Electronic fence and collision body construction method and system based on point cloud and triangle network fusion, and computer readable storage medium

Abstract

The present application relates to an electronic fence and a collision body construction method, a construction system and a computer readable storage medium based on point cloud and triangle network fusion. Taking the hoisting operation of the transformer substation as the exclusive application scene, a full-coverage three-dimensional columnar forbidden area is constructed, two sets of collision models that can be linked for detection are generated, including octomap and BVH triangular grid, the special-shaped polygon is adaptively divided, the model failure automatic bottom switching mechanism is established, and the basic process is: obtaining electronic fence / construction area polygon information, generating three-dimensional columnar point cloud, constructing octomap octree model and adaptively triangular division BVH triangular grid model, double model fault tolerance selection output collision object, and outputting the model to the boom collision warning calculation module, so as to provide a high-reliability three-dimensional forbidden area model for high-precision positioning boom posture restoration and real-time calculation of the distance between the boom and the live body.

Description

Technical Field

[0001] The present invention relates to a method for constructing electronic fences and collision objects based on the fusion of point clouds and triangular networks, an electronic fence and collision object construction system for implementing the method, and a computer-readable storage medium for storing the corresponding computer program. Background Technology

[0002] The high-voltage equipment inside the substation is dense and the high and low voltage structures are intertwined. The hoisting operation is an extremely high-risk live cross operation. The movement of the boom covers the entire high-altitude area, which can easily lead to the boom approaching or intruding into the live restricted area, causing short circuits and electric shocks, which are major safety accidents. The existing electronic fences are mostly two-dimensional planar polygon constraints, which are difficult to adapt to the needs of the hoisting operation in the substation. The main problems are: (1) Only two-dimensional fences are used, which cannot match the three-dimensional spatial movement of the boom; (2) The single modeling method has a low fault tolerance rate and is prone to failure in the complex terrain of the substation; (3) It is not compatible with Octomap and FCL, which increases the burden of subsequent data processing. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for constructing a three-dimensional electronic fence and a collision body in the construction area, which is suitable for safety management of hoisting in substations.

[0004] The technical solution of this invention is: a method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network, comprising the following steps: S1. Obtain the coordinates of the vertices of the two-dimensional polygon at the bottom of the restricted area (e.g., the high-voltage live restricted area of ​​a substation, or a temporary construction area), set the protection height to be adapted to the high-altitude hoisting of the substation, and uniformly sample along the edge of the polygon with the set edge step length and the vertical height (z-axis) with the set vertical step length to generate a closed three-dimensional columnar point cloud. S2. Construct an octomap octree collision volume based on a 3D columnar point cloud (closed 3D columnar point cloud). According to the set fixed voxel resolution, batch fill 3D occupied voxels to generate the first set of 3D restricted area collision models, which are called the octomap model. S3. Based on the 3D columnar point cloud, extract the bottom and top vertices of the columnar fence, correct the polygon vertex wrapping order, and perform triangulation on the polygon using the ear-cutting method. If the ear-cutting method fails to traverse the triangulation, automatically switch to the fan-shaped triangulation for bottom support. Piece together the bottom, top, and side triangular patches to construct a complete closed BVH triangular mesh, and generate the second set of 3D restricted area collision model. The second set of 3D restricted area collision model is called the BVH triangular mesh model. S4. Implement a scenario-specific dual-model fault-tolerant switching mechanism. When the octomap model is valid, the octomap model is given priority as the model for collision detection (e.g., crane, personnel, vehicle collision detection) (corresponding input to the relevant detection modules). When the octomap model fails to build (e.g., due to missing point clouds, electromagnetic interference, etc.), seamlessly switch to the BVH triangular mesh model as the model for collision detection, so as to continue to implement and complete the subsequent no-go zone distance calculation. S5 outputs a scene-specific dual-model fault-tolerant switching mechanism to determine the model (the model used for collision detection), which is then used for subsequent processing such as calculating the safe distance between the boom's three-dimensional attitude and the electrified exclusion zone.

[0005] Preferably, in step S5, the output model native docking collision detection module (e.g., substation boom collision early warning module, especially substation boom layered collision early warning module).

[0006] Preferably, the collision detection module is a collision detection / collision warning module based on FCL (Flexible Collision Library).

[0007] Preferably, the model metadata output in step S5 includes the minimum XYZ boundary, the maximum XYZ boundary, and the coordinates of the geometric center point, which is directly supplied to the AABB pre-filtering stage of collision detection without the need for secondary format conversion.

[0008] Preferably, in the octomap octree collision body construction in step S2, after batch insertion of point clouds, the number of occupied leaf nodes is counted. If the number of effectively occupied leaf nodes is ≤2, the octomap is reconstructed using the point-by-point updateNode method to ensure that the octomap model can be used for collision distance calculation.

[0009] Preferably, the method for correcting the polygon vertex wrapping order in step S3 is to calculate the oriented area of ​​the polygon's XY projection. If the area is negative, the vertex array is reversed to ensure that the triangulation facets are oriented in a uniform manner.

[0010] Preferably, in step S3, an upper limit is set for the number of iterations of the ear-cutting method (for example, 10,000 times). If the complete subdivision is not completed after reaching the upper limit of the number of iterations, a sector subdivision strategy with the first vertex of the polygon as the common vertex is started, and the corresponding sector subdivision is performed.

[0011] Preferably, in step S1, the set edge step length is 0.2m and the set vertical step length is 0.5m.

[0012] Preferably, in step S2, the fixed voxel resolution is set to 0.25m.

[0013] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of any of the substation electronic fence and construction area collision body construction methods disclosed in this invention based on the fusion of point cloud and triangular network.

[0014] The electronic fence and collision body construction system based on point cloud and triangulation network fusion includes a processor and a memory. The memory stores a computer program that can run on the processor. When the processor executes the computer program, it implements the steps of any of the substation electronic fence and construction area collision body construction methods based on point cloud and triangulation network fusion disclosed in this invention.

[0015] The present invention has the following beneficial effects: 1) Achieve full-height three-dimensional fencing, completely eliminate blind spots in the high-altitude monitoring of traditional two-dimensional fencing booms, and fully constrain the full-amplitude extension, retraction, and rotation of the booms; 2) The dual-model automatic fault-tolerant architecture adapts to the complex electromagnetic environment of substations. When point cloud data is abnormal, the triangular mesh provides seamless fallback, preventing missed reports and early warning failures when the crane enters the restricted area. 3) The adaptive double triangulation scheme is suitable for irregular maintenance and live irregular polygons in substations, avoiding distance calculation distortion caused by modeling holes and missing patches; 4) The modeling products are natively compatible with Octomap and FCL collision calculation libraries, and can be directly input into the boom layer detection process without additional format conversion, ensuring the accuracy of distance calculation between the boom and charged body.

[0016] 5) The entire process is customized for safety management and control of substation hoisting. The entire set of coupled technical routes brings comprehensive effects such as zero blind spots at high altitudes, no failure in abnormalities, and accurate modeling of irregular areas. Attached Figure Description

[0017] Figure 1 This is a schematic diagram of the construction method involved in the present invention. Detailed Implementation

[0018] See Figure 1 This invention is adapted to high-risk hoisting operation scenarios in substations. It relies on high-precision positioning terminals to collect the three-dimensional coordinates of personnel, engineering vehicles, and cranes, and fully restores the full posture of boom extension / swing in three-dimensional space. It calculates the safe distance between the boom and live conductors such as high-voltage busbars and disconnect switches in real time. This three-dimensional collision zone bottom-level modeling scheme can provide stable three-dimensional model support for boom intrusion warning and live distance measurement.

[0019] The basic process of this invention is as follows: acquire polygon information of electronic fence / construction area → generate three-dimensional columnar point cloud → construct octomap octree model and adaptive triangulation to construct BVH triangular mesh model → dual-model fault-tolerant selection of output collision object → output model to crane collision early warning calculation module (existing technology can be used, such as FCL-based collision detection / anti-collision module with corresponding native interface).

[0020] The specific operating steps of this invention include: S1. Initialization of polygon coordinates and sampling parameters Read the vertex array of the substation restricted area polygon (vertices ≥ 3); fix the parameters according to the power high-altitude protection standard: set the total height of the fence, the sampling step of the horizontal edge can be 0.2m, and the sampling step of the vertical layer can be 0.5m; traverse each edge of the polygon, insert sampling points evenly, and generate three-dimensional points layer by layer along the Z-axis from the reference ground to the target height, forming a closed columnar point cloud without gaps. The point cloud is stored in the PCL standard PointXYZ format.

[0021] S2, Octomap Octree Model Construction Initialize the octomap::OcTree object, with a resolution that can be set to 0.25m; create an octomap::Pointcloud container to import all column sampling points in batches; call insertPointCloud to insert points in batches, and execute updateInnerOccupancy to update the voxel occupancy status; traverse all leaf nodes to count the number of valid occupied voxels. If there are too few valid voxels (≤2), use updateNode to rebuild the octree point by point to ensure that the model is valid and detectable.

[0022] S3, Adaptive Triangulation and BVH Mesh Construction 1) Vertex processing: Copy the bottom face vertices, add the target height to the Z-axis to generate the top face vertices; calculate the XY projection oriented area of ​​the bottom polygon, if the area is negative, reverse the vertex array order to ensure the uniformity of the partitioning direction; 2) Ear-cutting method: Loop through the vertices to find the "ear triangle", cut off the legal convex triangle and store it in the face array; if the subdivision cannot be completed after 10,000 loops (the specific number of loops can be set according to actual needs), it is determined to be an irregular complex polygon, and automatically switch to sector subdivision, splitting all triangles with the first vertex as the common vertex; 3) Complete cylindrical mesh: The bottom and top surfaces use bifurcated triangles, and two side triangles are generated for each pair of adjacent bottom / top vertices on the sides of the cylinder; all triangular faces are added to the fcl::BVHModelfcl::OBBRSSf model, and endModel is executed to complete the construction of the closed mesh; 4) Calculate the minimum / maximum XYZ boundary and geometric center point of the overall AABB of BVH, and store the metadata for subsequent boom AABB pre-filtering.

[0023] S4, Dual-model fault-tolerant switching logic The system verifies the number of valid voxels occupied by the octomap model in real time: if there are enough valid voxels (>2), the modeling is considered normal and the detection link prioritizes calling the octomap; if the number of valid voxels is ≤2 due to missing point clouds, scanning interference, etc., the BVH triangular mesh is automatically switched as the only collision model.

[0024] S5, Model Output The octomap binary .ot file, BVH mesh metadata, and bounding box center point information are stored synchronously. This output data format is natively compatible with the front-end substation equipment point cloud processing module (or system) and the rear-end boom layered collision warning module (or system), and can be directly loaded for high-precision positioning boom attitude reconstruction and boom and safe distance calculation between boom and live restricted area.

[0025] Uninterrupted monitoring of the crane's restricted area throughout the entire process.

[0026] This invention has the following characteristics: 1) Targeting the specific pain points of substation hoisting, we have developed a complete technical roadmap by reconstructing various general algorithms such as octomap, ear-cutting method, and BVH under scenario constraints, coupling and linking processes, and inventing a dual-model fault-tolerant architecture. 2) Specific scene parameter binding: There is no limit to the fence height, and suitable parameters such as 0.2m edge sampling step, 0.5m height layering step, and 0.25m octree resolution can be used. These parameters match the high-altitude protection requirements of power safety specifications, and random parameters are generally not suitable. 3) Dual-model linkage fault-tolerant overall architecture: Two sets of models that can be input into the same FCL collision detection link are generated simultaneously. When the point cloud is normal, the octree is given priority and the triangular mesh is automatically switched when the data is abnormal, thus realizing a two-layer fallback architecture. 4) Polygon adaptive dual subdivision mechanism: ear-cutting method as the main method and fan-shaped subdivision as the backup method, which is specially adapted to the irregular maintenance and live irregular contours of substations, and overcomes the failure defects of conventional subdivision. 5) The modeling output is natively compatible with the boom layered collision detection module, and the model metadata is directly supplied to the boom distance calculation, forming a complete data link from fence modeling to boom attitude distance measurement.

[0027] 6) The entire set of technical solutions are mutually supportive and inseparable, achieving high-altitude full coverage, high fault tolerance, and no missed reports, which cannot be achieved by simply stacking the individual technologies together.

[0028] This invention provides a three-dimensional restricted area collision body modeling solution for substation hoisting operations as a specific application scenario. It constructs a fully covered three-dimensional columnar restricted area and generates two sets of collision models, octomap and BVH triangular mesh, that can be linked for detection. It adaptively subdivides irregular polygons and establishes an automatic fallback switching mechanism for model failure. This provides a highly reliable three-dimensional restricted area model for high-precision positioning of boom attitude restoration and real-time calculation of the distance between the boom and live parts, and provides corresponding support for boom intrusion early warning and live distance measurement.

[0029] This invention is applicable to substation hoisting safety management scenarios that rely on high-precision positioning terminals to reconstruct the three-dimensional posture of the boom and calculate the safe distance between the boom and high-voltage live conductors.

[0030] Unless otherwise specified, the preferred and optional technical means disclosed in this invention can be arbitrarily combined to form several different specific embodiments when one preferred or optional technical means is a further limitation of another technical means.

Claims

1. A method for constructing electronic fences and collision bodies based on the fusion of point clouds and triangulation networks, including the following steps: S1. Obtain the coordinates of the vertices of the two-dimensional polygon at the bottom of the restricted area, set the protection height to adapt to the high-altitude hoisting of the substation, and uniformly sample along the edge of the polygon with the set edge step length and the vertical height with the set vertical step length to generate a closed three-dimensional columnar point cloud. S2. Construct an octomap octree collision volume based on a 3D columnar point cloud. According to the set fixed voxel resolution, batch fill 3D occupied voxels to generate the first set of 3D restricted area collision models, which are referred to as the octomap model. S3. Based on the 3D columnar point cloud, extract the bottom and top vertices of the columnar fence, correct the polygon vertex wrapping order, and perform triangulation on the polygon using the ear-cutting method. If the ear-cutting method fails to traverse the triangulation, automatically switch to the fan-shaped triangulation for bottom support. Piece together the bottom, top, and side triangular patches to construct a complete closed BVH triangular mesh, and generate the second set of 3D restricted area collision model. The second set of 3D restricted area collision model is called the BVH triangular mesh model. S4. Implement a scenario-specific dual-model fault-tolerant switching mechanism. When the octomap model is valid, the octomap model is used as the collision detection model first. When the octomap model fails to build, seamlessly switch to the BVH triangular mesh model as the collision detection model. S5 outputs a model determined by a scenario-specific dual-model fault-tolerant switching mechanism.

2. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to claim 1, characterized in that... In step S5, the output model native docking collision detection module is generated.

3. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to claim 2, characterized in that... The collision detection module is an FCL-based collision detection / collision warning module.

4. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to claim 3, characterized in that... The model metadata output in step S5 includes the minimum XYZ boundary, the maximum XYZ boundary, and the coordinates of the geometric center point, which are directly supplied to the AABB pre-filtering stage of collision detection without the need for secondary format conversion.

5. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to any one of claims 1-4, characterized in that... In the octomap octree collision body construction in step S2, after batch insertion of point clouds, the number of occupied leaf nodes is counted. If the number of effectively occupied leaf nodes is ≤2, the octomap is reconstructed using the point-by-point updateNode method to ensure that the octomap model can be used for collision distance calculation.

6. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to any one of claims 1-4, characterized in that... In step S3, the method for correcting the polygon vertex wrapping order is to calculate the oriented area of ​​the polygon's XY projection. If the area is negative, the vertex array is reversed to ensure that the triangulation facets face the same direction.

7. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to any one of claims 1-4, characterized in that... In step S3, an upper limit is set for the number of iterations of the ear-cutting method. If the complete subdivision is not completed after reaching the upper limit of the number of iterations, the sector subdivision strategy with the first vertex of the polygon as the common vertex is started, and the corresponding sector subdivision is performed.

8. The method for constructing an electronic fence and collision body based on the fusion of point cloud and triangular network according to any one of claims 1-4, characterized in that... In step S1, the set edge step length is 0.2m and the set vertical step length is 0.5m; in step S2, the set fixed voxel resolution is 0.25m.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that... When the computer program is executed by the processor, it implements the steps of the substation electronic fence and construction area collision body construction method based on point cloud and triangular network fusion as described in any one of claims 1-8.

10. An electronic fence and collision body construction system based on the fusion of point cloud and triangulation network, characterized in that... It includes a processor and a memory, the memory storing a computer program that can run on the processor, and the processor executing the computer program to implement the steps of the substation electronic fence and construction area collision body construction method based on point cloud and triangular network fusion as described in any one of claims 1-8.

Images