Robotic welding teaching system and method based on multi-level binocular vision

Abstract

This invention discloses a robot welding teaching system and method based on multi-level binocular vision, belonging to the field of robot welding technology. The system includes a macroscopic positioning system, a mobile transfer positioning device, and a 6D pose teaching pendant. The 6D pose teaching pendant is used to capture critical path points on the workpiece to be welded, and its surface is equipped with reflective markers. The mobile transfer positioning device is a movable binocular vision device used to detect the spatial pose of the 6D pose teaching pendant, and its surface is also equipped with reflective markers. The macroscopic positioning system is fixedly installed on the welding robot and is used to measure the pose of the mobile transfer positioning device. The macroscopic positioning system, the mobile transfer positioning device, and the 6D pose teaching pendant are all connected to a host computer. This invention improves anti-occlusion capabilities, enhances the system's adaptability in complex welding scenarios, and ensures stable operation in various environments. It significantly improves teaching efficiency. The high-precision binocular vision system and marker point positioning ensure welding quality.

Description

Technical Field

[0001] This invention belongs to the field of robotic welding technology, specifically relating to a robotic welding teaching system and method based on multi-level binocular vision. Background Technology

[0002] As the heavy industry accelerates its digital transformation and industrial upgrading, robotic welding systems have become core equipment for improving welding quality, efficiency, and ensuring production safety. Currently, in aerospace, automotive, and other fields, robots have replaced traditional manual welding, significantly improving welding production efficiency and quality. In the heavy industry, adopting machines to perform welding operations is an inevitable path to drive the industry towards intelligence and automation. However, in complex production environments, existing robotic welding path planning technologies still face many challenges, especially in the welding of unstructured workpieces, where the limitations of traditional path planning methods become increasingly apparent.

[0003] Traditional welding path planning technologies are mainly divided into three types: teach-and-playback, offline programming, and intelligent teach-free, with the following specific characteristics:

[0004] The teach-and-reproduce technique requires operators to manually control the teach pendant to mark the welding path point by point. Although it is widely used, in the welding of multi-layer, multi-pass, and dispersed weld structures, the teaching cycle is long, the efficiency is low, and it is easily affected by human error, which cannot meet the requirements of large-scale production.

[0005] 2. Offline programming technology performs path planning through a computer simulation environment. Although it does not occupy production time, its path planning accuracy depends entirely on the standard model of the workpiece. In the heavy industry and shipbuilding industry, many workpieces lack complete standard models and generally have processing errors and assembly deviations, causing the simulated path to be out of sync with the actual working conditions. This requires a lot of manual correction, which greatly reduces the effectiveness of practical applications.

[0006] 3. Intelligent teach-free welding technology attempts to automatically identify weld features through sensors to achieve autonomous path planning and welding. However, the technology is still immature in the identification of weld features and adaptation to complex bevels in unstructured scenarios, and it is difficult to cope with the complex environment on site.

[0007] In recent years, vision-based teaching methods have received widespread attention. These methods utilize industrial cameras or laser sensors to acquire the pose information of the teaching tool or weld seam, aiming to simplify the teaching process. For example, a fixed-mounted industrial camera can be used to photograph a teaching pendant with QR codes or reflective markers. The trajectory of the teaching pendant can be calculated through image processing, and then real-time correction can be performed using a laser weld seam tracking sensor. However, when teaching heavy-duty workpieces with internal cavities, deep grooves, or complex curved surfaces, fixed-mounted macroscopic cameras are easily obstructed by the workpiece itself or the operator, resulting in the incomplete capture of feature points on the teaching pendant, leading to teaching failure or a significant decrease in accuracy. Another example is the integration of binocular vision sensors and structured light lasers into the front end of the welding torch. While this enables local real-time tracking, its field of view is limited, failing to acquire the absolute pose of the teaching pendant in global space, still requiring reliance on pre-taught or model data.

[0008] Therefore, although existing technologies have been applied in specific scenarios, they still have significant limitations in welding unstructured workpieces such as heavy industry and shipbuilding. Teach-and-playback technology is inefficient, offline programming relies on standard models and struggles to adapt to workpiece errors, while intelligent teach-free technology is not yet mature. Furthermore, existing visual teaching solutions are limited to fixed viewpoints or partial fields of view, making it difficult to address the occlusion problems of complex workpieces.

[0009] To address these issues, there is an urgent need to develop a welding robot path planning technology that adapts to unstructured scenarios and balances efficiency and accuracy, in order to meet more complex production needs. Summary of the Invention

[0010] The purpose of this invention is to address the shortcomings of existing technologies by providing a robot welding teaching system and method based on multi-level binocular vision. This addresses the problems of existing fixed vision teaching systems being easily obscured and having low teaching efficiency in unstructured scenarios, thereby enabling rapid and high-precision planning of welding paths for unstructured workpieces.

[0011] To achieve the above objectives, the technical solution adopted by this invention is: a robot welding teaching system based on multi-level binocular vision, including a macroscopic positioning system, a mobile transfer positioning device, and a 6D pose teaching pendant; the 6D pose teaching pendant is used to capture critical path points on the workpiece to be welded, and its surface is provided with reflective markers; the mobile transfer positioning device is a movable binocular vision device, the front of which is used to detect the spatial pose of the 6D pose teaching pendant, and its surface is provided with reflective markers; the macroscopic positioning system is fixedly installed on the welding robot and is used to measure the pose of the mobile transfer positioning device; the macroscopic positioning system, the mobile transfer positioning device, and the 6D pose teaching pendant are all connected to a host computer, which calculates coordinate transformations, records teaching points, plans paths, and generates robot programs.

[0012] Furthermore, the 6D pose teaching gun includes an upper positioning part and a lower grip part; the upper positioning part adopts an asymmetric polyhedral structure, with reflective marking points distributed on its surface.

[0013] Furthermore, the reflective marking point has a white core and black ring structure.

[0014] Furthermore, the host computer stores the transformation matrix from the macroscopic positioning system coordinate system to the robot base coordinate system, and calculates the pose of the 6D pose teaching gun in the robot base coordinate system in real time according to the following formula. ;

[0015] ;

[0016] In the formula:

[0017] For the macroscopic positioning system coordinate system O w To the robot base coordinate system O robot The coordinate transformation matrix;

[0018] For the coordinate system O of the relay positioning instrument m To the macroscopic positioning system coordinate system O w The coordinate transformation matrix;

[0019] For the teaching gun coordinate system O t To the coordinate system O of the transfer positioning instrument m The coordinate transformation matrix;

[0020] This is the welding torch attitude matrix;

[0021] This is the weld position matrix.

[0022] The present invention also provides a teaching method based on the above system, comprising the following steps:

[0023] S1. System Calibration: Calibrate the intrinsic and extrinsic parameters of the binocular camera of the macro positioning system and the mobile relay positioning device, as well as the transformation matrix between the macro positioning system coordinate system and the robot base coordinate system;

[0024] S2. Place the mobile relay positioning device: Place the mobile relay positioning device in the work area so that it can be observed by the macro positioning system and its front field of view covers the 6D pose teaching gun.

[0025] S3. Teaching Operation: The operator holds the 6D pose teaching gun and points it to the key points of the welding path; the host computer captures the spatial position and attitude data of the 6D pose teaching gun in real time through the binocular vision system of the mobile transfer positioning device and completes precise positioning; the position and attitude information of each key point of the welding path is automatically collected and stored in the host computer as the basic data for path planning.

[0026] S4. Pose Calculation and Path Generation: Based on the stored pose data of key points in the welding path, the host computer automatically calculates the accurate coordinates of each key point in the robot base coordinate system through a pre-calibrated coordinate transformation matrix, and performs path planning based on these accurate coordinates; based on the path planning results and user-set parameters, it automatically generates a complete welding robot execution program.

[0027] Furthermore, in S4, user-configurable parameters include interpolation type parameters and welding process parameters. The welding process parameters include welding speed, current, and voltage.

[0028] The beneficial effects of this invention are as follows: This invention introduces a movable mobile relay positioning device as an intermediary between the macroscopic positioning system and the 6D pose teaching gun. The macroscopic system is responsible for global positioning of the mobile relay positioning device, providing initial positioning and ensuring wide-area coverage. The mobile relay positioning device is responsible for positioning the 6D pose teaching gun, providing local fine positioning and ensuring high accuracy. The two work together to solve the problem that the macroscopic positioning system cannot completely capture the pose of the 6D pose teaching gun in complex welding scenarios due to occlusion. This allows for flexible handling of complex workpieces, ensuring accurate positioning and data transmission of the 6D pose teaching gun in complex welding scenarios, enhancing the system's adaptability to complex welding scenarios, and ensuring stable operation in various environments.

[0029] Compared to traditional teach-and-playback techniques, this solution eliminates the need to operate the robot via a teach pendant, significantly simplifying the operation process. Operators only need to use a 6D pose teach pendant to intuitively calibrate key feature points of the weld, avoiding the tedious point-by-point adjustment process and greatly improving teaching efficiency.

[0030] This invention employs a high-precision positioning method that combines a high-precision binocular vision system with circular marker points, ensuring that the accuracy of position and attitude data meets industrial welding requirements throughout the entire teaching and path planning process. This results in high positioning accuracy and stable and controllable welding quality. Attached Figure Description

[0031] Figure 1 This is a three-dimensional structural diagram of the multi-level binocular vision robot 6D teaching system invented in this patent.

[0032] Figure 2 Schematic diagram of a 6D pose teaching gun;

[0033] Figure 3 This is a schematic diagram of a mobile relay positioning device.

[0034] In the diagram, 1-macro positioning system, 2-mobile transfer positioning device, 21-crossbeam, 22-tripod, 3-6D pose teaching gun, 31-upper positioning part, 32-lower grip part, 300-reflective marker, 310-LED indicator light, 4-welding robot. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0036] like Figure 1 As shown, the robot welding teaching system based on multi-level binocular vision includes a macro positioning system 1, a mobile transfer positioning device 2, a 6D pose teaching gun 3, and a welding robot 4.

[0037] Among them, welding robot 4 is an existing device, which usually has its own controller and robotic arm, with a real welding torch for welding installed at the end of the robotic arm.

[0038] The structure of the 6D pose teaching gun 3 is as follows: Figure 2 As shown, drawing inspiration from the design of a real welding torch, the overall structure is divided into an upper positioning section 31 and a lower grip section 32. The upper positioning section 31 adopts an asymmetrical polyhedral structure, with reflective markers 300 distributed on its surface. Each reflective marker 300 has a white core and a black ring structure; the center of each marker is made of highly reflective white material, while the outer ring is made of light-absorbing black material. This enhances the robustness of the binocular vision system in feature extraction under complex lighting conditions, ensuring high-precision pose detection. The upper positioning section 31 has at least four reflective markers 300, and these four markers are not coplanar, with no two points being equidistant from each other. The lower grip section 32 replicates the grip area and end shape of a real welding torch, facilitating quick selection of key weld feature points by the operator and improving operational flexibility.

[0039] The 6D pose teaching pendant 3 integrates a microcontroller, a wireless communication module, an LED indicator 310, and a trigger button. The LED indicator and trigger button are located on the outer wall of the lower grip portion 32. The wireless communication module, LED indicator, and trigger button are all electrically connected to the microcontroller. The microcontroller communicates in real-time with a host computer (such as a computer) via the wireless communication module, transmitting pose data and teaching point recording instructions. The LED indicator, controlled by the microcontroller, provides feedback to the operator on the validity status of the current teaching point. For example, a solid green light indicates a valid and compliant teaching point, while a solid red light indicates a pose exceeding limits or a path problem, reminding the operator to adjust. This 6D pose teaching pendant 3 is wireless and flexible, overcoming the inefficiency of traditional teaching methods. The operator only needs to hold the 6D pose teaching pendant 3 and point it directly at the target position to complete the teaching, achieving stable communication with the welding robot 4 and the macroscopic positioning system 1.

[0040] The mobile relay positioning device 2 is a mobile binocular vision device, and its specific structure is as follows: Figure 3 As shown, the system includes a crossbeam 21 mounted on a movable tripod 22. Two industrial cameras are mounted on either end of the crossbeam 21 facing the workpiece to be welded (i.e., the front), forming a binocular vision system for capturing images of the reflective markers 300 on the 6D pose teaching pendant 3. The industrial cameras used must support external triggering mode to ensure synchronized acquisition; they must also support global exposure to reduce motion blur caused by handheld shaking of the 6D pose teaching pendant 3; the field of view must be greater than 120 degrees; the baseline distance between the two industrial cameras must be no less than 500 mm; and the radial distortion of the lenses must be <0.5%, and the tangential distortion <0.1% to ensure accuracy.

[0041] The surface of the mobile relay positioning device 2 is covered with reflective markers for identification and capture by the macro positioning system 1. The reflective markers are usually arranged on the crossbeam 21, with a minimum of 4 markers. The 4 reflective markers are not coplanar, and the distance between any two of the 4 markers is not the same.

[0042] The mobile relay positioning device 2 is equipped with an embedded processor, which is used to control the binocular camera to acquire images, perform feature point extraction and stereo matching, and calculate the 6D pose of the 6D pose teaching gun 3 in the coordinate system of the relay positioning device.

[0043] The mobile relay positioning device 2 acquires image information from the 6D pose teaching pendant 3 using binocular vision, extracts feature points on the 6D pose teaching pendant 3, and performs stereo matching to obtain high-precision three-dimensional coordinates of the feature points on the 6D pose teaching pendant 3. It then calculates the three-dimensional coordinates and attitude of the 6D pose teaching pendant 3 in the relay positioning device coordinate system, i.e., the pose of the 6D pose teaching pendant 3 in the relay positioning device coordinate system. The mobile relay positioning device 2 communicates with the host computer wirelessly.

[0044] The mobile relay positioning device 2 acts as an "intermediary" between the macro camera and the 6D pose teaching gun 3, moving flexibly to achieve precise positioning and spatial pose calculation of the 6D pose teaching gun 3. Its design effectively solves the occlusion problem of traditional fixed macro cameras, flexibly adapts to different workpiece shapes, and ensures precise positioning and data transmission of the 6D pose teaching gun 3 in complex environments.

[0045] The macroscopic positioning system 1 consists of two macroscopic industrial cameras fixedly mounted on the welding robot 4, and it communicates with the host computer. It has two main functions. First, it can detect the arbitrarily positioned mobile transfer positioning device 2, thereby obtaining the spatial pose of the 6D pose teaching pendant 3 in the macroscopic positioning system coordinate system. Second, it is relatively fixed to the welding robot 4, and through a pre-calibrated transformation matrix, it can easily convert the spatial position of the 6D pose teaching pendant 3 in the macroscopic positioning system coordinate system into its position in the robot base coordinate system, facilitating robot task planning.

[0046] Based on the principle of binocular vision measurement, in order to calculate the spatial pose of the 6D pose teaching gun 3 and the calibration reference, it is necessary to ensure that the target object can be captured and imaged simultaneously by two industrial cameras.

[0047] In this robotic welding teaching system, the industrial camera of the macro-positioning system 1 is fixed and cannot be flexibly adjusted. Considering the limited space in the weld seam area inside some workpieces, direct observation of the 6D pose teaching pendant 3 using the macro-positioning system 1 results in field-of-view obstruction, making direct teaching positioning difficult. The added mobile transfer positioning device 2 acts as an intermediary. This device can be flexibly and freely deployed, easily adjusted to the field of view of the macro-positioning system 1, and simultaneously achieves precise observation of the 6D pose teaching pendant 3. By using the mobile transfer positioning device 2 to complete the line-of-view transfer and pose transfer, the spatial obstruction problem is effectively avoided, significantly expanding the system's applicable working conditions and comprehensively improving the versatility and adaptability of welding teaching scenarios.

[0048] The teaching method based on this teaching system has the following steps:

[0049] S1. System Calibration: Calibrate the intrinsic and extrinsic parameters of the binocular camera in macroscopic positioning system 1, the intrinsic and extrinsic parameters of the binocular camera in mobile relay positioning device 2, and the coordinate transformation matrix between the macroscopic positioning system coordinate system and the robot base coordinate system. Specifically, the intrinsic parameters of the binocular camera can be calibrated using the Zhang's planar checkerboard calibration method; the extrinsic parameters of the binocular camera can be calibrated using a planar target-based binocular extrinsic parameter calibration method; and the coordinate system of the macroscopic positioning system and the robot base coordinate system can be calibrated using the Tsai-Lenz hand-eye calibration method. All calibration methods used are existing industry methods and will not be described in detail. Once the binocular cameras are installed, and the camera and lens have not been replaced, only one calibration is required. The intrinsic and extrinsic parameters of the binocular camera in macro positioning system 1 are the same as those of the binocular camera in mobile relay positioning device 2. The purpose of intrinsic parameter calibration is to calculate the degree of lens distortion and construct the intrinsic parameter matrix. The parameters mainly include normalized focal length, principal point coordinates, distortion coefficients, etc. The purpose of extrinsic parameter calibration is to establish the spatial positional relationship between the two cameras. The parameters mainly include orthogonal rotation matrix, translation vector, etc.

[0050] S2. Positioning the mobile transfer positioning device 2: Adjust the position, height, and attitude of the mobile transfer positioning device 2 so that it is within the monitoring area of ​​the macro positioning system 1, can be simultaneously observed by the binocular camera of the macro positioning system 1, and its field of view covers the 6D pose teaching pendant 3, ensuring accurate recognition of the pose of the 6D pose teaching pendant 3 and ensuring that the workpiece weld falls within the effective monitoring range of the mobile transfer positioning device 2. Alignment can be quickly assisted by the LED indicator 310 of the 6D pose teaching pendant 3. When the system successfully calculates the 6D pose teaching pendant 3, the indicator light turns green.

[0051] S3. Teaching Operation: The operator holds the 6D pose teaching pendant 3 and sequentially points it at key points of the welding path on the workpiece in a reasonable posture. Key points of the welding path include the welding preparation point, arc initiation point, trajectory key points, arc extinguishing point, and safe withdrawal point. During operation of the 6D pose teaching pendant 3, the host computer uses the binocular vision system of the mobile relay positioning device 2 to capture the spatial position and posture data of the 6D pose teaching pendant 3 in real time and complete precise positioning. The position and posture information of each key point of the welding path is automatically collected and stored in the host computer as the basic data for path planning. Specifically: clicking the trigger button automatically collects and stores the position and posture information of the end of the 6D pose teaching pendant 3 at that key point in the host computer as the basic data for path planning. When clicking the point, the indicator light of the 6D pose teaching pendant 3 must be green. As is well known, in the welding process, in addition to accurately determining the welding position, the reasonable selection of the welding gun posture is also crucial to ensuring welding quality. For example, when welding overhead, a push-welding technique is used to prevent the molten pool from collapsing; and when welding narrow corners, the welding torch needs to be tilted to avoid collision and interference. The 6D pose teaching torch 3 used in this invention has a lower grip 32 designed according to the shape of the welding torch, allowing the operator to intuitively and conveniently specify the welding torch posture during the teaching process. This allows the invention to comprehensively consider welding process parameters, welding material type, and path, as well as the welding torch posture, thereby ensuring welding quality.

[0052] 6D pose teaching gun 3 in robot base coordinate system O robot Lower position It can be obtained through the following formula:

[0053] ;

[0054] In the formula:

[0055] For the macroscopic positioning system coordinate system O w To the robot base coordinate system O robot The coordinate transformation matrix;

[0056] For the coordinate system O of the relay positioning instrument m To the macroscopic positioning system coordinate system O w The coordinate transformation matrix;

[0057] For the teaching gun coordinate system O t To the coordinate system O of the transfer positioning instrument m The coordinate transformation matrix;

[0058] This is the welding torch attitude matrix;

[0059] This is the weld position matrix.

[0060] It can be obtained through calibration. , The calculation principles and steps are completely identical. The following is given... Detailed calculation steps:

[0061] I. Image Acquisition

[0062] In macro positioning system 1, the calibrated left and right cameras are used to photograph the marker points on the mobile relay positioning device 2, and their pixel coordinate set is obtained:

[0063] ;

[0064] in, This is the set of coordinates of the marker points on the mobile relay positioning device 2, captured by the left camera. The coordinates of the nth marker point on the mobile relay positioning device 2, captured by the left camera; This is the set of coordinates of the marker points on the mobile relay positioning device 2, captured by the left camera. The coordinates of the nth marker point on the mobile relay positioning device 2, as captured by the right camera.

[0065] II. Three-dimensional reconstruction

[0066] By utilizing the camera's intrinsic and extrinsic parameters and the principle of binocular stereo matching, the pixel coordinates of the left and right cameras are registered to obtain the marker point in the macroscopic positioning system coordinate system O. w The following is a set of three-dimensional coordinates:

[0067] ;

[0068] in, The marker point on the mobile relay positioning device 2 is located in the macroscopic positioning system coordinate system O. w The set of three-dimensional coordinates below For the nth marker point on the mobile relay positioning device 2, in the macroscopic positioning system coordinate system O w The three-dimensional coordinates below.

[0069] III. Calculation of Coordinate System Transformation Matrix

[0070] According to the CAD design model of the transfer positioning device, the marker point is in the coordinate system of the mobile transfer positioning device 2 (i.e., the coordinate system O of the transfer positioning device). m The set of three-dimensional coordinates under:

[0071] ;

[0072] in, The marker point on the mobile relay positioning device 2 is located in the relay positioning device coordinate system O.m The set of three-dimensional coordinates below For the nth marker point on the mobile relay positioning device 2 in the relay positioning device coordinate system O m The three-dimensional coordinates below.

[0073] use The transformation matrix of the two sets of coordinates is obtained by using the classic SVD-based rigid body registration algorithm. .

[0074] S4. Pose Calculation and Path Generation: After teaching all key points of the welding path, the host computer automatically calculates the accurate coordinates of each key point in the robot base coordinate system based on the stored pose data of the key points and a pre-calibrated coordinate transformation matrix. Path planning is then performed based on these accurate coordinates. Based on the path planning results and the parameters set by the user according to actual needs, a complete welding robot execution program is automatically generated. The parameters set by the user according to actual needs include the interpolation type (such as linear, circular, or spline interpolation) and welding process parameters (such as welding speed, current, voltage, etc.), which generate the robot welding program.

[0075] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A robot welding teaching system based on multi-level binocular vision, characterized in that, It includes a macro positioning system (1), a mobile relay positioning device (2), and a 6D pose teaching gun (3). The 6D pose teaching gun (3) is used to capture critical path points on the workpiece to be welded, and reflective markers (300) are set on the surface. The mobile relay positioning device (2) is a movable binocular vision device used to detect the spatial pose of the 6D pose teaching gun (3), and its surface is provided with reflective markers (300). The macro positioning system (1) is fixedly installed on the welding robot (4) and is used to measure the pose of the mobile transfer positioning device (2); The macro positioning system (1), the mobile relay positioning device (2) and the 6D pose teaching gun (3) are all connected to the host computer. The host computer calculates coordinate transformation, records teaching points, plans paths and generates robot programs. The host computer stores the transformation matrix from the macroscopic positioning system coordinate system of the macroscopic positioning system (1) to the robot base coordinate system, and calculates the pose of the 6D pose teaching gun (3) in the robot base coordinate system in real time according to the following formula. ; ； In the formula: For the macroscopic positioning system coordinate system O w To the robot base coordinate system O robot The coordinate transformation matrix; For the coordinate system O of the relay positioning instrument m To the macroscopic positioning system coordinate system O w The coordinate transformation matrix; For the teaching gun coordinate system O t To the coordinate system O of the transfer positioning instrument m The coordinate transformation matrix; This is the welding torch attitude matrix; This is the weld position matrix; Obtained through calibration. , Calculation principles, steps and completely consistent; Calculation steps: I. Image Acquisition In the macro positioning system (1), the calibrated left and right cameras are used to photograph the marker points on the mobile relay positioning device (2) to obtain their pixel coordinate set: ； in, The set of coordinates of the marker points on the mobile relay positioning device (2) captured by the left camera. The coordinates of the nth marker point on the mobile relay positioning device (2) captured by the left camera; The set of coordinates of the marker points on the mobile relay positioning device (2) captured by the left camera. The coordinates of the nth marker point on the mobile relay positioning device (2) captured by the right camera; II. Three-dimensional reconstruction By utilizing the camera's intrinsic and extrinsic parameters and the principle of binocular stereo matching, the pixel coordinates of the left and right cameras are registered to obtain the marker point in the macroscopic positioning system coordinate system O. w The following is a set of three-dimensional coordinates: ； in, The marker point on the mobile relay positioning device (2) is located in the macroscopic positioning system coordinate system O. w The set of three-dimensional coordinates below For the nth marker point on the mobile relay positioning device (2) in the macroscopic positioning system coordinate system O w The three-dimensional coordinates below; III. Calculation of Coordinate System Transformation Matrix According to the CAD design model of the transfer positioning device, the set of three-dimensional coordinates of the marker point in the coordinate system of the mobile transfer positioning device (2) is as follows: ； in, The marker point on the mobile relay positioning device (2) is located in the coordinate system O of the relay positioning device. m The set of three-dimensional coordinates below For the nth marker point on the mobile relay positioning device (2) in the relay positioning device coordinate system O m The three-dimensional coordinates below; use The transformation matrix of the two sets of coordinates is obtained by using the classic SVD-based rigid body registration algorithm. .

2. The robot welding teaching system based on multi-level binocular vision according to claim 1, characterized in that, The 6D pose teaching gun (3) includes an upper positioning part (31) and a lower grip part (32); the upper positioning part (31) adopts an asymmetric polyhedral structure, and the reflective marking points (300) are distributed on its surface.

3. The robot welding teaching system based on multi-level binocular vision according to claim 2, characterized in that, The reflective marker (300) has a white core and black ring structure.

4. A teaching method based on the robot welding teaching system based on multi-level binocular vision as described in any one of claims 1-3, characterized in that, Includes the following steps: S1: Calibrate the binocular camera intrinsic and extrinsic parameters of the macroscopic positioning system (1) and the mobile relay positioning device (2), as well as the transformation matrix between the macroscopic positioning system coordinate system and the robot base coordinate system; S2: Place the mobile relay positioning device (2) in the work area so that it can be observed by the macro positioning system (1) and its front field of view covers the 6D pose teaching gun (3). S3: The operator holds the 6D pose teaching gun (3) and points it to the key points of the welding path; the host computer captures the spatial position and attitude data of the 6D pose teaching gun (3) in real time through the binocular vision system of the mobile transfer positioning device (2) and completes the precise positioning; the position and attitude information of each key point of the welding path is automatically collected and stored in the host computer as the basic data for path planning. S4: The host computer automatically calculates the accurate coordinates of each key point in the robot base coordinate system based on the stored welding path key point pose data and a pre-calibrated coordinate transformation matrix, and performs path planning based on these accurate coordinates; based on the path planning results and user-set parameters, it automatically generates a complete welding robot execution program.

5. The teaching method according to claim 4, characterized in that, In S4, user-configurable parameters include interpolation type parameters and welding process parameters.

Images