Three-dimensional laser vision automated assembly and welding system

Abstract

The present application relates to the field of automatic welding technology, in particular to a three-dimensional laser vision automatic assembly and welding system, the system comprises a point cloud area analysis module, a density anomaly compensation module, a posture offset detection module, a clamping trajectory linkage module and a path closed loop calibration module. Through global space capture of high-density point cloud data, combined with partition feature density analysis and automatic anomaly compensation, dynamic identification of workpiece space structure parameters is realized, multi-region space posture monitoring is adopted, and the cooperative correction mechanism between clamping action and trajectory path is matched, realizing multi-level closed loop adjustment in the whole assembly process. In the face of complex structural parts, deformed parts and multi-clamping path interference working conditions, error detection, parameter linkage and adaptive adjustment can be carried out synchronously, further realizing the depth cooperation of space features, trajectory planning and assembly path, promoting the improvement of automatic assembly precision and consistency, and improving the flexibility and adaptability of the system.

Description

Technical Field

[0001] This invention relates to the field of automated welding technology, and in particular to an automated assembly and welding system based on three-dimensional laser vision. Background Technology

[0002] Automated welding falls under the category of mechanical manufacturing and automation. It encompasses the automated control of the welding process, automatic planning of welding paths, automatic setting of welding parameters, workpiece positioning, and intelligent identification and adjustment of clamping and assembly. It is widely used in industries such as automotive manufacturing, shipbuilding, rail transportation, and heavy machinery. Traditional automated welding systems utilize robotic arms or welding equipment to weld workpieces using preset trajectories and welding parameters. They typically employ two-dimensional vision sensors for workpiece position detection, linear guides for torch movement, standard fixtures for workpiece clamping, and conventional arc welding or laser welding for the welding operation.

[0003] Existing systems are limited by two-dimensional sensing methods and can only collect workpiece position information on a single plane. They lack the ability to analyze multi-dimensional data of complex spatial structures. Under conditions such as irregular workpieces, multi-area clamping, and irregular trajectory, problems such as missing spatial features, slow path response, and accumulation of assembly errors are likely to occur. The lack of multi-level dynamic parameter adjustment and real-time compensation links leads to unstable positioning and insufficient path adaptability, which affects production efficiency and assembly accuracy. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the existing technology and propose an automated assembly and welding system based on three-dimensional laser vision.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: a three-dimensional laser vision-based automated assembly and welding system, the system comprising:

[0006] The point cloud region analysis module is used to collect the point cloud coordinates of the workpiece based on a 3D laser vision sensor, compare the number and volume of feature points according to structural segments, optimize the point cloud data structure, and obtain spatial partition density features.

[0007] The density anomaly compensation module is used to analyze the spatial distance changes of adjacent nodes based on the spatial partition density characteristics, compare the point cloud distribution of the abnormal area and the normal area, adjust the laser head scanning angle and step range, and obtain the point cloud coverage balance index.

[0008] The attitude offset detection module is used to compare the point cloud coordinates with the target assembly attitude points based on the point cloud coverage equalization index, analyze the differences in node spatial coordinates, determine the sections where the welding reference positioning block has shifted, and obtain the spatial offset distribution characteristics.

[0009] The clamping trajectory linkage module is used to analyze the correspondence between offset nodes and fixture distribution based on the spatial offset distribution characteristics, determine the influence of clamping state changes on workpiece posture, adjust fixture output parameters and trajectory commands, and obtain trajectory clamping linkage parameters.

[0010] The path closed-loop calibration module is used to compare the spatial differences between path nodes and target trajectory based on the trajectory clamping linkage parameters, identify nodes that do not meet the standards, determine whether the calibrated path segments meet the pairing requirements, and obtain the path pairing calibration results.

[0011] The present invention improves upon this invention by including the following: the spatial partition density features include density distribution uniformity, partition filling integrity, and boundary feature stability; the point cloud coverage balance index includes compensation area continuity and overall coverage consistency; the spatial offset distribution features include the number of offset areas, offset trend direction, and maximum offset of key nodes; the trajectory clamping linkage parameters include clamp response sensitivity, trajectory adjustment amplitude, and linkage synchronization; and the path pair calibration results include node calibration ratio, path compliance, and error convergence degree.

[0012] The present invention is improved in that the point cloud region parsing module includes:

[0013] The data stream receiving submodule is used to analyze the collected point cloud coordinate data of the workpiece surface based on the 3D laser vision sensor, optimize the classification labels of spatial point coordinates and reflection intensity, filter invalid points caused by positioning errors in the annotation, and adjust the spatial arrangement and label relationship of the effective coordinate group to obtain the spatial structure annotation dataset.

[0014] The spatial segment construction submodule is used to determine the spatial boundaries of each label coordinate group based on the spatial structure annotation dataset, compare the boundary morphology of the label corresponding region, analyze the uniformity of the distribution of point groups within the spatial range, optimize the sub-segment division logic and boundary coherence, and obtain the segment spatial distribution sequence.

[0015] The density discrimination calculation submodule is used to calculate the spatial range of each sub-segment based on the spatial distribution sequence of the segment, filter feature points with spatial variation, analyze the degree of distribution concentration of feature points in each region, compare the spatial coverage ratio of feature points in each segment, judge the distribution differences, and obtain the spatial partition density features.

[0016] The present invention is improved in that the density anomaly compensation module includes:

[0017] The density continuity judgment submodule is used to determine the continuity of density change between sub-segments based on the density characteristics of the spatial partition, compare the density difference distribution of adjacent segments, screen the spatial locations with density jumps, determine density breaks and distribution void areas, and obtain a spatial discontinuity positioning sequence.

[0018] The point cloud characteristic comparison submodule is used to analyze the point cloud coordinates and distances between points in abnormal sections based on the spatial discontinuous positioning sequence, compare the spatial distribution patterns of normal sections, filter the differences in point cloud distribution characteristics between abnormal and normal sections, determine the range of spatial coverage changes, and obtain point cloud distribution difference feature groups.

[0019] The scanning parameter control submodule is used to adjust the scanning angle and scanning path of the laser head according to the point cloud distribution difference feature group, optimize the step interval setting, calibrate the spatial movement sequence of the laser head, and regenerate the scanning command covering the sparse area to obtain the point cloud coverage balance index.

[0020] The present invention is improved in that the attitude offset detection module includes:

[0021] The attitude coordinate comparison submodule is used to compare the spatial coordinates of each node with the target assembly attitude coordinates based on the point cloud coverage equalization index, calculate the coordinate differences in the three-dimensional direction, determine the spatial offset trend of each node, optimize the node position distribution, and obtain the node three-dimensional offset vector set.

[0022] The spatial offset recognition submodule is used to determine the distribution of the three-dimensional offset amplitude of the node in the spatial segment based on the node's three-dimensional offset vector set, analyze the continuous offset changes in the node sequence, compare the spatial characteristics of the node with the assembly reference segment, mark the associated spatial segments, and obtain a continuous identification group of the offset region.

[0023] The offset segment filtering submodule is used to calculate the node statistical characteristics of the main offset direction of the segment based on the continuous identifier group of the offset region, determine the aggregation trend of node offset changes, optimize the consistency of the main direction data, and obtain the spatial offset distribution characteristics.

[0024] The present invention is improved in that the clamping trajectory linkage module includes:

[0025] The spatial mapping analysis submodule is used to determine the spatial correspondence between offset nodes and fixture distribution based on the spatial offset distribution characteristics, compare the three-dimensional arrangement between nodes and fixtures, identify areas with overlapping or abnormal coverage, and adjust the corresponding node and fixture distribution parameters to obtain the fixture spatial correspondence parameters.

[0026] The attitude influence assessment submodule is used to analyze the force changes of the clamping state of the fixture under the distribution of each node based on the corresponding parameters of the fixture space, compare the correlation between the changes in clamping parameters and the spatial attitude of the workpiece, and determine the attitude response changes within the path disturbance range to obtain the attitude disturbance response parameters.

[0027] The linkage parameter generation submodule is used to optimize the force direction and clamping sequence of the fixture based on the attitude disturbance response parameters, adjust the synchronization mode of the trajectory control command, calibrate the triggering conditions of the linkage between the clamping action and the path command, and obtain the trajectory clamping linkage parameters.

[0028] The present invention is improved in that the path closed-loop calibration module includes:

[0029] The distance difference calculation submodule is used to analyze the difference between path nodes and target trajectory nodes in three-dimensional spatial coordinates based on the trajectory clamping linkage parameters, calculate the direction and magnitude of the spatial coordinate difference between nodes, compare the distribution changes of all nodes in space, determine the spatial consistency between the path and the target trajectory, and obtain the node spatial offset feature group.

[0030] The path node filtering submodule is used to filter nodes with continuous changes based on the node spatial offset feature group, analyze the node distribution in the trajectory direction, optimize the node grouping criteria, and determine the abnormal distribution of nodes in the trajectory to obtain the abnormal distribution set of path nodes.

[0031] The path adjustment judgment submodule is used to determine the spatial distribution pattern of abnormal nodes based on the abnormal distribution set of path nodes, analyze the distribution relationship between abnormal nodes and normal nodes, compare the coherence of node distribution, optimize the path adjustment process, and obtain the path pair calibration result.

[0032] The present invention is improved in that the workpiece point cloud coordinate data refers to the set of three-dimensional X, Y, Z coordinates of each point in space obtained after the three-dimensional laser vision sensor scans the workpiece surface. The spatial distance change refers to the distance change between adjacent feature points in three-dimensional space. Abnormalities reflect uneven scanning coverage or abnormalities on the workpiece surface. The correspondence refers to the mapping relationship between the point cloud offset nodes and the fixture in the spatial layout.

[0033] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0034] In this invention, global spatial capture of high-density point cloud data is combined with partitioned feature density analysis and automatic anomaly compensation to achieve dynamic identification of workpiece spatial structural parameters. Multi-region spatial attitude monitoring is adopted, along with a collaborative correction mechanism between clamping actions and trajectory paths, to achieve multi-level closed-loop adjustment throughout the assembly process. In the face of complex structural parts, deformable parts, and interference from multiple clamping paths, error detection, parameter linkage, and adaptive adjustment can be performed simultaneously, further realizing deep collaboration between spatial features, trajectory planning, and assembly paths, promoting the improvement of automated assembly accuracy and consistency, and enhancing the system's flexibility and adaptability. Attached Figure Description

[0035] Figure 1This is a system flowchart of the present invention;

[0036] Figure 2 This is a flowchart of the point cloud region parsing module in this invention;

[0037] Figure 3 This is a flowchart of the density anomaly compensation module in this invention;

[0038] Figure 4 This is a flowchart of the attitude offset detection module in this invention;

[0039] Figure 5 This is a flowchart of the clamping trajectory linkage module in this invention;

[0040] Figure 6 This is a flowchart of the path closed-loop calibration module in this invention. Detailed Implementation

[0041] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0042] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Example 1

[0043] Please see Figure 1 This invention provides a technical solution: a three-dimensional laser vision-based automated assembly and welding system comprising:

[0044] The point cloud region analysis module is used to analyze the collected workpiece point cloud coordinate data based on a 3D laser vision sensor, compare the number and volume of feature points in each structural segment one by one, calculate the feature point density of each segment, determine whether the density distribution pattern of the spatial partition meets the pair recognition requirements, and optimize the data structure and spatial partitioning data of the point cloud information to obtain the spatial partition density features.

[0045] The density anomaly compensation module is used to determine whether there is discontinuity in the density distribution of each segment based on the spatial partition density characteristics, analyze the changes in the spatial distance between adjacent nodes in each segment, compare the point cloud distribution characteristics of the abnormal area and the normal area, adjust the scanning angle parameters of the laser head, and reset the step range to obtain the point cloud coverage balance index.

[0046] The attitude offset detection module is used to compare the position of the point cloud coordinates with the corresponding point of the target assembly attitude based on the point cloud coverage equalization index, analyze the difference between the actual spatial coordinates of the node and the target spatial coordinates, determine the segment where the corresponding point of the welding reference positioning block has spatial offset, identify the area where the offset amplitude is higher than the preset offset threshold, and obtain the spatial offset distribution characteristics.

[0047] The clamping trajectory linkage module is used to analyze the spatial correspondence between offset nodes and the current fixture distribution based on the spatial offset distribution characteristics, determine the impact of clamping state changes on the workpiece spatial posture, compare the changes in clamping data with the trajectory disturbance range, adjust the output parameters of the fixture and the trajectory planning instructions, optimize the linkage configuration of clamping action and path control, and obtain the trajectory clamping linkage parameters.

[0048] The path closed-loop calibration module is used to compare the spatial distance difference between path nodes and target trajectories based on trajectory clamping linkage parameters, identify nodes that do not meet the standards, statistically analyze the distribution characteristics of various types of nodes, determine whether the calibrated path segments can meet the pairing requirements, summarize qualified nodes and path adjustment information, and obtain path pairing calibration results.

[0049] Spatial partition density characteristics include density distribution uniformity, partition filling integrity, and boundary feature stability; point cloud coverage balance indicators include compensation area continuity and overall coverage consistency; spatial offset distribution characteristics include the number of offset areas, offset trend direction, and maximum offset of key nodes; trajectory clamping linkage parameters include clamp response sensitivity, trajectory adjustment amplitude, and linkage synchronization; path pair calibration results include node calibration ratio, path compliance, and error convergence degree.

[0050] In the point cloud region analysis module, the workpiece point cloud coordinate data refers to the set of three-dimensional (X, Y, Z) coordinates of each point in space obtained after the 3D laser vision sensor scans the workpiece surface; it is a digital representation of the workpiece surface structure. Each structural segment refers to each local region of the workpiece after spatial division according to preset structural or functional partitions, used for refined analysis of the point cloud characteristics of different parts. Individual comparison refers to comparing and calculating data for each structural segment separately, rather than processing the entire structure, highlighting the analytical granularity of each unit region. Feature point density refers to the ratio of the number of feature points (points reflecting significant geometric and morphological changes) in the point cloud to the volume of that segment within each structural segment; it measures the density of the point cloud distribution. Key parameters include: the density distribution pattern of spatial partitions, which refers to the spatial variation and distribution pattern of feature point density within each segment, used to determine the integrity and validity of point cloud data; pairing recognition requirements, which refer to the recognition requirements for the uniformity and integrity of point cloud data spatial distribution during automated pairing processes, to meet the accuracy of subsequent posture detection, trajectory planning, and other processes; the data structure of point cloud information, which refers to the organization method (such as octree structure, KD tree, etc.) used to store and retrieve the collected point cloud data in order to improve computing efficiency and data processing speed; and spatial partitioning data, which refers to the structured data obtained by labeling or numbering the overall point cloud data of the workpiece according to preset spatial regions, used for subsequent regional feature analysis and parameter retrieval.

[0051] In the density anomaly compensation module, discontinuity refers to the phenomenon that feature points are sparse, broken, or missing in a certain partition or local area of ​​the point cloud density distribution, affecting the continuity and integrity of the overall point cloud data; spatial distance change refers to the change in distance between adjacent feature points in three-dimensional space, and anomalies reflect uneven scanning coverage or abnormal workpiece surface; anomaly area refers to local areas found to have density discontinuity, sparse point cloud, or abnormal distribution after analysis, which are the key areas for subsequent compensation scanning; normal area refers to workpiece areas with uniform point cloud distribution and density that meets recognition requirements, which can be used as a reference benchmark for anomaly area compensation comparison; point cloud distribution characteristics refer to the spatial distribution pattern of point cloud in a certain area, such as dense, sparse, uniform, etc., which directly affect the recognition and compensation effect; laser head refers to the component in the three-dimensional laser vision sensor used to emit and receive laser beams for spatial scanning, and is the execution end of the scanning action; scanning angle parameters refer to parameters such as the pitch angle and azimuth angle of the laser head when performing spatial scanning, which determine the spatial coverage range of point cloud data acquisition; step range refers to the distance moved by the laser head or scanning mechanism in each step when moving in space, used to adjust scanning accuracy and resolution.

[0052] In the attitude deviation detection module, the target assembly attitude corresponding point refers to the spatial coordinates of each key point (reference point, assembly point, etc.) in the target attitude model defined by the system in the process design stage, which is the reference data for judging the actual assembly accuracy; the actual spatial coordinates refer to the spatial point positions obtained by the three-dimensional laser vision sensor and combined with the actual state of the current workpiece; the target spatial coordinates refer to the theoretical spatial positions of each reference point or feature point in the target assembly attitude model, which is the benchmark for error analysis; the welding reference positioning block refers to the structural reference component used for spatial positioning during the workpiece assembly and welding process, which facilitates attitude detection and alignment; the spatial deviation segment refers to the workpiece area where significant positional deviation is found by comparing the actual point cloud with the target model, which needs further correction.

[0053] In the clamping trajectory linkage module, spatial correspondence refers to the one-to-one mapping relationship between the actual point cloud offset nodes and the fixture in spatial layout, which facilitates the analysis of the impact of the clamping process on the workpiece posture; clamping state change refers to the force state exerted on the workpiece by the fixture during the clamping process and its dynamic changes in the assembly process; workpiece spatial posture refers to the actual position and orientation of the workpiece in three-dimensional space, which is the core parameter for assembly accuracy and subsequent welding path planning; change and trajectory disturbance range refers to the range of influence on the predetermined welding trajectory when the clamping data changes with time or space; output parameters refer to control parameters such as the magnitude of the clamping force applied by the fixture to the workpiece; trajectory planning instructions refer to the path control parameters generated and issued by the automated control system to the motion actuator, guiding the end effector to complete the welding or assembly action; linkage configuration refers to the joint adjustment and control settings between the fixture control parameters and the path motion parameters to achieve coordinated operation between the two.

[0054] In the path closed-loop calibration module, spatial distance difference refers to the difference in distance between the actual acquired path nodes and the target trajectory nodes in three-dimensional spatial coordinates, which is the core basis for judging path accuracy; unqualified nodes refer to path nodes whose deviation exceeds the acceptable range after spatial distance comparison, and need further correction; the distribution characteristics of various nodes refer to the spatial distribution pattern of different path nodes (qualified, unqualified, etc.), used to analyze the overall calibration effect; calibrated path segments refer to the path parts that have passed error calibration and meet the process accuracy requirements; assembly requirements refer to the comprehensive requirements standards of automated assembly process for spatial positioning, assembly accuracy, and welding trajectory; qualified nodes refer to path nodes that have met the spatial error tolerance requirements after calibration; path adjustment information: records the parameter changes and control commands generated during error calibration and trajectory correction, used for subsequent data traceability and optimization.

[0055] Please see Figure 2 The point cloud region parsing module includes:

[0056] The data stream receiving submodule is used to analyze the collected point cloud coordinate data of the workpiece surface based on the 3D laser vision sensor, optimize the classification labels of spatial point coordinates and reflection intensity, filter invalid points caused by positioning errors in the annotation, and adjust the spatial arrangement and label relationship of the effective coordinate group to obtain the spatial structure annotation dataset.

[0057] The workpiece surface is scanned using a 3D laser vision sensor to acquire initial point cloud data. The 3D coordinates and reflection intensity values ​​within the scan results are read point by point. First, points with reflection intensity values ​​between 20 and 60 are grouped into label group L1, those between 61 and 140 into label group L2, and those exceeding 140 into label group L3. Then, the spatial positions of each point and its three nearest neighbors are extracted within each label group, and the average distance between points is calculated. Points with an average distance exceeding 5.0 mm are considered invalid and directly discarded. Simultaneously, fluctuation detection is performed on the reflection intensity data of points within each label group. If the fluctuation range of the reflection value is large... Points with an intensity of 35 or less are marked as abnormal and removed. After removing invalid points, the remaining valid points are numbered and updated, and then spatially rearranged according to their Z-axis coordinate values ​​from smallest to largest. This ensures that the point sequence within each label group reflects the height and hierarchy of the workpiece structure. After the arrangement is complete, the center coordinate values ​​of each label group are recalculated and compared with the original center positions. When the offset is greater than 3 mm, a local reclassification process is triggered, and the distribution detection and intensity filtering operations are performed on the local point set again until the center coordinate offset of all labels is less than 3 mm. This ensures the spatial structure continuity and intensity consistency within each label group, while providing an accurate spatial structure annotation dataset for subsequent processing.

[0058] The spatial segment construction submodule is used to determine the spatial boundaries of each label coordinate group based on the spatial structure labeled dataset, compare the boundary morphology of the corresponding label regions, analyze the uniformity of point group distribution within the spatial range, optimize the sub-segment division logic and boundary coherence, and obtain the segment spatial distribution sequence.

[0059] For each label group, a 3D boundary range is constructed using the valid point data. The boundary size is determined based on the difference between the maximum and minimum coordinate points. If the size exceeds 300mm in a certain direction, it is defined as the primary dividing direction. During the boundary analysis, if abrupt changes occur on the boundary surface in a certain area, causing the overall boundary shape to change beyond the specified standard, interpolation is used to fill in the missing points and reconstruct the boundary shape, maintaining the continuity of the boundary contour. The boundary range is then divided into spatial units with a side length of 2mm. The number of points contained in each unit is counted. When the maximum difference in the number of points between units within a region exceeds 80... When identifying points, if the distribution of points in the area is uneven, a subdivision logic is initiated, dividing the original label segment into 2 to 5 sub-segments in the dense direction. The segmentation threshold range is determined based on the actual number of points. After subdivision, the coordinate jump value at the boundary between two adjacent sub-segments is analyzed. When the jump value exceeds 4mm, a fine-tuning operation is performed on the boundary position to reduce the jump value to within 3mm, ensuring smooth connection between the boundaries of the sub-segments. Based on the adjusted boundaries and the re-division results, a complete spatial distribution sequence of segments is output. Each sub-segment has a unique number for subsequent density calculation and recognition processes.

[0060] The density discrimination calculation submodule is used to calculate the spatial range of each sub-segment based on the spatial distribution sequence of the segment, filter feature points with spatial variation, analyze the distribution concentration of feature points in each region, compare the spatial coverage ratio of feature points in each segment, judge the distribution differences, and obtain the spatial partition density features.

[0061] The boundary dimensions of each sub-segment are read from the spatial distribution sequence, and the number of valid points contained therein is counted. The spatial density value of each sub-segment is analyzed based on the calculation results. When the density value is less than 0.02 points per cubic millimeter, it is initially identified as a sparse region. Then, feature points within each segment are identified, and areas of change are selected by analyzing curvature variations. When the proportion of feature points in a sub-segment to the total number of valid points exceeds 30%, it indicates that the segment has strong structural features. Subsequently, spatial clustering is detected in the point cloud region. The sub-segment is divided into a 3D grid with a side length of 3mm. When the number of feature points in a certain grid cell exceeds 45% of the total number of feature points in that sub-segment, it is considered that the sub-segment is sparse. Feature points are concentrated within a segment, or if they are scattered, they are considered to be evenly distributed. Then, the actual distribution range of feature points on the boundary of the sub-segment is measured. When the coverage area is less than 60% of the entire sub-segment boundary, the area is considered to have incomplete coverage. Then, the density value changes between any two sub-segments are compared. When the difference between the two exceeds 40%, it is recorded as a density abrupt boundary region. The above calculation results are combined and statistically analyzed to summarize all sub-segments with low feature point density, severe feature point concentration, insufficient boundary coverage, or abrupt density changes. The spatial partition density feature results with structural characteristics are output for subsequent attitude detection and trajectory optimization.

[0062] Please see Figure 3 The density anomaly compensation module includes:

[0063] The density continuity judgment submodule is used to determine the continuity of density changes between sub-segments based on the density characteristics of spatial partitions, compare the density difference distribution of adjacent segments, screen spatial locations with density jumps, identify density breaks and distribution void areas, and obtain spatial discontinuity positioning sequences.

[0064] Based on the density values ​​of each sub-segment in the spatial partition density features, the density data of adjacent sub-segments are first read sequentially from left to right according to their numbers. The density value of each sub-segment is calculated by the ratio of the number of feature points it contains to the spatial volume of the region. Then, the density values ​​of any two adjacent sub-segments are calculated one-to-one. When the difference exceeds 40% of the average density value of the current workpiece sample, it is marked as a density mutation point. Then, a difference distribution curve is established according to the scanning path direction of the entire workpiece structure. The density jump trend of three or five consecutive points is analyzed by sliding window method. If there is a jump phenomenon in two or more consecutive segments, it is judged as a density discontinuity area. At the same time, the distance between adjacent jump points is calculated. When the distance is less than 20mm, it is aggregated into a single density fracture zone. Otherwise, it is divided into multiple fracture zones. Then, the empty point intervals around each fracture zone in the entire point cloud data are analyzed. If the number of empty points in a certain segment is less than 10, the segment is considered to be a distribution void area. Finally, the segment numbers of all density mutation areas, fracture zones and void areas are combined to form a spatial discontinuity positioning sequence.

[0065] The point cloud characteristic comparison submodule is used to analyze the point cloud coordinates and distances between points in abnormal sections based on spatial discontinuous positioning sequences, compare the spatial distribution patterns of normal sections, filter the differences in point cloud distribution characteristics between abnormal and normal areas, determine the range of spatial coverage changes, and obtain point cloud distribution difference feature groups.

[0066] Based on the abnormal segments identified in the spatially discontinuous localization sequence, the spatial coordinate sets of the corresponding point clouds are extracted one by one. These are compared with the point cloud data of the same structure in normal segments. First, the 3D distance distribution characteristics between all points within the abnormal segments are statistically analyzed. The sparsity of the point group is determined by the average and maximum distances between adjacent points. When the average distance exceeds 4.5mm and the maximum distance exceeds 12mm, the area is identified as a sparse abnormal area. Then, the fluctuation of the point distribution density within the spatial boundary of this sub-segment is detected. The entire area is divided into a 3D grid with a side length of 3mm, and the number of points in each grid is counted. If the difference in the number of points between two adjacent grids exceeds 20 points, it is considered a local uneven distribution. At the same time, point cloud regions with the same structural characteristics in the normal section are extracted for grid statistics. If the grid difference in this region is within 10 points, it is recorded as a reference. Then, the grid difference range, maximum spacing value, and grid position of the minimum number of points in the abnormal area are compared with those in the normal area one by one, and the feature point distribution difference value is recorded. When more than 20% of the grid cells show an abnormal situation, it is determined that there is a spatial coverage change in the abnormal section. By integrating the above feature comparison results, a point cloud distribution difference feature group is formed.

[0067] The scanning parameter control submodule is used to adjust the scanning angle and scanning path of the laser head according to the point cloud distribution difference feature group, optimize the step interval setting, calibrate the spatial movement sequence of the laser head, and regenerate the scanning command covering the sparse area to obtain the point cloud coverage balance index.

[0068] Each anomaly in the point cloud distribution difference feature group is read. First, the spatial position of each anomaly segment on the workpiece is located. The laser head adjustment direction is set according to its directional characteristics. If the points in the anomaly area are concentrated in the horizontal extension direction, the laser head pitch angle is adjusted to the range of 5 to 10 degrees. If the points are concentrated at local corners in the vertical plane, the azimuth angle is adjusted to the range of 10 to 20 degrees. Then, the scanning path is redefined based on the grid point distribution data of the anomaly segment. An edge coverage path is added to the original scanning trajectory, and the step interval is reduced to 70% of the original set value, from 1.0 mm to 0.7 mm. For the three-dimensional grid boundary of the sparse area, 3 to 5 compensation scanning lines are set for each path according to the density of missing points. At the same time, the laser head movement order is reordered so that the compensation path scans the tail section of the original path first. Then, a new scanning instruction data packet is generated. After each path change, the scanning area reconstruction verification is performed. If the number of newly added points meets the coverage requirement of more than 80% of the original missing area, the adjustment is recorded as effective and included in the scanning instruction set. The point cloud coverage balance index for sparse coverage repair is output.

[0069] Please see Figure 4 The attitude offset detection module includes:

[0070] The attitude coordinate comparison submodule is used to compare the spatial coordinates of each node with the target assembly attitude coordinates based on the point cloud coverage equalization index, calculate the coordinate differences in the three-dimensional direction, determine the spatial offset trend of each node, optimize the node position distribution, and obtain the node three-dimensional offset vector set.

[0071] Based on the workpiece key node data located by the point cloud coverage equalization index, the current actual spatial coordinates of each node are extracted sequentially. The target assembly posture coordinates defined in the process design file are read as a comparison reference. A correspondence between the current node coordinates and the target node coordinates is established one-to-one according to the node number. Then, the difference of the three-dimensional coordinate values ​​of each node is calculated, and the positional differences in the X, Y, and Z directions are compared. In each direction, when the offset value exceeds 3.5mm, the node is recorded as a single-axis offset anomaly. When any two directions exceed the threshold simultaneously, it is marked as a dual-axis offset node. If all three directions exceed the threshold, it is recorded as a tri-axis offset node. Then, the offset trend of all nodes in each direction is statistically analyzed. When more than 60% of the nodes have a positive offset value in the Z-axis direction, it is marked as an overall Z-axis offset. The trend is then analyzed, and the offset values ​​are clustered. Nodes with offset values ​​between 1.5mm and 3.5mm are classified as mild offset segments, those between 3.5mm and 6.0mm as moderate offset segments, and those greater than 6.0mm as severe offset segments. Each node is classified and grouped according to its offset level. The spatial distribution of the grouped nodes in the workpiece structure is then further examined. When more than 5 nodes of the same offset level are found to be arranged consecutively in the same plane area, the segment is determined to be a continuous spatial offset segment. The node numbers in the areas with offset clusters are re-planned so that nodes with the same offset direction and similar offset magnitude are arranged continuously in the data structure. Finally, a set of three-way offset vectors for nodes is formed. Each vector contains the node number, the offset values ​​in the X, Y, and Z directions, and the offset level information.

[0072] The spatial offset recognition submodule is used to determine the distribution of the three-dimensional offset amplitude of nodes in the spatial segment based on the node three-dimensional offset vector set, analyze the continuous offset changes in the node sequence, compare the spatial characteristics of nodes and assembly reference segments, mark the associated spatial segments, and obtain the continuous identification group of offset regions.

[0073] The offset of each node in the X, Y, and Z directions is read one by one. The spatial segment in which each node is located is analyzed, and the nodes are divided into different spatial segments according to their coordinate range. The distribution characteristics of the three-dimensional offset values ​​of the nodes in each spatial segment are statistically analyzed. When more than 90% of the nodes in a certain segment have an offset value greater than 4.0 mm in the X direction, that spatial segment is judged as the X-axis offset concentration area. Then, the node numbers in each segment are sorted, and the difference in the offset value of adjacent nodes is checked. When the difference in the offset value of adjacent nodes is within 2.0 mm and the offset direction is consistent, three or more consecutive nodes are marked as offset. For continuous segments, combined with the segment data of the assembly reference point, the spatial distance between the offset segment and the reference segment is analyzed. When the shortest distance between the offset segment and the reference segment is less than 15mm and the offset direction is towards the reference segment, it is determined that the offset segment has a spatial relationship with the assembly reference. Among all segments, the areas that meet the conditions of continuous offset and proximity to the reference are selected. The total number of nodes, the offset level ratio, and the spatial position overlap range of each group of segments are counted. The results are formed into a continuous offset area identification group. Each group of identification records includes the start node number, end node number, offset direction, average offset amount, and relative distance to the reference segment.

[0074] The offset segment filtering submodule is used to calculate the node statistical characteristics in the main offset direction of the segment based on the continuous identifier group of the offset region, determine the aggregation trend of node offset changes, and optimize the consistency of the main direction data, using the following formula:

[0075] ;

[0076] The spatial offset distribution characteristics are obtained, where, Representing the Spatial offset distribution characteristics of each offset segment Indicates the first The number of nodes in each offset segment Indicates the first Within the offset segment, the first Offset data of each node in the main offset direction. Indicates the first The average of all node offset data in the main offset direction within each offset segment (i.e. );

[0077] Spatial offset distribution characteristics refer to the overall distribution characteristics of spatial offset phenomena of nodes in the main offset direction within each spatial segment. It is an aggregated statistical characteristic of the main offset behavior of nodes in a specific spatial segment, reflecting the comprehensive criterion of spatial attitude consistency and offset risk in the assembly or welding process of that segment. It is a parameter used for subsequent spatial calibration, assembly error analysis and trajectory adjustment.

[0078] Call the spatial segment number of each segment in the continuous identifier group of the offset region For the corresponding node set, extract each node sequentially along the determined main offset direction. raw offset data Statistical processing is performed to evaluate the spatial offset aggregation characteristics of the segment. The original data is standardized using linear normalization (Min-Max normalization) for a given segment. The original node offset data is [2.5mm, 2.9mm, 3.3mm, 3mm, 3.6mm], and the corresponding normalized values ​​are [0, 0.3636, 0.7273, 0.4545, 1], respectively. Then, the mean of the normalized offset data is calculated:

[0079] ;

[0080] Perform the summation of the absolute values ​​of the offset data for each node:

[0081] ;

[0082] Calculate the variance components of the offset data They are respectively:

[0083] ;

[0084] ;

[0085] ;

[0086] ;

[0087] ;

[0088] The sum of squares is 0.571.

[0089] The standard deviation is obtained by taking the square root of the average value.

[0090] ;

[0091] Substitute the above values ​​into the calculation formula:

[0092] ;

[0093] Based on the spatial offset aggregation reference interval set in the process design stage, the numerical range is divided into the following three segments according to the statistical behavior of the offset distribution characteristics in actual production:

[0094] when The segment is identified as a dispersed segment of offset behavior, indicating that the node offset fluctuations within this segment are small and the overall attitude is stable.

[0095] when The segment is determined to be a balanced segment of offset behavior, indicating that there is a certain degree of structural offset in the segment, but its distribution is not significant and is within the normal process error range.

[0096] when If the offset behavior is determined to be an aggregate segment, it means that the node offset direction is consistent and the fluctuation is focused in a specific space, resulting in welding positioning error or abnormal structural stress.

[0097] Calculation results Compared with these three intervals, the value falls into the third interval, that is... Therefore, it can be directly determined that the spatial segment belongs to the aggregation segment of the offset behavior. This result shows that the nodes in the segment have shown an overall focusing trend in the main offset direction. The error in the space no longer shows a random distribution, but forms a directional superposition trend, indicating that the segment has structural instability and belongs to the abnormality in the assembly process.

[0098] Please see Figure 5 The clamping trajectory linkage module includes:

[0099] The spatial mapping analysis submodule is used to determine the spatial correspondence between offset nodes and fixture distribution based on spatial offset distribution characteristics, compare the three-dimensional arrangement between nodes and fixtures, identify areas with overlapping or abnormal coverage, and adjust the corresponding node and fixture distribution parameters to obtain the fixture spatial correspondence parameters.

[0100] Based on the three-dimensional offset coordinates of each offset node in the spatial offset distribution characteristics, the node number and spatial coordinate values ​​are extracted point by point, and a spatial layout mapping relationship with the fixture assembly in the three-dimensional coordinate system is established. First, the installation position of the fixture fixing point is calibrated, with each fixture represented by its center point position. The three-dimensional coordinate error is controlled within ±1.0mm. By calculating the spatial distance between the offset node and the center point of the fixture, the nearest fixture number is determined. If the distance from the node to the fixture is less than 15mm, it is determined that there is a direct clamping relationship. Then, the number of fixtures within a 25mm range around the offset node is counted. If three or more fixtures overlap and act on the same node area, it is determined to be a fixture overlap area. Next, the coverage area of ​​the clamping area is projected onto the node point cloud area and compared. When the coverage ratio is greater than 80%, it is considered good coverage, and when the coverage ratio is less than 60%, it is considered abnormal coverage. Then, the node numbers with overlap or abnormal coverage and their corresponding clamp numbers are recorded, and the spatial installation direction of the clamps in the area is adjusted. The original clamp output direction is extended outward by 5 degrees to make a slight tilt adjustment. At the same time, the position of the clamp center point is corrected within ±2.0mm to make the clamp action point closer to the actual offset center of the node. The matching status, coverage ratio, clamp correction angle and position adjustment amount contained in the distribution relationship between each offset node and the clamp are output to form the corresponding parameters of the clamp space.

[0101] The attitude influence assessment submodule is used to analyze the force changes of the clamping state of the fixture under the distribution of each node based on the corresponding parameters of the fixture space, compare the correlation between the changes in clamping parameters and the spatial attitude of the workpiece, and determine the attitude response changes within the path disturbance range to obtain the attitude disturbance response parameters.

[0102] After receiving the corresponding parameters of the fixture space, the relationship between the output direction, clamping force level, and node coordinates of each fixture is analyzed one by one. The current set output direction angle, clamping force value, and control sequence of the fixture are read. By establishing a timing reference table between the clamping path and the node number, the coordinate changes of each group of nodes are loaded according to the fixture triggering sequence. The attitude changes of the first 10 groups of nodes in the X, Y, and Z directions after clamping are counted. When the displacement of a node in any direction exceeds 3.0 mm and the angle between the fixture output direction and that direction is less than 30 degrees, it is considered that the fixture has caused a direct attitude disturbance to the node. Then, the node space height before and after clamping is compared. If the height difference of three consecutive nodes exceeds 2.5mm along the entire path, it is marked as a concentrated disturbance area. The clamping force of the fixture is then divided into levels within the standard output value ±20N. The average value of node offset change is recorded for each level of force value to determine the correlation between clamping force change and attitude disturbance. When the increase in clamping force causes the node offset to increase by more than 1.5mm, it is recorded as a clamping force sensitive area. The correlation data between the control parameters of each fixture and the change in node attitude response are summarized to identify the degree and range of disturbance to the attitude of nodes in the path caused by the fixture control strategy. The attitude disturbance response parameters of each set of fixtures and nodes during the clamping process are output.

[0103] The linkage parameter generation submodule is used to optimize the force direction and clamping sequence of the fixture based on the attitude disturbance response parameters, adjust the synchronization method of the trajectory control command, calibrate the triggering conditions of the linkage between the clamping action and the path command, and obtain the trajectory clamping linkage parameters.

[0104] Based on the attitude change patterns of the fixture control variables and path nodes recorded in the attitude disturbance response parameters, the output direction angle of each fixture is adjusted to form an angle of more than 60 degrees with the attitude disturbance direction. When the node offset direction is the Z-axis, the fixture force application angle is set to deflect to within 45 degrees in the YZ plane. At the same time, the clamping sequence of the fixtures is rearranged, and the fixtures in the original sequence that overlap with the disturbance node are executed 1 to 2 positions earlier or later to avoid the superimposed offset caused by synchronous clamping. Then, for the trajectory control commands in the path execution module, the delay synchronization condition between path execution and fixture triggering is set. After the node clamping is completed, the path segment execution is triggered after a delay of 80ms to ensure that the trajectory is effective under stable attitude. If multiple clamps act on the path segment at the same time in the disturbance sensitive area, the synchronization interference is mitigated by increasing the execution interval to 150ms. In addition, the path segment execution flag is set to be triggered only when the node offset recovers to within ±1.0mm. This judgment logic is embedded in the logic judgment module between the clamp control command and the trajectory control, generating linkage trigger parameters such as the force direction of each clamp, clamping sequence number, path command delay time and node error trigger threshold, and outputting trajectory clamping linkage parameters.

[0105] Please see Figure 6 The path closed-loop calibration module includes:

[0106] The distance difference calculation submodule is used to analyze the differences between path nodes and target trajectory nodes in three-dimensional spatial coordinates based on trajectory clamping linkage parameters, calculate the direction and magnitude of spatial coordinate differences between nodes, compare the distribution changes of all nodes in space, determine the spatial consistency between the path and the target trajectory, and obtain node spatial offset feature groups.

[0107] Based on the clamping force adjustment value, trajectory node execution order, attitude trigger delay, and other data recorded in the trajectory clamping linkage parameters, the three-dimensional spatial coordinate values ​​of the path nodes and the target trajectory nodes are compared one by one. The coordinate groups of the actual scanned nodes are numbered and aligned with the target trajectory coordinates set in the CAD model. The coordinate difference of each pair of nodes in the X, Y, and Z directions is calculated, and the positive and negative changes in each direction are recorded to form a difference vector. Then, the directional distribution of node offset is statistically analyzed for each path segment to determine whether there is concentrated offset. For example, if more than 90% of the nodes in path segment P3 have an offset greater than 2.5mm in the Y direction and the direction is consistent, it is recorded as a Y-axis offset trend segment. Then, the centroid of node distribution in all path segments is calculated. If the path centroid is evenly distributed... If the object is displaced relative to the target trajectory, the center offset and offset direction are statistically analyzed. At the same time, the adjacent spacing sequence between nodes is extracted. When the spacing change between three consecutive nodes exceeds 30% of the average node spacing of the original trajectory, it is marked as an abnormal dispersion area between nodes. In actual workpiece scanning cases, for example, in path segment P5 nodes N201 to N208, the X-direction coordinate offset value is continuously between +3.8mm and +4.6mm, while the Z-direction remains within ±0.5mm. It is determined that there is a single-axis concentrated offset in this segment, forming a set of spatial difference parameters between each node and the target trajectory, including the three-axis coordinate difference, difference directionality mark, node spacing change value, centroid offset vector, and other parameters. The output is a node spatial offset feature group.

[0108] The path node filtering submodule is used to filter nodes with continuous changes based on the node spatial offset feature group, analyze the node distribution along the trajectory direction, optimize the node grouping criteria, and determine the abnormal distribution of nodes in the trajectory to obtain the abnormal distribution set of path nodes.

[0109] After reading the node spatial offset feature group, the spatial difference sequence of nodes is analyzed one by one according to the path segment number. Nodes with continuous changing trends are identified and screened. It is determined whether the node offset value in a certain direction shows a monotonically increasing or decreasing distribution. When the difference of 5 or more consecutive nodes in a certain direction does not exceed 0.8mm and the changing trend is consistent, they are classified as continuous offset segments. Then, a node distribution map of the trajectory direction is established by projecting the nodes of all path segments according to their coordinate axes. The degree of deviation of each node from the target trajectory line on the projection line is measured. When the deviation angle exceeds 10 degrees and the offset distance continuously exceeds 20mm, the node is recorded as a trajectory offset node. In specific implementation, the path segment P4 node N150 to N155 exhibits a continuous downward trend in the Z direction, with each point changing by approximately -1.1mm to -1.4mm. Furthermore, the arrangement angle along the path direction deviates by 10 degrees, meeting the screening criteria. Next, the grouping logic for all nodes is redefined. Instead of grouping them at equal intervals, path segments are adaptively divided based on spatial offset trends and continuous changes. When multiple small regions within a path segment exhibit significantly different offset types—for example, an initial rise followed by a fall in the X direction without buffer nodes—it is determined that the path segment has an abnormal distribution. All nodes satisfying the above abnormal structural characteristics are aggregated, and the abnormal distribution set of path nodes is output. Each data set records the node number range, offset direction trend, offset continuity level, and trajectory offset angle.

[0110] The path adjustment judgment submodule is used to determine the spatial distribution pattern of abnormal nodes based on the abnormal distribution set of path nodes, analyze the distribution relationship between abnormal nodes and normal nodes, compare the coherence of node distribution, optimize the path adjustment process, and obtain the path pair calibration results.

[0111] Upon receiving the abnormal distribution set of path nodes, the positional relationship of each abnormal node group in the overall spatial coordinate system is determined sequentially. The start and end numbers, average values ​​of the three coordinates, and the path segment number of the node group are extracted. The spatial location region in the workpiece structure coordinate system is then marked. The abnormal nodes are then compared with normally distributed nodes in the path, analyzing the coordinate change patterns of the two types of nodes in the X, Y, and Z directions to determine if spatial continuity exists. If there is an overlap of less than 3mm between the abnormal node group and the normal node group in the coordinate directions, and the node order is continuous, it is marked as a continuous and repairable area. Conversely, if the spatial jump exceeds 5mm and the node numbers are not continuous, it is defined as a path break zone. Further calculations are made of the distribution length and coverage ratio of each abnormal node segment within the path segment. When the abnormal segment length... If the degree exceeds 30% of the original path segment and covers the starting or ending point of the path, it is given priority in the candidate area for path structure adjustment. Based on this, a new path segment division logic is constructed, and the original path segment is divided according to the position of abnormal node segments. For each adjusted path segment, the trajectory node coordinate repositioning operation is performed. The correction is performed by keeping the starting point, fine-tuning the ending point, and linear interpolation of intermediate nodes. For example, if the original path segment P2 has abnormal segments N80 to N88, which account for about 35%, and the starting point N80 is exactly the first node of this segment, P2 is divided into two segments, P2a and P2b. After the nodes in the path segment are redistributed, the path pointer update operation in the trajectory control command is executed, and the changes in the number of path nodes and the changes in the overall path direction correction angle before and after the adjustment are recorded. The path pairing calibration result is output.

[0112] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A three-dimensional laser vision-based automated assembly and welding system, characterized in that, The system includes: The point cloud region analysis module is used to collect the point cloud coordinates of the workpiece based on a 3D laser vision sensor, compare the number and volume of feature points according to structural segments, optimize the point cloud data structure, and obtain spatial partition density features. The point cloud region parsing module includes: The data stream receiving submodule is used to analyze the collected point cloud coordinate data of the workpiece surface based on the 3D laser vision sensor, optimize the classification labels of spatial point coordinates and reflection intensity, filter invalid points caused by positioning errors in the annotation, and adjust the spatial arrangement and label relationship of the effective coordinate group to obtain the spatial structure annotation dataset. The spatial segment construction submodule is used to determine the spatial boundaries of each label coordinate group based on the spatial structure annotation dataset, compare the boundary morphology of the label corresponding region, analyze the uniformity of the distribution of point groups within the spatial range, optimize the sub-segment division logic and boundary coherence, and obtain the segment spatial distribution sequence. The density discrimination calculation submodule is used to calculate the spatial range of each sub-segment based on the spatial distribution sequence of the segment, filter feature points with spatial changes, analyze the degree of distribution concentration of feature points in each region, compare the spatial coverage ratio of feature points in each segment, judge the distribution differences, and obtain the spatial partition density features. The density anomaly compensation module is used to analyze the spatial distance changes of adjacent nodes based on the spatial partition density characteristics, compare the point cloud distribution of the abnormal area and the normal area, adjust the laser head scanning angle and step range, and obtain the point cloud coverage balance index. The density anomaly compensation module includes: The density continuity judgment submodule is used to determine the continuity of density change between sub-segments based on the density characteristics of the spatial partition, compare the density difference distribution of adjacent segments, screen the spatial locations with density jumps, determine density breaks and distribution void areas, and obtain a spatial discontinuity positioning sequence. The point cloud characteristic comparison submodule is used to analyze the point cloud coordinates and distances between points in abnormal sections based on the spatial discontinuous positioning sequence, compare the spatial distribution patterns of normal sections, filter the differences in point cloud distribution characteristics between abnormal and normal sections, determine the range of spatial coverage changes, and obtain point cloud distribution difference feature groups. The scanning parameter control submodule is used to adjust the scanning angle and scanning path of the laser head according to the point cloud distribution difference feature group, optimize the step interval setting, calibrate the spatial movement sequence of the laser head, and regenerate the scanning command covering the sparse area to obtain the point cloud coverage balance index. The attitude offset detection module is used to compare the point cloud coordinates with the target assembly attitude points based on the point cloud coverage equalization index, analyze the differences in node spatial coordinates, determine the sections where the welding reference positioning block has shifted, and obtain the spatial offset distribution characteristics. The clamping trajectory linkage module is used to analyze the correspondence between offset nodes and fixture distribution based on the spatial offset distribution characteristics, determine the influence of clamping state changes on workpiece posture, adjust fixture output parameters and trajectory commands, and obtain trajectory clamping linkage parameters. The path closed-loop calibration module is used to compare the spatial differences between path nodes and target trajectory based on the trajectory clamping linkage parameters, identify non-compliant nodes, determine whether the calibrated path segments meet the pairing requirements, and obtain the path pairing calibration results. The spatial partition density features include density distribution uniformity, partition filling integrity, and boundary feature stability; the point cloud coverage balance index includes compensation area continuity and overall coverage consistency; the spatial offset distribution features include the number of offset areas, offset trend direction, and maximum offset of key nodes; the trajectory clamping linkage parameters include clamp response sensitivity, trajectory adjustment amplitude, and linkage synchronization; and the path pair calibration results include node calibration ratio, path compliance, and error convergence degree. The workpiece point cloud coordinates refer to the set of three-dimensional X, Y, and Z coordinates of each point obtained by the three-dimensional laser vision sensor after scanning the workpiece surface. The spatial distance change refers to the change in distance between adjacent feature points in three-dimensional space. Abnormalities reflect uneven scanning coverage or abnormalities on the workpiece surface. The correspondence refers to the mapping relationship between the point cloud offset nodes and the fixture in the spatial layout.

2. The automated assembly and welding system based on three-dimensional laser vision according to claim 1, characterized in that, The attitude offset detection module includes: The attitude coordinate comparison submodule is used to compare the spatial coordinates of each node with the target assembly attitude coordinates based on the point cloud coverage equalization index, calculate the coordinate differences in the three-dimensional direction, determine the spatial offset trend of each node, optimize the node position distribution, and obtain the node three-dimensional offset vector set. The spatial offset recognition submodule is used to determine the distribution of the three-dimensional offset amplitude of the node in the spatial segment based on the node's three-dimensional offset vector set, analyze the continuous offset changes in the node sequence, compare the spatial characteristics of the node with the assembly reference segment, mark the associated spatial segments, and obtain a continuous identification group of the offset region. The offset segment filtering submodule is used to calculate the node statistical characteristics of the main offset direction of the segment based on the continuous identifier group of the offset region, determine the aggregation trend of node offset changes, optimize the consistency of the main direction data, and obtain the spatial offset distribution characteristics.

3. The automated assembly and welding system based on three-dimensional laser vision according to claim 1, characterized in that, The clamping trajectory linkage module includes: The spatial mapping analysis submodule is used to determine the spatial correspondence between offset nodes and fixture distribution based on the spatial offset distribution characteristics, compare the three-dimensional arrangement between nodes and fixtures, identify areas with overlapping or abnormal coverage, and adjust the corresponding node and fixture distribution parameters to obtain the fixture spatial correspondence parameters. The attitude influence assessment submodule is used to analyze the force changes of the clamping state of the fixture under the distribution of each node based on the corresponding parameters of the fixture space, compare the correlation between the changes in clamping parameters and the spatial attitude of the workpiece, and determine the attitude response changes within the path disturbance range to obtain the attitude disturbance response parameters. The linkage parameter generation submodule is used to optimize the force direction and clamping sequence of the fixture based on the attitude disturbance response parameters, adjust the synchronization mode of the trajectory control command, calibrate the triggering conditions of the linkage between the clamping action and the path command, and obtain the trajectory clamping linkage parameters.

4. The automated assembly and welding system based on three-dimensional laser vision according to claim 1, characterized in that, The path closed-loop calibration module includes: The distance difference calculation submodule is used to analyze the difference between path nodes and target trajectory nodes in three-dimensional spatial coordinates based on the trajectory clamping linkage parameters, calculate the direction and magnitude of the spatial coordinate difference between nodes, compare the distribution changes of all nodes in space, determine the spatial consistency between the path and the target trajectory, and obtain the node spatial offset feature group. The path node filtering submodule is used to filter nodes with continuous changes based on the node spatial offset feature group, analyze the node distribution in the trajectory direction, optimize the node grouping criteria, and determine the abnormal distribution of nodes in the trajectory to obtain the abnormal distribution set of path nodes. The path adjustment judgment submodule is used to determine the spatial distribution pattern of abnormal nodes based on the abnormal distribution set of path nodes, analyze the distribution relationship between abnormal nodes and normal nodes, compare the coherence of node distribution, optimize the path adjustment process, and obtain the path pair calibration result.

Images