Data acquisition method based on three-dimensional reconstruction system

Abstract

The invention relates to the technical field of data storage, and particularly discloses a data acquisition method based on a three-dimensional reconstruction system, and the method comprises the following steps: S1, data acquisition: carrying out the data preprocessing of an original data set # imgabs0 #; s2, data preprocessing: preprocessing the data # imgabs 1 # received in real time, and converting the data # imgabs 1 # into a standardized data set # imgabs 2 #; s3, importance score calculation: aiming at the preprocessed data # imgabs3 #, calculating an importance score # imgabs4 # of the preprocessed data # imgabs3 #; and S4, based on the importance score # imgabs5 #, implementing a dynamic caching strategy. The block chain technology realizes data traceability and tampering prevention, and in combination with the space-time prediction model based on the graph convolutional network, the objectivity of data credibility evaluation and the accuracy of prediction importance are enhanced, so that a more reliable decision basis is provided for a three-dimensional reconstruction system, and the industrial accident risk caused by data distortion is reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of data storage, and particularly to a data acquisition method for a three-dimensional reconstruction system. Background Art

[0002] With the rapid development of industrial Internet of Things (IIoT) and intelligent manufacturing technologies, three-dimensional reconstruction systems have been widely used in industrial equipment condition monitoring, fault diagnosis, and production process optimization. Three-dimensional reconstruction relies on the efficient acquisition and processing of multi-source heterogeneous data (such as sensor data, image information, equipment logs, etc.). However, the existing technologies still have the following deficiencies in data acquisition and cache management: 1. Difficulty in integrating multi-source heterogeneous data. There are various types of sensors and devices deployed in industrial sites, with significant differences in data formats, protocols, and frequencies. Traditional methods are difficult to efficiently integrate these data, resulting in low data utilization and high analysis latency.

[0003] 2. Insufficient requirements for real-time performance and reliability. Three-dimensional reconstruction requires real-time processing of dynamic data, but the existing systems lack effective mechanisms in data capture rate, time alignment, and integrity verification, which are prone to data loss or misalignment, affecting the reconstruction accuracy and decision-making reliability.

[0004] 3. Static cache strategy and low resource utilization. Traditional cache mechanisms are based on fixed rules (such as First In First Out, FIFO), without fully considering the differences in data importance. For data with high frequency of access, high change rate, or strong correlation with key performance indicators (KPIs), static strategies cannot achieve optimal allocation of cache resources, resulting in low cache hit rate and high system latency.

[0005] 4. Single evaluation of data importance and poor adaptability. Existing methods mostly rely on a single indicator (such as access frequency) to evaluate the value of data, ignoring multi-dimensional factors such as the urgency, predictive importance, and credibility of data, and are difficult to adapt to the dynamic requirements of complex industrial scenarios.

[0006] 5. Lack of multi-level cache and intelligent scheduling capabilities. Industrial data is characterized by large scale and strong timeliness, but traditional systems lack a multi-level cache architecture and intelligent scheduling strategies based on data characteristics, and cannot achieve a balance between access speed and storage cost.

[0007] In view of the above problems, the present invention proposes a three-dimensional reconstruction data acquisition method based on dynamic importance scoring. By means of innovative data preprocessing, multi-dimensional scoring models, and adaptive cache strategies, the processing efficiency of multi-source heterogeneous data and the utilization rate of cache resources are significantly improved, meeting the requirements of industrial scenarios for real-time performance, reliability, and intelligence. Summary of the Invention

[0008] The purpose of the present invention is to provide a data acquisition method based on a three-dimensional reconstruction system to solve the problems raised in the above-mentioned background technology.

[0009] To achieve the above purpose, the present invention provides the following technical solution: A data acquisition method based on a three-dimensional reconstruction system includes the following steps: S1. Data acquisition: Configure multiple data sources for the three-dimensional reconstruction system, and these data sources are used to collect raw data related to three-dimensional reconstruction; S2. Data preprocessing: Preprocess the data received in real time and convert it into a standardized data set This preprocessing process includes: Data cleaning: Eliminate abnormal, duplicate, and incorrect data; Format unification: Convert data in different formats into a unified standardized format; Feature extraction: Use the PCA algorithm to extract key feature vectors ; Data compression: Implement lossless or lossy compression technology to reduce storage and transmission costs; S3. Importance score calculation: Calculate the importance score for the preprocessed data The calculation formula is as follows: ; ; Where: is the data access frequency; is the urgency of the data; is the data change rate, and the calculation formula is: ; where is the time interval; is the predicted importance based on the deep learning prediction model; is the correlation between the data and the key business indicators, and the calculation formula is: ; represents the correlation coefficient, is the feature vector, is the key performance indicator KPI; is the data credibility, reflecting the reliability of the data source and the integrity of the data itself; , , , , , is the weight coefficient; S4. Based on the importance score implement a dynamic caching strategy.

[0010] Preferably, in S1, the specific steps of data acquisition are: S11. Configure multiple data sources, and the total number of data sources is , where is a positive integer, and each data source corresponds to an industrial device or sensor, numbered ; S12. For each data source at a specific time stamp capture the raw data in real time through the industrial Internet of Things network , the includes the following information: Data source identification code , used to uniquely distinguish data sources ; Measurement value or data content , representing the raw information collected by the data source at time ; Data format or type description , indicating the structure, type or format of the data generated by the data source to facilitate the identification and processing of data from multiple sources and formats; Additional information , covering additional information related to the data; S13. Aggregate the raw data collected from all data sources in step S12 to construct a raw data set at time , that is: ; Wherein: represents the data collection activity at time ; represents taking all data sources at time to generate the raw data at time of the collection; ; S14. Perform time stamp alignment and data integrity verification on the raw data set ; S15. Calculate the data capture rate to evaluate the effectiveness of data collection at time , and the data capture rate is defined as: ; S16. Based on the data capture rate and the result of data integrity verification, determine whether the data collection process meets the standards of real-time performance and reliability. If the conditions are met, the original data set is preprocessed.

[0011] Preferably, the prediction importance uses a spatio-temporal prediction model based on a graph convolutional network.

[0012] Preferably, the data credibility is evaluated through blockchain technology. The distributed ledger is used to record the data source information to achieve data traceability and anti-tampering.

[0013] Preferably, based on the importance score , the specific steps for implementing the dynamic caching strategy are as follows: S41. Cache storage: Store the data with an importance score exceeding the threshold in the high-speed cache area; S42. Cache eviction: When the cache capacity exceeds the threshold , use the LVD algorithm to remove the data in ascending order; S43. Multi-level caching: According to the differences, the data can be stored in different levels of cache media for multi-level caching; S44. Intelligent data scheduling: Apply the reinforcement learning algorithm to intelligently schedule the cached data according to the system state and network conditions, and optimize the data acquisition and transmission path; S45. Adaptive weight update: Use Bayesian optimization technology to adjust the weight coefficient in real time according to the cache hit rate H, system latency L, and data loss rate D.

[0014] Preferably, the multi-level cache includes at least three levels of cache media: memory cache, solid-state drive cache, and disk cache, and the data is allocated to the corresponding cache level according to the importance score of the data

[0015] Preferably, the data allocation rule for the cache level is as follows: When , store the data in the memory cache; When , store the data in the solid-state drive cache; When , store the data in the disk cache; Among them, and​​ is a preset importance scoring threshold, and satisfies .

[0016] Preferably, the calculation formula for real-time adjustment of the weight coefficient is:

[0017] Wherein, , , are weight parameters, and .

[0018] According to the above-mentioned data acquisition method based on a three-dimensional reconstruction system, a data storage system based on Internet technology is proposed, including: a data acquisition module, a data preprocessing module, a scoring calculation module, a dynamic cache management module, an intelligent scheduling module, and an adaptive optimization module; the data acquisition module obtains real-time data of industrial equipment through a sensor network and IoT devices . The data preprocessing module is used to preprocess the real-time data to obtain preprocessed data ; the scoring calculation module is used to calculate the importance score of the data

[0019] The dynamic cache management module is used to execute a dynamic cache policy, including data storage, elimination, and multi-level cache management; the intelligent scheduling module uses a reinforcement learning algorithm to achieve intelligent scheduling of data and optimization of the transmission path; the adaptive optimization module uses Bayesian optimization to achieve real-time update of the weight coefficient. The present invention provides a data storage method based on Internet technology, and has the following beneficial effects in view of the deficiencies of the prior art:

[0020] 1. High-efficiency integration ability of multi-source heterogeneous data: Through a standardized preprocessing process, effectively eliminate the format differences of multi-device and multi-protocol data in the industrial field, improve data compatibility and analyzability, and combine the data source identification code and additional information to enhance the system's recognition and parsing ability of heterogeneous data, and significantly improve data utilization.

[0021] 3. Optimized Allocation of Caching Resources: A dynamic caching strategy based on multi-dimensional importance scoring breaks through the limitations of traditional static caching. Through the LVD elimination algorithm and a multi-level caching architecture (memory / solid-state drive / disk), it realizes hierarchical storage of data according to value, significantly improving the cache hit rate of high-frequency, high-change-rate, and KPI-related data, and reducing system latency.

[0022] 4. Adaptive Dynamic Optimization Capability: Bayesian optimization technology adjusts the scoring weight coefficients in real time and combines with reinforcement learning for intelligent scheduling, enabling the system to dynamically optimize data transmission paths and storage strategies based on network status, cache hit rate, etc. This mechanism balances storage costs and resource utilization while ensuring the access speed of critical data, enhancing the system's robustness.

[0023] 5. Improvement of Data Quality and Credibility: Blockchain technology enables data traceability and anti-tampering. Combined with a spatio-temporal prediction model based on graph convolutional networks, it enhances the objectivity of data credibility assessment and the accuracy of predicting importance, providing a more reliable decision-making basis for the 3D reconstruction system and reducing the risk of industrial accidents caused by data distortion.

[0024] 6. Enhanced Adaptability to Industrial Scenarios: The configurable design of weight coefficients and the dynamically adjustable multi-level cache thresholds enable the system to flexibly adapt to the different requirements of different industrial fields (such as scenarios prioritizing real-time performance or sensitive to storage costs), expanding the application scope of the technical solution. Description of the Drawings

[0025] Figure 1 It is a flowchart of a data acquisition method for a 3D reconstruction system according to the present invention; Figure 2 It is a flowchart of implementing the dynamic caching strategy of the present invention; Figure 3 It is a schematic block diagram of a data storage system based on Internet technology according to the present invention. Detailed Embodiments

[0026] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] As Figure 1 shown, the present invention provides a technical solution: A data acquisition method for a 3D reconstruction system includes the following steps: S1. Data Acquisition: Configure multiple data sources for the 3D reconstruction system, and these data sources are used to collect raw data related to 3D reconstruction; S11. Configure multiple data sources, the total number of data sources being , where is a positive integer, and each data source corresponds to an industrial device or sensor, numbered ; S12. For each data source , at a specific time stamp , capture the raw data in real time through the industrial Internet of Things network. The contains the following information: Data source identification code , used to uniquely distinguish data sources ; Measurement value or data content , representing the raw information collected by the data source at time ; Data format or type description , indicating the structure, type, or format of the data generated by the data source to facilitate the identification and processing of data from multiple sources and formats; Additional information , covering additional information related to the data, such as units, precision, device status, and other necessary descriptive details; S13. Aggregate the raw data collected from all data sources in step S12 to construct a raw data set at time , that is: ; Where: represents the data collection activity at time ; represents taking the collection of the raw data generated by all data sources at time S14. Perform time stamp alignment and data integrity verification on the raw data set to ensure: Time alignment: All data is associated with an accurate time stamp to meet the requirements of real-time data processing; Data integrity: Check the set of critical data sources to ensure that at each time , the corresponding raw data can be obtained from all critical data sources; S15. Calculate the data capture rate to evaluate the effectiveness of data collection at a certain time The data capture rate is defined as follows: ; S16. Based on the data capture rate and the result of data integrity verification, determine whether the data collection process meets the standards of real-time and reliability. If the conditions are met, preprocess the original data set for data preprocessing.

[0028] Through S11 to S16, the real-time collection and integration of data from multiple sources and in multiple formats are realized, providing a comprehensive and reliable raw data basis for subsequent data preprocessing, analysis and decision-making; by introducing data format or type descriptions and additional information the system's ability to identify and process heterogeneous data is enhanced; through time stamp alignment and data integrity verification, the real-time and reliability of data are ensured, meeting the requirements of the industrial Internet of Things environment for efficient data processing; the calculation of the data capture rate helps to monitor and evaluate the effectiveness of the data collection process in real time, providing a basis for the adaptive adjustment and optimization of the system.

[0029] S2. Data preprocessing: Preprocess the data received in real time to convert it into a standardized data set This preprocessing process includes: Data cleaning: Eliminate abnormal, duplicate and incorrect data; Format unification: Convert data in different formats into a unified standardized format; Feature extraction: Use the PCA algorithm to extract key feature vectors ; Data compression: Implement lossless or lossy compression techniques to reduce storage and transmission costs; Through these preprocessing steps, the original data is cleaned, standardized, characterized and compressed, and transformed into a more consistent and usable standardized data set providing strong support for subsequent data analysis and decision-making.

[0030] S3. Importance score calculation: Calculate the importance score for the preprocessed data The calculation formula is as follows: ; Where: represents the data Frequency of access: Frequently accessed data is usually of high importance in the system. Caching such data first can improve system response speed and user experience. The number of data accesses within the statistical period can be used to measure this indicator; The urgency of the data; that is, the necessity of timely processing of data. For data that requires real-time response or is associated with emergency events, a higher urgency value should be assigned. This indicator can be set according to business needs, event levels or preset rules; is the data change rate, which is used to measure the change amplitude of data in unit time. The calculation formula is: ;in, is the time interval; data with a high rate of change indicates that the system state or environment has changed significantly, and it is of great significance to obtain and process such data in a timely manner; In order to improve the prediction accuracy based on the prediction importance of deep learning prediction models, a spatiotemporal prediction model based on graph convolutional networks is used to consider the correlation of data in time and space; is the correlation between data and key business indicators, and the calculation formula is: ; represents the correlation coefficient, is the feature vector, It is the key performance indicator KPI; this indicator reflects the impact of data on key business indicators. The higher the relevance, the more important the data is, and it should be cached first. To ensure data credibility, the reliability of the data source and the integrity of the data itself are reflected through blockchain technology. The distributed ledger is used to record data source information, achieve data traceability and tamper-proof, and ensure the authenticity and reliability of the data. , , , , , is the weight coefficient; =1, and each coefficient is greater than or equal to zero. Each weight coefficient is used to balance the influence of different evaluation indicators on the importance of data; Through the comprehensive calculation of the above indicators, the overall importance score of the data is obtained. In actual applications, each weight coefficient can be adjusted according to specific business requirements and system conditions. , , , , , To optimize the cache strategy. For example, for a system with high real-time requirements, you can increase and For applications with high data credibility requirements, the weight of The weight of Importance Rating It is the basis for the implementation of the subsequent dynamic caching strategy. In step S4, the system will determine the data caching, elimination and scheduling strategies based on the scoring results to ensure the efficient acquisition of key data and the optimal use of system resources.

[0031] S4. Based on importance scoring , implement dynamic caching strategies; like Figure 2 As shown in the figure, the specific steps to implement the dynamic cache strategy are: S41. Cache storage: scoring importance Exceeding the threshold The data is stored in the cache area; that is, Data ,For the data that meets the requirements, it will be saved in the cache first to ensure the fast access and processing of key data and improve the system response speed.

[0032] S42, cache elimination: when the cache capacity Threshold exceeded When using the LVD algorithm, press Remove data from low to high order; the specific steps are as follows: S421. Score the data in the cache according to importance Sort in ascending order; S422, starting with the data with the lowest score, remove them in sequence until the cache capacity is reached. ; S423. Ensure that data with high importance scores is always retained in the cache to ensure efficient access to key data by the system.

[0033] S43, multi-level cache: according to The data can be stored in different levels of cache media for multi-level caching. The multi-level cache includes at least three levels of cache media: memory cache, solid-state drive cache, and disk cache, and is scored based on the importance of the data. Allocate data to the corresponding cache level; Level 1 cache (memory cache): used to store data with the highest importance score and provide the fastest access speed; Secondary cache (SSD cache): used to store data with medium importance scores, taking into account both access speed and storage capacity; Tertiary cache (disk cache): Used to store data with a lower importance score, providing a larger storage capacity; The data allocation rules for cache levels are as follows: When , store the data in the primary cache; When , store the data in the secondary cache; When , store the data in the tertiary cache; Among them, and are preset importance score thresholds, and satisfy .

[0034] S44, Intelligent data scheduling: Apply reinforcement learning algorithms to intelligently schedule cached data based on the system state and network conditions, optimizing the data acquisition and transmission paths; The specific implementation is as follows: S441, State space definition: Construct a state space from parameters such as the cache state, network bandwidth, and latency of the system, and monitor the system environment in real time; S442, Action space definition: Include operations such as migration, replication, deletion of cached data, and selection of data transmission paths; S443, Reward function design: Design a reward function based on system performance metrics such as cache hit rate, data access latency, and network load, guiding the optimization direction of the reinforcement learning model; S444, Policy update: Adopt reinforcement learning algorithms such as Q-learning to continuously update the scheduling policy, so that in different system states, select the optimal scheduling actions to improve the overall performance; Through the adaptive adjustment of reinforcement learning, the system can dynamically optimize the data scheduling policy in a complex and changing network environment, maximizing the data access efficiency and network resource utilization rate.

[0035] S45, Adaptive weight update: Use Bayesian optimization technology to adjust the weight coefficients in real time according to the cache hit rate H, system latency L, and data loss rate D; The calculation formula for adjusting the weight coefficients in real time is:

[0036] Cache miss rate : Reflects the probability of cache misses, where H is the cache hit rate; System latency L: Refers to the average latency time of data access; Data loss rate D: Refers to the proportion of data loss caused by cache eviction or network reasons; Among them, , , are weight parameters, and , which is used to balance the importance of different performance metrics.

[0037] Bayesian optimization steps: Build a surrogate model: Use Gaussian process or other regression models to establish the mapping relationship between weight coefficients and performance metrics; Collect sample data: Measure the corresponding performance metric values under different combinations of weight coefficients to enrich the training data of the surrogate model; Select the optimal point: Through the prediction of the surrogate model, find the weight combination that minimizes the objective function in the weight coefficient space; Update the weight coefficients: Apply the new optimal weight coefficients to the importance score calculation and implement a new caching strategy; Through the above process, the system can adaptively adjust the weight coefficients of various factors, keep the caching strategy in the optimal state all the time, adapt to the real-time changing system requirements and network environment, improve the cache hit rate, and reduce the system latency and data loss rate.

[0038] Through the implementation of the above dynamic caching strategy, the system can achieve multi-level cache storage and intelligent data scheduling according to the importance score of the data, make full use of the cache resources, and improve the efficiency and reliability of data access; at the same time, based on the adaptive weight update mechanism of Bayesian optimization, it ensures the dynamic adjustment of the importance score, enables the caching strategy to adapt to the complex changes in the industrial Internet environment, and continuously optimizes the system performance.

[0039] As Figure 3 shown, the data storage system based on Internet technology is applied to a data acquisition method for a three-dimensional reconstruction system, including: a data acquisition module, a data preprocessing module, a scoring calculation module, a dynamic cache management module, an intelligent scheduling module, and an adaptive optimization module; the data acquisition module obtains the real-time data of industrial equipment through a sensor network and IoT devices ; the data preprocessing module is used to preprocess the real-time data to obtain preprocessed data ; the scoring calculation module is used to calculate the importance score of the data ; the dynamic cache management module is used to execute the dynamic caching strategy, including data storage, elimination, and multi-level cache management; the intelligent scheduling module uses a reinforcement learning algorithm to achieve intelligent data scheduling and transmission path optimization; the adaptive optimization module uses Bayesian optimization to achieve real-time update of weight coefficients.

[0040] The dynamic cache management module is used to execute the dynamic caching strategy, including data storage, elimination, and multi-level cache management: Cache storage: Store the data with importance score exceeding the threshold into the cache area; in this way, key data can be preferentially cached, improving the system's response speed and data processing efficiency; Cache eviction: When the cache capacity exceeds the threshold , use the LVD (Least Valuable Data) algorithm to remove data in ascending order of importance score ; this strategy ensures the caching of high-importance data while releasing space to store new data; Multi-level cache: According to the difference in importance score , data can be stored in cache media at different levels to implement a multi-level cache. The multi-level cache includes at least three levels of cache media: memory cache, solid-state drive cache, and disk cache. Specifically: Memory cache: Used to store data with the highest importance score, providing the fastest access speed.

[0041] Solid-state drive cache: Used to store data with medium importance score, balancing speed and capacity.

[0042] Disk cache: Used to store data with relatively low importance score, providing a large storage capacity.

[0043] Based on the data acquisition device of the 3D reconstruction system, there is a program of the 3D reconstruction system data acquisition method stored in the data acquisition device of the 3D reconstruction system. When a program of the 3D reconstruction system data acquisition method is executed by a processor, the steps of a 3D reconstruction system data acquisition method are implemented.

[0044] Data acquisition and preprocessing: Control the data acquisition module and the data preprocessing module through program instructions to complete the acquisition and preprocessing of real-time data; Importance score calculation: Call the program of the score calculation module to implement the calculation of the importance score of the data ; Dynamic cache policy execution: Use the program to control the dynamic cache management module to execute cache storage, eviction, and multi-level cache operations; Intelligent data scheduling: Schedule the intelligent scheduling module through the program, apply the reinforcement learning algorithm, and realize the intelligent scheduling and transmission optimization of cache data; Adaptive weight adjustment: Call the program of the adaptive optimization module and use the Bayesian optimization technique to adjust the weight coefficient in the score calculation in real time.

[0045] In summary, a data acquisition method for a three-dimensional reconstruction system provided by this embodiment realizes the dynamic optimization and adaptive adjustment of the data caching strategy through the organic combination and coordinated operation of each functional module, can efficiently integrate multi-source heterogeneous data, improve data compatibility and utilization rate, ensure data real-time performance and reliability, reduce three-dimensional reconstruction errors, optimize cache resource allocation, increase cache hit rate, reduce system latency, has the ability of adaptive dynamic optimization, balance storage cost and resource utilization rate, improve data quality and credibility, provide a reliable decision-making basis for three-dimensional reconstruction, can flexibly adapt to the requirements of different industrial scenarios, and expand the application scope.

[0046] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A method for data acquisition based on a three-dimensional reconstruction system, characterized in that, It includes the following steps: S1. Data acquisition: Configure multiple data sources for the 3D reconstruction system, and these data sources are used to collect the raw data related to 3D reconstruction; S2. Data preprocessing: The data received in real time is preprocessed and converted into a standardized data set , and this preprocessing process includes: Data cleaning: Eliminate abnormal, duplicate, and error data; Format unification: Convert data in different formats into a unified standard format; Feature extraction: Use the PCA algorithm to refine the key feature vectors ; Data compression: Implement lossless or lossy compression techniques to reduce storage and transmission costs; S3. Importance score calculation: For the preprocessed data , calculate its importance score , and the calculation formula is as follows: ； Among them: is the data access frequency; is the urgency level of the data; is the data change rate, and its calculation formula is: ; where is the time interval; is the prediction importance of the deep learning-based prediction model; For the correlation between data and key business indicators, the calculation formula is: ; represents the correlation coefficient, is the eigenvector, is the key performance indicator KPI; For data credibility, which reflects the reliability of the data source and the integrity of the data itself; , , , , , are weight coefficients; S4. Based on the importance score , implement a dynamic caching strategy.

2. The data acquisition method of a three-dimensional reconstruction system according to claim 1, wherein In S1, the specific steps of data acquisition are: S11. Configure multiple data sources, and the total number of data sources is , where is a positive integer, and each data source corresponds to an industrial device or sensor, numbered ; S12. For each data source , at a specific time mark , capture the raw data in real time through the industrial Internet of Things network , which contains the following information: Data source identification code , used to uniquely distinguish data sources ; Measured value or data content , representing the data source at time The original information collected Data format or type description , indicating the data source The structure, type, or format of the generated data, to facilitate the identification and processing of data from multiple sources and in multiple formats; Additional information , covering additional information related to the data; S13. Aggregate the raw data collected from all data sources in step S12 to construct a raw data set at a time , that is ; ​​ Wherein: represents the data collection activity at time ; represents taking all data sources at time to generate the raw data collection; S14. Perform time marker alignment and data integrity verification on the original data set ; S15. Calculate the data capture rate , to evaluate the effectiveness of data collection at time . The data capture rate is defined as: ; S16. Based on the data capture rate and the result of data integrity verification, determine whether the data collection process meets the standards of real-time performance and reliability. If the conditions are met, then the original data set is subjected to data preprocessing.

3. The data acquisition method of a three-dimensional reconstruction system according to claim 2, wherein Prediction Importance Use a spatio-temporal prediction model based on a graph convolutional network.

4. The data acquisition method of a three-dimensional reconstruction system according to claim 3, characterized in that, Data credibility Evaluate through blockchain technology, utilize a distributed ledger to record data source information, and achieve data traceability and anti-tampering.

5. A data acquisition method based on a three-dimensional reconstruction system according to claim 4, characterized in that Based on importance scoring , the specific steps for implementing the dynamic caching policy are as follows: S41. Cache storage: Store the data with an importance score exceeding the threshold in the cache area; S42. Cache Eviction: When the cache capacity exceeds the threshold the LVD algorithm is adopted to remove data in ascending order from low to high; S43. Multi-level caching: Depending on the differences, data can be stored in different levels of cache media for multi-level caching; S44. Intelligent data scheduling: Apply the reinforcement learning algorithm, and according to the system state and network conditions, intelligently schedule the cached data to optimize the data acquisition and transmission paths; S45. Adaptive weight update: Utilize the Bayesian optimization technique to adjust the weight coefficient in real time according to the cache hit rate H, system latency L, and data loss rate D.

6. The data acquisition method of a three-dimensional reconstruction system according to claim 5, wherein, The multi-level cache includes at least three levels of cache media: memory cache, solid-state drive cache, and disk cache, and allocates data to the corresponding cache level according to the importance score of the data. Allocate data to the corresponding cache level.

7. A data acquisition method based on a three-dimensional reconstruction system according to claim 6, characterized in that, The data allocation rules at the cache level are as follows: When store the data in the memory cache; When the data is stored in the solid state drive cache; When store the data in the disk cache; Among them, and are preset importance scoring thresholds, and satisfy .

8. A data acquisition method for a three-dimensional reconstruction system according to claim 7, characterized in that The calculation formula for adjusting the weight coefficient in real time is: ； Among them, , , are weight parameters, and .

9. A data acquisition system based on a three-dimensional reconstruction system, which is applied to a data acquisition method based on a three-dimensional reconstruction system as described in claim 8, characterized in that, It includes: Data acquisition module: Obtain real-time data of industrial equipment through sensor networks and IoT devices ; Data preprocessing module, used for real-time data to perform preprocessing and obtain preprocessed data ; Scoring calculation module, used to calculate the importance score of data ; A dynamic cache management module, which is used to execute dynamic cache policies, including data storage, elimination, and multi-level cache management; An intelligent scheduling module, which uses the reinforcement learning algorithm to achieve intelligent data scheduling and transmission path optimization; An adaptive optimization module, which adopts Bayesian optimization to achieve real-time update of the weight coefficient.

Images