CN-122007738-A - Visual intelligent method for planning complex welding track

CN122007738ACN 122007738 ACN122007738 ACN 122007738ACN-122007738-A

Abstract

The invention relates to the technical field of industrial robot vision and discloses a vision intelligent method for planning a complex welding track, which comprises the steps of obtaining a real-time electric characteristic signal of an interference source of an operation environment, identifying a periodic phase of low radiation intensity of the interference source through derivation, controlling a vision sensor to collect three-dimensional characteristic point coordinates of an operation target area and synchronously latching a position and a posture of an operation unit in a sampling window determined by the periodic phase, converting a collection result into a space semantic coordinate system by utilizing a coordinate conversion model, retrieving isomorphic nodes in a logic topology database, extracting ideal tangential vectors and normal constraints, and reconstructing expected coordinates of a current track according to the obtained coordinates.

Inventors

YUE XIANGYU
AI JIAWEI
LI MING

Assignees

江西拓智自动化科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260325

Claims (9)

1. A visual intelligence method for complex welding trajectory planning, comprising the steps of: Step 101, acquiring a real-time electric characteristic signal representing physical characteristics of an interference source of an operation environment, generating a first derivative by performing first derivative operation on the real-time electric characteristic signal, and determining a periodic phase representing that the interference source is in low radiation intensity based on the first derivative meeting a preset signal jump criterion; 102, outputting an integral exposure pulse signal in a sampling window with a determined period phase, and synchronously latching real-time pose feedback data of an operation unit to construct a reference pose matrix with a time stamp attribute; Step 103, based on the reference pose matrix, mapping the three-dimensional feature point coordinates to a space logic coordinate system by utilizing a coordinate conversion model to generate an original position vector used for representing initial position information of a workpiece; And 104, carrying out track smoothing extrapolation on the original position vector according to the ideal tangential vector and normal constraint data by combining the control hysteresis cycle of the system, and reconstructing the logic expected coordinates of the current track.
2. The visual intelligent method for complex welding trajectory planning according to claim 1, wherein determining the periodic phase indicative of the interference source being at a low radiation intensity in step 101 comprises monitoring a sequence of magnitudes of real-time electrical signature signals, determining that the interference source enters a physical signature blanking period in a state in which the real-time magnitudes in the sequence of magnitudes drop to a preset energy transition threshold and the absolute value of a first derivative exceeds a preset phase jump criterion, and taking the starting time of the physical signature blanking period as a trigger zero of the periodic phase so that a sampling period of the visual sensor covers a low background light intensity window of the interference source.
3. The visual intelligent method for planning a complex welding track according to claim 1, wherein the parameterized logic topology database comprises a pre-imported controlled object model, logically parameterized target path data and associated local curvature distribution characteristics, and the logic parameter nodes are used for defining ideal motion state parameters at feature points in a target path space topology.
4. The visual intelligent method for planning complex welding tracks according to claim 1 is characterized by comprising the following steps of extracting local point cloud density distribution around three-dimensional feature point coordinates in a step 401, and implementing neighborhood retrieval in a parameterized logic topology database in a working condition that the local point cloud density is lower than a preset distribution threshold in a step 402, and implementing geometric completion processing on the current discrete three-dimensional feature point coordinates by utilizing geometric constraint relations of adjacent logic parameter nodes.
5. The visual intelligent method for complex welding track planning according to claim 1, wherein the coordinate transformation model is synthesized in real time by hand-eye calibration parameters of the visual sensor, the reference pose matrix and the kinematic equations of the working unit.
6. The visual intelligent method for complex welding trajectory planning according to claim 1, comprising the steps of monitoring the Euclidean distance deviation between the coordinates of the three-dimensional feature points and the corresponding nodes in the parameterized logical topology database, step 701, outputting a path deviation warning signal and stopping the current movement of the work unit when the Euclidean distance deviation continuously exceeds a preset safety boundary value, step 702.
7. A visual intelligence method for complex welding trajectory planning as claimed in claim 1, characterized in that the ideal tangential vector is used to define the direction of the motion vector of the work unit at the current spatial position, and the normal constraint data is used to define the geometrical pose constraint of the work unit with respect to the controlled object surface.
8. The visual intelligent method for planning complex welding trajectories according to claim 1, wherein the visual sensor comprises an industrial camera with an external hardware trigger interface and a narrow-band filter assembly, and the transmission center wavelength of the narrow-band filter assembly and the characteristic radiation spectrum main peak of the interference source are staggered.
9. A visual intelligence method for complex welding trajectory planning as claimed in claim 1, wherein the logically desired coordinates are used to generate control commands to be sent to a motion controller to drive the work cell to perform continuous trajectory tracking motions along paths pointed by ideal tangential vectors.

Description

Visual intelligent method for planning complex welding track Technical Field The invention relates to a visual intelligent method for planning a complex welding track, and belongs to the technical field of industrial robot vision. Background The current industrial robot adopts a visual guidance technology to realize welding automation in the field of heavy machinery manufacturing, a mainstream technical scheme utilizes a visual sensor to acquire a welding line image, geometric features are extracted to plan a motion track of an end effector, the mode can ensure the stability of a welding process under a standard working condition, besides the fact that perception hardware is easy to be interfered, software control logic also has defects, for example, china patent with an authorized bulletin number of CN111230869B discloses a collaborative planning method for a complex space curve welding line motion track and a welding process, a welding area is divided to construct a process knowledge base, path self-adaption is realized on the whole planning level, under a heavy current welding working condition, bottom logic depends on global visual measurement and does not touch an arc discharge physical characteristic surface phase, a perception window is interfered by strong radiation, and sampling certainty is insufficient. In order to cope with the deviation, the prior art adopts a Kalman filtering and other time sequence prediction method, the method carries out pose extrapolation towards future time by depending on the local curvature of a historical track point, and when a variable curvature path such as a space tube plate intersecting line is processed, the prior guiding of global geometric information of a workpiece is lacked, the predicted coordinate generated by time sequence prediction generates chord interception deviation with a real physical track, and the accumulated error in the track tracking process is increased. Therefore, how to realize real-time planning and accurate compensation of complex space tracks through physical phase association and geometric topology retrieval becomes the technical problem to be solved by the invention. Disclosure of Invention In order to solve the problems in the background technology, the technical scheme of the invention is as follows, a visual intelligent method for complex welding track planning, which comprises the following steps: Step 101, acquiring a real-time electric characteristic signal representing physical characteristics of an interference source of an operation environment, generating a first derivative by performing first derivative operation on the real-time electric characteristic signal, and determining a periodic phase representing that the interference source is in low radiation intensity based on the first derivative meeting a preset signal jump criterion; 102, outputting an integral exposure pulse signal in a sampling window with a determined period phase, and synchronously latching real-time pose feedback data of an operation unit to construct a reference pose matrix with a time stamp attribute; Step 103, based on the reference pose matrix, mapping the three-dimensional feature point coordinates to a space logic coordinate system by utilizing a coordinate conversion model to generate an original position vector used for representing initial position information of a workpiece; And 104, carrying out track smoothing extrapolation on the original position vector according to the ideal tangential vector and normal constraint data by combining the control hysteresis cycle of the system, and reconstructing the logic expected coordinates of the current track. Preferably, in step 101, determining the periodic phase representing that the interference source is at low radiation intensity includes monitoring an amplitude sequence of the real-time electrical characteristic signal, determining that the interference source enters a physical characteristic extinction period when the real-time amplitude in the amplitude sequence is reduced to a preset energy transition threshold and an absolute value of a first derivative exceeds a preset phase mutation criterion, and taking a starting time of the physical characteristic extinction period as a trigger zero point of the periodic phase so that a sampling period of the vision sensor covers a low background light intensity window of the interference source. Preferably, the parameterized logic topology database comprises a pre-imported controlled object model, target path data after logic parameterization and associated local curvature distribution characteristics, and logic parameter nodes are used for defining ideal motion state parameters at all characteristic points in a target path space topology structure. Preferably, the method comprises the steps of extracting local point cloud density distribution around three-dimensional feature point coordinates in a step 401, and implementing neighborhood retrieval in a p