and the curvature consistency loss function; The device further includes: The template acquisition module is used to acquire a preset rebar point cloud template and save the preset rebar point cloud template locally on the rebar binding robot. The rebar tying module specifically includes: The template determination unit is used to determine the target point cloud template from all the preset rebar point cloud templates stored locally by the rebar tying robot based on the completed point cloud data and the region type of the target tying area. The first point cloud alignment unit is used to perform preliminary alignment of the completed point cloud data and the target point cloud template based on a preset principal component analysis method to obtain a corresponding preliminary aligned point cloud. The second point cloud alignment unit is used to register the completed point cloud data with the target point cloud template based on the preliminary aligned point cloud using the target iterative nearest point algorithm and the target displacement correction strategy to obtain the aligned point cloud data corresponding to the target binding area.

6. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the rebar tying method as described in any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that, Used to store a computer program, wherein the computer program, when executed by a processor, implements the rebar tying method as described in any one of claims 1 to 4.

Images