Real-time rendering image generation method and system for 3D customized product

Abstract

The invention relates to the technical field of 3D customized products, and discloses a 3D customized product real-time rendering image generation method and system, and the method comprises the steps: achieving the preprocessing of sensor data through the local deployment of a lightweight model, and transmitting the processed data to a cloud server; the data transmission process is monitored in real time to judge the data interruption condition; extracting core geometric information data and core texture information data before interruption from a cache of the local lightweight model, and calculating a geometric feature change rate and a texture feature change rate; calculating coordinate parameters and a curved surface equation of the missing region by adopting a linear regression model based on the geometric feature change rate and the texture feature change rate; overlapping the calculated coordinate parameters of the missing area and the curved surface equation with the model before interruption; and step 7, evaluating the overlapped model, and generating an evaluation result. The problem of data missing caused by disconnection is solved, and the continuity of real-time image rendering is guaranteed.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of 3D customized products, in particular to a 3D customized product real-time rendering image generation method and system. BACKGROUND

[0002] The real-time rendering of 3D customized products is a dynamic visualization process relying on advanced computer graphics technology, which can quickly generate high-precision and high-fidelity virtual model images or animation previews at the moment when the user adjusts the design parameters. Through GPU acceleration algorithms and core technologies such as ray tracing, the system converts complex three-dimensional data into visual presentations with realistic lighting effects, physical textures, and environmental interactions, allowing designers, customers, or producers to intuitively perceive the appearance details and functional characteristics of the final product without waiting for long offline calculations. This instant feedback mechanism not only greatly improves the efficiency of creative iteration, but also supports cross-platform collaboration and immersive experiences, providing efficient solutions for personalized manufacturing, virtual try-on / try-out, and marketing displays, and becoming a key technology bridge connecting digital design and physical production.

[0003] In the existing 3D customized product real-time rendering image generation, the real-time mapping of sensor data in the digital twin scene is prone to disconnection due to network latency or protocol compatibility issues, affecting the response speed of the fault diagnosis system. SUMMARY

[0004] (I) Technical problems solved To solve the problems of the prior art, the present application provides a 3D customized product real-time rendering image generation method and system, which solves the problem of data loss caused by disconnection, fills in the missing area through local caching and calculation, ensures the continuity of real-time rendering images, avoids image discontinuity caused by data disconnection, and does not need to wait for network recovery or protocol repair. The local quickly completes the missing data calculation and model coincidence, greatly improves the response speed of the fault diagnosis system, and has other advantages.

[0005] (II) Technical solutions To achieve the above-mentioned purpose, the present application provides the following technical solutions: a 3D customized product real-time rendering image generation method, comprising the following steps: Step 1: Deploy a lightweight model locally to achieve preliminary processing and caching of sensor data before transmission, and transmit the processed data to a cloud server; Step 2: Real-time monitor the data transmission process and execute the next step during data interruption; Step 3: Extract the core geometric information data and core texture information data before interruption from the cache of the local lightweight model; Step 4: Calculate the geometric feature change rate and texture feature change rate based on the core geometric information data and core texture information data; Step five, based on the rate of change of geometric features and the rate of change of texture features, a linear regression model is used to calculate the coordinate parameters and surface equation of the missing area; Step six, the calculated coordinate parameters and surface equation of the missing area are superimposed with the model before the interruption; Step seven, the superimposed model is evaluated, and the evaluation result is generated.

[0006] Preferably, in step one, the processed data needs to be transmitted to the cloud server with time-sensitive network to ensure the consistency of sensor data transmission timestamp.

[0007] Preferably, in step two, real-time monitoring of the data transmission process includes: Deploying a bidirectional monitoring probe between the local edge node and the cloud server, real-time collecting network key indicators and data key indicators to form network parameter data set and data parameter data set respectively, and judging whether there is a data interruption risk based on the network parameter data set and the data parameter data set.

[0008] Preferably, the network parameter data set includes packet loss rate, round-trip delay rate, bandwidth fluctuation rate, and packet retransmission rate. The data parameter data set includes sensor data frame integrity, time sequence continuity, message interaction success rate, and data field integrity.

[0009] Preferably, the data interruption risk is judged based on the data risk index, and the calculation formula of the data risk index is:

[0010] In the formula, represents the data risk index, represents the normalized value of the th parameter in the network parameter data set, represents the weight of the th parameter in the network parameter data set; represents the normalized value of the th parameter in the data parameter data set, represents the weight of the th parameter in the data parameter data set; represents the scene adaptation coefficient.

[0011] Preferably, when the calculation result of the data interruption risk exceeds 0.6, it is determined that there is a data interruption risk.

[0012] Preferably, the calculation formula of the rate of change of geometric features is:

[0013] In the formula, Represents the rate of change of geometric characteristics. Represents the average rate of change of dimensional parameters. Represents the size of the sliding window. Represents the preprocessed Time dimension parameters, This represents the sensor sampling time interval; Represents the instantaneous rate of change of position of coordinate parameters. represent The vector representing the direction of change of the time coordinate. Represents the coordinates of the preprocessed point cloud. represent Three-dimensional coordinate data at time [time]. represent Time and The spatial Euclidean distance between the time coordinates; Represents the instantaneous rate of change of curvature-type parameters. The angle representing the instantaneous change in the direction of curvature.

[0014] Preferably, the formula for calculating the texture feature change rate is:

[0015] In the formula, Represents the rate of change of texture features. Represents the preprocessed Time-based numerical texture parameters This represents the time interval for sampling texture data; This represents the texture coordinate parameters at time t after preprocessing. This represents the texture direction vector at time t after preprocessing. This represents the texture repetition period parameter at time t after preprocessing. Represents the angle of change in texture direction. The average rate of change of numerical texture parameters, Represents the instantaneous rate of change of texture coordinates. The instantaneous rate of change representing the texture repetition period Preferably, the coordinate parameters of the missing region are as follows: The geometric and texture feature change rates from the last normal cycle, extracted from the local cache, are used as input features. ; Will Substituting the values ​​into the three coordinate regression models respectively, we obtain... 、 、 , that is, the missing area coordinates; The surface equation of the missing area is specifically: Substitute Into the 4 plane coefficient regression models respectively, to obtain 、 、 、

[0016] Substitute 、 、 、 Into the plane equation.

[0017] A 3D customized product real-time rendering image generation system applies a 3D customized product real-time rendering image generation method, including a local deployment module, a real-time monitoring module, a data calculation module, a model recombination module, The local deployment module is used for locally deploying a lightweight model, realizing preliminary processing and caching before sensor data transmission, and transmitting the processed data to a cloud server; The real-time monitoring module is used for real-time monitoring of the data transmission process and determining the data interruption risk; The data calculation module calculates the geometric feature change rate and the texture feature change rate based on the extracted core geometric information data and core texture information data before interruption, and calculates the coordinate parameters and the surface equation of the missing area based on the linear regression model of the geometric feature change rate and the texture feature change rate; The model recombination module is used for superimposing the calculated coordinate parameters and surface equation of the missing area with the model before interruption.

[0018] A 3D customized product real-time rendering image generation method and system, including the following steps:.

[0019] Compared with the prior art, the present application provides a 3D customized product real-time rendering image generation method and system, which has the following beneficial effects: The application realizes the solution of the sensor data disconnection problem in the 3D customized product digital twin scene through the complete closed loop of local preprocessing cache, real-time monitoring disconnection, local data extraction calculation, model coincidence evaluation. First, a lightweight model is deployed locally to complete preliminary processing and caching of sensor data in advance, which not only reduces the transmission pressure of the cloud, but also provides data basis for disconnection, and at the same time, the processed data is transmitted to the cloud. Then, the transmission process is monitored in real time. Once the disconnection caused by network delay or protocol compatibility occurs, the local processing flow is triggered immediately, the core geometry and texture information data before the disconnection are extracted from the cache, the geometry and texture feature change rate is calculated based on these data, and then the coordinate parameters and surface equation of the missing area are quickly calculated through the linear regression model. Subsequently, the calculation results are coincided with the model before the disconnection, and finally the coincided model is evaluated and the results are generated.

[0020] The data missing problem caused by disconnection is solved, the missing area is filled by local caching and calculation, the continuity of real-time rendering image is ensured, the image discontinuity caused by data disconnection is avoided, and at the same time, without waiting for network recovery or protocol repair, the missing data calculation and model coincidence are quickly completed locally, the response speed of the fault diagnosis system is greatly improved, the abnormality in the 3D customized product digital twin model can be found in time, and the lightweight model and simplified calculation method reduce the operation burden of the local device, balance the processing efficiency and device performance, and at the same time, the evaluation link can ensure the accuracy of the coincided model, and provide reliable data support for the production and quality inspection of subsequent 3D customized products. BRIEF DESCRIPTION OF DRAWINGS

[0021] Fig. 1 The method steps of the application are shown in the figure; Fig. 2 The system flowchart of the application is shown in the figure. DETAILED DESCRIPTION

[0022] The technical solutions in the embodiments of the application will be clearly and completely described below with reference to the drawings in the embodiments of the application. Obviously, the described embodiments are only part of the embodiments of the application, not all the embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the application.

[0023] Please refer to Figs. 1-2 A 3D customized product real-time rendering image generation method, comprising the following steps: Step one, deploy a lightweight model locally to realize preliminary processing and caching before sensor data transmission, and transmit the processed data to the cloud server, and at the same time, match the time-sensitive network to ensure that the sensor data transmission time stamp is consistent; By deploying a lightweight model locally as an edge computing node, pre-processing the original sensor data, and temporarily storing the core geometry and texture information of recent key frames through a caching mechanism, the bandwidth pressure caused by directly uploading massive raw data is reduced. At the same time, precise clock synchronization is achieved using a time-sensitive network to ensure that the data streams of different sensors have uniform high-precision timestamps, meeting the stringent real-time requirements of industrial scenarios and providing a reliable data foundation for subsequent calculations. Step two, real-time monitoring of the data transmission process, deploying bidirectional monitoring probes between the local edge node and the cloud server, real-time collection of network key indicators and data key indicators to form network parameter data sets and data parameter data sets respectively, and based on the network parameter data sets and the data parameter data sets, determining whether there is a data interruption risk, and executing the next step during the data interruption period; The network parameter data set includes packet loss rate, round-trip delay rate, bandwidth fluctuation rate, and packet retransmission rate. The data parameter data set includes sensor data frame integrity, time sequence continuity, message interaction success rate, and data field integrity. The data interruption risk is determined based on the data risk index, and the calculation formula of the data risk index is:

[0024] In the formula, represents the data risk index, represents the normalized value of the th parameter in the network parameter data set, represents the weight of the th parameter in the network parameter data set; represents the normalized value of the th parameter in the data parameter data set, represents the weight of the th parameter in the data parameter data set; represents the scene adaptation coefficient. When the calculation result of the data interruption risk exceeds 0.6, it is determined that there is a data interruption risk. The deployment of bidirectional monitoring probes between the local edge node and the cloud server can realize transparent control of the whole data transmission process, synchronously collect network parameter data sets and data parameter data sets to build a multi-dimensional health evaluation system, facilitate the prediction of potential risks, and enable the system to change from passive response to active defense, providing deterministic communication guarantee for industrial real-time applications. Step three, extracting the core geometry information data and core texture information data before interruption from the cache of the local lightweight model. Step four, calculating the geometric feature change rate and the texture feature change rate based on the core geometric information data and the core texture information data; The formula for calculating the geometric feature change rate is:

[0025] In the formula, represents the geometric feature change rate, represents the average change rate of size parameters, represents the size of the sliding window, represents the pre-processed size parameter at time t, represents the sensor sampling time interval; represents the instantaneous position change rate of coordinate parameters, represents the coordinate change direction vector at time t, represents the pre-processed point cloud coordinates, represents the three-dimensional coordinate data at time t, represents the pre-processed coordinate at time t, represents the spatial Euclidean distance between the coordinate at time t and the coordinate at time t-1; represents the instantaneous numerical change rate of curvature parameters, represents the instantaneous change angle of curvature direction; The multi-dimensional quantitative index based on the core geometric information data realizes accurate characterization of the dynamic deformation of the 3D model, the sliding window average algorithm effectively smooths noise and captures periodic small deformation trends, combined with the Euclidean distance normalized coordinate velocity field and the curvature direction angle difference, it can identify the non-uniform deformation mode of complex curved surfaces under tensile and bending combined stress, significantly improving the spatial resolution and time sequence traceability of fault diagnosis; The formula for calculating the texture feature change rate is:

[0026] In the formula, represents the texture feature change rate, represents the pre-processed numerical texture parameter at time t, represents the texture data sampling time interval; represents the pre-processed texture coordinate parameter at time t, represents the pre-processed texture direction vector at time t, represents the pre-processed texture repetition period parameter at time t, represents the texture direction change angle, ​average rate of change of a numerical texture parameter, instantaneous rate of change of a texture coordinate, instantaneous rate of change of a texture repetition period; The texture feature change rate calculation formula realizes dynamic analysis of the material surface characteristics through multi-dimensional quantitative indicators, effectively suppresses random noise interference based on sliding window statistics, accurately captures the gradual change trend, and uses the instantaneous displacement of UV coordinates and the direction angle double constraint mechanism to separate the rigid translation and topological deformation two different nature deformation modes. The relative change rate of the period parameter directly reflects the evolution law of the texture scale, and provides traceable quantitative basis for product quality judgment; Step five, based on the geometric feature change rate and the texture feature change rate, a linear regression model is used to calculate the coordinate parameters and the surface equation of the missing area; The coordinate parameters of the missing area are: The geometric feature change rate and the texture feature change rate of the last normal period extracted based on local cache are used as input features ; Substitute into the three coordinate regression models respectively, and get , , , which are the coordinates of the missing area; The surface equation of the missing area is: Substitute into the four plane coefficient regression models respectively, and get , , ,

[0027] Combine , , , into a plane equation; By encoding the dual time evolution law of geometry and texture into the input vector of the linear regression model, a closed-loop reasoning engine from multi-modal features to spatial structure is constructed, which converts abstract feature change rate into specific three-dimensional coordinate increment and plane equation coefficient. Based on the data of the last normal period, extrapolation prediction is carried out to ensure that the virtual repair during interruption strictly follows the original deformation trend of the object, avoids topological distortion caused by jump interpolation, and independently trains three regression models for the coordinate axes and four plane parameter estimators., not only retains the physical properties of each degree of freedom, but also realizes collaborative optimization, so that the system can still maintain sub-millisecond response speed when the network is interrupted, the error of the compensation model generated is controlled within the engineering allowable range, and the spatial fidelity and time sequence continuity of the digital twin are effectively maintained; Step six, the calculated coordinate parameters of the missing area and the surface equation are coincided with the model before interruption; Step seven, the coincided model is evaluated, and an evaluation result is generated.

[0028] The above-mentioned 3D customized product real-time rendering image generation method has the following application system: The application system comprises a local deployment module, a real-time monitoring module, a data calculation module, and a model recombination module. The local deployment module is used for deploying a lightweight model locally, and realizing preliminary processing and caching before sensor data transmission, and transmitting the processed data to a cloud server. The real-time monitoring module is used for real-time monitoring of the data transmission process, and determining a data interruption risk. The data calculation module calculates a geometric feature change rate and a texture feature change rate based on the extracted core geometric information data and core texture information data before interruption, and calculates coordinate parameters of a missing area and a surface equation based on the linear regression model of the geometric feature change rate and the texture feature change rate. The model recombination module is used for coinciding the calculated coordinate parameters of the missing area and the surface equation with the model before interruption.

[0029] Although the embodiments of the present application have been shown and described, it can be understood by those of ordinary skill in the art that various changes, modifications, replacements and variations can be made to the embodiments without departing from the principles and spirits of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A 3D customized product real-time rendering image generation method, characterized in that, The method comprises the following steps: Step 1: locally deploying a lightweight model to realize preliminary processing and caching of sensor data before transmission, and transmitting the processed data to a cloud server; Step 2: real-time monitoring of the data transmission process, and performing the next step during data interruption; Step 3: extracting the core geometric information data and core texture information data before interruption from the cache of the local lightweight model; Step 4: calculating the geometric feature change rate and the texture feature change rate based on the core geometric information data and the core texture information data; Step 5: calculating the coordinate parameters and surface equation of the missing area based on the linear regression model of the geometric feature change rate and the texture feature change rate; Step 6: superimposing the calculated coordinate parameters and surface equation of the missing area with the model before interruption; Step 7: evaluating the superimposed model and generating an evaluation result. 2.The 3D customized product real-time rendering image generation method of claim 1, wherein: In step 1, when transmitting the processed data to the cloud server, a time-sensitive network is required to ensure consistency of the sensor data transmission timestamp. 3.The 3D customized product real-time rendering image generation method of claim 1, wherein, In step 2, real-time monitoring of the data transmission process includes: Deploying a bidirectional monitoring probe between the local edge node and the cloud server to collect network key indicators and data key indicators in real time to form a network parameter data set and a data parameter data set, respectively, and determining whether there is a data interruption risk based on the network parameter data set and the data parameter data set.

4. The method of claim 3, wherein: The network parameter data set includes packet loss rate, round-trip delay rate, bandwidth fluctuation rate, and packet retransmission rate. The data parameter data set includes sensor data frame integrity, time sequence continuity, message interaction success rate, and data field integrity.

5. The method of claim 3, wherein the method further comprises: The data interruption risk is determined based on a data risk index, and the calculation formula of the data risk index is: ； In the formula, represents a data risk index, represents a normalized value of the i-th parameter in the network parameter data set, represents a normalized value of the i-th parameter in the network parameter data set, represents a weight of the i-th parameter in the network parameter data set, represents a weight of the i-th parameter in the network parameter data set, a normalized value representing a data parameter in the data parameter set, a normalized value representing a data parameter in the data parameter set, a weight representing a data parameter in the data parameter set, a weight representing a data parameter in the data parameter set, Represents the scene adaptation coefficient.

6. The method of claim 5, wherein: When the calculation result of the data interruption risk exceeds 0.6, it is determined that there is a data interruption risk.

7. The method of claim 1, wherein: The calculation formula of the geometric feature change rate is: ； in the formula, represents the rate of change of the geometric feature, represents the average rate of change of the size-related parameter, represents the size of the sliding window, represents the pre-processed time instant size parameter, represents the sensor sampling time interval; a rate of change of an instantaneous position representing a coordinate class parameter, representing a time coordinate change direction vector, representing pre-processed point cloud coordinates, representing three-dimensional coordinate data at a time, representing a time coordinate and a spatial Euclidean distance of a time coordinate; a rate of change of the instantaneous value representative of the curvature class parameter, an instantaneous change angle representative of the curvature direction. 8.The method of claim 7, wherein: The calculation formula of the texture feature change rate is: ； In the formula, represents a texture feature change rate, represents a pre-processed time numerical texture parameter, represents a texture data sampling time interval; a parameter representing a t-time texture coordinate after preprocessing, a parameter representing a t-time texture direction vector after preprocessing, a parameter representing a t-time texture repetition period after preprocessing, a parameter representing a texture direction change angle, a parameter representing an average change rate of a numerical texture parameter, a parameter representing an instantaneous position change rate of a texture coordinate, a parameter representing an instantaneous change rate of a texture repetition period. 9.The method of claim 8, wherein: The coordinate parameters of the missing area are: extracting, based on the local cache, a rate of change of geometric features and a rate of change of texture features of the last normal period as input features ; Will Substituting the values ​​into the three coordinate regression models respectively, we obtain... , , , which represents the coordinates of the missing region; The surface equation of the missing area is: Will Substituting each of the four plane coefficient regression models, we obtain... , , , ; The , , , are combined as a plane equation.

10. A 3D customized product real-time rendering image generation system applied to the 3D customized product real-time rendering image generation method of any one of claims 1-9, characterized in that: The system comprises a local deployment module, a real-time monitoring module, a data calculation module, and a model recombination module. The local deployment module is used to locally deploy a lightweight model, realize preliminary processing and caching of sensor data before transmission, and transmit the processed data to a cloud server. The real-time monitoring module is used to real-time monitor the data transmission process and determine the data interruption risk. The data calculation module calculates the geometric feature change rate and the texture feature change rate based on the extracted core geometric information data and core texture information data, and calculates the coordinate parameters and surface equation of the missing area based on the linear regression model of the geometric feature change rate and the texture feature change rate. The model recombination module is used to superimpose the calculated coordinate parameters and surface equation of the missing area with the model before interruption.

Images