A method for automatic processing of multi-specification workpieces based on three-dimensional vision

Abstract

The present invention discloses a method for automatic processing of multi-specification workpieces based on three-dimensional vision. First, a theoretical kinematic model of an industrial robot is established and identified to obtain the identified kinematic model; then, the industrial robot and the three-dimensional camera, as well as the industrial robot and the displacement device are calibrated respectively to obtain the corresponding homogeneous transformation matrix; then, the three-dimensional camera is used to collect visual information of the workpiece to be processed clamped on the displacement device, and then three-dimensional reconstruction is performed to obtain a three-dimensional point cloud of the workpiece to be processed; then the initial processing path of the workpiece to be processed is determined; finally, the initial processing path is corrected to obtain the corrected processing path, thereby controlling the processing operation of the industrial robot. The present invention can be widely used in the automatic processing of a variety of multi-specification small-batch workpieces, such as welding, gluing, laser processing, etc., has a very broad market application prospect, and has extremely important practical significance for improving the digitalization and intelligent level of my country's manufacturing industry.

Description

Technical Field

[0001] The present invention relates to an automatic workpiece processing method, in particular to an automatic workpiece processing method with multiple specifications based on three-dimensional vision. Background Art

[0002] Multi-specification, small-batch production involves producing a wide range of product specifications within a specified production cycle, but with smaller quantities of each specification. Compared to mass production, this approach is inefficient, costly, difficult to automate, and involves complex production processes. However, with rising labor costs and advances in automation technology, the processing of multi-specification, small-batch products is becoming increasingly important.

[0003] Take the processing of angle steel tower feet as an example. Angle steel tower feet are welded steel plates and are generally customized based on factors such as the terrain and the angle steel tower structure. The specific processing steps for the tower feet are: 1) Design the foot shape based on factors such as the terrain; 2) Cut the steel plates according to the drawings to complete the steel plate blanking; 3) Manual spot welding to complete the steel plate assembly; 4) Overall arc welding to complete all welds on the tower foot workpiece. There are many types of angle steel tower feet, but due to the strong demand for customization, the external dimensions of each angle steel tower vary significantly. Currently, the blanking process (steel plate cutting) for the tower foot workpiece is fully automated, but the diverse specifications of the tower feet make it difficult to develop a unified robot welding program, and all welding must be completed manually. Similar to the tower foot, the base of a wind turbine also has similar requirements.

[0004] In summary, the current production of small batches of workpieces with multiple specifications has the following problems: 1) The production method is inefficient, costly, difficult to automate, and the production organization process is complex; 2) The labor cost is high, and most links still require manual intervention; 3) It is difficult to recruit workers, and it is difficult to use robots to replace manual labor in harsh processing environments such as welding. Summary of the Invention

[0005] In order to solve the problems existing in the background technology, the present invention provides a method for automatic processing of multi-specification small batch workpieces based on three-dimensional vision, which uses three-dimensional vision to extract visual information of the workpiece surface, further generates the workpiece processing path, and completes the automatic processing of multi-specification small batch workpieces.

[0006] The technical solution adopted in the present invention is as follows:

[0007] 1) Establish and identify the theoretical kinematic model of the industrial robot to obtain the identified kinematic model;

[0008] 2) After hand-eye calibration of the industrial robot and 3D camera, the homogeneous transformation matrix of the 3D camera relative to the industrial robot base coordinate system is obtained. Denoted as the camera-robot calibration matrix; after calibrating the industrial robot and the displacement device, the homogeneous transformation matrix of the displacement device relative to the industrial robot base coordinate system is obtained Denoted as the displacement device-robot calibration matrix;

[0009] 3) Using a 3D camera to collect visual information of the workpiece to be processed, which is clamped on the positioner, and performing 3D reconstruction on the collected visual information of the workpiece to be processed based on the camera-robot calibration matrix and the positioner-robot calibration matrix to obtain a 3D point cloud of the workpiece to be processed;

[0010] 4) determining an initial processing path of the workpiece to be processed based on the three-dimensional point cloud of the workpiece to be processed;

[0011] 5) Correcting the initial machining path of the workpiece to be machined according to the identified kinematic model to obtain a corrected machining path;

[0012] 6) The corrected processing path is input into the robot control system, and the robot control system controls the industrial robot to process the workpiece.

[0013] The specific embodiment of 1) is:

[0014] 1.1) Install a target ball at the end of the industrial robot and establish a theoretical kinematic model of the industrial robot;

[0015] 1.2) After changing the actual rotation angles of each axis of the industrial robot, use a laser tracker to record the actual position information of the target ball relative to the coordinate system of the industrial robot base;

[0016] 1.3) Repeat step 1.2) several times to obtain the actual rotation angles of each axis of several sets of industrial robots and the corresponding position information of the target sphere relative to the industrial robot base coordinate system;

[0017] 1.4) Based on the actual axis rotation angles of several groups of industrial robots and the actual position information of the corresponding target balls relative to the industrial robot base coordinate system, and the optimized identification function, the theoretical kinematic model of the industrial robots is optimized using the least squares method to obtain the identified kinematic model, denoted as DH′.

[0018] The formula of the optimized identification function is as follows:

[0019]

[0020] in, The actual rotation angle of each axis of the industrial robot is θ i The position information of the target ball relative to the robot base coordinate system is calculated based on the theoretical kinematic model, p i is the actual position information of the target ball relative to the industrial robot base coordinate system.

[0021] Said 3) is specifically:

[0022] 3.1) Clamp the workpiece to be processed on the positioner and record the initial posture of the positioner;

[0023] 3.2) using a 3D camera to collect an initial point cloud of the workpiece to be processed at the current viewing angle, and preprocessing the initial point cloud of the workpiece to be processed at the current viewing angle to obtain a preprocessed point cloud;

[0024] 3.3) Use the camera-robot calibration matrix to transform the preprocessed point cloud to obtain the point cloud in the base coordinate system of the industrial robot at the current viewing angle;

[0025] 3.4) Control the rotation of the positioner and record the positioner's posture so that the posture of the workpiece to be processed is adjusted. Repeat 3.2)-3.3) to obtain point clouds in the base coordinate system of the industrial robot at different viewing angles until point clouds in the base coordinate system of the industrial robot are obtained at all viewing angles.

[0026] 3.5) Based on the positioner-robot calibration matrix, the point clouds of all positioner postures and the industrial robot's base coordinate system at the corresponding perspectives are spliced ​​and coordinate-converted to obtain a 3D point cloud of the workpiece to be processed under the positioner.

[0027] The specific embodiment of 4) is:

[0028] Extract point, line and surface information from the three-dimensional point cloud of the workpiece to be processed, determine the processing position according to the actual processing requirements, and then obtain the initial processing path of the workpiece to be processed.

[0029] Said 5) is specifically:

[0030] According to the position of the initial machining path of the workpiece to be machined, the actual axis rotation angle values ​​of the industrial robot in the initial machining path are calculated using the identified kinematic model. The actual axis rotation angle values ​​of the industrial robot in the initial machining path are used to replace the corresponding original axis rotation angle values ​​in the initial machining path and the machining path is updated to obtain the corrected machining path.

[0031] In said 6), the corrected processing path is sent to the robot control system via TCP / IP protocol or UDP protocol. The beneficial effects of the present invention are:

[0032] 1. This invention proposes a method for automated machining of small batches of workpieces with multiple specifications based on 3D vision. This method uses a 3D camera and a displacement device to perform 3D reconstruction of the workpiece, identifies point cloud information about points, lines, and surfaces, generates a machining path, and then modifies the machining path based on the identified kinematic model parameters, ultimately achieving automated machining of the workpiece. This method is widely applicable to automated machining of a variety of small batches of workpieces with multiple specifications, such as welding, gluing, and laser machining. It has broad market application prospects and is of great practical significance for enhancing the digitalization and intelligentization of my country's manufacturing industry.

[0033] 2. The present invention realizes the automatic processing of small-batch workpieces of multiple specifications, solves the pain points of related enterprises such as difficulty in recruiting workers and high labor costs, greatly improves the production efficiency of small-batch workpieces of multiple specifications, and at the same time provides a practical and effective solution to the automatic processing needs of small-batch workpieces of multiple specifications. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is an overall schematic diagram of the present invention;

[0035] Figure 2 It is the foot of the angle steel tower.

[0036] Figure 3 Flow chart of the method of the present invention.

[0037] In the figure: 1. Positioning device, 2. Tower foot of angle steel tower, 3. Base, 4. Industrial robot, 5. Welding power supply, 6. Steel cylinder, 7. Robot control cabinet, 8. 3D camera, 9. Camera bracket. DETAILED DESCRIPTION

[0038] The present invention will be further described below with reference to the accompanying drawings and examples.

[0039] The schematic diagram of the automatic workpiece processing equipment of the present invention is shown in the attached figure. Figure 1 As shown, the automatic workpiece processing equipment includes a positioning device 1, a base 3, an industrial robot 4, a welding power supply 5, a steel cylinder 6, a robot control cabinet 7, a 3D camera 8, a camera bracket 9, and a host computer system. The camera bracket 9 and base 3 are respectively positioned on the planes on either side of the positioning device 1. The 3D camera 8 is mounted on the camera bracket 9 and connected to the host computer system. The industrial robot 4 is mounted on the base 3. The welding power supply 5 is connected to the industrial robot 4. The steel cylinder 6, which contains welding shielding gas, is located on one side of the industrial robot 4. The tower foot 2 of the angle steel tower is mounted on the positioning device 1. The 3D camera 8 is used to photograph the angle steel tower, and the industrial robot 4 is used to operate on the angle steel tower. The host computer system processes the visual information captured by the 3D camera 8, generates a machining path code, and transmits it to the positioning device 1 and the industrial robot 4.

[0040] like Figure 3 As shown, the present invention includes the following steps:

[0041] 1) Calibrate the zero point and tool center point of the industrial robot 4, which in this embodiment is the tip of the welding wire of the welding gun. Establish a theoretical kinematic model (e.g., MDH model) of the industrial robot 4 and identify it to obtain the identified kinematic model;

[0042] 1) Specifically:

[0043] 1.1) Install a target ball at the end of the industrial robot 4 and establish a theoretical kinematic model of the industrial robot 4;

[0044] 1.2) After changing the actual rotation angles of each axis of the industrial robot 4, use a laser tracker to record the actual position information of the target ball relative to the base coordinate system of the industrial robot 4;

[0045] 1.3) Repeat step 1.2) 100 times to obtain 100 sets of actual rotation angles of each axis of the industrial robot 4 and the corresponding position information of the target ball relative to the base coordinate system of the industrial robot 4;

[0046] 1.4) Based on several groups of actual axis rotation angles of the industrial robot 4 and the corresponding actual position information of the target ball relative to the base coordinate system of the industrial robot 4, and the optimized identification function, the theoretical kinematic model of the industrial robot 4 is optimized using the least squares method to obtain the identified kinematic model, denoted as DH′.

[0047] The formula for optimizing the identification function is as follows:

[0048]

[0049] in, The actual rotation angle of each axis of the industrial robot 4 is θ i The position information of the target ball relative to the robot base coordinate system is calculated based on the theoretical kinematic model, p i is the actual position information of the target ball relative to the base coordinate system of the industrial robot 4.

[0050] 2) After hand-eye calibration of the industrial robot 4 and the 3D camera 8, the homogeneous transformation matrix of the 3D camera 8 relative to the base coordinate system of the industrial robot 4 is obtained Denoted as the camera-robot calibration matrix; after calibrating the industrial robot 4 and the displacement device 1, the homogeneous transformation matrix of the displacement device 1 relative to the base coordinate system of the industrial robot 4 is obtained Denoted as the displacement device-robot calibration matrix;

[0051] 3) If Figure 2As shown, a 3D camera 8 is used to collect visual information of the tower foot 2 of the angle steel tower clamped on the displacement device 1. The collected visual information of the tower foot 2 of the angle steel tower is 3D reconstructed according to the camera-robot calibration matrix and the displacement device-robot calibration matrix to obtain a 3D point cloud of the tower foot 2 of the angle steel tower;

[0052] 3) Specifically:

[0053] 3.1) Clamp the tower foot 2 of the angle steel tower onto the positioner 1 and record the initial posture of the positioner 1;

[0054] 3.2) Using the 3D camera 8 to collect an initial point cloud of the tower foot 2 of the angle steel tower at the current viewing angle, and preprocessing the initial point cloud of the tower foot 2 at the current viewing angle to obtain a preprocessed point cloud;

[0055] 3.3) Using the camera-robot calibration matrix, the pre-processed point cloud is transformed to obtain the point cloud in the base coordinate system of the industrial robot 4 at the current viewing angle;

[0056] 3.4) Control the rotation of the positioner 1 and record the posture of the positioner 1 so that the posture of the tower foot 2 of the angle steel tower is adjusted. Repeat 3.2)-3.3) to obtain point clouds in the base coordinate system of the industrial robot 4 at different viewing angles until point clouds in the base coordinate system of the industrial robot 4 are obtained at all viewing angles.

[0057] 3.5) Based on the positioner-robot calibration matrix, the point clouds of all positioner 1 postures and the corresponding viewpoints in the base coordinate system of industrial robot 4 are spliced ​​and transformed to obtain a 3D point cloud of the angle tower foot 2 under the positioner 1. Because the angle tower foot 2 is clamped to the positioner, the 3D point cloud of the angle tower foot 2 under the positioner 1 does not include the point cloud of the bottom surface of the angle tower foot 2.

[0058] 4) According to actual needs, determine the initial processing path of the tower foot 2 of the angle steel tower based on the three-dimensional point cloud of the tower foot 2 of the angle steel tower;

[0059] 4) Specifically:

[0060] Point, line and surface information is extracted from the three-dimensional point cloud of the tower foot 2 of the angle steel tower, and the processing position is determined according to the actual processing requirements, thereby obtaining the initial processing path of the tower foot 2 of the angle steel tower.

[0061] In this embodiment, the RANSAC algorithm is used to extract plane information from the three-dimensional point cloud of the tower foot 2 of the angle steel tower. Then, the intersection line is obtained based on the intersection information of the two planes. The intersection line information is used as the weld information of the tower foot 2 of the angle steel tower, and the initial welding path of the tower foot 2 of the angle steel tower is further generated.

[0062] 5) Correcting the initial welding path of the tower foot 2 of the angle steel tower according to the identified kinematic model to obtain a corrected welding path;

[0063] 5) Specifically:

[0064] According to the position of the initial processing path of the tower foot 2 of the angle steel tower, the actual axis rotation angle values ​​of the industrial robot 4 in the initial processing path are calculated using the identified kinematic model. The actual axis rotation angle values ​​of the industrial robot 4 in the initial processing path are used to replace the corresponding original axis rotation angle values ​​in the initial processing path and the processing path is updated to obtain the corrected processing path.

[0065] In this embodiment, according to the positions of the arc starting point and the arc ending point in the initial welding path of the tower foot 2 of the angle steel tower, the actual axis rotation angle values ​​of the industrial robot 4 at the arc starting point and the arc ending point are calculated using the identified kinematic model, and the actual axis rotation angle values ​​of the industrial robot 4 at the arc starting point and the arc ending point are used to replace the corresponding original axis rotation angle values ​​at the arc starting point and the arc ending point, and the welding path is updated to obtain a corrected welding path.

[0066] 6) The corrected processing path is input into the robot control system via the TCP / IP protocol or the UDP protocol, and the robot control system controls the industrial robot 4 to perform welding operations on the tower foot 2 of the angle steel tower.

Claims

1. A method for automatic processing of multi-specification workpieces based on three-dimensional vision, characterized in that: The following steps are involved: 1) Establish and identify the theoretical kinematic model of the industrial robot to obtain the identified kinematic model; 2) After hand-eye calibration of the industrial robot and 3D camera, the homogeneous transformation matrix of the 3D camera relative to the industrial robot base coordinate system is obtained. Denoted as the camera-robot calibration matrix; after calibrating the industrial robot and the displacement device, the homogeneous transformation matrix of the displacement device relative to the industrial robot base coordinate system is obtained Denoted as the displacement device-robot calibration matrix; 3) Using a 3D camera to collect visual information of the workpiece to be processed, which is clamped on the positioner, and performing 3D reconstruction on the collected visual information of the workpiece to be processed based on the camera-robot calibration matrix and the positioner-robot calibration matrix to obtain a 3D point cloud of the workpiece to be processed; 4) determining an initial processing path of the workpiece to be processed based on the three-dimensional point cloud of the workpiece to be processed; 5) Correcting the initial machining path of the workpiece to be machined according to the identified kinematic model to obtain a corrected machining path; 6) The corrected processing path is input into the robot control system, and the robot control system controls the industrial robot to process the workpiece.

2. The method for automatic processing of multi-specification workpieces based on three-dimensional vision according to claim 1, characterized in that: The specific embodiment of 1) is: 1.1) Install a target ball at the end of the industrial robot and establish a theoretical kinematic model of the industrial robot; 1.2) After changing the actual rotation angles of each axis of the industrial robot, use a laser tracker to record the actual position information of the target ball relative to the coordinate system of the industrial robot base; 1.3) Repeat step 1.2) several times to obtain the actual rotation angles of each axis of several sets of industrial robots and the corresponding position information of the target sphere relative to the industrial robot base coordinate system; 1.4) Based on the actual axis rotation angles of several groups of industrial robots and the actual position information of the corresponding target balls relative to the industrial robot base coordinate system, and the optimized identification function, the theoretical kinematic model of the industrial robots is optimized using the least squares method to obtain the identified kinematic model, denoted as DH′.

3. The method for automatic processing of multi-specification workpieces based on three-dimensional vision according to claim 2, characterized in that: The formula of the optimized identification function is as follows: in, The actual rotation angle of each axis of the industrial robot is θ i The position information of the target ball relative to the robot base coordinate system is calculated based on the theoretical kinematic model, p i is the actual position information of the target ball relative to the industrial robot base coordinate system.

4. The method for automatic processing of multi-specification workpieces based on three-dimensional vision according to claim 1, characterized in that: Said 3) is specifically: 3.1) Clamp the workpiece to be processed on the positioner and record the initial posture of the positioner; 3.2) using a 3D camera to collect an initial point cloud of the workpiece to be processed at the current viewing angle, and preprocessing the initial point cloud of the workpiece to be processed at the current viewing angle to obtain a preprocessed point cloud; 3.3) Use the camera-robot calibration matrix to transform the preprocessed point cloud to obtain the point cloud in the base coordinate system of the industrial robot at the current viewing angle; 3.4) Control the rotation of the positioner and record the positioner's posture so that the posture of the workpiece to be processed is adjusted. Repeat 3.2)-3.3) to obtain point clouds in the base coordinate system of the industrial robot at different viewing angles until point clouds in the base coordinate system of the industrial robot are obtained at all viewing angles. 3.5) Based on the positioner-robot calibration matrix, the point clouds of all positioner postures and the industrial robot's base coordinate system at the corresponding perspectives are spliced ​​and coordinate-converted to obtain a 3D point cloud of the workpiece to be processed under the positioner.

5. The method for automatic processing of multi-specification workpieces based on three-dimensional vision according to claim 1, characterized in that: The specific embodiment of 4) is: Extract point, line and surface information from the three-dimensional point cloud of the workpiece to be processed, determine the processing position according to the actual processing requirements, and then obtain the initial processing path of the workpiece to be processed.

6. The method for automatic processing of multi-specification workpieces based on three-dimensional vision according to claim 1, characterized in that: Said 5) is specifically: According to the position of the initial machining path of the workpiece to be machined, the actual axis rotation angle values ​​of the industrial robot in the initial machining path are calculated using the identified kinematic model. The actual axis rotation angle values ​​of the industrial robot in the initial machining path are used to replace the corresponding original axis rotation angle values ​​in the initial machining path and the machining path is updated to obtain the corrected machining path.

7. The method for automatically processing multi-specification workpieces based on three-dimensional vision according to claim 1, characterized in that: In the above 6), the corrected processing path is sent to the robot control system via TCP / IP protocol or UDP protocol.

Images