Digital twinborn scene adaptive optimization method based on point cloud data

Abstract

The invention relates to the technical field of digital twinning, particularly provides a digital twinning scene adaptive optimization method based on point cloud data, and solves the problems that smoothness and detail presentation cannot be balanced during network fluctuation, and a scene prediction optimization mechanism based on space-time semantic analysis is not established. The method comprises the following steps: data preprocessing and hierarchical construction: carrying out preprocessing of abnormal point elimination, missing value complementation and density adjustment on original point cloud data, dividing key and common regions according to scene requirements, and constructing a multi-scale and multi-resolution hierarchical point cloud data system; the method can deeply analyze the spatio-temporal dynamic characteristics of the point cloud data through the technologies of spatio-temporal dynamic semantic segmentation, dynamic change monitoring, prediction optimization and the like, processes the continuous time sequence point cloud data after adaptive transmission, models a time dependency relationship and extracts local and global characteristics in time and space dimensions respectively, and improves the accuracy of point cloud data processing. Accurate identification of object types, positions and dynamic change information is realized.

Description

Technical Field

[0001] This invention relates to the field of digital twin technology, and in particular to an adaptive optimization method for digital twin scenarios based on point cloud data. Background Technology

[0002] With the widespread application of digital twin technology in various fields, point cloud data plays a crucial role as the core foundation for building virtual models. Point cloud data can accurately describe the geometry, position, and spatial relationships of objects in real-world scenarios, providing rich information for the construction of digital twin scenarios. However, point cloud data is characterized by its massive data volume, posing numerous challenges in transmission and processing. How to efficiently process, transmit, and utilize point cloud data to achieve accurate simulation and optimization of digital twin scenarios has become a key issue that urgently needs to be addressed in the current development of digital twin technology, which is mainly applied in power grid systems.

[0003] Existing technologies have significant shortcomings in point cloud data processing, with most focusing on single-function implementation, such as semantic segmentation of static point clouds or single compression and transmission processing, lacking the comprehensive application of multiple key technologies.

[0004] The first prior art is a Chinese patent, "An Automatic Semantic Modeling Method for Point Clouds in Digital Twin Scenarios," application number CN116844004A. It improves the semantic segmentation accuracy of point clouds through multi-dimensional adaptive convolution and attention mechanisms, including steps such as point cloud downsampling, neighborhood feature extraction, and local shape encoding. However, it only focuses on the semantic segmentation of static point clouds and does not involve adaptive compression transmission of point cloud data or dynamic scene prediction optimization.

[0005] The second prior art is a Chinese patent, "Dynamic Point Cloud Compression Method and Apparatus Based on Time-Sequence Awareness," application number CN119379821A. This method extracts inter-frame features through sparse convolution, generates motion vectors for residual compression, and improves the efficiency of dynamic point cloud compression. However, it only solves the problem of point cloud compression efficiency and does not take into account the regional difference accuracy requirements of digital twin scenarios. Furthermore, it does not integrate semantic analysis and dynamic prediction functions, and therefore cannot support scenario optimization decisions.

[0006] The third existing technology is a Chinese patent, "A Method and System for Fast Compression of Dynamic Point Clouds Based on ORB Feature Extraction and IMU Motion Estimation," application number: CN118172386A. However, it only uses 2D image ORB feature mapping and IMU pose estimation to accelerate point cloud registration and reduce motion estimation time. It does not construct a multi-scale hierarchical structure and is not linked with semantic segmentation and dynamic prediction in digital twin scenarios.

[0007] Currently, existing technologies 1, 2, and 3 have the problem that they do not achieve network state adaptive multi-resolution transmission, cannot balance smoothness and detail presentation when the network fluctuates, and do not establish a scene prediction and optimization mechanism based on spatiotemporal semantic analysis, making it difficult to adjust the actual system parameters in advance. Therefore, this invention provides a digital twin scene adaptive optimization method based on point cloud data. Summary of the Invention

[0008] To achieve the above objectives, this invention adopts the following technical solution: Through data preprocessing and hierarchical construction, the original point cloud data is preprocessed, and based on the accuracy requirements of different areas in the digital twin scene, the point cloud data is divided into a multi-scale, multi-resolution hierarchical structure. Data resources are rationally allocated. In terms of adaptive compression and transmission control, network bandwidth, latency, and other parameters, as well as the scene's operating status, are monitored in real time. An adaptive transmission strategy is formulated, and the transmission strategy is flexibly adjusted according to the network environment and scene requirements to improve transmission efficiency and reduce transmission costs. Spatiotemporal dynamic semantic analysis processes the continuous time-series point cloud data after adaptive transmission to achieve dynamic semantic analysis and segmentation, accurately identifying object categories, locations, and dynamic change information. Dynamic prediction and scene optimization are achieved by establishing a real-time monitoring system to compare the semantically segmented point cloud data at adjacent times, predicting the future motion trajectory and state changes of objects, and adjusting the digital twin scene production parameters and equipment operating strategies in advance. Finally, through model feedback and continuous optimization mechanisms, the optimized digital twin scene is compared and verified with the actual scene. The prediction model parameters or structure are continuously adjusted, the digital twin model is updated, and a closed-loop optimization mechanism is formed to continuously improve model accuracy and scene matching.

[0009] This invention provides an adaptive optimization method for digital twin scenes based on point cloud data, the method comprising the following steps: Data preprocessing and hierarchy construction: The raw point cloud data is preprocessed by removing outliers, filling in missing values, and adjusting density. Based on the needs of the scenario, key and ordinary regions are divided to construct a multi-scale, multi-resolution hierarchical point cloud data system. Adaptive compression and transmission control: Real-time monitoring of network status and scene operation, and adaptive compression and transmission control of point cloud data by adjusting data compression methods and transmission order according to network conditions and scene requirements; Spatiotemporal dynamic semantic analysis: Utilizing network models to process continuous time series point cloud data, by analyzing the dynamic changes in the time dimension and the geometric structural features in the spatial dimension, spatiotemporal dynamic semantic analysis and point cloud data annotation are completed; Dynamic prediction and scenario optimization: Compare point cloud data after semantic segmentation at adjacent time points to monitor dynamic changes, extract spatiotemporal features and input them into the prediction model, and adjust the production parameters and operation strategies of the digital twin scenario based on the prediction results; Model feedback and continuous optimization: Compare the optimized digital twin scenario with the actual scenario, analyze the reasons for the deviation, and improve the model from the aspects of data, parameters and structure, update the scenario information, and form a closed-loop optimization mechanism.

[0010] Furthermore, in the data preprocessing and hierarchy construction steps, based on the functional requirements and visual focus of the digital twin scene, key areas requiring high detail and ordinary areas for general display are divided using a hierarchy division formula. For the key areas, a multi-level point cloud data structure from low to high resolution is constructed, with each layer gradually adding detailed information, and the layers are connected through specific spatial subdivision rules. For the ordinary areas, a point cloud data structure with fewer layers and lower resolution is constructed, with different data sampling intervals and detail retention levels set for each layer, forming a complete multi-scale, multi-resolution hierarchical point cloud data system.

[0011] Furthermore, in the data preprocessing and hierarchy construction steps, a hierarchy division formula is used to separate key areas requiring high detail and general display areas. The formula is as follows: ,in To adjust the weighting coefficients for different levels, set them according to the key aspects of the scenario. For the first Point cloud data density of each region It is the maximum point cloud data density across all regions. The maximum number of levels is determined by considering the overall scene complexity and resource constraints.

[0012] Furthermore, in the adaptive compression and transmission control step, network bandwidth values ​​and data transmission latency status information are acquired in real time. Combined with the current loading progress and display content operation status of the digital twin scene, when the network bandwidth is lower than a set threshold and the latency is high, point cloud data of key areas in the low-resolution layer is selected first. Compression processing is performed by reducing the storage precision of attribute information in the point cloud data and merging adjacent similar points to reduce the data volume before transmission, ensuring that the point cloud data of key areas can arrive first. As the network bandwidth increases and the latency decreases, point cloud data of higher resolution layers are transmitted sequentially according to the hierarchical order to supplement the details of the already displayed low-resolution scene. During the transmission process, the position changes of objects in the scene, the user's perspective switching in the scene, and the interactive operation of area zooming are continuously monitored. Once the attention demand of a specific area is detected, the current transmission queue is immediately interrupted, and higher resolution point cloud data of that area is transmitted first. After the transmission is completed, the original transmission order is restored.

[0013] Furthermore, in the spatiotemporal dynamic semantic analysis step, a network model composed of multiple data processing layers is constructed. Continuous time-series point cloud data, after adaptive transmission, is input into the model sequentially in chronological order. In the time dimension processing section, point cloud data from adjacent moments are arranged in chronological order to construct a data sequence. By calculating the differences between corresponding points in the point cloud data at different moments, including coordinate differences and attribute changes, a temporal correlation is established to capture the dynamic change trend of objects over time. In the spatial dimension processing section, neighboring points within a certain range are selected with each point as the center. The spatial geometric relationships between points are analyzed to extract the local geometric features of the object. By integrating the distribution information of all points in the scene, the overall structural features of the object are obtained. Through multi-layered time dimension correlation calculations, spatial dimension feature extraction, and feature fusion operations, the continuous time-series point cloud data is analyzed point by point. Each point is assigned a corresponding object category label to determine the specific position of the object at different moments and to record the changes in coordinates and attribute information of the object during its movement and changes.

[0014] Furthermore, in the spatiotemporal dynamic semantic analysis step, a temporal correlation is established by calculating the differences between corresponding location points in the point cloud data at different times. The calculation formula is as follows: ,in, This represents point cloud data difference calculation, which constructs a difference set by calculating the coordinate difference vector and attribute difference values ​​of corresponding points. It is the first Time and the Difference sets of point cloud data at different times It is the first The point cloud data set at each moment, It is the first A collection of point cloud data at any given time.

[0015] Furthermore, in the spatiotemporal dynamic semantic analysis step, the local geometric features of the extracted object are extracted using the following formula: ,in, It is a point Local geometric features, It is a point Neighborhood point set The number of elements, It is a point Points within the neighborhood, From point Time The vector, It is a vector The length of the mold, It is a point The normal vector is used to measure the relationship between neighboring points and points. The relationship between the normal directions.

[0016] Furthermore, in the dynamic prediction and scene optimization steps, a real-time monitoring process is established. Point cloud data after semantic segmentation at adjacent time points are compared point by point to calculate the comprehensive difference value of each object's corresponding point. A change threshold is set, and when the difference exceeds the threshold, it is determined that the area where the object is located has changed. For the area that has changed, point cloud data of the area at multiple consecutive time points are collected. The coordinate change rate and direction change of the object in the time dimension, as well as the shape deformation characteristics and size change information in the spatial dimension, are analyzed. The extracted spatiotemporal feature data is input into a prediction model composed of multiple data processing units. The model establishes a mapping relationship between the spatiotemporal features of the object and its future state by learning from a large amount of historical spatiotemporal feature data. Based on this, the currently extracted spatiotemporal features are analyzed and predicted to obtain the future state information of the object at multiple time points in the point cloud scene, including motion trajectory coordinates, posture change angles, and shape evolution information. Based on the prediction results, the production parameters and equipment operation strategies involved in the digital twin scene are adjusted accordingly.

[0017] Furthermore, in the dynamic prediction and scene optimization step, the comprehensive difference value of the corresponding point for each object is calculated, and the calculation formula is as follows: ,in, , , This is the coordinate difference. The difference between attributes. It is the combined difference value of corresponding points of an object at adjacent times.

[0018] Furthermore, in the dynamic prediction and scene optimization step, the currently extracted spatiotemporal features are analyzed and predicted based on this to obtain the future state information of objects in the point cloud scene at multiple future time points. The prediction formula is as follows: ,in, It is the predicted number Point cloud data at any given moment, These are weighting coefficients for historical variance data, reflecting the degree of influence of data at different times on the prediction. It is the first The set of differences in point cloud data between the current time step and the previous time step. It is the bias term of the prediction model.

[0019] Furthermore, in the model feedback and continuous optimization step, the object position and state information in the optimized digital twin scene are compared one by one with the corresponding information collected by sensors in the actual scene. The comprehensive deviation value between the two is calculated, including coordinate error and attribute difference. An error tolerance range is set, and when the difference exceeds the range, it is determined that there is a deviation. The accuracy of the input data of the prediction model, the rationality of the model parameter settings, and the integrity of the model structure are investigated and analyzed. If it is found that there is noise interference or missing data in the input data, the data is reprocessed. If the parameter settings are unreasonable, they are corrected by adjusting the parameter value range and changing the parameter calculation method. If there are defects in the model structure, the data processing layer is added or adjusted, and the data transmission method between layers is changed to improve it. At the same time, according to the changes in the actual scene, the object attribute data and scene layout information in the digital twin model are updated. The operation of scene comparison, deviation analysis, model adjustment and data update is repeated to continuously optimize the prediction model and the digital twin model, forming a closed-loop optimization mechanism.

[0020] Furthermore, in the model feedback and continuous optimization step, the comprehensive deviation value between the two is calculated, and the calculation formula is as follows: ,in, For the coordinates and attributes of points in a digital twin scenario, The coordinates and attributes of the corresponding points in the actual scene. It is the comprehensive deviation value between the digital twin scenario and the actual scenario. It refers to the number of points being compared. It is the first in the digital twin scenario The coordinates of the points In real-world scenarios, the first The coordinates of the points It is the first in the digital twin scenario The attribute values ​​of each point, In real-world scenarios, the first The attribute values ​​of each point.

[0021] Compared with the prior art, the present invention has the following advantages and beneficial effects: I. This invention utilizes spatiotemporal dynamic semantic segmentation, dynamic change monitoring, and predictive optimization technologies to deeply analyze the spatiotemporal dynamic characteristics of point cloud data. By processing continuous time-series point cloud data after adaptive transmission, it models temporal dependencies and extracts local and global features in both time and space dimensions, achieving accurate identification of object categories, locations, and dynamic change information. Based on the extracted spatiotemporal features, it accurately predicts the future motion trajectory and state changes of objects, allowing for advance adjustment of digital twin scene production parameters and equipment operation strategies, significantly improving the simulation and optimization capabilities of digital twin scenes for dynamic processes.

[0022] Second, this invention achieves efficient transmission and precise resource allocation of point cloud data by using multi-scale, multi-resolution hierarchical adaptive compression and transmission technology, and by dividing point cloud data into levels according to the accuracy requirements of different areas of the scene. When the network environment is poor, it can prioritize the compression and transmission of low-resolution key area point cloud data to ensure the basic operation of the digital twin scene and avoid slow scene loading or even lag caused by poor data transmission. This effectively improves the smoothness and real-time performance of the scene. When the network conditions improve, the system can gradually transmit higher resolution data to refine scene details and provide users with a richer visual experience. Attached Figure Description

[0023] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart provided in Embodiment 2 of the present invention; Figure 3 This is a flowchart provided in Embodiment 3 of the present invention; Figure 4 This is a flowchart provided in Embodiment 4 of the present invention; Figure 5 This is the flowchart provided in Embodiment 5 of the present invention; Figure 6 This is a flowchart provided in Embodiment 6 of the present invention. Detailed Implementation

[0024] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0025] Hereinafter, the terms "first," "second," etc., are used for descriptive convenience only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0026] In this invention, unless otherwise explicitly specified and limited, the term "connection" should be interpreted broadly. For example, "connection" can be a fixed mechanical connection, a detachable mechanical connection, or an integral part; or, "connection" can be a direct connection or an indirect connection through an intermediate medium. Furthermore, unless otherwise explicitly specified and limited, the term "coupling" should be interpreted broadly. For example, "coupling" can be a direct electrical connection, such as physical contact and electrical conduction between two components; it can also be understood as an electrical connection between different components in a circuit structure through physical lines capable of transmitting electrical signals, such as copper foil or wires on a printed circuit board (PCB), to transmit electrical signals; or, "coupling" can be an indirect electrical connection between two components through an intermediate medium; or, "coupling" can be an electrical connection between two components in a non-contact manner, such as an electrical connection between two components using capacitive coupling to transmit electrical signals.

[0027] In this embodiment of the invention, directional terms such as "up," "down," "left," and "right" may be defined relative to the orientation of the components shown in the accompanying drawings. It should be understood that these directional terms can be relative concepts, used for relative description and clarification, and can change accordingly depending on the orientation of the components in the accompanying drawings.

[0028] The key technical features of the embodiments of the present invention are as follows: First, through multi-dimensional data processing, the quality of point cloud data is guaranteed and resources are allocated reasonably. The adaptive transmission strategy dynamically adjusts data transmission according to network and scene requirements to ensure smooth scene operation. Spatiotemporal dynamic analysis deeply mines the spatiotemporal characteristics of the data, accurately simulates scene dynamics, and accurately predicts and optimizes scene changes based on features and optimizes decision-making. The closed-loop optimization mechanism continuously improves the matching degree and prediction accuracy between the digital twin scene and the actual scene through comparison, analysis and adjustment. The synergistic effect of various technical features achieves efficient optimization of the digital twin scene.

[0029] Example 1: like Figure 1 As shown, an embodiment of the present invention provides a method comprising the following steps: Data preprocessing and hierarchy construction: The raw point cloud data is preprocessed by removing outliers, filling in missing values, and adjusting density. Based on the needs of the scenario, key and ordinary regions are divided to construct a multi-scale, multi-resolution hierarchical point cloud data system. Adaptive compression and transmission control: Real-time monitoring of network status and scene operation, and adaptive compression and transmission control of point cloud data by adjusting data compression methods and transmission order according to network conditions and scene requirements; Spatiotemporal dynamic semantic analysis: Utilizing network models to process continuous time series point cloud data, by analyzing the dynamic changes in the time dimension and the geometric structural features in the spatial dimension, spatiotemporal dynamic semantic analysis and point cloud data annotation are completed; Dynamic prediction and scenario optimization: Compare point cloud data after semantic segmentation at adjacent time points to monitor dynamic changes, extract spatiotemporal features and input them into the prediction model, and adjust the production parameters and operation strategies of the digital twin scenario based on the prediction results; Model feedback and continuous optimization: Compare the optimized digital twin scenario with the actual scenario, analyze the reasons for the deviation, and improve the model from the aspects of data, parameters and structure, update the scenario information, and form a closed-loop optimization mechanism.

[0030] In the above embodiments, the original point cloud data is preprocessed through data preprocessing and hierarchical construction. Based on the accuracy requirements of different areas in the digital twin scene, the point cloud data is divided into a multi-scale, multi-resolution hierarchical structure, and data resources are rationally allocated. In terms of adaptive compression and transmission control, network bandwidth, latency and other parameters and scene operation status are monitored in real time to formulate adaptive transmission strategies. The transmission strategies are flexibly adjusted according to the network environment and scene requirements to improve transmission efficiency and reduce transmission costs. Spatiotemporal dynamic semantic analysis processes the continuous time series point cloud data after adaptive transmission to achieve dynamic semantic analysis and segmentation, accurately identifying object categories, locations and dynamic change information. Dynamic prediction and scene optimization are achieved by establishing a real-time monitoring system to compare the semantically segmented point cloud data at adjacent times to predict the future motion trajectory and state changes of objects, and to adjust the production parameters and equipment operation strategies of the digital twin scene in advance. Finally, through model feedback and continuous optimization mechanisms, the optimized digital twin scene is compared and verified with the actual scene. The prediction model parameters or structure are continuously adjusted, the digital twin model is updated, and a closed-loop optimization mechanism is formed to continuously improve the model accuracy and scene matching degree.

[0031] Example 2: like Figure 2 As shown in Example 1, in the data preprocessing and hierarchical construction steps provided in this embodiment of the invention, after obtaining the original point cloud data, the data is first denoised. By calculating the distance from each point to its neighboring points, points whose distance exceeds a certain multiple of the average distance of the neighboring points are identified as noise points and removed. This effectively eliminates abnormal data points caused by equipment interference and environmental factors, significantly improving the purity of the point cloud data. Next, the problem of missing data is addressed. Based on the spatial distribution and attributes of the points around the missing area, new points are generated at the missing locations, successfully filling the data gaps caused by scanning blind spots and occlusion. For cases of uneven point cloud density, in high-density areas, points are selected at fixed intervals, and redundant points within the intervals are removed. In low-density areas, new points are inserted at appropriate positions between adjacent points to achieve density uniformity and significantly improve data quality. After preprocessing, the entire point cloud data is divided into regions, and the importance of each region is assessed. A hierarchical division formula is used to separate critical regions requiring high detail from general display regions. The formula is as follows: ,in To adjust the weighting coefficients for different levels, set them according to the key aspects of the scenario. For the first Point cloud data density of each region It is the maximum point cloud data density across all regions. The maximum number of levels was determined by considering scene complexity and resource constraints. For key areas, a high-resolution point cloud data structure with multiple sub-levels was constructed, with details gradually added to each sub-level. For general areas, a structure with fewer levels and lower resolution was constructed, forming a multi-scale, multi-resolution hierarchical point cloud data system. This improved the data detail retention rate in key areas while reducing the amount of data in general areas. While ensuring high-precision modeling of key areas, this significantly reduced the overall data processing and storage costs, providing an efficient and high-quality data foundation for subsequent processing.

[0032] Example 3: like Figure 3 As shown, based on Embodiment 1, the adaptive compression and transmission control steps provided in this embodiment of the invention monitor the network status in real time, obtain the current network bandwidth and data transmission delay data. When the network bandwidth is low and the delay is high, the point cloud data of the key area in the low resolution layer is selected for processing first. Compression methods such as reducing the storage precision of data attributes and merging similar points are adopted to reduce the data volume before transmission, ensuring that the point cloud data of the key area can arrive first. In the low bandwidth environment, the loading time of the digital twin scene is shortened, maintaining the basic display of the scene and the visualization of key information. As network conditions improve, point cloud data of higher resolution levels are transmitted sequentially according to hierarchical order, gradually refining the scene. During transmission, the system continuously monitors scene changes and user operations. If a user focuses on a specific area, the current transmission queue is immediately interrupted, and higher resolution data for that area is transmitted first. After transmission is complete, the original transmission order is restored. This mechanism improves the interactive response speed of the scene even in environments with fluctuating network conditions. Users can quickly obtain detailed information about areas of interest, achieving adaptive matching between point cloud data transmission and network and scene requirements. This greatly enhances the smoothness of using digital twin scenarios and the user experience.

[0033] Example 4: like Figure 4As shown, based on Example 1, in the spatiotemporal dynamic semantic analysis step provided by this embodiment of the invention, the continuous time-series point cloud data after adaptive transmission is input into a specially constructed network model. In terms of time dimension processing, the point cloud data at adjacent times are compared, the differences between corresponding points in the point cloud data at different times are calculated, and the correlation relationship in the time dimension is established. The calculation formula is as follows: ,in, This represents point cloud data difference calculation, which constructs a difference set by calculating the coordinate difference vector and attribute difference values ​​of corresponding points. It is the first Time and the Difference sets of point cloud data at different times It is the first The point cloud data set at each moment, It is the first The point cloud data set at any given time can accurately capture the dynamic change trend of an object over time, reducing the error rate of capturing the object's motion trajectory.

[0034] In terms of spatial dimension processing, each point is taken as the center, and neighboring points within a certain range are selected. The spatial geometric relationships between these points are analyzed, including the distribution pattern, angle, distance, etc., to extract the local geometric features of the object. The extraction formula is as follows: ,in, It is a point Local geometric features, It is a point Neighborhood point set The number of elements, It is a point Points within the neighborhood, From point Time The vector, It is a vector The length of the mold, It is a point The normal vector is used to measure the relationship between neighboring points and points. By analyzing the relationship between normals and integrating the distribution information of all points in the entire scene, the overall structural features of the object are obtained. After multi-layered temporal and spatial dimension processing and feature fusion operations, the continuous time series point cloud data is analyzed point by point. Each point is assigned a corresponding object category label, the specific position of the object at different times is determined, and the information changes of the object during the motion and change process are recorded. The semantic segmentation accuracy of object categories is improved, and different objects in the scene and their dynamic changes can be clearly distinguished. This provides rich and accurate semantic information for digital twin scenes, making the simulation of actual dynamic processes in the scene more realistic and detailed.

[0035] Example 5: likeFigure 5 As shown, based on Example 1, the dynamic prediction and scene optimization steps provided in this embodiment of the invention establish a real-time monitoring mechanism to compare point cloud data after semantic segmentation at adjacent time points point by point, and calculate the comprehensive difference value of each object's corresponding point, including coordinate differences and attribute change values. The calculation formula is as follows: ,in, , , This is the coordinate difference. The difference between attributes. It is the comprehensive difference value of the corresponding point of an object at adjacent time moments. A change threshold is set. When the difference exceeds the threshold, it is determined that the area where the object is located has changed. For the area that has changed, point cloud data of it at multiple consecutive time moments are collected to analyze the object's motion speed and acceleration in the time dimension, as well as spatiotemporal feature information such as shape deformation and size change in the spatial dimension. The extracted spatiotemporal feature data is input into a specific prediction model. By learning the relationship between a large amount of historical spatiotemporal feature data and the actual changes of objects, the model analyzes and predicts the currently extracted spatiotemporal features to obtain the future state information of objects in the point cloud scene, including motion trajectory, posture changes, shape evolution, and other state change information. The prediction formula is as follows: ,in, It is the predicted number Point cloud data at any given moment, These are weighting coefficients for historical variance data, reflecting the degree of influence of data at different times on the prediction. It is the first The set of differences in point cloud data between the current time step and the previous time step. It is the bias term of the prediction model. The average error of the prediction is within an acceptable range. Based on these prediction results, the production parameters and equipment operation strategies in the digital twin scenario are adjusted accordingly, so that the operating efficiency of the equipment is improved and the resource scheduling is more reasonable when simulating the dynamic production process. This effectively enhances the adaptability and optimization capability of the digital twin scenario to the dynamic process.

[0036] Example 6: like Figure 6 As shown, based on Example 1, in the model feedback and continuous optimization steps provided in this embodiment of the invention, the information such as the position and state of objects in the optimized digital twin scene is compared one by one with the corresponding information collected by sensors in the actual scene, and the comprehensive deviation value between the two is calculated, including coordinate error, attribute difference, and other values. The calculation formula is as follows: ,in, For the coordinates and attributes of points in a digital twin scenario, The coordinates and attributes of the corresponding points in the actual scene. It is the comprehensive deviation value between the digital twin scenario and the actual scenario. It refers to the number of points being compared. It is the first in the digital twin scenario The coordinates of the points In real-world scenarios, the first The coordinates of the points It is the first in the digital twin scenario The attribute values ​​of each point, In real-world scenarios, the first For each attribute value, an error tolerance range is set. When the difference exceeds this range, it is determined that there is a deviation between the digital twin scene and the actual scene.

[0037] The analysis and investigation process involves examining the accuracy of the input data, the rationality of the model parameter settings, and the integrity of the model structure. If noise interference or missing data is found in the input data, the data is reprocessed. If the parameter settings are unreasonable, the parameter value range is adjusted or the parameter calculation method is changed. If there are defects in the model structure, improvements are made by adding or adjusting data processing layers or changing the data transmission method between layers. At the same time, based on changes in the actual scene, the object attribute data and scene layout information in the digital twin model are updated. Through multiple iterations and optimizations, the matching degree between the digital twin scene and the actual scene is improved, and the model's prediction accuracy for dynamic changes in the actual scene is increased. This forms a closed-loop optimization mechanism, continuously improving the model's accuracy and scene matching degree, ensuring that the digital twin scene can stably and accurately reflect the actual scene situation in the long term.

[0038] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital twin scene adaptive optimization method based on point cloud data, characterized in that, Includes the following steps: Data preprocessing and hierarchy construction: The raw point cloud data is preprocessed by removing outliers, filling in missing values, and adjusting density. Based on the needs of the scenario, key and ordinary regions are divided to construct a multi-scale, multi-resolution hierarchical point cloud data system. Adaptive compression and transmission control: Real-time monitoring of network status and scene operation, and adaptive compression and transmission control of point cloud data by adjusting data compression methods and transmission order according to network conditions and scene requirements; Spatiotemporal dynamic semantic analysis: Utilizing network models to process continuous time series point cloud data, by analyzing the dynamic changes in the time dimension and the geometric structural features in the spatial dimension, spatiotemporal dynamic semantic analysis and point cloud data annotation are completed; Dynamic prediction and scenario optimization: Compare point cloud data after semantic segmentation at adjacent time points to monitor dynamic changes, extract spatiotemporal features and input them into the prediction model, and adjust the production parameters and operation strategies of the digital twin scenario based on the prediction results; Model feedback and continuous optimization: Compare the optimized digital twin scenario with the actual scenario, analyze the reasons for the deviation, and improve the model from the aspects of data, parameters and structure, update the scenario information, and form a closed-loop optimization mechanism.

2. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 1, characterized in that, In the data preprocessing and hierarchy construction steps, based on the functional requirements and visual focus of the digital twin scene, the key areas requiring high detail and the general display areas are divided using a hierarchy division formula. For the key areas, a multi-level point cloud data structure from low to high resolution is constructed, with each layer gradually adding detailed information, and the layers are connected by specific spatial subdivision rules. For the general areas, a point cloud data structure with fewer layers and lower resolution is constructed, with different data sampling intervals and detail retention levels set for each layer, forming a complete multi-scale, multi-resolution hierarchical point cloud data system.

3. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 2, characterized in that, In the data preprocessing and hierarchy construction steps, key areas requiring high detail and general display areas are separated using a hierarchy division formula, which is as follows: ,in To adjust the weighting coefficients for different levels, set them according to the key aspects of the scenario. For the first Point cloud data density of each region It is the maximum point cloud data density across all regions. The maximum number of levels is determined by considering the overall scene complexity and resource constraints.

4. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 1, characterized in that, In the adaptive compression and transmission control steps, network bandwidth values ​​and data transmission latency status information are acquired in real time. Combined with the current loading progress and display content operation status of the digital twin scene, when the network bandwidth is lower than a set threshold and the latency is high, point cloud data of key areas in the low-resolution layer is selected first. Compression processing is performed by reducing the storage precision of attribute information in the point cloud data and merging adjacent similar points to reduce the data volume before transmission, ensuring that the point cloud data of key areas can arrive first. As the network bandwidth increases and the latency decreases, point cloud data of higher resolution layers are transmitted in hierarchical order to supplement the details of the already displayed low-resolution scene. During the transmission process, the position changes of objects in the scene, the user's perspective switching in the scene, and the interactive operation of area zooming are continuously monitored. Once the attention demand of a specific area is detected, the current transmission queue is immediately interrupted, and higher resolution point cloud data of that area is transmitted first. After the transmission is completed, the original transmission order is restored.

5. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 1, characterized in that, In the spatiotemporal dynamic semantic analysis step, a network model composed of multiple data processing layers is built. The continuous time series point cloud data after adaptive transmission is input into the model in chronological order. In the time dimension processing part, the point cloud data at adjacent times are arranged in chronological order to construct a data sequence. By calculating the differences of corresponding points in the point cloud data at different times, including coordinate differences and attribute changes, the correlation in the time dimension is established to capture the dynamic change trend of the object over time. In the spatial dimension processing section, each point is taken as the center, and neighboring points within a certain range are selected to analyze the spatial geometric relationship between the points, extract the local geometric features of the object, and obtain the overall structural features of the object by integrating the distribution information of all points in the scene. After multi-layer time dimension correlation calculation, spatial dimension feature extraction and feature fusion operations, the continuous time series point cloud data is analyzed point by point, each point is assigned a corresponding object category label, the specific position of the object at different times is determined, and the changes in coordinates and attribute information of the object during the motion and change process are recorded.

6. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 5, characterized in that, In the spatiotemporal dynamic semantic analysis step, a temporal correlation is established by calculating the differences between corresponding locations in the point cloud data at different times. The calculation formula is as follows: ,in, This represents point cloud data difference calculation, which constructs a difference set by calculating the coordinate difference vector and attribute difference values ​​of corresponding points. It is the first Time and the Difference sets of point cloud data at different times It is the first The point cloud data set at each moment, It is the first A collection of point cloud data at any given time.

7. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 5, characterized in that, In the spatiotemporal dynamic semantic analysis step, the local geometric features of the object are extracted, and the extraction formula is as follows: ,in, It is a point Local geometric features, It is a point Neighborhood point set The number of elements, It is a point Points within the neighborhood, From point Time The vector, It is a vector The length of the mold, It is a point The normal vector is used to measure the relationship between neighboring points and points. The relationship between the normal directions.

8. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 1, characterized in that, In the dynamic prediction and scene optimization steps, a real-time monitoring process is established. Point cloud data after semantic segmentation at adjacent time points are compared point by point to calculate the comprehensive difference value of each object's corresponding point. A change threshold is set. When the difference exceeds the threshold, it is determined that the area where the object is located has changed. For the area that has changed, point cloud data of the area at multiple consecutive time points are collected. The coordinate change rate and direction change of the object in the time dimension, as well as the shape deformation characteristics and size change information in the spatial dimension, are analyzed. The extracted spatiotemporal feature data is input into a prediction model composed of multiple data processing units. The model establishes a mapping relationship between the spatiotemporal features of the object and its future state by learning from a large amount of historical spatiotemporal feature data. Based on this, the currently extracted spatiotemporal features are analyzed and predicted to obtain the future state information of the object at multiple time points in the point cloud scene, including motion trajectory coordinates, posture change angle, and shape evolution information. Based on the prediction results, the production parameters and equipment operation strategies involved in the digital twin scene are adjusted accordingly.

9. The adaptive optimization method for digital twin scenes based on point cloud data as described in claim 8, characterized in that, In the dynamic prediction and scene optimization step, the comprehensive difference value of each object's corresponding point is calculated using the following formula: ,in, , , This is the coordinate difference. The difference between attributes. It is the combined difference value of corresponding points of an object at adjacent time points; In the dynamic prediction and scene optimization step, the extracted spatiotemporal features are analyzed and predicted to obtain the future state information of objects in the point cloud scene at multiple future time points. The prediction formula is as follows: ,in, It is the predicted number Point cloud data at any given moment, These are weighting coefficients for historical variance data, reflecting the degree of influence of data at different times on the prediction. It is the first The set of differences in point cloud data between the current time step and the previous time step. It is the bias term of the prediction model.

Images