Three-dimensional reconstruction method, simulation method and device of mechanical equipment

Abstract

The invention provides a three-dimensional reconstruction method, simulation method and device for mechanical equipment. The three-dimensional reconstruction method comprises the following steps: constructing an initial digital twinborn body of target mechanical equipment based on a preset three-dimensional reconstruction technology; according to the mechanical structure topology of the target mechanical equipment, segmenting the initial digital twin body to obtain a plurality of joint structures; and motion node information representing the joint motion mode is generated for each joint structure, and a target digital twinborn body of the target mechanical equipment is obtained. By means of the joint structure segmentation result and the bound motion node information, the target digital twin has the flexible attitude adjustment capability, and the simulation requirements for diversified working conditions and extreme operation attitudes in mine unmanned driving simulation can be fully met. Besides, the digital twin supports free attitude design and scene expansion, does not need to depend on full-time-sequence data acquisition, can flexibly generate equipment attitudes in different operation scenes, and remarkably improves the diversity and application suitability of simulation data.

Description

Technical Field

[0001] This disclosure relates to the fields of 3D reconstruction, smart mining, and autonomous driving, and in particular to a 3D reconstruction method, simulation method, and device for mechanical equipment. Background Technology

[0002] 3D Gaussian Splatting (3DGS), a cutting-edge technology in 3D reconstruction, has been widely applied in numerous scenarios due to its efficient data processing capabilities and hyper-realistic reconstruction effects. Its core principle is to discretize the target scene into a massive Gaussian-distributed point cloud, and then use the smoothness and continuity of the Gaussian function to complete scene rendering, enabling the rapid generation of high-precision 3D models. However, in the specific industrial scenario of unmanned mining operations, the limitations of this technology are becoming increasingly apparent.

[0003] Mechanical equipment in mining environments (such as excavators and loaders) undergoes drastic and complex posture changes during actual operation. Taking excavators as an example, the extension and rotation of the boom, and the opening, closing, and tilting of the bucket all involve the coordinated movement and posture transformation of multiple components. During the training of unmanned mining driving algorithms, if the data collected from real-world scenarios is insufficient in coverage or lacks information on extreme working conditions, simulation data is needed to supplement and expand the data. Traditional manual modeling methods not only require significant investment of manpower, time, and money, but are also limited by the subjectivity and accuracy bottlenecks of manual modeling, making it difficult to guarantee the consistency between the simulation model and the real mechanical equipment in terms of appearance details and kinematic characteristics, thus affecting the effectiveness of algorithm training.

[0004] While 3DGS technology can quickly reconstruct 3D mining machinery, the reconstructed model is essentially a collection of static point clouds, unable to support dynamic posture adjustments and multi-dimensional transformations. This makes it difficult to meet the simulation requirements of diverse postures and extreme working conditions of machinery in real-world scenarios. Improved techniques for time-series dynamic scenes, such as Deformable 3D Gaussians, can handle some time-varying issues, but they have the following core shortcomings: 1. The data acquisition threshold is extremely high, requiring the acquisition of full-attitude time-series data of the equipment, which is difficult to achieve comprehensive coverage in the complex operating environment of mines; 2. Attitude control relies entirely on the coverage of the collected data, making it impossible to design free attitudes and extend to extreme conditions according to simulation requirements, which greatly limits the diversity of simulation data and the applicability of scenarios.

[0005] In summary, existing 3D reconstruction and simulation technologies cannot effectively solve the problems of insufficient simulation data supply, incomplete attitude coverage, and poor scene adaptability in unmanned driving scenarios in mines. Summary of the Invention

[0006] This disclosure provides a method, simulation method, and apparatus for the three-dimensional reconstruction of mechanical equipment, which addresses the problems of insufficient simulation data supply, incomplete attitude coverage, and poor scene adaptability in existing three-dimensional reconstruction and simulation technologies for unmanned driving scenarios in mines.

[0007] In view of the above problems, in a first aspect, the present disclosure provides a method for three-dimensional reconstruction of mechanical equipment, including: An initial digital twin of the target mechanical equipment is constructed based on a pre-defined 3D reconstruction technology; Based on the mechanical structure topology of the target mechanical equipment, the initial digital twin is segmented to obtain multiple joint structures; For each joint structure, motion node information representing the joint's movement mode is generated to obtain a target digital twin of the target mechanical device.

[0008] In conjunction with the first aspect, in one possible implementation, the motion node information includes: joint identifier, joint rotation center, and rotation axis; The process of generating motion node information representing the motion mode of each joint structure includes: For each joint structure, a joint identifier is generated, and the joint rotation center and rotation axis of that joint structure are determined; and Store the joint identifiers along with the joint rotation center and rotation axis.

[0009] In conjunction with the first aspect, in one possible implementation, the motion node information further includes: motion constraints; The motion constraints include: The lower limit and upper limit of joint rotation angle determined for each joint structure; and The limiting boundaries are determined by collision volume detection for each joint structure.

[0010] In conjunction with the first aspect, in one possible implementation, the target digital twin is represented by a Gaussian ellipsoid; the motion node information is stored as attribute information of each ellipsoid in the corresponding joint structure Gaussian ellipsoid.

[0011] In conjunction with the first aspect, in one possible implementation, the initial digital twin of the target mechanical equipment constructed based on a preset 3D reconstruction technology includes: Acquire video data containing the target mechanical equipment; Based on a preset 3D reconstruction technology, the video data is processed to obtain target point cloud data corresponding to the initial digital twin of the target mechanical equipment; The step of segmenting the initial digital twin based on the mechanical structure topology of the target mechanical equipment includes: Based on the mechanical structure topology of the target mechanical equipment, the target point cloud data is segmented to obtain sub-point cloud data corresponding to multiple joint structures.

[0012] Secondly, a simulation method for mechanical equipment is provided, including: Determine the joint rotation parameters of the target joint structure of the target mechanical equipment; Based on the joint rotation parameters and the motion node information corresponding to the target joint, the posture of the target digital twin of the target mechanical equipment is transformed to obtain a deformable body with a transformed posture. The target digital twin of the target mechanical equipment is obtained based on the first aspect, or in combination with any possible implementation of the first aspect, the three-dimensional reconstruction method of the mechanical equipment.

[0013] In conjunction with the second aspect, in one possible implementation, the step of transforming the posture of the target digital twin of the target mechanical device according to the joint rotation parameters and the motion node information corresponding to the target joint to obtain a deformable body with a transformed posture includes: Based on the joint identifier corresponding to the target joint structure, determine the joint rotation center and rotation axis corresponding to the target joint structure; The pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, the pose of the target digital twin is transformed to obtain a deformable body with a transformed pose.

[0014] In conjunction with the second aspect, in one possible implementation, determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: When the target joint structure includes joint structures other than the terminal joint structure, the pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. The method further includes: when the target joint structure is the end joint structure, transforming the posture of the target digital twin according to the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters, to obtain a deformable body with a transformed posture.

[0015] In conjunction with the second aspect, in one possible implementation, the pose matrix includes: position offset and rotation angle parameters; determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: Determine the last joint structure in the linked joint group where the target joint structure is located, and take the last joint structure as the current joint structure. Perform the following steps until the current joint structure has no parent joint structure: Step 1: Based on the joint rotation parameters of the parent joint structure, determine the position offset of the joint rotation center of the current joint structure relative to the joint rotation center of the parent joint structure, and the rotation angle parameter of the rotation axis of the current joint structure relative to the rotation axis of the parent joint structure; Step 2: Take the parent joint structure of the current joint structure as the new current joint structure, and perform Step 1. For each target joint structure and all its subordinate joint structures, the pose matrix corresponding to the joint structure is determined based on the position offset and rotation angle parameters determined for the joint structure and all its superior joint structures.

[0016] In conjunction with the second aspect, in one possible implementation, the pose of the target digital twin is transformed based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters to obtain a deformable body with a transformed pose, including: Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, determine the transformed joint rotation center and the transformed rotation axis of the target joint structure. Based on the transformed joint rotation center, the transformed rotation axis, and the joint rotation parameters, the posture of the target digital twin is transformed to obtain a deformable body with the transformed posture of the target. Preferably, the method further includes: Based on the deformable body, multimedia data of the target mechanical equipment changing posture is generated.

[0017] Thirdly, a three-dimensional reconstruction device for mechanical equipment is provided, comprising: The 3D reconstruction module is used to construct an initial digital twin of the target mechanical equipment based on preset 3D reconstruction technology. The joint structure segmentation module is used to segment the initial digital twin according to the mechanical structure topology of the target mechanical equipment to obtain multiple joint structures; The target digital twin determination module is used to generate motion node information representing the motion mode of each joint structure, thereby obtaining the target digital twin of the target mechanical device.

[0018] Fourthly, a simulation device for mechanical equipment is provided, comprising: The joint rotation parameter input module is used to determine the joint rotation parameters of the target joint structure of the target mechanical equipment. The attitude transformation module is used to transform the attitude of the target digital twin of the target mechanical equipment according to the joint rotation parameters and the motion node information corresponding to the target joint, so as to obtain a deformable body with the target attitude transformation. The target digital twin of the target mechanical equipment is obtained based on the first aspect, or in combination with any possible implementation of the first aspect, the three-dimensional reconstruction method of the mechanical equipment.

[0019] The beneficial effects of the embodiments disclosed herein include: This disclosure provides a method, simulation method, and apparatus for 3D reconstruction of mechanical equipment, comprising: constructing an initial digital twin of the target mechanical equipment based on a preset 3D reconstruction technology; segmenting the initial digital twin according to the mechanical structure topology of the target mechanical equipment to obtain multiple joint structures; generating motion node information representing the motion mode of each joint structure to obtain a target digital twin of the target mechanical equipment. The 3D reconstruction method for mechanical equipment provided in this disclosure, without changing the existing data acquisition method, can directly use video or images of the mechanical equipment captured by an image acquisition device as the data source for 3D reconstruction. Compared with traditional manual modeling methods, this design significantly reduces manpower input, time costs, and capital consumption, while cleverly avoiding the complex data acquisition process problems of methods such as deformable 3D Gaussian methods. Relying on preset 3D reconstruction technology, an initial digital twin can be quickly built, effectively solving the core contradiction of high cost and low accuracy in traditional manual modeling. The segmentation process strictly follows the mechanical structure topology of the mechanical equipment, ensuring that the division of joint structures completely matches the motion mechanism of the real equipment, fundamentally guaranteeing the motion authenticity and mechanical rationality of the target digital twin, making it accurately fit the operating characteristics of the real mechanical equipment. By leveraging the segmentation results of joint structures and the bound motion node information, the target digital twin possesses flexible attitude adjustment capabilities, fully meeting the simulation needs of diverse working conditions and extreme operating postures in mine unmanned driving simulation. Furthermore, this digital twin supports free attitude design and scene expansion, without relying on full-time data acquisition. It can flexibly generate equipment postures under different operating scenarios, successfully solving the industry problems of insufficient simulation data supply and incomplete scene coverage, significantly improving the diversity and application adaptability of simulation data. The generated large amount of simulation data with varying postures provides rich and diverse training samples for autonomous driving perception algorithms, effectively solving the problems of traditional 3DGS's inability to control posture changes and the limited posture control of deformable 3D Gaussian. The rich and diverse simulation data can cover more mechanical equipment posture change scenarios, allowing autonomous driving perception algorithms to encounter more comprehensive situations during training, thereby improving the algorithm's accuracy in recognizing different types of mechanical equipment, reducing the recognition error rate caused by mechanical equipment posture changes in practical applications, and improving the safety and reliability of mine unmanned driving systems. Attached Figure Description

[0020] Figure 1 A flowchart of a three-dimensional reconstruction method for mechanical equipment provided in this embodiment of the disclosure; Figure 2 A flowchart of a simulation method for mechanical equipment provided in this embodiment of the disclosure; Figure 3 A structural diagram of the three-dimensional reconstruction device for mechanical equipment provided in the embodiments of this disclosure; Figure 4A structural diagram of the simulation device for mechanical equipment provided in the embodiments of this disclosure. Detailed Implementation

[0021] This disclosure provides a method, simulation method, and apparatus for three-dimensional reconstruction of mechanical equipment. Preferred embodiments of this disclosure are described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustrative and explanatory purposes only and are not intended to limit this disclosure. Furthermore, the embodiments and features described in this application can be combined with each other unless otherwise specified.

[0022] This disclosure provides a method for three-dimensional reconstruction of mechanical equipment, such as... Figure 1 As shown, it includes: S101. Construct an initial digital twin of the target mechanical equipment based on preset 3D reconstruction technology; S102. Based on the mechanical structure topology of the target mechanical equipment, the initial digital twin is segmented to obtain multiple joint structures; S103. Generate motion node information representing the motion mode of each joint structure to obtain the target digital twin of the target mechanical equipment.

[0023] In this embodiment, a static 3D reconstruction model is transformed into a target digital twin that can freely adjust its posture and adapt to simulation requirements, solving the problem that the posture of a 3D reconstruction model cannot be freely changed in traditional 3D reconstruction technologies. Using mature 3D reconstruction technology, physical information such as the appearance, size, and key component morphology of the target mechanical equipment (e.g., mining excavators, loaders) is collected to construct a 3D model that matches the real equipment at a 1:1 scale, i.e., the initial digital twin. For example, images of the target mechanical equipment are acquired using image acquisition devices, such as videos or pictures of the excavator taken with a mobile phone. High-efficiency and high-precision technologies such as 3DGS are used to quickly acquire the appearance details (e.g., bucket tooth shape, boom outline) and physical dimensions of the target mechanical equipment, ensuring that the geometric accuracy of the initial 3D model is consistent with the real mechanical equipment. The initial digital twin is essentially a complete 3D model, containing basic information such as the overall structure, appearance texture, and positional relationships of key components, serving as the foundation for subsequent joint structure segmentation and motion. Based on the mechanical structure topology of the target mechanical equipment, the initial digital twin is segmented to obtain multiple joint structures. For example, by combining the actual mechanical design logic of the target machinery (such as component connection relationships and degrees of freedom of motion), the initial digital twin can be modularly decomposed into independently movable joint structures, such as the excavator's base, turntable joint, boom joint, forearm joint, and bucket joint. A segmentation algorithm based on geometric features and topological relationships can be used to accurately segment the joint structures by identifying key features such as connection surfaces and motion axes between components; alternatively, based on engineering experience, the initial digital twin can be manually segmented to obtain multiple joint structures. After segmentation, the relative positional relationships and connection interfaces of each joint structure can be preserved, laying the foundation for the subsequent association of motion nodes.

[0024] Furthermore, motion node information representing the movement mode of each joint structure is generated, resulting in a target digital twin of the target mechanical equipment. For each joint structure, key motion-related parameters are defined, endowing the joint structure with independent movement capabilities, ultimately integrating to form a target digital twin capable of dynamically adjusting posture. For example, motion node information may include one of the following: motion type (e.g., rotational motion, telescopic motion, opening / closing motion), degrees of freedom (e.g., 1 degree of freedom rotation), range of motion (e.g., rotation angle 0-360°, telescopic length 0-5m), motion constraints (e.g., mechanical limits, mechanical load limits), comprehensively representing the joint's motion logic and boundaries. The motion node information of each joint is associated through motion axes and connection interfaces to form a complete motion chain, such as the motion chain formed by the excavator base, turntable, boom, forearm, and bucket. Each joint structure can move independently or collaboratively based on the motion node information, resulting in a target digital twin of the target mechanical equipment capable of posture transformation. The target digital twin possesses both static fidelity and dynamic controllability. It retains the geometric consistency between the initial digital twin and the real mechanical equipment, and can freely generate different operating postures (such as bucket lifting, boom extension, turntable rotation), and even special postures under extreme conditions, by adjusting the motion node information of each joint structure. By endowing the joint structure with motion capabilities, the target digital twin is upgraded from a static initial digital twin to a dynamically controllable simulation carrier. It can be directly used for diverse scene simulations in the training of unmanned driving algorithms in mining, solving the problem that traditional 3D reconstruction technology cannot support posture transformation.

[0025] This application embodiment does not require changes to existing data acquisition methods. Videos or images of the excavator captured by image acquisition equipment can serve as the data source for reconstruction, significantly reducing manpower, time, and financial costs compared to traditional manual modeling. Furthermore, it avoids the complex data acquisition challenges of methods such as deformable 3D Gaussian methods. An initial digital twin is rapidly constructed based on preset 3D reconstruction technology. Joint structure segmentation balances modeling efficiency and practicality, avoiding the high cost and low accuracy issues of traditional manual modeling. Segmentation based on the mechanical structure topology ensures that the joint structure division is consistent with the actual motion logic of the mechanical equipment, guaranteeing the motion realism and mechanical rationality of the target digital twin, and conforming to real mechanical characteristics. Through joint structure segmentation and motion node information, the resulting target digital twin possesses dynamic attitude adjustment capabilities, meeting the simulation needs of diverse working conditions and extreme attitudes in mine unmanned driving simulation. The target digital twin supports free attitude design and expansion, without relying on full-time data acquisition, and can flexibly generate equipment attitudes under different operating scenarios, solving the problems of insufficient and incomplete simulation data supply and improving the diversity of simulation data.

[0026] In another embodiment of this disclosure, the motion node information includes: joint identifier, joint rotation center, and rotation axis.

[0027] In step S103 above, motion node information representing the motion mode of each joint structure is generated, including: For each joint structure, a joint identifier is generated, and the joint rotation center and rotation axis of that joint structure are determined; and Store the joint identifiers along with the joint rotation center and rotation axis.

[0028] In this embodiment, for each joint structure, a standardized construction of motion node information is achieved by generating a unique joint identifier, accurately determining the joint rotation center and rotation axis, and establishing associated storage, providing core data support for the dynamic motion control of the joint structure. For each joint structure, a joint identifier is generated. A unique identification code, i.e., a joint identifier, is assigned to each independent joint structure to distinguish different joint structures. For example, the joint identifier of the excavator's base is 01, and the joint identifier of the turntable joint is 02. Based on the actual mechanical structure topology and motion principle of the target mechanical equipment, the core physical parameters for each joint structure to achieve motion are accurately determined: the joint rotation center (three-dimensional spatial coordinate point) and the rotation axis (spatial straight line). The joint rotation center can refer to the fixed spatial point around which the joint structure rotates; it is the center point of the joint structure's motion, essentially a physical reference point with precise coordinates in three-dimensional space, determining the axis position of the joint rotation and directly affecting the trajectory and posture accuracy of the joint structure's motion. The rotation axis can refer to the spatial straight line passing through the joint rotation center and determining the direction of joint rotation; it is the directional reference for the joint structure's motion, essentially a straight line in three-dimensional space, determining the direction around which the joint can rotate. Establish a relationship between joint identifiers and the corresponding joint rotation center and rotation axis, and store this relationship in a structured format to form complete motion node information. Ensure that each joint identifier uniquely corresponds to a set of joint rotation center and rotation axis data, with no duplication or omissions. Extended fields (such as motion angle range and constraints) can be reserved to provide space for subsequent addition of motion parameters. This achieves structured management of motion node information, enabling the target digital twin to quickly access the corresponding rotation parameters through joint identifiers, supporting dynamic motion calculations of the joint structure (such as angle adjustment around the rotation axis and simulation of multi-joint cooperative motion), and facilitating the maintenance and updating of subsequent motion node information.

[0029] In another embodiment of this disclosure, the motion node information further includes: motion constraints; Motion constraints include: The lower limit and upper limit of joint rotation angle determined for each joint structure; and The limiting boundaries are determined by collision volume detection for each joint structure.

[0030] In this embodiment, in addition to joint identifiers, joint rotation centers, and rotation axes, a lower limit and an upper limit for joint rotation angles, as well as a limiting boundary determined by collision volume detection, are added. Motion constraints are associated with and stored with existing motion node information to form a complete motion node information system, defining safe and compliant boundaries for the dynamic movement of joint structures and ensuring the physical realism of the target digital twin's movement. Based on the mechanical structure topology of the target mechanical equipment, a minimum allowable value for the rotation angle is set as the lower limit for the joint rotation angle, and a maximum allowable value for the rotation angle is set as the upper limit for the joint rotation angle, limiting the range of motion of the joint structure around the rotation axis and matching the limiting structure of the real machinery. The limiting boundary is determined by collision volume detection for each joint structure. For example, a three-dimensional volume model is performed on each joint structure and its adjacent joint structures (such as the excavator boom and turntable, bucket and boom), simulating the spatial position changes of each joint structure during movement, detecting whether there is volume overlap (collision), and determining the limiting boundary in the form of a collision-prohibited spatial region, thus limiting the movement space of the joint structure. For example, high-precision 3D volumetric models are constructed for each joint structure and its adjacent joint structures. These models, such as those based on the geometry of an initial digital twin, generate simplified convex hull volumes or precise mesh volumes to ensure a realistic representation of the spatial occupancy of the joint structures. Collision detection algorithms (such as bounding box hierarchical tree algorithms and GJK algorithms) are used to simulate the full range of joint structure motion from the lower limit to the upper limit of joint rotation angles, calculating the volume overlap between the joint structure and its adjacent joint structures in real time. For instance, during the simulation of an excavator boom rotating from 0° to 60°, the system detects whether the boom overlaps with the side of the turntable. Scenarios involving the coordinated motion of multiple joint structures (such as the coordinated movement of the boom and forearm) are considered, detecting collision risks under combined motion of multiple joint structures. This avoids interference when a single joint structure has no collision but multiple joints move in concert, ensuring the comprehensiveness of the boundary constraints. Thus, motion constraints are supplemented from the spatial relationship dimension of multiple joint structures, avoiding collisions and interference between the joint structure and its adjacent joint structures during motion. This prevents simulation scenarios that violate physical laws (such as the forearm penetrating the turntable), ensuring the spatial rationality of the simulated posture. Setting motion constraints can help automatically intercept out-of-bounds motion and collision risks, eliminating the need for manual verification of posture rationality, reducing the manpower and time costs of simulation debugging, and avoiding distortion of algorithm training data due to incorrect postures.

[0031] In another embodiment of this disclosure, the target digital twin is represented by a Gaussian ellipsoid; motion node information is stored as attribute information of each ellipsoid in the corresponding joint structure Gaussian ellipsoid.

[0032] In this embodiment, the target digital twin of the target mechanical equipment is represented in the form of a Gaussian ellipsoid. The motion node information (joint identifier, joint rotation center, rotation axis, motion constraints, etc.) corresponding to each joint structure is then bound and stored as attribute information of the Gaussian ellipsoid to which that joint structure belongs. This lays the foundation for high-precision dynamic rendering and flexible posture adjustment of the target digital twin. A Gaussian ellipsoid can refer to the core unit in 3DGS technology that represents the geometric shape of a scene or object. By flexibly adjusting parameters, it can accurately fit various geometric structures in the real world, from simple regular shapes to complex curved surfaces, and is particularly suitable for high-precision modeling of target digital twins of mining machinery. Based on 3D Gaussian splashing technology, the target digital twin of the target mechanical equipment is represented in the form of a Gaussian ellipsoid. Each Gaussian ellipsoid can represent the geometric shape and visual attributes of the target digital twin through its position, Gaussian parameters, and the various orders of the spherical harmonic functions, thereby ensuring the geometric fidelity and rendering realism of the target digital twin. Gaussian ellipsoids inherit the high-efficiency rendering advantages of 3D Gaussian splashing technology. Combined with the motion node information stored in the attributes, it can quickly calculate the position and attitude changes of the Gaussian ellipsoid after the joint structure moves, without the need to re-perform complex geometric modeling, reducing the amount of calculation during dynamic rendering and ensuring a real-time interactive experience.

[0033] In another embodiment of this disclosure, step S103 above, constructing an initial digital twin of the target mechanical equipment based on a preset three-dimensional reconstruction technology, includes: Step 1: Acquire video data containing the target mechanical equipment; Step 2: Based on the preset 3D reconstruction technology, process the video data to obtain the target point cloud data corresponding to the initial digital twin of the target mechanical equipment; In step S102 above, the initial digital twin is segmented according to the mechanical structure topology of the target mechanical equipment, including: Step 3: Based on the mechanical structure topology of the target mechanical equipment, segment the target point cloud data to obtain sub-point cloud data corresponding to multiple joint structures.

[0034] In this embodiment, video data containing the target mechanical equipment is acquired, and then processed based on a preset 3D reconstruction technique to generate target point cloud data corresponding to the initial digital twin. Finally, the target point cloud data is segmented according to the topology of the mechanical structure to obtain sub-point cloud data corresponding to each joint structure, laying the geometric data foundation for the subsequent generation of the target digital twin. For step one above, video information of the target mechanical equipment (such as a mining excavator or loader) in a static or weakly dynamic state can be recorded using an image acquisition device to obtain the visual and spatial data required to construct the initial digital twin, which is the basic data source for subsequent 3D reconstruction. For step two above, mature 3D reconstruction techniques (such as 3D Gaussian splashing technology) are used to extract features, estimate camera pose, and calculate depth from the images in the video data, ultimately generating target point cloud data representing the 3D geometric shape of the target mechanical equipment. The point cloud data can be a dataset composed of 3D spatial points, each point containing 3D coordinate information (x, y, z). The target point cloud data can also carry additional attributes such as color and reflection intensity. The target point cloud data is the core geometric carrier of the initial digital twin. Regarding step three above, based on the mechanical structure topology of the target machinery (such as component assembly relationships, joint positions, and structural function division), a point cloud segmentation algorithm can be used to modularly decompose the target point cloud data, dividing it into sub-point cloud data corresponding to the real joint structures (such as base, turntable joints, upper arm joints, and forearm joints), thus achieving structured division of the point cloud. Video data can be acquired through image acquisition devices (such as cameras and mobile phones), eliminating the need for expensive equipment like LiDAR. The acquisition process is flexible and convenient, adapting to complex mining environments and significantly reducing the data acquisition cost and difficulty of initial digital twin construction. 3D reconstruction technology based on video data can fully utilize the texture and spatial information of images, generating target point cloud data that accurately restores the appearance and size proportions of the target machinery, providing high-precision geometric support for the initial digital twin and reducing subsequent modeling errors. Segmenting the target point cloud data according to the mechanical structure topology (such as component connection relationships and motion logic) ensures a one-to-one correspondence between sub-point cloud data and the real joint structures. The generated sub-point cloud data can be directly used as the basis for Gaussian ellipsoid modeling of joint structures without additional data format conversion, thus shortening the construction cycle of the target digital twin.

[0035] Based on the target digital twin of the target mechanical equipment provided in the above embodiments, this disclosure also provides a simulation method for mechanical equipment, such as... Figure 2 As shown, it includes the following steps: S201. Determine the joint rotation parameters of the target joint structure of the target mechanical equipment; S202. Based on the joint rotation parameters and the motion node information corresponding to the target joint, the posture of the target digital twin of the target mechanical equipment is transformed to obtain the deformable body of the target with transformed posture. The target digital twin of the target mechanical equipment can be obtained using the three-dimensional reconstruction method of the mechanical equipment provided in any of the above embodiments.

[0036] In this embodiment, the joint rotation parameters (such as rotation angle and rotation direction) of the target joint structure are clearly defined. Combined with the stored motion node information (joint identifier, joint rotation center, rotation axis, motion constraints, etc.) of the target joint structure, the posture of the target digital twin is dynamically adjusted, ultimately obtaining a target transformed posture deformable body that meets the requirements. This achieves simulation of different working postures of the target mechanical equipment, providing diverse posture data support for scenarios such as training algorithms for unmanned driving in mines. Based on the simulation scenario requirements (such as mining excavation, unloading, and turning), the joint rotation parameters can be the key parameters set for controlling posture changes of the joint structure (i.e., the target joint structure) in the target mechanical equipment that requires posture adjustment. The joint rotation parameters clearly define how the joint structure rotates. For example, the joint rotation parameters can include rotation angle and rotation direction, which are the basis for subsequent posture changes. The rotation angle clearly defines the specific angle value of the joint structure's rotation around the joint rotation center and rotation axis, and needs to be set in conjunction with motion constraints. The rotation direction can be clockwise or counterclockwise, for example, the bucket joint rotates 15° clockwise around the rotation axis. Joint rotation parameters can be manually input through the simulation platform's visual interface, or automatically loaded based on preset working condition templates (such as standard excavation posture templates). Alternatively, random, compliant joint rotation parameters can be generated through algorithms to batch generate diverse simulation postures. Based on the determined joint rotation parameters and combined with the stored motion node information of the target joint structure, algorithms can adjust the spatial position and shape of the Gaussian ellipsoid corresponding to the target joint structure in the target digital twin, achieving dynamic transformation of the overall posture and ultimately generating a deformable body that meets the requirements. This transforms joint rotation parameters into actual posture changes of the target digital twin, generating a physically realistic deformable body that meets simulation needs. This provides a core platform for the simulation application of mining machinery and equipment, while ensuring the accuracy and usability of the simulation results.

[0037] In another embodiment of this disclosure, in step S202 above, the posture of the target digital twin of the target mechanical device is transformed according to the joint rotation parameters and the motion node information corresponding to the target joint to obtain a deformable body with a transformed posture, including: Step (1): Determine the joint rotation center and rotation axis corresponding to the target joint structure based on the joint identifier corresponding to the target joint structure; Step (2): Determine the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters; Step (3): Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, transform the pose of the target digital twin to obtain the deformable body with the transformed pose.

[0038] In this embodiment, the joint rotation center and rotation axis of the target joint structure are located by joint identifiers, and a pose matrix representing the posture change is calculated by combining the joint rotation parameters (establishing transformation rules). Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, the posture of the target digital twin is adjusted to generate a deformable body with the target posture change. For step (i) above, using the joint identifier of the target joint structure as an index, the joint rotation center and rotation axis of the target joint structure are retrieved from the motion node information of the target digital twin to clarify the motion reference for posture change. If the posture of multiple target joint structures (such as the upper arm joint, forearm joint, and bucket joint) needs to be adjusted, the corresponding joint rotation center and rotation axis are retrieved according to their respective joint identifiers to avoid confusion of motion references between different target joint structures. For step (ii) above, the pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters. The pose matrix may include the position offset of the joint rotation center and the rotation angle parameters of the rotation axis. The pose matrix reflects the changes in the joint rotation center and rotation axis of the target joint structure after posture change relative to before posture change. The pose matrix can transform motion references and rotation requirements into standardized mathematical expressions, providing unified calculation rules for subsequent posture transformations. Regarding step (iii) above, using the pose matrix as the core calculation, combined with the original joint rotation center, rotation axis, and joint rotation parameters, the spatial position of the Gaussian ellipsoid to which the target joint structure belongs is updated, and motion constraints are verified, ultimately generating a deformable body that meets the requirements for target posture transformation. Considering the characteristics of mining machinery equipment with multiple joints and high loads, it can support synchronous transformation of multiple joint structures (such as coordinated adjustment of the boom joint, forearm joint, and bucket joint), and the parameterization characteristics of the pose matrix can quickly adapt to different working conditions (such as excavation, unloading, and steering), generating diverse deformable bodies to meet the simulation data requirements of unmanned driving algorithm training for extreme postures and complex movements.

[0039] In another embodiment of this disclosure, step (ii) above, determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: When the target joint structure includes joint structures other than the terminal joint structure, the pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. The method further includes: when the target joint structure is the most distal joint structure, transforming the posture of the target digital twin according to the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters, to obtain a deformable body with a transformed posture.

[0040] In this embodiment, if the target joint structure includes joint structures other than the most distal joint structure (such as the boom joint and forearm joint of an excavator), the pose matrix is ​​determined based on its joint rotation center, rotation axis, and joint rotation parameters. If the target joint structure is the most distal joint structure (such as the bucket joint of an excavator), the pose transformation of the target digital twin is directly combined with its joint rotation center, rotation axis, and rotation parameters to generate a deformable body. Because the joint movements of mining machinery are hierarchically related (e.g., boom joint movement drives synchronous displacement of the forearm joint and bucket joint, while bucket joint movement does not affect boom joint movement), by distinguishing between non-most distal joint structures and most distal joint structures, a pose matrix is ​​determined for non-most distal joint structures (used to drive the transformation of higher-level joint structures), while the most distal joint directly performs pose transformation to conform to the motion transmission logic of real machinery.

[0041] In another embodiment of this disclosure, the pose matrix includes: position offset and rotation angle parameters; in step (ii) above, determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: Determine the last joint structure in the linked joint group where the target joint structure is located, and take the last joint structure as the current joint structure. Perform the following steps until the current joint structure has no parent joint structure: Step 1: Based on the joint rotation parameters of the parent joint structure, determine the position offset of the current joint structure's joint rotation center relative to the parent joint structure's joint rotation center, and the rotation angle parameter of the current joint structure's rotation axis relative to the parent joint structure's rotation axis; Step 2: Take the parent joint structure of the current joint structure as the new current joint structure, and perform Step 1. For each target joint structure and all its subordinate joint structures, the pose matrix corresponding to the joint structure is determined based on the position offset and rotation angle parameters determined for the joint structure and all its superior joint structures.

[0042] In this embodiment, within the linked joint group containing the target joint structure, the last joint structure is locked, and the process traces backward from it to the top-level joint structure. The positional offset of the current joint structure's rotation center relative to the rotation center of the previous-level joint structure, as well as the rotation angle parameter of the current joint structure's rotation axis relative to the rotation axis of the previous-level joint structure, are calculated level by level. Then, for each target joint structure and its subordinate joint structures, the positional offset and rotation angle parameters relative to all superior joint structures are integrated to determine the pose matrix corresponding to that joint structure. The entire process, through hierarchical parameter transmission, ensures that the pose matrices of each joint structure within the linked joint group are interconnected, accurately reflecting the transmission logic of the coordinated motion of multiple joint structures. Linked joint groups are divided according to the motion transmission relationships of mining machinery. For example, in an excavator, all joint structures driven by the base and involved in adjusting the working posture (turntable joint, boom joint, forearm joint, bucket joint) constitute a linked joint group; similarly, in a mining truck, the frame joint, cab joint, and bucket joint constitute a linked joint group, which can be automatically divided according to the mechanical structure topology. In the linkage group, the joint structure that has no downstream joint structure and directly participates in the operation (such as excavator bucket and loader bucket) is selected as the last joint structure. Taking the excavator bucket structure (the last joint structure is the current joint structure), boom structure (the next level), arm structure (the next level), turntable structure (the next level), and base (the top level) as an example, parameters are calculated level by level. First, the positional offset of the bucket structure's joint rotation center relative to the boom structure's joint rotation center, and the rotation angle parameter of the bucket structure's rotation axis relative to the boom structure's rotation axis are determined. Then, taking the boom structure as the new current joint structure, the positional offset of the boom structure's joint rotation center relative to the arm structure's joint rotation center, and the rotation angle parameter of the boom structure's rotation axis relative to the arm structure's rotation axis are determined. This continues until the current joint structure is the base, and there is no next-level joint structure. For each target joint structure and all its subordinate joint structures within the linked joint group, the position offsets and rotation angle parameters determined for the joint structure itself and with all its superior joint structures are integrated to obtain position offsets and rotation angle parameters that encompass the effects of motion across all levels. This serves as the pose matrix corresponding to the joint structure. For example, for the forearm structure, the position offsets of the joint rotation centers of the forearm structure relative to the upper arm structure, the upper arm structure relative to the turntable structure, and the turntable structure relative to the base, along with the rotation angle parameters of the rotation axes, are integrated to form the pose matrix for the forearm structure. For the upper arm structure, the position offsets of the joint rotation centers of the upper arm structure relative to the turntable structure and the turntable structure relative to the base, along with the rotation angle parameters of the rotation axes, are integrated to form the pose matrix for the upper arm structure.The joint structure motion of mining machinery and equipment has a strict hierarchical transmission relationship. By tracing the parameters of each joint structure relative to the previous level joint structure in reverse, this transmission effect can be fully captured, so that the pose matrix can truly reflect the physical laws of the coordinated motion of multiple joint structures.

[0043] In another embodiment of this disclosure, in step (iii) above, the pose of the target digital twin is transformed according to the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters to obtain a deformable body with the transformed pose of the target, including: Step (1): Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, determine the joint rotation center and rotation axis of the target joint structure after transformation; Step (2): Based on the transformed joint rotation center, the transformed rotation axis, and the joint rotation parameters, transform the posture of the target digital twin to obtain the deformable body with the transformed posture of the target. Preferably, the method further includes: Based on the deformable body, generate multimedia data showing the changing posture of the target mechanical equipment.

[0044] In this embodiment, based on the pose matrix and the original joint rotation center and rotation axis, the target joint structure's transformed joint rotation center and transformed rotation axis are calculated. Then, combining the transformed joint rotation center and transformed rotation axis with the joint rotation parameters, the target digital twin's posture is adjusted to generate a deformable body, along with corresponding multimedia data. For step (1) above, based on the joint rotation center and rotation axis of the target joint structure, the position offset of the joint rotation center and the rotation angle parameters of the rotation axis included in the pose matrix are used. Matrix operations are used to calculate the target joint structure's rotation center and transformed rotation axis after posture transformation, providing a precise dynamic reference for subsequent posture adjustment. For step (2) above, based on the determined transformed joint rotation center and transformed rotation axis, and joint rotation parameters (such as rotation angle and rotation direction), the spatial position and shape of the Gaussian ellipsoid to which the target joint structure belongs are adjusted, and the motion constraints are verified, ultimately generating a target transformed posture deformable body that meets the requirements. Based on the generated target transformed posture deformable body, multimedia data such as images, videos, and 3D animations are generated through rendering and export functions, transforming abstract deformable body data into intuitive and visual results, expanding simulation application scenarios. Thus, in multi-target joint structure scenarios, each joint structure updates its baseline and performs transformations in sequence according to its level, ensuring that the linkage posture conforms to the mechanical logic of upstream driving downstream, which is suitable for the simulation needs of various mining machinery such as excavators and loaders.

[0045] Based on the same disclosed concept, this disclosure also provides a three-dimensional reconstruction device for mechanical equipment. Since the principle of solving the problem by these devices is similar to that of the aforementioned three-dimensional reconstruction method for mechanical equipment, the implementation of this device can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0046] This disclosure provides a three-dimensional reconstruction device for mechanical equipment, such as... Figure 3 As shown, it includes: The 3D reconstruction module 301 is used to construct an initial digital twin of the target mechanical equipment based on a preset 3D reconstruction technology; The joint structure segmentation module 302 is used to segment the initial digital twin according to the mechanical structure topology of the target mechanical equipment to obtain multiple joint structures; The target digital twin determination module 303 is used to generate motion node information representing the motion mode of each joint structure, thereby obtaining the target digital twin of the target mechanical device.

[0047] In another embodiment of this disclosure, the motion node information includes: joint identifier, joint rotation center, and rotation axis; The target digital twin determination module 303 is used to generate a joint identifier for each joint structure and determine the joint rotation center and rotation axis of the joint structure; and Store the joint identifiers along with the joint rotation center and rotation axis.

[0048] In another embodiment of this disclosure, the motion node information further includes: motion constraints; The motion constraints include: The lower limit and upper limit of joint rotation angle determined for each joint structure; and The limiting boundaries are determined by collision volume detection for each joint structure.

[0049] In another embodiment of this disclosure, the target digital twin is represented by a Gaussian ellipsoid; the motion node information is stored as attribute information of each ellipsoid in the corresponding joint structure Gaussian ellipsoid.

[0050] In another embodiment of this disclosure, the target digital twin determination module 303 is used for Acquire video data containing the target mechanical equipment; Based on a preset 3D reconstruction technology, the video data is processed to obtain target point cloud data corresponding to the initial digital twin of the target mechanical equipment; The joint structure division module 302 is used for Based on the mechanical structure topology of the target mechanical equipment, the target point cloud data is segmented to obtain sub-point cloud data corresponding to multiple joint structures.

[0051] Based on the same disclosed concept, this disclosure also provides a simulation device for mechanical equipment. Since the principle of solving the problem by these devices is similar to that of the aforementioned simulation method for mechanical equipment, the implementation of this device can refer to the implementation of the aforementioned method, and the repeated parts will not be described again.

[0052] The mechanical equipment simulation device provided in the embodiments of this disclosure, such as... Figure 4 As shown, it includes: The joint rotation parameter input module 401 is used to determine the joint rotation parameters of the target joint structure of the target mechanical equipment. The attitude transformation module 402 is used to transform the attitude of the target digital twin of the target mechanical equipment according to the joint rotation parameters and the motion node information corresponding to the target joint, so as to obtain a deformable body with the target attitude transformation. The target digital twin of the target mechanical equipment can be obtained using the three-dimensional reconstruction method of the mechanical equipment provided in any of the above embodiments.

[0053] In another embodiment of this disclosure, the attitude transformation module 402 is used for Based on the joint identifier corresponding to the target joint structure, determine the joint rotation center and rotation axis corresponding to the target joint structure; The pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, the pose of the target digital twin is transformed to obtain a deformable body with a transformed pose.

[0054] In another embodiment of this disclosure, the attitude transformation module 402 is used for When the target joint structure includes joint structures other than the terminal joint structure, the pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. The attitude transformation module 402 is further configured to transform the attitude of the target digital twin according to the joint rotation center and rotation axis corresponding to the target joint structure and the joint rotation parameters when the target joint structure is the end joint structure, so as to obtain a deformable body with the target changed attitude.

[0055] In another embodiment of this disclosure, the pose matrix includes: position offset and rotation angle parameters; the pose transformation module 402 is used for... Determine the last joint structure in the linked joint group where the target joint structure is located, and take the last joint structure as the current joint structure. Perform the following steps until the current joint structure has no parent joint structure: Step 1: Based on the joint rotation parameters of the parent joint structure, determine the position offset of the joint rotation center of the current joint structure relative to the joint rotation center of the parent joint structure, and the rotation angle parameter of the rotation axis of the current joint structure relative to the rotation axis of the parent joint structure; Step 2: Take the parent joint structure of the current joint structure as the new current joint structure, and perform Step 1. For each target joint structure and all its subordinate joint structures, the pose matrix corresponding to the joint structure is determined based on the position offset and rotation angle parameters determined for the joint structure and all its superior joint structures.

[0056] In another embodiment of this disclosure, the attitude transformation module 402 is used for Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, determine the transformed joint rotation center and the transformed rotation axis of the target joint structure. Based on the transformed joint rotation center, the transformed rotation axis, and the joint rotation parameters, the posture of the target digital twin is transformed to obtain a deformable body with the transformed posture of the target. Preferably, the attitude transformation module 402 is further used for Based on the deformable body, multimedia data of the target mechanical equipment changing posture is generated.

[0057] Through the above description of the embodiments, those skilled in the art can clearly understand that the embodiments of this disclosure can be implemented in hardware or by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solutions of the embodiments of this disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (such as a CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this disclosure.

[0058] Those skilled in the art will understand that the accompanying drawings are merely schematic diagrams of a preferred embodiment, and the modules or processes in the drawings are not necessarily essential for implementing this disclosure.

[0059] Those skilled in the art will understand that the modules in the apparatus of the embodiments can be distributed in the apparatus of the embodiments as described in the embodiments, or they can be located in one or more devices different from this embodiment with corresponding changes. The modules of the above embodiments can be combined into one module, or they can be further divided into multiple sub-modules.

[0060] The sequence numbers of the embodiments disclosed above are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0061] Obviously, those skilled in the art can make various modifications and variations to this disclosure without departing from its spirit and scope. Therefore, if such modifications and variations fall within the scope of the claims of this disclosure and their equivalents, this disclosure is also intended to include such modifications and variations.

Claims

1. A method for three-dimensional reconstruction of mechanical equipment, characterized in that, include: An initial digital twin of the target mechanical equipment is constructed based on a pre-defined 3D reconstruction technology; Based on the mechanical structure topology of the target mechanical equipment, the initial digital twin is segmented to obtain multiple joint structures; For each joint structure, motion node information representing the joint's movement mode is generated to obtain a target digital twin of the target mechanical device.

2. The method as described in claim 1, characterized in that, The motion node information includes: joint identifier, joint rotation center, and rotation axis; The process of generating motion node information representing the motion mode of each joint structure includes: For each joint structure, a joint identifier is generated, and the joint rotation center and rotation axis of that joint structure are determined; and Store the joint identifiers along with the joint rotation center and rotation axis.

3. The method as described in claim 2, characterized in that, The motion node information also includes: motion constraints; The motion constraints include: The lower limit and upper limit of joint rotation angle determined for each joint structure; and The limiting boundaries are determined by collision volume detection for each joint structure.

4. The method according to any one of claims 1-3, characterized in that, The target digital twin is represented by a Gaussian ellipsoid; the motion node information is stored as the attribute information of each ellipsoid in the corresponding joint structure Gaussian ellipsoid.

5. The method as described in claim 1, characterized in that, The initial digital twin of the target mechanical equipment constructed based on the preset three-dimensional reconstruction technology includes: Acquire video data containing the target mechanical equipment; Based on a preset 3D reconstruction technology, the video data is processed to obtain target point cloud data corresponding to the initial digital twin of the target mechanical equipment; The step of segmenting the initial digital twin based on the mechanical structure topology of the target mechanical equipment includes: Based on the mechanical structure topology of the target mechanical equipment, the target point cloud data is segmented to obtain sub-point cloud data corresponding to multiple joint structures.

6. A simulation method for mechanical equipment, characterized in that, include: Determine the joint rotation parameters of the target joint structure of the target mechanical equipment; Based on the joint rotation parameters and the motion node information corresponding to the target joint, the posture of the target digital twin of the target mechanical equipment is transformed to obtain a deformable body with a transformed posture. The target digital twin of the target mechanical equipment is obtained based on the three-dimensional reconstruction method of the mechanical equipment according to any one of claims 1-5.

7. The method as described in claim 6, characterized in that, The step of transforming the posture of the target digital twin of the target mechanical device according to the joint rotation parameters and the motion node information corresponding to the target joint to obtain a deformable body with a transformed posture includes: Based on the joint identifier corresponding to the target joint structure, determine the joint rotation center and rotation axis corresponding to the target joint structure; The pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, the pose of the target digital twin is transformed to obtain a deformable body with a transformed pose.

8. The method as described in claim 7, characterized in that, The step of determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: When the target joint structure includes joint structures other than the terminal joint structure, the pose matrix of the target joint structure is determined based on the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters. The method further includes: when the target joint structure is the end joint structure, transforming the posture of the target digital twin according to the joint rotation center and rotation axis corresponding to the target joint structure, as well as the joint rotation parameters, to obtain a deformable body with a transformed posture.

9. The method as described in claim 7 or 8, characterized in that, The pose matrix includes: position offset and rotation angle parameters; determining the pose matrix of the target joint structure based on the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, includes: Determine the last joint structure in the linked joint group where the target joint structure is located, and take the last joint structure as the current joint structure. Perform the following steps until the current joint structure has no parent joint structure: Step 1: Based on the joint rotation parameters of the parent joint structure, determine the position offset of the joint rotation center of the current joint structure relative to the joint rotation center of the parent joint structure, and the rotation angle parameter of the rotation axis of the current joint structure relative to the rotation axis of the parent joint structure; Step 2: Take the parent joint structure of the current joint structure as the new current joint structure, and perform Step 1. For each target joint structure and all its subordinate joint structures, the pose matrix corresponding to the joint structure is determined based on the position offset and rotation angle parameters determined for the joint structure and all its superior joint structures.

10. The method as described in claim 7, characterized in that, Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, and the joint rotation parameters, the pose of the target digital twin is transformed to obtain a deformable body with the transformed pose, including: Based on the pose matrix, the joint rotation center and rotation axis corresponding to the target joint structure, determine the transformed joint rotation center and the transformed rotation axis of the target joint structure. Based on the transformed joint rotation center, the transformed rotation axis, and the joint rotation parameters, the posture of the target digital twin is transformed to obtain a deformable body with the transformed posture of the target. Preferably, the method further includes: Based on the deformable body, multimedia data of the target mechanical equipment changing posture is generated.

11. A three-dimensional reconstruction device for mechanical equipment, characterized in that, include: The 3D reconstruction module is used to construct an initial digital twin of the target mechanical equipment based on preset 3D reconstruction technology. The joint structure segmentation module is used to segment the initial digital twin according to the mechanical structure topology of the target mechanical equipment to obtain multiple joint structures; The target digital twin determination module is used to generate motion node information representing the motion mode of each joint structure, thereby obtaining the target digital twin of the target mechanical device.

12. A simulation device for mechanical equipment, characterized in that, include: The joint rotation parameter input module is used to determine the joint rotation parameters of the target joint structure of the target mechanical equipment. The attitude transformation module is used to transform the attitude of the target digital twin of the target mechanical equipment according to the joint rotation parameters and the motion node information corresponding to the target joint, so as to obtain a deformable body with the target attitude transformation. The target digital twin of the target mechanical equipment is obtained based on the three-dimensional reconstruction method of the mechanical equipment according to any one of claims 1-5.

Images