Intelligent text annotation system, method and device for standard semantic knowledge extraction

Abstract

This invention discloses an intelligent text annotation system, method, and device for extracting standard semantic knowledge, relating to the field of electronic digital data processing technology. The intelligent text annotation system for extracting standard semantic knowledge includes: a data access synchronization verification module, a boundary recognition analysis module, and an annotation accuracy verification module. This invention performs data synchronization verification analysis by accessing synchronized data to determine whether structural filtering optimization is needed. Then, based on the recognition time of artifact feature points, it analyzes the accuracy of boundary recognition to determine whether segmentation parameter optimization is needed. Finally, based on deformation monitoring data, it determines whether resolution and frequency optimization are needed. This improves the accuracy of semantic annotation of corresponding identified content in the extraction of information from precious prints in 3D digital preservation, solving the problem of semantic annotation accuracy of corresponding identified content in the extraction of information from precious prints in 3D digital preservation in existing technologies.

Description

Technical Field

[0001] This invention relates to the field of electronic digital data processing technology, and in particular to an intelligent text annotation system, method and device for extracting standard semantic knowledge. Background Technology

[0002] The standardized information extraction process involves labeling and classifying data into multiple levels and types based on target attributes and application requirements. These include: basic physical attribute labeling, content semantic labeling, multimodal association labeling, and domain knowledge enhancement labeling. Identification and labeling of the extraction object are crucial for achieving the core pre-processing of standardized information extraction. Existing technologies utilize multi-source data collection and fusion, employing optical scanning of surface textures to determine color and specific details, while adhering to operational safety requirements to minimize damage to valuable cultural relics and texts. Secondly, the scanned data needs to be validated and registered in 3D space, aligning data from different sources to a unified coordinate system. Marker points are used to adapt the data to the coordinate system, while maintaining accuracy. Boundary segmentation of multimodal semantics is performed, and content is divided based on handwriting, ink, and color edges. This information is then applied to the 3D modeling space to determine modeling vectors, colors, star-shaped parameters, etc. Finally, point cloud verification is conducted to confirm the modeling data and point cloud, reducing modeling errors.

[0003] For example, Chinese invention patent with publication number CN114091406A discloses an intelligent text annotation method and system for knowledge extraction, which includes: annotating a portion of text data in unstructured text content to generate annotated data, and inputting this data as a training set into a deep learning model for training; using the trained deep learning model to predict and evaluate unannotated text; if the evaluation result is reliable, it is added to the training set; if the evaluation result is marked as unreliable, a new round of annotation is performed on representative samples from the unreliable samples, and the process is repeated until all identification is completed.

[0004] For example, Chinese invention patent CN112149423B discloses a corpus annotation method and system for joint extraction of domain entity relationships. The method includes: using the Scrapy crawler framework to crawl initial data that conforms to the specified adjustment from web pages and using regular expressions to preprocess the initial data. The process is as follows: filtering out the information to be filtered in the initial data to obtain filtered data, and using manual or regular expressions to supplement the missing parts to obtain noise-free text data; defining the relationship set between entities according to the characteristics of the domain corpus; and using the YEDDA text annotation tool to annotate the noise-free text data to obtain annotated data.

[0005] The above-mentioned technology has at least the following technical problems:

[0006] In existing technologies, the annotation classification for standardized information extraction is divided into text layer, image layer, and symbol layer based on the semantic annotation of the content. The text layer includes annotation of the coordinates and content of text blocks; the image layer classifies illustrations and seals by theme and identifies art styles; and the symbol layer annotates the semantics of special symbols, including annotations, seals, and punctuation marks, as well as the characteristics of writing tools. However, cultural relics have complex physical characteristics. For example, the material of paper has non-uniform characteristics, and the fiber orientation of different papers is random. For example, Xuan paper has fine fibers, which will form regular light bands along the grain direction. The light transmittance is relatively low due to the fluctuation of fiber arrangement, but it will cause the ink to seep and extend along the grain direction, resulting in errors in edge recognition. On the other hand, paper with coarse fibers will form irregular diffuse reflection on the surface, and the light transmittance of different parts will vary greatly due to the surface arrangement. The ink will exhibit anisotropic-star-shaped radial diffusion. This leads to the problem of low accuracy of semantic annotation of the corresponding content in the process of extracting information from the three-dimensional digital preservation of precious prints. Summary of the Invention

[0007] To address the technical problem of low accuracy in semantic annotation of corresponding content during the extraction of information from valuable printed materials in 3D digitization, this invention provides an intelligent text annotation system, method, and device for extracting standard semantic knowledge. The technical solution is as follows:

[0008] On one hand, an intelligent text annotation system for standard semantic knowledge extraction is provided. This system includes: a data access synchronization verification module, a boundary recognition analysis module, and an annotation accuracy verification module. The data access synchronization verification module performs data synchronization verification analysis on the digitization process of corresponding 3D modeling parameters during data stream access based on the acquired access synchronization data. This determines whether structural filtering optimization is needed, which involves adjusting the filter intensity and raster spacing to improve the synchronization response speed of text feature generation. The boundary recognition analysis module, after access synchronization is completed, performs boundary recognition accuracy analysis on the identification and annotation process of the extracted object based on the acquired artifact feature point recognition time. This determines whether segmentation parameter optimization is needed, which involves adjusting the boundary segmentation area to improve the recognition accuracy of text boundary determination. The annotation accuracy verification module, after boundary recognition is completed, performs data accuracy verification analysis on the annotation process of point cloud data and texture mapping based on the acquired deformation monitoring data. This determines whether resolution frequency optimization is needed, which involves adjusting the resolution and repetition frequency to improve the annotation accuracy of 3D modeling digitization parameters.

[0009] On the other hand, an intelligent text annotation method for standard semantic knowledge extraction is provided. This method includes: S1, performing data synchronization verification analysis on the digitization process of corresponding 3D modeling parameters during the data stream access process based on the acquired access synchronization data to determine whether structural filtering optimization is required; S2, after access synchronization is completed, performing boundary recognition accuracy analysis on the recognition and annotation process of the extracted object based on the acquired cultural relic feature point recognition time to determine whether segmentation parameter optimization is required; S3, after boundary recognition is completed, performing data accuracy verification analysis on the annotation process of point cloud data and texture mapping based on the acquired deformation monitoring data to determine whether resolution frequency optimization is required.

[0010] On the other hand, an intelligent text annotation device for standard semantic knowledge extraction is provided, which includes: a 3D laser scanner, a built-in timer, an interferometer, a spectrophotometer, and a laser confocal microscope; the 3D laser scanner is used to monitor the mean interval of the point cloud network; the built-in timer is used to monitor the synchronization time of the texture-point cloud and the recognition time of the artifact feature points; the interferometer is used to monitor the displacement range of the paper surface features; the spectrophotometer is used to monitor the variation range of the paper reflectivity; and the laser confocal microscope is used to monitor the variation range of the halo area ratio.

[0011] The beneficial effects of the technical solutions provided in the embodiments of the present invention include at least the following:

[0012] 1. By analyzing real-time access and synchronous data, the recognition time of cultural relic feature points and deformation monitoring data, the system triggers structural filter grating interval optimization, segmentation parameter optimization and resolution frequency optimization respectively. During scanning, it can automatically suppress spot jitter, optimize the point cloud raster vector alignment to reduce the risk of edge tearing, compensate for the boundary blurring phenomenon caused by the sudden change in the reflectivity of the object surface during high-speed rotating scanning, maintain the point cloud topology, suppress the amount of ghosting caused by the inertial displacement of the scanning head, achieve low latency guarantee for frame-by-frame alignment of texture semantic vectorization annotation during multimodal data concurrent processing, accelerate the seamless connection of point cloud patches in the high-precision scanning process, and form a composite digital solution that takes into account both scanning integrity guarantee and distortion correction.

[0013] 2. By dynamically integrating the cumulative effects of the paper surface feature displacement change rate, reflectivity fluctuation, and halo diffusion ratio in the time dimension, a synchronous influence index is generated. This accurately quantifies the distortion variables of high reflectivity deformation and bright spot interference on the reconstruction of the 3D modeling surface. The integration process effectively smooths short-term local noise abnormal fluctuations, enhances the matching reliability of the nonlinear extension law of feature point displacement trajectory with the historical spectrum of the database, and performs real-time calibration filtering to ensure that the sub-pixel level edge smoothness meets the rubbing authenticity verification standard, thereby improving the alignment accuracy of the multispectral tomography data stream.

[0014] 3. Accuracy impact index is constructed by dynamically verifying the deviation of the mean of point cloud network intervals and the efficiency deviation of texture-point cloud synchronization time in deformation monitoring data, accurately quantifying the bidirectional impact of deformation and point cloud vectorization reconstruction; dynamically suppressing texture coordinate drift, and adaptively optimizing the resolution projection compensation intensity based on the mean score feedback of point cloud network intervals to reduce the risk of texture boundary misalignment; achieving detail enhancement in the sub-pixel level annotation process, simultaneously improving color fidelity and engineering accuracy, and forming an accuracy that takes into account the reconstruction of complex materials and irregular surfaces. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0016] Figure 1 A schematic diagram of the structure of an intelligent text annotation system for extracting standard semantic knowledge provided in an embodiment of this application;

[0017] Figure 2 The flowchart of the data access synchronization verification module provided in the embodiments of this application is as follows;

[0018] Figure 3 The flowchart of the boundary recognition and analysis module provided in the embodiments of this application is as follows;

[0019] Figure 4 The flowchart is shown below for the annotation accuracy verification module provided in the embodiments of this application.

[0020] Figure 5 A flowchart of an intelligent text annotation method for extracting standard semantic knowledge provided in an embodiment of this application. Detailed Implementation

[0021] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0022] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0023] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0024] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0025] This invention provides an intelligent text annotation system for extracting standard semantic knowledge, such as... Figure 1 The diagram shows the structure of an intelligent text annotation system for standard semantic knowledge extraction. The system's processing flow can include the following modules: a data access synchronization verification module, a boundary recognition analysis module, and an annotation accuracy verification module. The data access synchronization verification module performs data synchronization verification analysis on the digitization process of corresponding 3D modeling parameters during data stream access based on the acquired access synchronization data. This determines whether structural filtering optimization is needed, which involves adjusting the filter intensity and raster spacing to improve the synchronization response speed of text feature generation. The boundary recognition analysis module, after access synchronization is complete, performs boundary recognition accuracy analysis on the object's identification and annotation process based on the acquired artifact feature point recognition time. This determines whether segmentation parameter optimization is needed, which involves adjusting the boundary segmentation area to improve the accuracy of text boundary determination. The annotation accuracy verification module, after boundary recognition is complete, performs data accuracy verification analysis on the annotation process of point cloud data and texture mapping based on the acquired deformation monitoring data. This determines whether resolution and frequency optimization are needed, which involves adjusting the resolution and repetition frequency to improve the annotation accuracy of 3D modeling digitization parameters.

[0026] In this embodiment, the data access synchronization verification module analyzes the access synchronization data and performs structural filtering optimization to effectively improve the synchronization response speed of the data stream generating text features, making the 3D modeling data access smoother and more efficient, and reducing data latency and errors. The boundary recognition analysis module analyzes the accuracy of boundary recognition based on the recognition time of cultural relic feature points and optimizes the segmentation parameters, which can accurately determine the text boundary, making the 3D model boundary clearer and more accurate, and improving the modeling detail. The annotation accuracy verification module uses deformation monitoring data to perform data accuracy verification analysis and resolution frequency optimization, which can improve the annotation accuracy of 3D modeling digital parameters, ensure accurate matching of model texture, point cloud and other data, and make the final generated 3D model of higher quality and stronger realism.

[0027] like Figure 2 The diagram shown is a flowchart of the data access synchronization verification module provided in this application embodiment. The specific design logic is as follows: 1. Determine whether the acquired data access synchronization impact index is greater than a preset value in the database. If it is not greater, perform boundary recognition accuracy analysis. If it is greater, dynamically adjust the filtering strength according to the acquired deviation, and perform data synchronization verification test on the adjusted filtering strength. 2. Determine whether the reacquired index is not greater than a preset value. If it is, perform boundary recognition accuracy analysis. If not, dynamically adjust the raster spacing and reacquire the index. Determine whether the new index is greater than a preset index. If it is, perform a structure processing warning; otherwise, perform boundary recognition accuracy analysis.

[0028] Furthermore, based on the acquired access synchronization data, a data synchronization verification analysis is performed on the digitization process of the corresponding 3D modeling parameters during the data stream access process. Specific steps include: within the defined data stream access time interval, performing integral processing on the acquired access synchronization data, and simultaneously performing coupled calculations on the area of ​​the acquired access synchronization data to obtain data access synchronization impact indicators. The data stream access time interval represents the unit time period used to analyze data synchronization efficiency; integral processing represents the quantification of the cumulative changes in access synchronization data within the data stream access time interval. The access synchronization data includes the amplitude of paper surface feature displacement change, the amplitude of paper reflectance change, and the amplitude of halo area ratio change. The amplitude of paper surface feature displacement change represents the difference between the maximum and minimum displacement of the valuable text at the end of the data stream access time interval; the amplitude of paper reflectance change represents the difference between the maximum and minimum reflectance of the valuable text at the end of the data stream access time interval; and the amplitude of halo area ratio change represents the difference between the maximum and minimum halo area ratio of the valuable text at the end of the data stream access time interval. It is important to understand that a data stream access time interval may contain only one maximum and one minimum value.

[0029] The area of ​​synchronized access data includes the area of ​​change in paper surface feature displacement, the area of ​​change in paper reflectivity, and the area of ​​change in halo area ratio. The area of ​​change in paper surface feature displacement represents the area under the integral curve corresponding to the change in paper surface feature displacement over time within the data stream access time interval, used to quantify the cumulative effect of feature displacement change. The area of ​​change in paper reflectivity represents the area under the integral curve corresponding to the change in reflectivity over time within the data stream access time interval, used to quantify the cumulative effect of reflectivity change. The area of ​​change in halo area ratio represents the area under the integral curve corresponding to the change in halo area ratio over time within the data stream access time interval, used to quantify the cumulative effect of halo area ratio change. The data access synchronization impact index represents the quantitative data on the degree of influence of synchronized access data on the data synchronization verification analysis of the digitization process of 3D modeling parameters.

[0030] In this embodiment, the data access synchronization impact index increases with the increase of the displacement amplitude of paper surface features, the change amplitude of paper reflectivity, and the change amplitude of halo area ratio. When the displacement amplitude of paper surface features increases, it means that the paper's position shifts more drastically during scanning or data acquisition. This directly interferes with the scanning device's accurate capture of paper surface features, causing the acquired data to be misaligned in the three-dimensional modeling coordinate system. Consequently, the corresponding positions of reflectivity data and halo area ratio data deviate, affecting the consistency and synchronization of these three data.

[0031] An increased variation in paper reflectance indicates a deterioration in the uniformity of the paper's surface material or color, leading to greater differences in reflectivity across different areas. This intensifies signal intensity fluctuations received by the scanning device, interfering with the accurate assessment of paper surface feature displacement. It also affects the accuracy of halo area ratio measurement, as changes in reflectance alter the scattering and reflection patterns of light on the paper surface. An increased variation in halo area ratio typically indicates non-uniformity in the scanning light source or localized optical anomalies on the paper surface, such as oil stains or wrinkles. These abnormal areas interfere with normal data acquisition, resulting in inaccurate reflectance measurements and errors in the measurement of paper surface feature displacement, as the halo may obscure or distort the true location of feature points.

[0032] By considering the aforementioned interrelationships, it is helpful to identify factors that may affect the accuracy and synchronization of data in advance during the data acquisition and synchronous verification stages of 3D modeling. At the same time, by constructing dynamic monitoring through multi-dimensional parameter integration, quality optimization can be achieved during the 3D modeling digitization process. Based on the data flow integration processing mechanism of time interval, the cumulative evolution characteristics of surface displacement, reflectivity fluctuations, and halo diffusion can be accurately captured. The parameter settings of scanning equipment can be optimized in a targeted manner. For example, the light source intensity and angle can be adjusted to reduce the influence of halo, and the scanning resolution and speed can be optimized to reduce the influence of paper displacement, thereby improving the accuracy and reliability of data synchronous verification analysis in the 3D modeling parameter digitization process.

[0033] Further, to determine whether structural filtering optimization should be performed, the specific steps are as follows: The acquired data access synchronization impact index is compared with the preset data access synchronization impact index in the database. If the acquired data access synchronization impact index is greater than the preset data access synchronization impact index in the database, it is recorded as unqualified data access synchronization, and structural filtering optimization and verification are performed. If the acquired data access synchronization impact index is not greater than the preset data access synchronization impact index in the database, it is recorded as qualified data access synchronization, and boundary identification accuracy analysis is performed. The specific steps of structural filtering optimization are as follows: Based on the deviation of the acquired data access synchronization impact index, a filter intensity adjustment value is mapped in the database to quantify the degree of dynamic adjustment of the filter intensity, thereby improving the noise processing speed in high-reflection areas.

[0034] The specific steps for structural filter optimization verification are as follows: After adjusting the filter intensity, a filter effect verification command is sent to preset personnel to conduct data synchronization verification tests based on the adjusted filter intensity. If the data synchronization impact index of the reacquired data after the data synchronization verification test is not greater than the preset data synchronization impact index, the structural filter optimization is completed and boundary recognition accuracy analysis is performed; otherwise, raster spacing optimization is performed. The specific steps for raster spacing optimization are as follows: Based on the deviation of the reacquired data synchronization impact index, the raster spacing adjustment value is mapped in the database to quantify the distance between the raster grids that needs to be dynamically adjusted to reduce the error of raster projection. After structural filter optimization, it is determined whether the reacquired data synchronization impact index is greater than the preset data synchronization impact index in the database. If not, the structural filter optimization is completed and boundary recognition accuracy analysis is performed; if so, a structural processing warning is issued.

[0035] In this embodiment, the deviation of the data access synchronization impact index represents the difference between the acquired data access synchronization impact index and the preset data access synchronization impact index in the database. The preset data access synchronization impact index is represented by the summation and averaging of historical data access synchronization impact indices within the historical data stream access time interval. The filter intensity is adjusted in a timely manner based on the acquired current filter intensity using a feature decomposition algorithm to improve the speed of noise processing. At the same time, the historical filter intensity and raster spacing are used as sample data and input into the hybrid regression model. The filter intensity-raster spacing hybrid regression model is trained based on the feature decomposition algorithm to obtain the filter intensity-raster spacing hybrid regression model. The acquired filter intensity and raster spacing are input into the filter intensity-raster spacing hybrid regression model to output the corresponding filter intensity adjustment value and raster spacing adjustment value.

[0036] A closed-loop adjustment mechanism using dynamic threshold judgment and progressive optimization is used to achieve real-time calibration of 3D modeling parameters: by comparing the difference between measured indicators and preset values ​​in real time, the structural filter adjustment intensity is intelligently triggered to prioritize the elimination of high reflectivity noise and quickly improve the effective signal-to-noise ratio of feature data capture; when the filter parameter optimization fails to meet the standard, it automatically switches to the grating spacing optimization mode, and reduces geometric information distortion caused by optical deformation by precisely reconfiguring the physical projection parameters of the grating; the bidirectional optimization path combined with the hybrid control mechanism of manual verification ensures sensitive adaptive parameter fine-tuning and forms a fault-tolerant control system with multi-dimensional parameter linkage.

[0037] like Figure 3 The diagram shown is a flowchart of the boundary recognition analysis module provided in this application embodiment. The specific design logic is as follows: determine whether the recognition time of the acquired cultural relic feature points is greater than the preset time. If not, it means that the boundary recognition is qualified and data accuracy verification analysis is performed. If yes, the segmentation parameters are optimized, the boundary adjustment value is mapped based on the time deviation, and the effect of the adjusted boundary is verified. determine whether the re-acquired time is not greater than the preset time in the database. If yes, the segmentation parameter optimization is completed and data accuracy verification analysis is performed. Otherwise, a boundary recognition warning is sent.

[0038] Further, to determine whether segmentation parameter optimization is necessary, the specific steps are as follows: compare the acquired artifact feature point recognition time with the preset artifact feature point recognition time in the database: if the acquired artifact feature point recognition time is greater than the preset artifact feature point recognition time in the database, it is recorded as boundary recognition failure and segmentation parameter optimization is performed; if the acquired artifact feature point recognition time is not greater than the preset artifact feature point recognition time in the database, it is recorded as boundary recognition success and data accuracy verification analysis is performed; the artifact feature point recognition time represents the time length from data acquisition to artifact print completion of feature point detection, used to quantify the processing efficiency of artifact features during digitization. The preset artifact feature point recognition time is represented by the sum and average of historical artifact feature point recognition times within the historical boundary recognition time interval.

[0039] The specific steps for optimizing segmentation parameters are as follows: Based on the deviation of the acquisition of cultural relic feature point recognition time, the segmentation boundary adjustment value is mapped in the database to quantify the degree to which the current boundary segmentation area needs to be reduced, so as to promote the recognition of blurred edges caused by ink smudging, and thus improve the accuracy of texture consistency optimization boundary determination; After the segmentation boundary adjustment test, the segmentation boundary effect verification instruction is sent to the preset personnel to perform boundary recognition specification verification based on the segmentation boundary after the segmentation parameter optimization adjustment. If the cultural relic feature point recognition time re-acquired after the boundary recognition specification test is not greater than the preset cultural relic feature point recognition time, the segmentation parameter optimization is completed and the data accuracy verification analysis is performed; otherwise, a boundary recognition warning is issued. The cultural relic feature point recognition time deviation represents the difference between the acquisition of cultural relic feature point recognition time and the preset cultural relic feature point recognition time in the database.

[0040] In this embodiment, the Conditional Random Field Post-Optimization Algorithm is used to dynamically adjust the segmentation boundary based on the current segmentation boundary, thereby promoting the recognition of edge blurring caused by ink smudging. At the same time, the segmentation boundary area is used as sample data and input into the convolutional neural network model. The boundary recognition-convolutional neural network model is trained based on the Conditional Random Field Post-Optimization Algorithm. The current segmentation boundary area is input into the boundary recognition-convolutional neural network model to adaptively adjust the input segmentation boundary adjustment value.

[0041] This example achieves dynamic parameter optimization by real-time monitoring of feature processing efficiency during the digitization of cultural relics. It employs a segmentation strategy initiated by analyzing recognition time: when the feature point detection time exceeds the limit, the segmentation range is automatically reduced to solve the problem of outline blurring caused by ink smudges and enhance the positioning accuracy of complex patterns. The quantization and narrowing of the segmentation boundary not only compresses invalid recognition areas to improve processing speed, but also strengthens the gradient contrast at the ink smudge boundary through area selection and focusing to curb texture discontinuity. The segmentation optimization effect is confirmed through testing to prevent information loss caused by over-segmentation.

[0042] like Figure 4 The diagram shows the workflow of the annotation accuracy verification module provided in this application embodiment. The specific design logic is as follows: determine whether the accuracy impact index of the acquired data is greater than the preset index. If not, complete the data accuracy verification analysis. If yes, dynamically adjust the resolution and perform data accuracy verification test on the adjusted data. Determine whether the reacquired index is not greater than the preset index. If yes, complete the data accuracy verification analysis. Otherwise, adjust the repetition frequency to reacquire the data accuracy impact index. Determine whether the new index is greater than the preset index. If yes, issue a data accuracy warning. Otherwise, complete the data accuracy verification analysis.

[0043] Furthermore, based on the obtained deformation monitoring data, the annotation process of point cloud data and texture mapping is analyzed for data accuracy verification. The specific steps are as follows: Obtain deformation monitoring data at the end of the accuracy verification period, simultaneously obtain the preset deformation monitoring data from the database, and perform proportional calculation to obtain the deformation monitoring data score. At the same time, the obtained deformation monitoring data score is harmonic averaged to obtain the data accuracy impact index.

[0044] The deformation monitoring data includes the mean point cloud network interval and the texture-point cloud synchronization time. The data accuracy impact index is used to quantify the impact of deformation monitoring data on the accuracy of data verification. The deformation monitoring data score includes the mean point cloud network interval score and the texture-point cloud synchronization time score. The mean point cloud network interval score represents the ratio of the obtained mean point cloud network interval to the preset mean point cloud network interval. The texture-point cloud synchronization time score represents the ratio of the obtained texture-point cloud synchronization time to the preset texture-point cloud synchronization time. The preset deformation monitoring data includes the preset mean point cloud network interval and the preset texture-point cloud synchronization time. The mean point cloud network interval represents the statistical mean of the distance between adjacent points in the 3D point cloud. The texture-point cloud synchronization time represents the time length between the texture data and the point cloud data in the point cloud space from standby to application during the synchronization process.

[0045] In this embodiment, the data accuracy impact index increases with the increase of the average point cloud network interval and the texture-point cloud synchronization time. When the average point cloud interval increases, the amount of data decreases and the transmission time is shortened, resulting in a reduction in synchronization time, but this can lead to a loss of detail. Conversely, when the average interval decreases, the increase in data density can enhance the possibility of feature matching, but the increase in load leads to memory bandwidth pressure and parallel computing bottlenecks, thus increasing the synchronization time of both.

[0046] By considering the aforementioned interrelationships, it is helpful to accurately balance the relationship between data volume, detail preservation, and synchronization efficiency during the data acquisition phase, based on the correlation between the mean point cloud network interval and the texture-point cloud synchronization time. This avoids problems in other aspects caused by excessively pursuing one metric. During system operation, changes in these key metrics can be monitored in real time, potential performance bottlenecks can be predicted in advance, and system parameters can be adjusted in a timely manner to ensure high-quality generation and efficient synchronization of 3D model data.

[0047] Furthermore, the specific steps for determining whether to perform resolution frequency optimization include: if the accuracy impact index of the acquired data is greater than the preset data accuracy impact index in the database, then the data is recorded as inaccurate and resolution frequency optimization is performed; if the accuracy impact index of the acquired data is not greater than the preset data accuracy impact index in the database, then the data is recorded as accurate and data accuracy verification analysis is completed.

[0048] The resolution frequency optimization process involves the following steps: First, based on the deviation of the acquired data accuracy impact index, a resolution adjustment value is mapped into the database. This value quantifies the degree to which the resolution of a specified object in the modeling area needs to be dynamically changed, thereby enhancing the ability to capture target details and balancing the energy consumption and detection accuracy of high resolution. After the resolution frequency adjustment, a resolution effect verification command is sent to pre-selected personnel to conduct data accuracy verification tests based on the optimized resolution. If the deviation of the data accuracy impact index after the data accuracy verification test is not greater than the pre-selected data accuracy impact index, then the resolution frequency optimization is completed and a smart text annotation completion command is sent; otherwise, the frequency optimization is repeated.

[0049] The specific steps of repetition frequency optimization include: mapping the deviation of the accuracy impact index of the reacquired data into the database to obtain the repetition frequency adjustment value, which is used to quantify the degree to which the scanning repetition frequency of the specified object in the modeling area needs to be dynamically changed, so as to improve the point cloud generation rate and thus balance the relationship between cultural relic damage and point cloud density; after resolution frequency optimization, it is determined whether the acquired data accuracy impact index is greater than the preset data accuracy impact index in the database. If so, a data accuracy warning is issued; otherwise, the resolution frequency optimization is completed and the intelligent text annotation instruction is sent.

[0050] In this embodiment, the resolution is adjusted in real time based on the size of the current specified region resolution by the iterative nearest parameter algorithm, which improves the detection accuracy. At the same time, the resolution adjustment value and the scan repetition frequency adjustment value are used as sample data and input into the neural architecture search model. The resolution-repetition frequency neural architecture search model is obtained by training based on the iterative nearest parameter algorithm. The obtained resolution size and repetition frequency are input into the resolution-repetition frequency neural architecture search model and finally output the corresponding resolution adjustment value and repetition frequency adjustment value.

[0051] The deviation of the data accuracy impact index represents the difference between the deviation of the acquired data accuracy impact index and the preset data accuracy impact index in the database. By establishing dynamic resolution adjustment, an intelligent balance between 3D modeling accuracy and system energy consumption is achieved. When insufficient capture of target details is detected, the regional resolution is automatically increased to enhance microcrack and geometric reproduction. The multi-dimensional parameter real-time joint adjustment mechanism prevents the risk of cumulative light damage in traditional high-precision scanning, enabling high-frequency scanning operations to maintain the point cloud density threshold while controlling the total laser projection dose. The closed-loop verification system integrates automatic calibration and test confirmation dual verification channels, and gradually approaches the optimal acquisition scheme through progressive resolution fine-tuning, which preserves the complete topological information of fragile textures and avoids temperature and humidity disturbances caused by over-scanning.

[0052] like Figure 5The diagram shows a flowchart of an intelligent text annotation method for standard semantic knowledge extraction provided in this application embodiment. The intelligent text annotation method for standard semantic knowledge extraction provided in this application embodiment includes the following steps: S1, performing data synchronization verification analysis on the digitization process of corresponding 3D modeling parameters during data stream access based on the acquired access synchronization data, to determine whether structural filtering optimization is required; S2, after access synchronization is completed, performing boundary recognition accuracy analysis on the identification and annotation process of the extracted object based on the acquired artifact feature point recognition time, to determine whether segmentation parameter optimization is required; S3, after boundary recognition is completed, performing data accuracy verification analysis on the annotation process of point cloud data and texture mapping based on the acquired deformation monitoring data, to determine whether resolution frequency optimization is required.

[0053] In this embodiment, firstly, structural filtering optimization eliminates the surface data jumps caused by scanning equipment vibration in real time, and the topological fidelity of surface texture is improved through dynamic noise reduction algorithm; secondly, adaptive segmentation parameters iteratively adjust the lens focal length correlation threshold to accurately determine the feature mixing boundary, and intelligently lock the critical point by combining curvature spectrum recognition technology; finally, the resolution dynamic modulation module automatically switches the multi-frequency laser sampling mode according to the surface characteristics of the cultural relic.

[0054] The intelligent text annotation device for standard semantic knowledge extraction provided in this application includes: a 3D laser scanner, a built-in timer, an interferometer, a spectrophotometer, and a laser confocal microscope; the 3D laser scanner is used to monitor the mean interval of the point cloud network; the built-in timer is used to monitor the synchronization time of texture-point cloud and the recognition time of cultural relic feature points; the interferometer is used to monitor the displacement variation of paper surface features; the spectrophotometer is used to monitor the variation of paper reflectivity; and the laser confocal microscope is used to monitor the variation of halo area ratio. The data received by the above devices is stored in a memory through a high-speed transmission channel for data temporary relay and data storage, so as to enable real-time retrieval and analysis, subsequent data tracing, report generation, and deep learning.

[0055] It should be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0056] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. This computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, all or part of the flow or function according to the embodiments of the present invention is generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0057] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0058] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0059] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0060] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0061] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0062] In the embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0063] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0064] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0065] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0066] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An intelligent text annotation system for extracting standard semantic knowledge, characterized in that, include: Data access synchronization verification module, boundary identification and analysis module, and annotation accuracy verification module; The data access synchronization verification module is used to perform data synchronization verification analysis on the digitization process of corresponding 3D modeling parameters during the data stream access process based on the acquired access synchronization data, so as to determine whether structural filtering optimization is required. The structural filtering optimization refers to improving the synchronization response speed of the data stream generating text features by adjusting the filtering intensity and the raster spacing. The boundary recognition analysis module is used to perform boundary recognition accuracy analysis on the recognition and annotation process of the extracted object based on the recognition time of the cultural relic feature points after the access synchronization is completed, so as to determine whether to optimize the segmentation parameters. The segmentation parameter optimization means improving the recognition accuracy of text boundary determination by adjusting the boundary segmentation area. The annotation accuracy verification module is used to perform data accuracy verification analysis on the annotation process of point cloud data and texture mapping based on the acquired deformation monitoring data after boundary recognition is completed, in order to determine whether resolution frequency optimization should be performed. Resolution frequency optimization means improving the annotation accuracy of 3D modeling digital parameters by adjusting the resolution and repetition frequency. The specific steps for determining whether to perform structural filtering optimization are as follows: The acquired data access synchronization impact indicators are compared with the preset data access synchronization impact indicators in the database: If the obtained data access synchronization impact index is greater than the preset data access synchronization impact index in the database, it is recorded as data access synchronization unqualified and structural filtering optimization and structural filtering optimization verification are performed. If the obtained data access synchronization impact index is not greater than the preset data access synchronization impact index in the database, then the data access synchronization is considered qualified and boundary identification accuracy analysis is performed. The specific steps for optimizing the structural filtering are as follows: Based on the deviation of the access synchronization impact index of the acquired data, the filter strength adjustment value is mapped in the database to quantify the degree of dynamic adjustment of the filter strength, so as to improve the noise processing speed in high-reflection areas. The specific steps for the structural filter optimization verification are as follows: After adjusting the filter strength, a filter effect verification command is sent to the preset personnel to conduct a data synchronization verification test based on the filter strength after structural filter optimization. If the data access synchronization impact index of the data re-acquired after the data synchronization verification test is not greater than the preset data access synchronization impact index, the structural filter optimization is completed and the boundary recognition accuracy analysis is performed; otherwise, the raster spacing optimization is performed. The specific steps for optimizing the grating spacing are as follows: Based on the deviation of the access synchronization impact index of the reacquired data, the raster grid spacing adjustment value is mapped in the database. This value is used to quantify whether the distance between the raster grids needs to be increased or decreased to reduce the error of the raster projection. After structural filtering optimization, it is determined whether the impact index of the newly acquired data access synchronization is greater than the preset impact index of the data access synchronization in the database. If not, the structural filtering optimization is completed and the accuracy analysis of boundary identification is performed. If so, a structural processing warning is issued. The process of verifying and analyzing the accuracy of the annotation of point cloud data and texture mapping based on the acquired deformation monitoring data includes the following steps: Obtain deformation monitoring data at the end of the accuracy verification period, simultaneously retrieve preset deformation monitoring data from the database, perform proportional calculations to obtain deformation monitoring data scores, and simultaneously perform harmonic averaging on the obtained deformation monitoring data scores to obtain data accuracy impact indicators. The deformation monitoring data includes the average point cloud network interval and the texture-point cloud synchronization time. The data accuracy impact index is used to quantify the degree of impact of deformation monitoring data on the accuracy of data verification. The specific steps for determining whether to perform resolution frequency optimization include: If the accuracy impact index of the acquired data is greater than the preset data accuracy impact index in the database, it is recorded as inaccurate data and resolution frequency optimization is performed. If the accuracy impact index of the acquired data is not greater than the preset data accuracy impact index in the database, then the data is considered accurate and the data accuracy verification analysis is completed.

2. The intelligent text annotation system for standard semantic knowledge extraction as described in claim 1, characterized in that, The process of digitizing the corresponding 3D modeling parameters during data stream access based on the acquired access synchronization data includes the following steps: Within the defined data stream access time interval, the acquired access synchronization data is integrally processed, and the area of ​​the acquired access synchronization data is coupled and calculated to obtain the data access synchronization impact index. The data stream access time interval represents a unit time period used to analyze data synchronization efficiency. The integral processing represents the cumulative changes of access synchronization data within the data stream access time interval. The access synchronization data includes the displacement amplitude of paper surface features, the amplitude of paper reflectivity, and the amplitude of halo area ratio. The data access synchronization impact index represents the quantitative data on the degree of impact of access synchronization data on the data synchronization verification analysis of the digitization process of 3D modeling parameters.

3. The intelligent text annotation system for standard semantic knowledge extraction as described in claim 1, characterized in that, The specific steps for determining whether to optimize the segmentation parameters are as follows: The identification time of the acquired cultural relic feature points is compared with the preset identification time of cultural relic feature points in the database: If the recognition time of the acquired cultural relic feature points is longer than the preset recognition time of cultural relic feature points in the database, it is recorded as unqualified boundary recognition and the segmentation parameters are optimized. If the recognition time of the acquired cultural relic feature points is not greater than the preset recognition time of cultural relic feature points in the database, it is recorded as qualified for boundary recognition and data accuracy verification analysis is performed. The artifact feature point recognition time refers to the length of time it takes for the artifact print to complete feature point detection after data collection.

4. The intelligent text annotation system for standard semantic knowledge extraction as described in claim 3, characterized in that, The specific steps for optimizing the segmentation parameters are as follows: Based on the deviation in the recognition time of the cultural relic feature points, the segmentation boundary adjustment value is mapped in the database to quantify the degree to which the boundary segmentation area needs to be reduced, so as to promote the recognition of blurred edges caused by ink smudging. After the segmentation boundary adjustment test, a segmentation boundary effect verification command is sent to the preset personnel to perform boundary recognition specification verification based on the optimized and adjusted segmentation boundary according to the segmentation parameters. If the recognition time of the cultural relic feature points re-acquired after the boundary recognition specification test is not greater than the preset cultural relic feature point recognition time, the segmentation parameter optimization is completed and the data accuracy verification analysis is performed; otherwise, a boundary recognition warning is issued.

5. The intelligent text annotation system for standard semantic knowledge extraction as described in claim 4, characterized in that, The resolution frequency optimization involves the following steps: Based on the accuracy of the acquired data, the deviation of the indicator is mapped in the database to obtain a resolution adjustment value, which is used to quantify the degree to which the resolution of a specified object in the modeling area needs to be dynamically changed, so as to enhance the ability to capture target details; After adjusting the resolution frequency, a resolution effect verification command is sent to the preset personnel to conduct a data accuracy verification test based on the resolution frequency optimization. If the deviation of the accuracy impact index of the data re-acquired after the data accuracy verification test is not greater than the preset data accuracy impact index, the resolution frequency optimization is completed and a smart text annotation completion command is sent; otherwise, the frequency optimization is repeated. The specific steps for optimizing the repetition frequency include: Based on the deviation of the indicator of the accuracy of the reacquired data, the repetition frequency adjustment value is obtained by mapping it in the database. This value is used to quantify the degree to which the scanning repetition frequency of the specified object in the modeling area needs to be dynamically changed, so as to improve the point cloud generation rate. After resolution and frequency optimization, determine whether the accuracy impact index of the acquired data is greater than the preset data accuracy impact index in the database. If so, issue a data accuracy warning; otherwise, complete the resolution and frequency optimization and send a command to complete the intelligent text annotation.

6. A method applied to an intelligent text annotation system for standard semantic knowledge extraction as described in any one of claims 1-5, characterized in that, Includes the following steps: S1. Based on the acquired access synchronization data, perform data synchronization verification analysis on the digitization process of the corresponding 3D modeling parameters during the data stream access process to determine whether structural filtering optimization is required. S2, after the access synchronization is completed, the accuracy of boundary recognition is analyzed for the identification and annotation process of the extracted objects based on the recognition time of the cultural relic feature points, so as to determine whether to optimize the segmentation parameters; S3. After boundary recognition is completed, the accuracy of the annotation process of point cloud data and texture mapping is verified based on the obtained deformation monitoring data to determine whether resolution frequency optimization should be performed.

7. A device for applying the intelligent text annotation system for standard semantic knowledge extraction as described in any one of claims 1-5, comprising: 3D laser scanner, built-in timer, interferometer, spectrophotometer and laser confocal microscope; The 3D laser scanner is used to monitor the average interval of the point cloud network. The built-in timer is used to monitor the texture-point cloud synchronization duration and the artifact feature point recognition duration; The interferometer is used to monitor the magnitude of displacement changes of characteristic features on the paper surface. The spectrophotometer is used to monitor the variation in paper reflectance. The laser co-concentration microscope is used to monitor the variation in the halo area ratio.

Images