Elevator accessory spraying system based on hierarchical graph optimization

Abstract

The invention relates to the technical field of spraying control, in particular to an elevator accessory spraying system based on hierarchical graph optimization, which comprises front-end acquisition equipment, a server and a spraying device, and is characterized in that the server comprises an image acquisition module, a three-dimensional modeling module, a hierarchical graph construction module, a track planning module and an execution module. The system collects multi-view images through front-end equipment to generate a three-dimensional grid model, a coarse-grained graph and a fine-grained graph are constructed in a layered mode, global path rapid searching is conducted on the coarse-grained graph through an A * algorithm, local path accurate fitting is conducted on the fine-grained graph through a Bezier curve, and a spraying device is controlled to execute operation according to the path. According to the system, the track planning efficiency can be improved, meanwhile, it is ensured that the spraying track is attached to the surface features of the accessories, and the track planning efficiency and the spraying quality are improved.

Description

Technical Field

[0001] This invention relates to the field of spraying control technology, and more specifically to a spraying system for elevator parts based on hierarchical graph optimization. Background Technology

[0002] In the field of elevator component manufacturing, the spraying process is a crucial step in ensuring the corrosion resistance and appearance quality of components such as guide rails and car panels. Traditional spraying methods mainly rely on manual operation. However, with the large-scale development of the elevator industry, higher requirements have been placed on the precision, efficiency, and automation of component spraying. On the one hand, modern elevator production lines require the spraying time of individual components to be minimized, and traditional manual or semi-automatic spraying methods can no longer meet the needs of mass production. On the other hand, the surface features of new elevator components are becoming increasingly complex (such as the toothed curved surface of T-shaped guide rails, the hollow grid structure of car panels, and the polyhedral edges of counterweights). Traditional spraying trajectory planning technology is prone to problems such as missed spraying, over-spraying, or uneven coating thickness.

[0003] The development of image recognition and automation technology has provided a technical foundation for intelligent spraying solutions. For example, Chinese invention patent application publication number CN119973984A, entitled "An Intelligent Spraying Method, Device and Electronic Equipment Based on Visual Recognition", acquires images of the duct to be sprayed through a visual sensor, extracts the contour and position coordinates through segmentation, determines the spatial direction using a key point detection algorithm, and then generates a three-dimensional parametric model. Based on this, a mesh model is constructed and the spray nozzle motion trajectory is planned using the Dijkstra algorithm, thereby controlling the robotic arm to complete the spraying. However, this scheme still has the following shortcomings when dealing with the complex geometry of elevator components: For detailed structures such as curved sections of curved guide rails and sharp corners of hollow panels, Dijkstra's path search based on the original mesh is prone to trajectory jumps, resulting in uneven coating thickness; Dijkstra's algorithm aims at the shortest global path, making it difficult to accurately fit local details. For example, in the toothed area of ​​the guide rail, shortening the path length may cause the nozzle trajectory to deviate from the toothed contour, resulting in local coatings that are too thin; in addition, when the component has polyhedral turns (such as counterweight corners) or branching structures (such as guide rail connectors), Dijkstra's algorithm may get stuck in local optima, leading to incomplete coating coverage. Therefore, there is an urgent need for a spray trajectory planning scheme that can balance fitting accuracy and planning efficiency. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an elevator component spraying system based on hierarchical graph optimization, which can improve trajectory planning efficiency while ensuring the adhesion of the spraying trajectory to the surface features of the component.

[0005] The basic solution provided by this invention is: an elevator component spraying system based on hierarchical graph optimization, comprising a front-end data acquisition device, a server, and a spraying device; the server includes: The image acquisition module is used to acquire multi-view images of elevator components captured by the front-end acquisition device; A 3D modeling module is used to generate a 3D mesh model of elevator components based on the multi-view images; The layered graph construction module is used to construct the three-dimensional mesh model into coarse-grained and fine-grained graphs. The coarse-grained graph is constructed by voxelizing the three-dimensional mesh model and is used for fast global path search. The fine-grained graph is constructed by extracting the original mesh model of the corresponding region with the nodes of the coarse-grained graph as the spatial center, preserving the original topological relationship of the mesh vertices and edges, and is used for accurate fitting of local paths. The trajectory planning module is used to search for a global path using the A* algorithm in the coarse-grained graph, and to optimize the local path by Bézier curve fitting in the fine-grained graph based on the global path. The execution module is used to control the spraying device to perform spraying operations according to the optimized path.

[0006] The principle of this invention is as follows: the image acquisition module acquires multi-view images of elevator components collected by the front-end acquisition device; the 3D modeling module generates a 3D mesh model of the components based on the multi-view images; the layered graph construction module performs layered processing on the 3D mesh model, constructs a coarse-grained graph through voxelization to achieve a fast global path search framework, and simultaneously extracts the original mesh with coarse-grained nodes as the center to construct a fine-grained graph to preserve surface detail topology; the trajectory planning module first searches for a global path in the coarse-grained graph using the A* algorithm, then maps the global path to the fine-grained graph, and smoothly fits the local vertex sequence using Bézier curves to generate a continuous trajectory; the execution module generates motion control commands based on the optimized trajectory to drive the spraying device to complete precise spraying.

[0007] The beneficial effects of this invention are as follows: 1. By discretizing complex 3D mesh models into regular voxel units through voxelization, a node network of a coarse-grained graph is constructed. The heuristic search of the A* algorithm in the coarse-grained graph can quickly locate the global optimal path direction, effectively avoiding the problem of traditional algorithms getting stuck in local optima in complex structures, shortening path planning time and improving trajectory planning efficiency.

[0008] 2. By preserving the original mesh topology of the elevator parts surface through fine-grained maps and combining it with the smooth fitting optimization of local paths using Bézier curves, the spraying trajectory can accurately fit the complex geometric features of the parts (such as curved surfaces, corners, and edges), effectively eliminating problems such as missed spraying, overspraying, or uneven coating thickness caused by trajectory jumps in traditional technologies, thereby improving spraying quality and trajectory accuracy.

[0009] 3. The hierarchical design of the layered map constructs a two-stage trajectory planning mode of global planning and local optimization. The coarse-grained map ensures the planning efficiency of the overall spraying path, while the fine-grained map and Bézier curves achieve accurate fitting of local details, solving the problem that traditional single algorithms cannot simultaneously balance efficiency and accuracy.

[0010] Furthermore, the process of constructing a coarse-grained graph by the layered graph construction module includes: voxelizing the 3D mesh model according to a preset voxel size, treating each voxel as a node in the coarse-grained graph, and establishing undirected edge connections between nodes only when they share a face along the coordinate axis in a 3D Cartesian coordinate system. By voxelizing the 3D mesh model, the complex 3D model is discretized into regular voxel units, each voxel serving as a node in the coarse-grained graph, and undirected edge connections are established only when nodes share a face along the coordinate axis in a 3D Cartesian coordinate system. This connection method conforms to the natural properties of adjacency relationships in 3D space, making the path more closely match the actual spatial motion trajectory during subsequent global path search in the coarse-grained graph, reducing unnecessary path detours, and further improving the efficiency and rationality of global path search.

[0011] Furthermore, the process of constructing the fine-grained map by the layered map construction module includes: using the nodes of the coarse-grained map as the center, extracting the original mesh model of the corresponding region as the fine-grained map, and preserving the original topological relationships of the mesh vertices and edges. Using the nodes of the coarse-grained map as the spatial center, extracting the original mesh model of the corresponding region as the fine-grained map, and preserving the original topological relationships of the mesh vertices and edges, ensures that the detailed features of the accessory surface (such as the toothed curved surface of the T-shaped guide rail and the hollowed-out mesh structure of the car panel) are accurately geometrically represented in the fine-grained map. This processing provides an accurate geometric basis for the subsequent Bezier curve fitting of the local path, effectively avoiding trajectory deviation caused by model simplification, and ensuring the coating uniformity and coverage integrity of the local sprayed area.

[0012] Furthermore, the trajectory planning module includes a jump pointer, used to identify and skip invalid nodes during the coarse-grained graph search for a global path, thereby shortening the search path. Invalid nodes include voxel nodes completely contained within the solid model, as well as voxel nodes belonging to a preset non-painted area. The process of determining whether a voxel node is completely contained within the elevator component solid model includes: emitting rays from the node center along the six axes of the three-dimensional Cartesian coordinate system; if the number of intersections between any ray and the solid model surface is odd, then the node is determined to be completely contained within the solid model. During the coarse-grained graph search, the jump pointer can identify invalid nodes corresponding to internal cavities and non-painted areas of the component in real time, and directly skip these nodes, avoiding redundant searches on invalid nodes by the A* algorithm. By directly connecting valid nodes through the jump pointer, the search length of the global path is shortened, further improving path planning efficiency. Especially when dealing with multi-faceted turning or branching structures such as counterweights, corners, and guide rail connectors, it can effectively prevent the algorithm from getting trapped in local optima, ensuring the shortest and most feasible global path.

[0013] Furthermore, the trajectory planning module's process of optimizing the local path using Bézier curve fitting includes: determining the corresponding node sequence in the coarse-grained graph based on the global path, extracting the vertex sequence in the fine-grained graph corresponding to each node; and smoothly fitting the vertex sequence using a cubic Bézier curve to generate a continuous spraying trajectory. Smoothing the vertex sequence in the fine-grained graph using a cubic Bézier curve transforms the discrete vertex sequence into a continuous and differentiable curve trajectory, eliminating trajectory jumps in areas such as curved guide rail surfaces and sharp corners of hollow panels, making the nozzle movement smoother and more continuous, and ensuring uniform coating thickness. Simultaneously, utilizing the parametric characteristics of Bézier curves, the curve shape can be dynamically adjusted according to the surface curvature of the accessory, and the trajectory can be dynamically optimized based on the surface curvature of the accessory, further improving the spraying accuracy in detailed areas.

[0014] Furthermore, the execution module controls the spraying device to perform spraying operations by sending motion control commands. These motion control commands are generated based on the Bézier curve control points output by the trajectory planning module. Through parametric control point design, the motion trajectory and speed parameters of the spraying device can be precisely controlled, enabling the spraying device to strictly follow the path planning of the Bézier curve and achieve precise spraying operations on complex and detailed areas such as the toothed curved surface of the T-shaped guide rail and the hollowed-out corners of the car panel. Attached Figure Description

[0015] Figure 1 This is a system module diagram of an embodiment of an elevator parts spraying system based on hierarchical graph optimization according to the present invention. Detailed Implementation

[0016] The following detailed description illustrates the specific implementation method: The basic implementation examples are as follows: Figure 1 As shown, an elevator component spraying system based on hierarchical graph optimization includes a front-end data acquisition device, a server, and a spraying device; the server includes: The image acquisition module is used to acquire multi-view images of elevator components captured by the front-end acquisition device; A 3D modeling module is used to generate a 3D mesh model of elevator components based on the multi-view images; The layered graph construction module is used to construct the 3D mesh model into coarse-grained and fine-grained graphs. The coarse-grained graph is constructed by voxelizing the 3D mesh model and is used for fast global path search. The fine-grained graph is constructed by extracting the original mesh model of the corresponding region with the nodes of the coarse-grained graph as the spatial center, and retaining the original topological relationship of the mesh vertices and edges for accurate local path fitting. The trajectory planning module is used to search for a global path using the A* algorithm in the coarse-grained graph, and to optimize the local path by Bézier curve fitting in the fine-grained graph based on the global path. The execution module is used to control the spraying device to perform spraying operations according to the optimized path.

[0017] In this embodiment, the front-end acquisition device uses a binocular 3D camera, which is paired with a rotating worktable to collect images around the elevator parts. A set of binocular images is collected at 15° intervals, for a total of 24 perspective image data, thereby comprehensively covering the surface features of the elevator parts.

[0018] The server's hardware configuration is as follows: CPU is AMD Ryzen 7 5800X; GPU is NVIDIA GeForce RTX 3060; memory is 64GB DDR4; storage consists of a 512GB NVMe SSD and a 2TB hard drive; operating system is Ubuntu 20.04 LTS; 3D modeling is based on Meshlab.

[0019] The specific implementation process is as follows: The elevator parts are fixed on the center of the rotating worktable. The worktable rotates clockwise once every 15°. The camera simultaneously acquires a set of binocular images, acquiring a total of 24 viewpoints and 48 images.

[0020] The process of constructing a coarse-grained graph by the layered graph construction module includes: performing voxelization on the three-dimensional mesh model according to the preset voxel size, taking each voxel as a node of the coarse-grained graph, and establishing undirected edge connections between the nodes only when they share a face along the coordinate axis in the three-dimensional Cartesian coordinate system.

[0021] The preset voxel size is determined according to the following formula:

[0022] in, This refers to the minimum feature size of the 3D model of elevator components (e.g., the minimum groove width of a T-shaped guide rail tooth). Taking 1 / 5 of the minimum feature size as the voxel size ensures that each feature forms a continuous sequence of nodes in the coarse-grained map, avoiding feature breakage caused by discretization. For example, when the minimum size of the component is... ,but ; The 3D mesh model is voxelized into an axis-aligned bounding box. The bounding box is the smallest cuboid capable of accommodating the components. After voxelization, each voxel has a side length of [missing information]. The cube is used, with each voxel as a node in the coarse-grained map, where nodes are only related if they are aligned in a three-dimensional Cartesian coordinate system. When a face is shared along an axis, an undirected edge connection is established.

[0023] The process of constructing a fine-grained graph by the layered graph construction module includes: taking the nodes of the coarse-grained graph as the center, extracting the original mesh model of the corresponding region as the fine-grained graph, and preserving the original topological relationship of the mesh vertices and edges.

[0024] Using the center of the coarse-grained graph node as the origin, extract the side length as The original mesh model within the cube region, if the number of mesh vertices within the region is less than 20, is dynamically expanded to a side length of... Cube region; Fine-grained graphs are stored as undirected graphs ,in To extract the set of grid vertices within the region; For the set of edges between vertices, only retain edges with lengths less than [a certain value] in the original mesh. To avoid loss of detail due to long-distance connections.

[0025] The trajectory planning module is used to search for a global path in the coarse-grained graph using the A* algorithm, and to optimize the local path in the fine-grained graph by Bézier curve fitting based on the global path. The trajectory planning module includes a jump pointer, which is used to identify and skip invalid nodes to shorten the search path during the global path search in the coarse-grained graph. The invalid nodes include voxel nodes that are completely contained by the solid model, as well as voxel nodes that belong to a preset non-painted area. The process of determining whether a voxel node is completely contained by the elevator component solid model includes: emitting rays from the center of the node to six axes. If the number of intersection points between the ray and the solid model in any direction is odd, then the node is determined to be completely contained by the solid model.

[0026] The pre-defined non-painted areas include accessory mounting holes, internal cavities, etc., which are annotated using the CAD model to form non-painted areas composed of coordinate points. For each voxel node, the six-axis ray method along the three-dimensional Cartesian coordinate system is used to determine whether it is a completely contained internal node: from the node center outwards... If a ray is emitted in any direction and the number of intersections between the ray and the model surface is odd, then it is considered an invalid node that is completely contained within the model. The state representation of global path search using the A* algorithm is as follows: coarse-grained graph nodes The state is three-dimensional coordinates The heuristic function uses Euclidean distance:

[0027] in The coordinates of the target node.

[0028] The process of skipping invalid nodes using the jump pointer optimization is as follows: During the A* algorithm's node expansion process, if the current node... If a node is invalid, then traverse all its neighboring nodes; if a valid node exists... Then, directly add it to the OPEN list of the A* algorithm. The parent node points to and skip Subsequent expansion; if three consecutive nodes satisfy and Adjacent and intermediate nodes If it is an invalid node, delete it. direct connection and .

[0029] The process of optimizing the local path by Bézier curve fitting in the trajectory planning module includes: determining the corresponding node sequence in the coarse-grained graph based on the global path, extracting the vertex sequence in the fine-grained graph corresponding to each node; and using a cubic Bézier curve to smoothly fit the vertex sequence to generate a continuous spraying trajectory.

[0030] The mathematical model of a cubic Bézier curve is as follows, given control points. The parametric equation of the curve is:

[0031] in Used to describe the position of a point on a curve. The value of affects the contribution of each control point to the curve through polynomial weights, such as... wait; The control point generation process is as follows: Fine-grained graph vertex sequence ; The coarse-grained node sequence corresponding to the global path ; in It is a fine-grained diagram An ordered arrangement of vertices generated according to the mesh topology order; step: For each coarse-grained node Extract its corresponding fine-grained vertex sequence Sort by grid topology; for Downsampling (sampling rate 1 / 3) is performed to obtain a point sequence. ; Calculate intermediate control points and :

[0032] in for The surface normal vector of the corresponding region is obtained by weighted averaging of the mesh vertex normal vectors; Generate Bézier curves ,connect and And smoothly transition between adjacent curve segments.

[0033] The trajectory smoothing optimization process is as follows: For adjacent Bézier curve segments and Ensure that the first derivative is continuous at the endpoints:

[0034] in for The first derivative.

[0035] The execution module controls the spraying device to perform the spraying operation by sending motion control commands, which are generated based on the Bézier curve control points output by the trajectory planning module.

[0036] The above are merely embodiments of the present invention. Commonly known structures and characteristics are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A coating system for elevator components based on hierarchical graph optimization, characterized in that: It includes front-end data acquisition equipment, a server, and a spraying device; the server includes: The image acquisition module is used to acquire multi-view images of elevator components captured by the front-end acquisition device; A 3D modeling module is used to generate a 3D mesh model of elevator components based on the multi-view images; The layered graph construction module is used to construct the three-dimensional mesh model into coarse-grained and fine-grained graphs. The coarse-grained graph is constructed by voxelizing the three-dimensional mesh model and is used for fast global path search. The fine-grained graph is constructed by extracting the original mesh model of the corresponding region with the nodes of the coarse-grained graph as the spatial center, preserving the original topological relationship of the mesh vertices and edges, and is used for accurate fitting of local paths. The trajectory planning module is used to search for a global path using the A* algorithm in the coarse-grained graph, and to optimize the local path by Bézier curve fitting in the fine-grained graph based on the global path. The execution module is used to control the spraying device to perform spraying operations according to the optimized path.

2. The elevator component spraying system based on hierarchical graph optimization according to claim 1, characterized in that: The process of constructing a coarse-grained graph by the layered graph construction module includes: performing voxelization on the three-dimensional mesh model according to the preset voxel size, taking each voxel as a node of the coarse-grained graph, and establishing undirected edge connections between the nodes only when they share a face along the coordinate axis in the three-dimensional Cartesian coordinate system.

3. The elevator component spraying system based on hierarchical graph optimization according to claim 2, characterized in that: The process of constructing a fine-grained graph by the layered graph construction module includes: taking the nodes of the coarse-grained graph as the center, extracting the original mesh model of the corresponding region as the fine-grained graph, and preserving the original topological relationship of the mesh vertices and edges.

4. The elevator component spraying system based on hierarchical graph optimization according to claim 3, characterized in that: The trajectory planning module includes a jump pointer, which is used to determine and skip invalid nodes in the process of searching the global path in the coarse-grained map in order to shorten the search path. The invalid nodes include voxel nodes that are completely contained by the entity model, as well as voxel nodes that belong to the preset non-painted area. The process of determining whether a voxel node is completely contained within the solid model of the elevator component includes: emitting rays from the center of the node along the six axes of the three-dimensional Cartesian coordinate system; if the number of intersections between the ray along any axis and the surface of the solid model is odd, then the node is determined to be completely contained within the solid model.

5. The elevator component spraying system based on hierarchical graph optimization according to claim 4, characterized in that: The process of optimizing the local path by Bézier curve fitting in the trajectory planning module includes: determining the corresponding node sequence in the coarse-grained graph based on the global path, extracting the vertex sequence in the fine-grained graph corresponding to each node; and using a cubic Bézier curve to smoothly fit the vertex sequence to generate a continuous spraying trajectory.

6. The elevator component spraying system based on hierarchical graph optimization according to claim 5, characterized in that: The execution module controls the spraying device to perform the spraying operation by sending motion control commands, which are generated based on the Bézier curve control points output by the trajectory planning module.

Images