Cable end point cloud automatic extraction method and system based on standard tool model matching

Abstract

The invention discloses a cable end point cloud automatic extraction method and system based on standard tool model matching, and the method achieves the full-automatic and high-precision extraction from original point cloud input to pure cable point cloud output through preprocessing, background elimination, point cloud separation based on standard tool model matching and verification optimization of full-process automatic module design. The extraction precision and efficiency are improved; the method comprises the following steps: point cloud preprocessing; background point cloud elimination; based on spatial distribution characteristics of point clouds, background point clouds far away from a cable-tool area are removed through area growth clustering and spatial range constraint; the tool and the cable point cloud are separated; separation is realized based on standard tool model matching, a preset standard tool model is loaded, a tool part in the scanning point cloud is registered with the standard model by adopting a registration algorithm, the tool point cloud is removed through distance screening, and the cable point cloud is reserved; verifying and optimizing a result; and performing boundary integrity check on the cable point cloud obtained by separation, and cleaning residual interference points.

Description

Technical Field

[0001] This invention relates to the technical field of three-dimensional laser scanning and point cloud data processing, and in particular to a method and system for automatic extraction of point clouds from cable ends based on standard tooling model matching. Background Technology

[0002] In power systems, cable accessories are critical components ensuring the safe and stable operation of cable lines, and their quality directly affects the reliability of power transmission. With the widespread use of high-voltage and ultra-high-voltage cables in power grids, stringent requirements have been placed on the manufacturing precision of cable accessories—especially the assembly gap between the cable end and the accessory, which must be strictly controlled. Otherwise, uneven electric field distribution may cause partial discharge, ultimately leading to serious accidents such as insulation breakdown. Therefore, during the manufacturing process of cable accessories, it is necessary to accurately obtain the geometry of the cable end using 3D modeling technology to provide data support for subsequent work.

[0003] Traditional cable termination modeling relies on 3D laser scanning technology. Its core steps include: attaching positioning markers to the cable termination surface, acquiring point cloud data containing these markers using a laser scanner, and then stitching and registering the point clouds from multiple perspectives based on the markers to generate a complete model. However, this process suffers from significant technical bottlenecks: First, the positioning markers have a potential impact on cable performance. The insulation surface of the cable termination needs to remain intact to ensure its corona resistance and aging resistance. Applying the positioning markers requires special adhesive, which may penetrate the microscopic pores of the insulation layer, damaging the material's dielectric properties. Removing the markers may cause scratches on the insulation surface, creating localized electric field concentration points. In 110kV and above high-voltage cables, such minor damage can become a potential long-term malfunction. Second, manual operation is inefficient. A standard cable termination requires 20-50 positioning markers, evenly distributed across different axial sections. Applying and removing markers for a single termination takes approximately 10-15 minutes, severely limiting efficiency in mass production scenarios. Third, the complexity of the scanning environment can cause interference. Cable accessory fabrication sites typically contain background objects such as metal supports, tool benches, and the ground, which laser scanners may capture point clouds of inconsequential objects. Furthermore, variations in ambient lighting can cause uneven reflectivity in the point cloud, further complicating subsequent data processing.

[0004] To address these issues, the industry has proposed using an "integrated positioning marker fixture" instead of directly affixing markers to the cable. This fixture, made of insulating material, has high-precision positioning markers pre-attached. During use, it is placed on the cable end, and after scanning, the fixture is removed directly, avoiding contact damage to the cable itself. While this technology significantly improves operational efficiency and protects the cable insulation, it introduces new technical challenges: there is a gap between the fixture and the cable end; the scanning point cloud records the fixture's geometry, and the point clouds of both highly overlap at the contact boundary; besides the fixture, the spatial distribution of the background point cloud (such as the ground or supports) may intersect with the cable point cloud; and the point cloud data contains noise, making traditional threshold segmentation and region growing methods prone to "oversegmentation" or "undersegmentation," failing to meet the accuracy and efficiency requirements of automated production.

[0005] Therefore, how to achieve fully automatic and high-precision extraction of the main point cloud of the cable end from the mixed point cloud containing tooling and background interference has become a key technology that urgently needs to be broken through in the field of digital manufacturing of cable accessories. It not only directly affects the manufacturing quality and production efficiency of cable accessories, but also relates to the long-term operational safety of high-voltage cable lines. Summary of the Invention

[0006] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by disclosing an automatic extraction method and system for cable end point clouds based on standard tooling model matching. Through preprocessing, background removal, point cloud separation based on standard tooling model matching, and verification optimization of the fully automated module design, the invention achieves fully automatic and high-precision extraction from raw point cloud input to clean cable point cloud output, thereby improving extraction accuracy and efficiency.

[0007] The technical solution adopted by this invention to solve its technical problem is: an automatic extraction method for cable end point clouds based on standard tooling model matching, comprising the following steps:

[0008] S1: Point cloud preprocessing; Denoising, downsampling and coordinate normalization are performed on the original point cloud containing cable ends, tooling and background.

[0009] S2: Background point cloud removal; Based on the spatial distribution characteristics of point clouds, background point clouds far from the cable-tooling area are removed through region growing clustering and spatial range constraints.

[0010] S3: Separation of tooling and cable point clouds; separation is achieved based on standard tooling model matching. A preset standard tooling model is loaded, and a registration algorithm is used to register the tooling part in the scanned point cloud with the standard model. Tooling point clouds are removed by distance filtering, while cable point clouds are retained.

[0011] S4: Result verification and optimization; perform boundary integrity checks on the separated cable point cloud and clean up residual interference points.

[0012] Furthermore,

[0013] In step S1, the original point cloud containing cable ends, tooling and background is denoised by using a statistical filtering algorithm to calculate the average distance between each point and its neighboring points, and outlier points whose average distance exceeds a preset multiple of the standard deviation are removed, as well as dust and random reflection noise in the background.

[0014] The downsampling uses voxel grid filtering to resample the point cloud according to the voxel size that matches the point cloud accuracy requirements.

[0015] The coordinate normalization uses the tooling positioning reference as the origin to establish a local coordinate system for the cable, so that the X-axis is along the cable axis and the Y / Z axes are radial, ensuring that the coordinate reference is consistent for each scan.

[0016] Furthermore,

[0017] In step S2, the region growing clustering divides the point cloud into multiple connected clusters according to the point spacing threshold that meets the clustering requirements. The largest cluster is retained as the cable-tool assembly point cloud, and the remaining small clusters are determined as background and deleted.

[0018] The spatial range constraint is based on the size parameters of the cable and tooling, with a preset three-dimensional spatial range, and point clouds that exceed the range are automatically cropped.

[0019] Furthermore,

[0020] In step S3, the standard tooling model is a pre-constructed tooling CAD model or a high-precision scanning model.

[0021] The registration algorithm adopts the ICP iterative nearest point algorithm, and the registration accuracy meets the accuracy requirements for cable accessory manufacturing.

[0022] The process involves filtering out tooling point clouds by distance, calculating the shortest distance from each point in the scanned point cloud to the surface of the standard tooling model, identifying points whose shortest distance is less than a preset distance threshold as tooling points and removing them, and leaving the remaining point cloud as the cable body.

[0023] Furthermore,

[0024] In step S4, residual interference points are cleaned up by using radius filtering. A neighborhood radius that is adapted to the cleaning requirements is set, and isolated points with fewer than a preset number of points in the neighborhood are removed.

[0025] Furthermore,

[0026] The result verification and optimization in step S4 also includes light smoothing of the cable point cloud, Gaussian filtering of the cable point cloud to eliminate edge jaggedness and retain the original geometric features.

[0027] This invention also discloses an automatic point cloud extraction system for cable ends based on standard tooling model matching, comprising:

[0028] Preprocessing module: used to perform noise reduction, downsampling, and coordinate normalization on the original point cloud;

[0029] Background removal module: used to remove background point clouds based on spatial distribution features;

[0030] Point cloud separation module: contains standard model matching units, used to load standard tooling models and execute...

[0031] ICP registration algorithm and distance filtering separation tooling and cable point cloud;

[0032] Verification and optimization module: used to perform boundary integrity checks, residual interference cleanup, and smoothing on cable point clouds;

[0033] Control module: Used to coordinate the work of each module to realize a fully automated processing flow from raw point cloud input to clean cable point cloud output.

[0034] Furthermore,

[0035] It also includes a tooling assembly, which has a receiving groove, into which the cable end is inserted.

[0036] Furthermore,

[0037] The system uses the PCL point cloud library to implement algorithm functions or integrates it into Geomagic and Artec Studio scanning software, and achieves automated operation through custom processing templates.

[0038] Compared with the prior art, the present invention has the following advantages:

[0039] 1. High degree of automation, significantly improving processing efficiency.

[0040] This invention achieves fully automated module design from raw point cloud input to clean cable point cloud output through preprocessing, background removal, point cloud separation based on standard tooling model matching, and verification optimization. Compared with traditional manual segmentation methods, the processing time for a single set of point clouds is significantly reduced, and the efficiency of cable end model acquisition is significantly improved, effectively adapting to the efficiency requirements of mass production scenarios for cable accessories.

[0041] 2. Excellent extraction accuracy, meeting the precision manufacturing requirements of high-voltage cable accessories.

[0042] This invention ensures stable and controllable accuracy of cable point cloud extraction by using a pre-set standard tooling model and a high-precision registration algorithm. It meets the precision manufacturing requirements of geometric data for 110kV and above high-voltage cable accessories, and provides reliable data support for the accurate assembly of accessories and cable ends and the guarantee of electric field uniformity.

[0043] 3. Highly robust and adaptable to complex field application scenarios.

[0044] This invention relies on the core logic of standard tooling model matching, combined with the fixed relative position design of the tooling and cable end. Even in the face of complex situations such as slight wear of the tooling, changes in on-site lighting and background noise interference, it can still stably retain the point cloud of the main cable body, reduce tooling residue, and ensure the stability and consistency of the point cloud extraction process.

[0045] 4. Ensure cable insulation performance and support digital manufacturing transformation.

[0046] This invention completely eliminates the need to directly paste positioning marks onto the surface of the cable end insulation layer using specialized tooling pre-installed with positioning marks. This avoids problems such as glue residue and insulation layer scratches from the source, reducing the risk of cable accessory failures caused by insulation defects. At the same time, the extracted pure cable point cloud can provide a data foundation for the digitalization of cable accessory manufacturing, promoting the industry's transformation from traditional experience-driven to data-driven, and providing key technical support for the long-term safe and reliable operation of high-voltage cable lines. Attached Figure Description

[0047] Figure 1 This is a schematic flowchart of the method of the present invention;

[0048] Figure 2 This is a schematic diagram of the module structure of the system of the present invention;

[0049] Figure 3 This is a schematic diagram of the original mixed point cloud;

[0050] Figure 4 This is a schematic diagram of the point cloud obtained from the extracted cable end. Detailed Implementation

[0051] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0052] like Figures 1-4 As shown, this invention discloses an automatic point cloud extraction method for cable ends based on standard tooling model matching, including the following steps:

[0053] S1: Point cloud preprocessing; Denoising, downsampling and coordinate normalization are performed on the original point cloud containing cable ends, tooling and background.

[0054] Specifically, in step S1, the original point cloud containing cable ends, tooling and background is denoised by using a statistical filtering algorithm to calculate the average distance between each point and its neighboring points, remove outliers whose average distance exceeds a preset multiple of the standard deviation, and remove dust and random reflection noise in the background.

[0055] The downsampling uses voxel grid filtering to resample the point cloud according to the voxel size that matches the point cloud accuracy requirements.

[0056] The coordinate normalization uses the tooling positioning reference as the origin to establish a local coordinate system for the cable, so that the X-axis is along the cable axis and the Y / Z axes are radial, ensuring that the coordinate reference is consistent for each scan.

[0057] S2: Background point cloud removal; Based on the spatial distribution characteristics of point clouds, background point clouds far from the cable-tooling area are removed through region growing clustering and spatial range constraints.

[0058] Specifically, in step S2, the region growing clustering divides the point cloud into multiple connected clusters according to the point spacing threshold that meets the clustering requirements. The largest cluster is retained as the cable-tool assembly point cloud, and the remaining small clusters are determined as background and deleted.

[0059] The spatial range constraint is based on the size parameters of the cable and tooling, with a preset three-dimensional spatial range, and point clouds that exceed the range are automatically cropped.

[0060] S3: Separation of tooling and cable point clouds; separation is achieved based on standard tooling model matching. A preset standard tooling model is loaded, and a registration algorithm is used to register the tooling part in the scanned point cloud with the standard model. Tooling point clouds are removed by distance filtering, while cable point clouds are retained.

[0061] Specifically, in step S3, the standard tooling model is a pre-constructed tooling CAD model or a high-precision scanning model;

[0062] The registration algorithm adopts the ICP iterative nearest point algorithm, and the registration accuracy meets the accuracy requirements for cable accessory manufacturing.

[0063] The process involves filtering out tooling point clouds by distance, calculating the shortest distance from each point in the scanned point cloud to the surface of the standard tooling model, identifying points whose shortest distance is less than a preset distance threshold as tooling points and removing them, and leaving the remaining point cloud as the cable body.

[0064] S4: Result verification and optimization; perform boundary integrity checks on the separated cable point cloud and clean up residual interference points.

[0065] Specifically, in step S4, residual interference points are cleaned up by using radius filtering, setting a neighborhood radius that matches the cleaning requirements, and removing isolated points whose number of points in the neighborhood is less than a preset number.

[0066] Specifically, the result verification and optimization in step S4 also includes performing a light smoothing process on the cable point cloud, applying Gaussian filtering to the cable point cloud to eliminate jagged edges and retain the original geometric features.

[0067] S1: Point cloud preprocessing

[0068] The Statistical Outlier Removal filter from the PCL library was used to denoise the original point cloud, with the number of neighboring points and the standard deviation factor set to a preset value.

[0069] For each point P in the point cloud i Search for its k nearest neighbors (e.g., k=50). This constitutes the local neighborhood of the point. Calculate P. i Find the distances to all k of its neighbors and calculate the mean μ of these distances. i and standard deviation σ i Set a global distance threshold, typically a preset multiple of the standard deviation, iterate through all points, and if a certain point P... i μ i Points exceeding this threshold are identified as outliers and removed. This method efficiently removes discrete noise points that do not constitute a surface, such as dust in the background or random reflection noise. These noise points can affect the accuracy of subsequent clustering and registration.

[0070] A VoxelGrid filter is used, with a voxel size tailored to the requirements for downsampling. The 3D space is divided into a uniform cubic grid (voxels) with side length 's'. For each non-empty voxel, all points within that voxel are replaced by its centroid (or the first point). While maintaining the overall shape, this reduces the point cloud data volume by 1-2 orders of magnitude, significantly accelerating subsequent calculations. Laser scanning point clouds typically have uneven density; voxel filtering can make their distribution more uniform, avoiding density deviations in subsequent algorithms. The voxel size 's' is a core parameter, slightly smaller than 1 / 3 to 1 / 2 of the cable's critical dimensions (such as insulation thickness) to ensure sufficient detail is retained for accurate analysis and measurement after downsampling.

[0071] Coordinate normalization establishes a local coordinate system for the cable based on the tooling positioning reference. With the center of the inner circle of the tooling as the origin and the X-axis along the cable axis, the entire point cloud is rotated and translated to coincide with this newly established local coordinate system. This ensures that point clouds scanned at different times and in different batches have a completely consistent coordinate reference, which is a prerequisite for subsequent automated registration and comparison. The X-axis along the cable axis is very convenient for subsequent straightness analysis, length measurement, or axial cross-sectional analysis of the cable. The Y / Z axes are radial, facilitating the measurement of cable diameter, eccentricity, etc.

[0072] S2: Background point cloud removal

[0073] Region-growing clustering is employed, with a point spacing threshold set to adapt to specific needs, retaining the largest connected cluster; a point spacing threshold d is set. cluster For example, 5mm. This threshold should be greater than the distance between points on the cable or tooling surface, but less than the gap between the cable and the distant background. Start with any unvisited point and use it as a seed for a new cluster. Recursively find all points whose distance to the seed point is less than the point spacing threshold and add them to the current cluster. When no more points can be added, the cluster growth is complete. Find the next unvisited point and start a new cluster. Calculate the volume of all clusters, keep the one or a few clusters with the largest volume, and delete the rest.

[0074] Cables and tooling, as a whole, are usually the largest and most cohesive objects in the scene. This method can effectively eliminate tools, supports, etc., scattered around them.

[0075] Spatial range constraints are implemented by pre-setting a three-dimensional spatial range based on the dimensions of the cable and tooling, and then clipping point clouds that exceed this range. This three-dimensional spatial range in the X, Y, and Z directions should be slightly larger than the maximum possible size of the cable-tooling assembly. This three-dimensional spatial range is applied to the point cloud after clustering and filtering to remove all points that exceed the range. Point clouds that are close together and form certain clusters can be removed, even if they clearly do not belong to the target object.

[0076] S3: Tooling and cable point cloud separation

[0077] Loading the standard tooling model (CAD modeling, with accuracy meeting the requirements for high-voltage cable accessory manufacturing) serves as the "benchmark" for the separation operation;

[0078] CAD model: error-free, perfect surface, can be directly used for distance calculation.

[0079] High-precision scanning model: obtained by scanning a clean, cable-free tooling multiple times and averaging the results, which can reflect the actual manufacturing deviation of the tooling.

[0080] The ICP algorithm is used for registration, with a preset number of iterations, and the registration accuracy meets the requirements.

[0081] Closest point search: For each point in the scanned point cloud, find the point with the closest Euclidean distance on the surface of the standard tooling model to form a corresponding point pair.

[0082] Transformation estimation: Calculate a rigid body transformation (rotation matrix R and translation vector T) that minimizes the mean square error between all corresponding point pairs. This can be achieved by methods such as SVD decomposition.

[0083] Apply transformation: Apply (R,T) to the entire scanned point cloud.

[0084] Iteration: Repeat the above steps until the change in mean square error is less than a preset convergence threshold ε, or the maximum number of iterations is reached.

[0085] The registration accuracy meets the accuracy requirements for cable accessory manufacturing. The accuracy of ICP depends on the point cloud quality, initial position, and parameter settings. For example, ε < 0.1mm is required to ensure that the separated cable shape is accurate enough to guide the automated processing or installation of subsequent accessories.

[0086] Set a distance threshold for the adapted scene, discard tooling point clouds, and retain cable point clouds. After registration, the tooling portion in the scanned point cloud should highly overlap with the standard model. Calculate the shortest directed distance from each point in the scanned point cloud to the triangular facet of the standard tooling model.

[0087] Set a distance threshold d threshold For example, if the distance is 1.0mm, all points with a distance of less than 1 are considered as tooling point clouds and are discarded. All points with a distance greater than 1 are considered as cable point clouds and are retained.

[0088] S4: Result Validation and Optimization

[0089] Inspect key areas such as the cable end face and insulation layer edges to confirm that there are no missing point clouds.

[0090] Radius filtering is used to set a neighborhood radius that is adapted to the cleaning requirements, and isolated points with fewer than the preset number of neighboring points are removed; for each point in the separated cable point cloud, the number of neighboring points n within its neighborhood radius r is counted.

[0091] Set a minimum number of points threshold n min If n <n min If the point is isolated, it is considered an isolated point and is removed.

[0092] It is mainly used to remove residual tooling noise points, which may be caused by minor registration errors or reflections from the paint layer on the tooling surface. These points usually exist in isolation in the cable point cloud or are random noise generated during the scanning process.

[0093] The radius r should be set relatively small (e.g., 2-3 times the average spacing of the point cloud), n min The value is usually set to 3-5. This effectively reduces noise without damaging the cable's fine structure (such as the gaps between strands).

[0094] Gaussian filtering was used for smoothing to obtain the final point cloud of the cable end.

[0095] For each point in the point cloud, its smoothed new position is determined by the weighted average of the positions of itself and all points in a certain neighborhood.

[0096] Neighborhood selection: For each point P in the cable point cloud i Search for all neighboring points of radius R around it.

[0097] Weight calculation: For each point P in the neighborhood j According to its destination P i distance d ij The weights w are calculated using a Gaussian function. j :

[0098] ,

[0099] σ is the standard deviation of the Gaussian function, which controls the distribution of the weights. The smaller σ is, the more concentrated the weights are near the center point; the larger σ is, the more evenly the weight distribution is distributed, and the stronger the smoothing effect.

[0100] Location update: Calculate point P i Smoothed new position :

[0101] ,

[0102] That is, the new location is the weighted average of the locations of all neighboring points. R determines the neighborhood range participating in the smoothing calculation. R is generally relatively small, set to 2 to 4 times the average point spacing of the point cloud. σ is set as a proportion of R, for example, σ=R / 2 or σ=R / 3.

[0103] Furthermore, smoothing must be performed after removing residual interference points (such as radius filtering). If smoothing is done first, noisy points will "pull" around good data points, thus contaminating the clean point cloud. On the point cloud before and after smoothing, measure the fixed position of the cable (such as a certain diameter) to ensure that the dimensional change is within the allowable tolerance (e.g., less than 0.05 mm).

[0104] The processing results of this embodiment show that the point cloud of the cable end is completely extracted, with no obvious tooling or background residue, and the processing time is short, which fully meets the efficiency and accuracy requirements of on-site scanning.

[0105] An automatic point cloud extraction system for cable ends based on standard tooling model matching includes:

[0106] Preprocessing module: used to perform noise reduction, downsampling, and coordinate normalization on the original point cloud;

[0107] Background removal module: used to remove background point clouds based on spatial distribution features;

[0108] Point cloud separation module: Includes a standard model matching unit, used to load standard fixture models, execute ICP registration algorithms, and use distance filtering to separate fixture and cable point clouds;

[0109] Verification and optimization module: used to perform boundary integrity checks, residual interference cleanup, and smoothing on cable point clouds;

[0110] Control module: Used to coordinate the work of each module to realize a fully automated processing flow from raw point cloud input to clean cable point cloud output.

[0111] Specifically, it also includes a tooling assembly, which has a receiving groove, into which the cable end is inserted.

[0112] This system achieves fully automated and high-precision extraction of cable end point clouds from complex scanning environments through deep collaboration between "hardware tooling design" and "software algorithm process".

[0113] It has the following advantages:

[0114] 1. Significantly improves automation and processing efficiency.

[0115] Full-process automation: The control module coordinates the sequential execution of each module, from the initial point cloud input to the final result output, without any manual intervention. This avoids the tedious operations required by traditional methods, such as manual selection of regions and adjustment of parameters by technicians.

[0116] "One-click" operation: Users only need to input the scan data and preset model into the system to quickly obtain a clean cable point cloud, which greatly reduces the operation threshold and technical requirements, improves the overall detection efficiency, and is very suitable for batch processing on the production line.

[0117] 2. Significantly improves extraction accuracy and reliability

[0118] The absolute reference role of the standard model: Using a high-precision CAD or scanned model as the registration reference fundamentally avoids the ambiguity and errors that may arise from relying solely on geometric feature segmentation.

[0119] ICP registration ensures high precision: Through precise iterative calculations, the scanning fixture is aligned with the standard model, ensuring the accuracy of the separation reference.

[0120] Guidance and constraints of physical tooling:

[0121] Precise positioning: The accommodating slots on the fixture physically ensure that the cable end is placed in the same position and orientation every time. This provides a stable and unique physical reference for the software's "coordinate normalization" module, making the initial conditions for each scan highly consistent.

[0122] Simplifying the separation problem: The presence of the receiving groove makes the contact relationship between the cable and the tooling regular and predictable. In the point cloud, the cable point cloud is mainly distributed inside and above the receiving groove, while the tooling point cloud constitutes a regular, known geometric structure (the model of the receiving groove). This makes separation by distance filtering exceptionally accurate and direct.

[0123] 3. Effectively enhances the robustness and stability of the system.

[0124] Insensitive to background and noise:

[0125] The background removal module effectively filters out interference that is not connected to the cable-tool assembly through spatial distribution and clustering.

[0126] The physical fixture itself isolates the cable from most of the background through the containment groove, reducing the complexity of the problem from the source.

[0127] Greater adaptability to changes in cable appearance: Regardless of variations in cable color, texture, or surface reflectivity, as long as the cable can be secured by the fixture's receiving slot, the system can reliably separate it by matching a known, unchanging fixture model. This approach shifts the focus from the changing target (cable) to the fixed environment (fixture), significantly improving system stability.

[0128] 4. The multiplier effect of synergy between physical tooling and algorithm modules

[0129] Preprocessing stage: The positioning reference of the tooling (such as specific corners or planes of the receiving slot) provides clear and automatically identifiable features for the "coordinate normalization" module, ensuring the accuracy and repeatability of the coordinate system.

[0130] Background Removal Stage: The physical isolation effect of the accommodating slot makes the cable-tool assembly naturally become a compact connected domain in space, making the removal algorithm based on spatial range and clustering more efficient and accurate.

[0131] Point cloud separation stage:

[0132] The inner surface of the receiving groove is an important component of the standard tooling model.

[0133] In the scanned point cloud, points that closely fit the inner wall of the accommodating groove are precisely registered and removed, while points in the space within the groove not occupied by the model are identified as cables. This "negative space" extraction method is highly efficient.

[0134] Distance threshold d threshold The setting thus has a clear physical basis (i.e., the average gap between the cable and the inner wall of the accommodating groove), making the parameter setting more scientific and the results more reliable.

[0135] Verification and optimization phase: Smoothing can be specifically optimized for the relatively regular cylindrical surface of the cable extracted from the regular accommodating groove. The parameter adjustment is more targeted and the effect is better.

[0136] 5. Clear practical engineering value

[0137] Standardization and repeatability: The system and tooling constitute a complete inspection unit that can produce consistent and comparable measurement results at different times and by different operators, which is crucial for quality control and process traceability.

[0138] Provides high-quality data for applications: The final output of clean, smooth, and coordinate-uniform cable end point clouds can be directly used for: parameter measurement, data backup, and defect detection.

[0139] The system's greatest innovation and advantage lies in the fact that it is not a purely software algorithm, but an integrated hardware and software solution that incorporates dedicated hardware (a fixture with accommodating slots). The physical fixture is not just a clamp, but also a "semantic marker" and "positioning reference" embedded in the real world. It greatly simplifies the complex vision problems that the software needs to solve, transforming uncertain and variable scenarios into a deterministic and structured computational model, thereby achieving a comprehensive improvement in efficiency, accuracy, and robustness.

[0140] Specifically, the system uses the PCL point cloud library to implement algorithm functions or integrates it into Geomagic or Artec Studio scanning software, and achieves automated operation through custom processing templates.

[0141] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Within the scope of knowledge possessed by those skilled in the art, various changes can be made without departing from the spirit of the present invention. These changes involve related technologies well known to those skilled in the art, and all of them fall within the protection scope of the present invention.

[0142] Many other changes and modifications can be made without departing from the concept and scope of this invention. It should be understood that this invention is not limited to the specific embodiments, and the scope of this invention is defined by the appended claims.

Claims

1. A cable end point cloud automatic extraction method based on standard tooling model matching, characterized in that, The method comprises the following steps: S1: point cloud preprocessing; the original point cloud containing cable end, tooling and background is denoised, down-sampled and coordinate normalized; S2: background point cloud elimination; based on the spatial distribution characteristics of the point cloud, the background point cloud far from the cable-tooling area is eliminated through region growing clustering and spatial range constraint; S3: tooling and cable point cloud separation; based on standard tooling model matching, the tooling part in the scanning point cloud is matched with the standard model by loading a preset standard tooling model and using a registration algorithm, the tooling point cloud is eliminated through distance screening, and the cable point cloud is reserved; S4: result verification and optimization; the boundary integrity of the separated cable point cloud is checked, and residual interference points are cleaned up.

2. The cable end point cloud automatic extraction method based on standard tooling model matching of claim 1, wherein, In the step S1, the original point cloud containing cable end, tooling and background is denoised, a statistical filtering algorithm is used to calculate the average distance of each point and the neighborhood points, and the outlier points with a distance average exceeding the preset standard deviation are eliminated to remove the dust and random reflection point noise in the background; The down-sampling uses voxel grid filtering to resample the point cloud according to the voxel size adapted to the point cloud precision requirement; The coordinate normalization takes the positioning reference of the tooling as the origin to establish a cable local coordinate system, so that the X axis is along the cable axial direction, and the Y / Z axis is radial, and the coordinate reference of each scanning is ensured to be consistent.

3. The cable end point cloud automatic extraction method based on standard tooling model matching of claim 2, wherein, In the step S2, the region growing clustering divides the point cloud into multiple connected clusters according to the point distance threshold adapted to the clustering requirement, the largest cluster is reserved as the cable-tooling combined point cloud, and the remaining small clusters are determined as background and deleted; The spatial range constraint presets a three-dimensional space range based on the size parameters of the cable and the tooling, and automatically clips the point cloud beyond the range.

4. The cable end point cloud automatic extraction method based on standard tooling model matching of claim 3, wherein, In the step S3, the standard tooling model is a tooling CAD model or a high-precision scanning model constructed in advance; The registration algorithm uses an ICP iterative closest point algorithm, and the registration accuracy meets the cable accessory manufacturing accuracy requirement; In the step S3, the tooling point cloud is eliminated through distance screening, the shortest distance of each point in the scanning point cloud to the surface of the standard tooling model is calculated, the points with a shortest distance less than a preset distance threshold are determined as tooling points and eliminated, and the remaining point cloud is the cable main body.

5. The cable end point cloud automatic extraction method based on standard tooling model matching of claim 4, wherein, In the step S4, the residual interference points are cleaned up by using radius filtering, setting a neighborhood radius adapted to the cleaning requirement, and eliminating the isolated points with a number of points in the neighborhood less than a preset number.

6. The cable end point cloud automatic extraction method based on standard tooling model matching according to any one of claims 1-5, characterized in that, In the step S4, the result verification and optimization further comprises a slight smoothing processing of the cable point cloud, a Gaussian filtering of the cable point cloud, an elimination of edge jaggies, and a reservation of original geometric features.

7. An automatic cable end point cloud extraction system based on standard tooling model matching, characterized in that, The cable end point cloud automatic extraction method based on standard tooling model matching comprises the following steps: A preprocessing module is used to denoise, down-sample and coordinate normalize the original point cloud; A background elimination module is used to eliminate the background point cloud based on the spatial distribution characteristics; A point cloud separation module comprises a standard model matching unit, which is used to load a standard tooling model, perform an ICP registration algorithm and distance screening to separate the tooling and cable point cloud; A verification and optimization module is used to check the boundary integrity of the cable point cloud, clean up residual interference and perform smoothing processing. ​ Control module: used for coordinating the work of each module, realizing the automatic processing flow from the original point cloud input to the pure cable point cloud output.

8. The cable end point cloud automatic extraction system based on standard tooling model matching of claim 7, wherein, The tool assembly is further provided with a containing groove, and the cable end is inserted into the containing groove.

9. The cable end point cloud automatic extraction method based on standard tooling model matching of claim 7, wherein, The system realizes algorithm functions by using a PCL point cloud library or is integrated in Geomagic or Artec Studio scanning software, and realizes automatic operation through a self-defined processing template.

Images