Method and system for checking metrological detection data based on the fusion of OCR and NLP

Abstract

The application discloses a kind of based on OCR and NLP fusion's metrological detection data verification method and system, wherein the method comprises: obtaining original image data containing metrological equipment identification mark and real-time reading, and synchronously collecting environmental parameter sensor data and detecting document text;Image recognition data set is obtained by using identification model to extract equipment nameplate and instrument value data;Text semantic data set is obtained by using NLP analysis engine to extract calibration procedure and historical detection parameters;The above multi-modal data is input into multi-source data fusion and anomaly detection model, and the spatiotemporal consistency measure index is calculated by combining preset principle and self-encoder model, to realize the intelligent identification of multi-modal data abnormal characteristics;The standard verification result after rectification is output by using deep learning error correction model and warning is given when triggering anomaly.The application solves the problems that manual transcription of metrological detection data is prone to error, heterogeneous data is difficult to fuse and it is difficult to trace the whole process, and effectively improves the detection accuracy and automation level.

Description

Technical Field

[0001] This invention relates to the fields of image processing and data verification technology, and more specifically, to a method and system for verifying metrological testing data based on the fusion of OCR and NLP. Background Technology

[0002] As the "eyes" of industrial production and quality improvement, metrology and testing play a core role in ensuring measurement values ​​in key industrial chains such as high-end equipment manufacturing, biomedicine, and new energy and new materials. With the increasing sophistication of modern industry, the demand for verification and calibration of batch equipment has exploded, placing higher demands on the real-time performance, accuracy, and traceability of metrology and testing data.

[0003] However, traditional metrology and data processing methods still heavily rely on manual operation, and existing solutions suffer from the following three key technical drawbacks: 1. Low data collection efficiency and backlog: Traditional verification processes rely on manual reading by inspectors, paper records, and manual transcription into the management system. Verification and data entry for a single device are time-consuming and cumbersome, and data backlog is severe in batch testing scenarios, failing to meet the high-frequency, real-time testing response requirements of industrial sites.

[0004] 2. High transcription error and significant subjective interference: Due to factors such as differences in the experience of testing personnel, visual fatigue, changes in ambient lighting, and damage to instrument markings (such as blurred metal stamps or worn laser engravings), manual reading and transcription are prone to misreading, omission, or misrecording, resulting in a persistently high data error rate that directly affects the authority and accuracy of the measured values.

[0005] 3. Heterogeneous Data Fragmentation and Traceability Chain Disruption: Metrological testing involves multi-source heterogeneous data, including images (such as nameplate / instrument readings), sensor stream data (such as temperature, humidity / vibration), and unstructured text (such as verification procedures / work orders). Existing technologies lack multimodal fusion verification mechanisms, and data is scattered and stored in paper documents or independent databases, resulting in fragmentation between systems. This leads to a lack of spatiotemporal consistency verification in the testing process, making it difficult to trace historical data throughout its entire lifecycle. Summary of the Invention

[0006] The purpose of this invention is to provide a measurement and testing data verification method and system based on the fusion of OCR and NLP, which solves the technical problems in the prior art such as low efficiency and error-proneness of artificial transcription, ineffective fusion and verification of fragmented multi-source heterogeneous data, and difficulty in quality traceability.

[0007] The first aspect of this invention provides a method for verifying metrological testing data based on the fusion of OCR and NLP, comprising the following steps: Acquire raw image data and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process. The raw image data includes at least the metrological equipment identification mark and real-time reading. The original image data is subjected to target detection and character localization using a preset recognition model. Key information from the equipment nameplate and the readings of pointer / digital displays are extracted to obtain an image recognition dataset. The detection document text is subjected to entity extraction and semantic parsing using a natural language processing parsing engine. Key indicators of the verification procedure and historical detection parameters are extracted to obtain a text semantic dataset. The image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features. The abnormal features of the multimodal data are input into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered.

[0008] In this solution, the acquisition of raw image data and the simultaneous collection of environmental parameter sensor data and detection document text during the metrological testing process specifically include: The raw image data, environmental parameter sensor data, and detection document text are acquired in parallel through distributed mobile acquisition terminals or sensor network interfaces. The environmental parameter sensor data includes temperature, humidity, and vibration parameters; The testing documentation includes the testing order, verification procedures, and historical testing record sheets; and When the network is offline, all types of data are encrypted and cached offline, and the cached data is automatically uploaded to the cloud server for two-way synchronization after the network is restored.

[0009] In this solution, the step of using a preset recognition model to perform target detection and character localization on the original image data, extracting key information from the equipment nameplate and the reading data from pointer / digital displays, and obtaining an image recognition dataset specifically includes: The recognition model includes an improved YOLOv8 recognition model, wherein an attention mechanism is introduced into the feature extraction network of the recognition model; The improved YOLOv8 recognition model is used to perform localization detection on the original image data, wherein the localization area includes at least the equipment nameplate, laser engraving, stamped markings, and instrument pointers / digital displays; The positioning area is segmented and recognized by an optical character recognition algorithm to extract key information of the device nameplate and the indicated value data in real time. The image recognition dataset is obtained by verifying the range rationality and format based on a preset metrology standard database.

[0010] In this solution, the step of using a natural language processing parsing engine to perform entity extraction and semantic parsing on the detected document text, extracting key indicators of the verification procedure and historical detection parameters, and obtaining a text semantic dataset specifically includes: Acquire a dictionary of specialized terminology and a named entity recognition model for a specific metrology field; The natural language processing parsing engine parses the unstructured paragraphs in the detected document text and extracts structured triplet information including at least the verification items, measurement upper limit, allowable error and equipment number; A time-series linked knowledge base is constructed based on the structured triple information and the detection parameters in historical data to obtain the text semantic dataset.

[0011] In this solution, the step of inputting the image recognition dataset, the environmental parameter sensor data, and the text semantic dataset into a multi-source data fusion and anomaly detection model to calculate a spatiotemporal consistency metric to identify multimodal data anomaly features specifically includes: The image recognition dataset, environmental parameter sensor data, and text semantic dataset are preprocessed, wherein the preprocessing mechanism includes at least filtering and noise reduction, missing value imputation, and standardization and alignment of multi-source heterogeneous data. Using preset statistical principles, the environmental parameter sensor data and the indicated data are initially screened for extreme value anomalies, and the first type of suspicious anomalies are marked. The aligned multi-source heterogeneous data is input into a deep learning neural network based on an autoencoder, and the reconstruction error and spatiotemporal correlation are calculated to obtain the spatiotemporal consistency metric. The multimodal data anomaly features are comprehensively identified based on the spatiotemporal consistency metric, wherein the multimodal data anomaly features include at least one of image reading jumps, human transcription errors, and environmental parameter mutations.

[0012] In this solution, the step of inputting the multimodal data anomaly features into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and generating an automatic early warning prompt when an abnormal threshold is triggered, specifically includes: The multimodal data anomaly features and historical error correction samples are input into the deep learning error correction model to complete the compensation calculation for indication error and environmental drift. Obtain the standard metrological testing and verification results generated by the deep learning error correction model, and bind and package the verification results, the test object ID, the operator information and the timestamp on the blockchain; When the abnormal characteristics exceed the preset safety threshold, an alarm signal is automatically sent to trigger a manual review process or restart the process.

[0013] A second aspect of the present invention also provides a metrological testing data verification system based on the fusion of OCR and NLP, comprising a memory and a processor. The memory includes a metrological testing data verification method program based on the fusion of OCR and NLP. When the processor executes the metrological testing data verification method program based on the fusion of OCR and NLP, it performs the following steps: Acquire raw image data and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process. The raw image data includes at least the metrological equipment identification mark and real-time reading. The original image data is subjected to target detection and character localization using a preset recognition model. Key information from the equipment nameplate and the readings of pointer / digital displays are extracted to obtain an image recognition dataset. The detection document text is subjected to entity extraction and semantic parsing using a natural language processing parsing engine. Key indicators of the verification procedure and historical detection parameters are extracted to obtain a text semantic dataset. The image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features. The abnormal features of the multimodal data are input into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered.

[0014] A third aspect of the present invention provides a computer-readable storage medium comprising a machine program for a measurement and testing data verification method based on the fusion of OCR and NLP. When the measurement and testing data verification method program based on the fusion of OCR and NLP is executed by a processor, it implements the steps of the measurement and testing data verification method based on the fusion of OCR and NLP as described in any of the preceding claims.

[0015] The fourth aspect of the present invention provides a computer program product, which includes computer program code. When the computer program code is run on a computer, the computer implements the steps of a metrological testing data verification method based on the fusion of OCR and NLP as described in any of the preceding claims.

[0016] The fifth aspect of the present invention provides an electronic device, the electronic device comprising: a processor and a memory; wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the steps of a metrological testing data verification method based on the fusion of OCR and NLP as described in any of the preceding claims.

[0017] The present invention discloses a method and system for verifying metrological testing data based on the fusion of OCR and NLP, which has the following beneficial effects: 1. By improving YOLOv8 and high-precision OCR technology, automated acquisition of equipment marking and indication data under complex working conditions (such as rusted steel stamps and laser engraving) has been achieved. The acquisition time for a single acquisition has been reduced to the second level, which has shortened the detection time of a single device and broken the efficiency bottleneck of traditional manual data entry.

[0018] 2. By combining NLP text parsing with sensor environmental data and employing the 3σ principle and an autoencoder deep learning model, a spatiotemporal consistency verification system for multi-source data was constructed. This system can automatically identify reading jumps, mistranscriptions, and environmental anomalies, reducing the human error rate and ensuring the authenticity and reliability of the detection data.

[0019] 3. By using a deep learning error correction model, automatic correction of environmental drift and indication error is achieved. At the same time, blockchain technology is introduced to package and bind bit detection ID and timestamp, reducing the traceability time from several hours to seconds. Attached Figure Description

[0020] Figure 1 The diagram illustrates the steps of a metrological testing data verification method based on the fusion of OCR and NLP according to the present invention. Figure 2 The flowchart illustrates a scenario application of the metrological testing data verification method based on the fusion of OCR and NLP according to the present invention. Figure 3 The diagram illustrates the detection process of a metrological testing data verification method based on the fusion of OCR and NLP according to the present invention. Figure 4 The diagram illustrates the abnormal feature identification process of a metrological testing data verification method based on the fusion of OCR and NLP according to the present invention. Figure 5 A block diagram of a metrology and testing data verification system based on the fusion of OCR and NLP is shown. Detailed Implementation

[0021] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments of the present invention and the features thereof can be combined with each other.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0023] Currently, the implementation schemes most similar to this invention are mainly the following: Scheme 1, Scheme 2, or Scheme 3.

[0024] Specifically, Option 1: Employing a traditional method of purely manual data entry combined with fixed-rule verification. This method requires inspectors to manually read the instrument data and type it into the system, which then performs only a simple verification using hard-coded threshold ranges. The disadvantages are: inability to identify human error in data entry, inability to handle interference from complex environments, and extremely low efficiency.

[0025] Specifically, Option 2: Employing a single optical character recognition (OCR) data acquisition scheme. This scheme directly performs OCR recognition on instrument images, obtains the displayed text, and stores it directly in the database. The drawbacks are: extremely low recognition rate for complex markings such as stamps and laser engravings, and a lack of correlation analysis between environmental parameters and procedural texts, making it unable to identify abnormal readings caused by drastic environmental changes.

[0026] Specifically, Option 3: Employing an independent LIMS / ERP system interface solution. This solution relies on standard communication interfaces provided by various equipment manufacturers for data transmission. The disadvantages are: many older instruments in the field lack communication interfaces, resulting in extremely high system upgrade costs; and severe data asynchrony issues across systems prevent full-process dynamic traceability.

[0027] Through in-depth analysis of the above-mentioned existing technical solutions, several key drawbacks of the existing technology can be clearly summarized: 1. The data acquisition mode is limited and has poor adaptability to complex field environments, resulting in low recognition rate and susceptibility to interference; 2. The lack of a correlation and verification mechanism for multi-source heterogeneous data (images, text, sensors) makes it impossible to accurately distinguish between value jumps and human transcription errors; 3. Insufficient automation and intelligence, making it impossible to achieve non-intrusive cross-system closed-loop interaction and full lifecycle blockchain traceability.

[0028] To address these shortcomings, the present invention aims to provide a method and system for verifying metrological testing data based on the fusion of OCR (Optical Character Recognition) and NLP (Natural Language Processing). The present invention aims to achieve the following objectives: (1) To achieve intelligent automatic acquisition and high-precision recognition of multimodal (image, text, sensor) data; (2) Realize spatiotemporal consistency measurement and automatic anomaly identification based on the 3σ principle and autoencoder model; (3) Achieve automated error correction, blockchain full-process anti-tampering traceability and non-intrusive RPA cross-system closed-loop interaction.

[0029] Therefore, the core innovation and key protection points are: 1. An improved method for high-precision automatic extraction of complex symbols and instrument readings by integrating YOLOv8 with OCR and incorporating an attention mechanism; 2. A model for detecting spatiotemporal consistency anomalies and dynamically correcting errors in multi-source heterogeneous data by combining statistical 3σ with deep learning autoencoders; 3. An unmanned verification system for the entire metrology and testing process based on NLP semantic extraction, blockchain packaging and on-chaining, and RPA non-intrusive interaction.

[0030] Specifically, Figure 1 The diagram illustrates the steps of a metrological testing data verification method based on the fusion of OCR and NLP according to the present invention.

[0031] like Figure 1 As shown, this invention discloses a method for verifying metrological testing data based on the fusion of OCR and NLP, comprising the following steps: S102, acquire raw image data, and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process, wherein the raw image data includes at least the metrological equipment identification mark and real-time reading; S104, using a preset recognition model to perform target detection and character localization on the original image data, extracting key information from the device nameplate and the reading data from the pointer / digital display instrument to obtain an image recognition dataset; S106, Entity extraction and semantic parsing of the detected document text are performed using a natural language processing parsing engine to extract key indicators of the verification procedure and historical detection parameters, and a text semantic dataset is obtained. S108, the image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features; S110, the abnormal features of the multimodal data are input into the deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered to an abnormality.

[0032] It should be noted that, in this embodiment, as Figure 2 As shown, this invention is applied to intelligent verification scenarios for metrological testing data and can be widely used in statutory metrological technical institutions, third-party metrological testing institutions, and metrological laboratories of industrial enterprises. During actual testing, testing personnel use a mobile data acquisition terminal to photograph and collect data from the nameplate, stamped markings, and instrument readings of the metrological equipment. Simultaneously, they acquire environmental parameters such as on-site temperature, humidity, and vibration, and import text information such as testing authorization forms, verification procedures, and historical testing records.

[0033] Specifically, in this embodiment, an OCR recognition model is first used to automatically extract equipment identification and indication information. Then, a natural language processing engine is used to parse the verification procedures and historical test data to form structured text semantic information. Subsequently, multi-source heterogeneous data fusion is performed by combining environmental sensor data, and spatiotemporal consistency analysis is completed using statistical rules and deep learning models to automatically identify and intelligently verify abnormal readings, human transcription errors, and environmental anomalies. Finally, a standard metrological test result is generated through an error correction model, and an early warning is automatically triggered when an anomaly occurs. At the same time, the test result, equipment number, operator information, and timestamp are written to the blockchain to achieve reliable data storage and full lifecycle traceability throughout the entire testing process. By adopting this invention, the workload of manual data entry can be significantly reduced, the level of automation in metrological testing can be improved, the accuracy of test results can be increased, and the security, integrity, and traceability of test data can be enhanced.

[0034] According to embodiments of the present invention, such as Figure 3 As shown, the acquisition of raw image data, and the simultaneous collection of environmental parameter sensor data and detection document text during the metrological testing process, specifically includes: The raw image data, environmental parameter sensor data, and detection document text are acquired in parallel through distributed mobile acquisition terminals or sensor network interfaces. The environmental parameter sensor data includes temperature, humidity, and vibration parameters; The testing documentation includes the testing order, verification procedures, and historical testing record sheets; and When the network is offline, all types of data are encrypted and cached offline, and the cached data is automatically uploaded to the cloud server for two-way synchronization after the network is restored.

[0035] It should be noted that, in this embodiment, the distributed mobile acquisition terminal can be a smart mobile terminal, an industrial tablet terminal, a handheld metering and testing terminal, or other portable devices with image acquisition capabilities; the sensor network interface can adopt a serial communication interface, Bluetooth communication interface, wireless LAN interface, industrial Ethernet interface, or IoT communication interface, used to establish data communication connections with temperature sensors, humidity sensors, and vibration sensors. All types of acquisition devices employ a unified timestamp mechanism to synchronously mark the original image data, environmental parameter sensor data, and detection document text, ensuring consistency of data from different sources in the time dimension and providing a unified data foundation for subsequent multi-source data fusion analysis.

[0036] Specifically, in this embodiment, the testing personnel use a distributed mobile acquisition terminal to capture images of the equipment nameplate, stamped markings, and instrument readings of the metrology equipment. Simultaneously, they acquire environmental parameters such as temperature, humidity, and vibration at the testing site via a sensor network interface, and import testing documents such as testing orders, verification procedures, and historical testing records. The acquired raw image data, environmental parameter sensor data, and testing document texts are then assigned unified data identifiers and timestamps, and packaged according to a preset data format to form a dataset to be processed. When the network connection is normal, all types of data are uploaded to the cloud server in real time. When the testing site is offline or network communication is abnormal, the mobile acquisition terminal uses an encrypted offline caching mechanism to store the acquired data in a local secure storage unit and record the cache index and synchronization status. After the network is restored, it automatically performs breakpoint resume and bidirectional synchronization operations based on the cache index, uploading the offline cached data to the cloud server and simultaneously acquiring updated data from the server, achieving data consistency between the terminal and the cloud, and ensuring the integrity, reliability, and continuity of data throughout the metrology testing process.

[0037] According to an embodiment of the present invention, the step of using a preset recognition model to perform target detection and character localization on the original image data, extracting key information from the equipment nameplate and the reading data of the pointer / digital display instrument to obtain an image recognition dataset specifically includes: The recognition model includes an improved YOLOv8 recognition model, wherein an attention mechanism is introduced into the feature extraction network of the recognition model; The improved YOLOv8 recognition model is used to perform localization detection on the original image data, wherein the localization area includes at least the equipment nameplate, laser engraving, stamped markings, and instrument pointers / digital displays; The positioning area is segmented and recognized by an optical character recognition algorithm to extract key information of the device nameplate and the indicated value data in real time. The image recognition dataset is obtained by verifying the range rationality and format based on a preset metrology standard database.

[0038] It should be noted that, in this embodiment, the improved YOLOv8 recognition model introduces an attention mechanism into the feature extraction network of the YOLOv8 target detection network to enhance the model's ability to represent the features of key targets such as equipment nameplates, laser engravings, stamped markings, and instrument reading areas, thereby improving target localization accuracy under complex backgrounds, low light, oblique shooting, and partial occlusion conditions. The optical character recognition algorithm can employ a deep learning-based OCR recognition model to perform character segmentation, character recognition, and character sequence reconstruction on the detected target area, and output corresponding structured recognition results, providing accurate image recognition data for subsequent data fusion and anomaly detection.

[0039] Specifically, in this embodiment, after receiving the original image data, the image is first preprocessed, including image denoising, brightness equalization, distortion correction, image enhancement, and size normalization, to improve image quality. Subsequently, the preprocessed image is input into the improved YOLOv8 recognition model to automatically locate target areas such as equipment nameplates, laser engravings, stamped markings, and instrument pointers or digital displays, and outputs the position coordinates and category information of the corresponding target areas. For the located target areas, further image cropping, perspective correction, and character region segmentation are performed, and optical character recognition algorithms are used to identify key information such as equipment model, equipment number, manufacturing number, measurement range parameters, and instrument readings.

[0040] Furthermore, in this embodiment, for pointer-type instruments, the corresponding indicated value can be calculated by combining the pointer positioning result; for digital display instruments, the digital display content is directly identified. After identification, the system performs consistency and rationality checks on the equipment model, measuring range, number format, and indicated value range based on a preset metrology standard database. When the identification result meets the preset rules, the identified key information of the equipment nameplate and the indicated value data are packaged into an image recognition dataset. When there are missing characters, abnormal formats, or indicated values ​​exceeding the preset range, the corresponding identification result is marked as abnormal and output to the subsequent data fusion and intelligent verification module for further analysis.

[0041] According to an embodiment of the present invention, the step of using a natural language processing parsing engine to perform entity extraction and semantic parsing on the detected document text, extracting key indicators of the verification procedure and historical detection parameters, and obtaining a text semantic dataset specifically includes: Acquire a dictionary of specialized terminology and a named entity recognition model for a specific metrology field; The natural language processing parsing engine parses the unstructured paragraphs in the detected document text and extracts structured triplet information including at least the verification items, measurement upper limit, allowable error and equipment number; A time-series linked knowledge base is constructed based on the structured triple information and the detection parameters in historical data to obtain the text semantic dataset.

[0042] It should be noted that, in this embodiment, the natural language processing parsing engine is built based on a professional terminology dictionary and a named entity recognition model for a specific metrology field. The professional terminology dictionary includes professional terms in fields such as instrument names, verification items, measurement range specifications, accuracy grades, permissible errors, verification cycles, and units of measurement. The named entity recognition model is used to identify key entities in the test document text, such as equipment names, equipment numbers, verification items, measurement parameters, and test results. It also combines contextual semantic relationships to complete entity disambiguation, relation extraction, and semantic association analysis, thereby improving the parsing accuracy of unstructured test documents and providing unified textual semantic information for subsequent multi-source data fusion.

[0043] Specifically, in this embodiment, after obtaining the text of testing documents such as testing request forms, verification procedures, and historical testing record sheets, the text content is first preprocessed by word segmentation, removal of invalid characters, text normalization, and syntactic analysis. Then, a professional terminology dictionary is called to match metrological terms in the text, and a named entity recognition model is used to identify key entity information such as verification items, measurement limits, permissible errors, equipment numbers, equipment models, verification dates, and testing personnel. After entity recognition, a semantic relation extraction algorithm is further combined to establish the relationships between entities and generate structured triple information including subject, relation, and object. For example, the equipment number is used as the subject, the verification item as the relation, and the permissible error or measurement limit as the object to form a corresponding structured semantic representation. Subsequently, the structured triple information is associated and fused with the testing parameters, testing time, and testing results in the historical testing records to establish a corresponding temporal linkage knowledge base in chronological order, forming a text semantic dataset. This provides semantic support and historical data basis for subsequent anomaly detection, error correction, and intelligent verification.

[0044] According to embodiments of the present invention, such as Figure 4 As shown, the step of inputting the image recognition dataset, the environmental parameter sensor data, and the text semantic dataset into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features specifically includes: The image recognition dataset, environmental parameter sensor data, and text semantic dataset are preprocessed, wherein the preprocessing mechanism includes at least filtering and noise reduction, missing value imputation, and standardization and alignment of multi-source heterogeneous data. Using preset statistical principles, the environmental parameter sensor data and the indicated data are initially screened for extreme value anomalies, and the first type of suspicious anomalies are marked. The aligned multi-source heterogeneous data is input into a deep learning neural network based on an autoencoder, and the reconstruction error and spatiotemporal correlation are calculated to obtain the spatiotemporal consistency metric. The multimodal data anomaly features are comprehensively identified based on the spatiotemporal consistency metric, wherein the multimodal data anomaly features include at least one of image reading jumps, human transcription errors, and environmental parameter mutations.

[0045] It should be noted that, in this embodiment, the multi-source data fusion and anomaly detection model is used to perform unified fusion analysis on image recognition datasets, environmental parameter sensor data, and text semantic datasets. The multi-source heterogeneous data includes structured data, semi-structured data, and unstructured data. Since data from different sources differ in acquisition frequency, data format, data precision, and time scale, unified data preprocessing operations need to be performed on all types of data before fusion analysis. These operations include filtering and noise reduction, missing value imputation, outlier removal, data normalization, and timestamp alignment to eliminate data bias between different data sources and improve the consistency and reliability of multi-source data fusion. The deep learning neural network based on an autoencoder is used to learn the latent feature distribution of normal detection data. By comparing the error between the input data and the reconstructed data, and combining the temporal and spatial correlations between different data sources, a spatiotemporal consistency metric reflecting the consistency of multi-source data is calculated, providing a quantitative basis for anomaly detection.

[0046] Specifically, in this embodiment, the device nameplate information, device number, and indication data in the image recognition dataset, the temperature, humidity, and vibration parameters in the environmental parameter sensor data, and the verification items, allowable errors, and historical detection parameters in the text semantic dataset are first subjected to data cleaning, filtering and noise reduction, missing value imputation, and standardization. Data alignment between different data sources is then completed based on a unified timestamp, forming a unified multi-source heterogeneous dataset. Subsequently, based on preset statistical principles, initial extreme value anomaly screening is performed on the environmental parameter sensor data and indication data. Data exceeding a preset statistical threshold are marked as first-type suspicious anomalies and retained for subsequent in-depth analysis.

[0047] Furthermore, in this embodiment, the standardized and aligned multi-source heterogeneous data is then input into a deep learning neural network based on an autoencoder. The encoder extracts the latent feature representations of the multi-source data, and the decoder reconstructs the input data. Based on the reconstruction error between the input data and the reconstructed data, combined with the correlation between historical detection data, environmental change trends, and detection time series, a corresponding spatiotemporal consistency metric is calculated. When the spatiotemporal consistency metric is lower than a preset consistency threshold or the reconstruction error exceeds a preset anomaly threshold, the corresponding detection data is determined to be abnormal. Further, a comprehensive analysis is performed combining image recognition results, text semantic information, and changes in environmental parameters to identify multimodal data anomaly features such as image reading jumps, human transcription errors, and sudden changes in environmental parameters. Corresponding anomaly feature data is generated as input to subsequent error correction and intelligent early warning modules, achieving intelligent anomaly identification throughout the entire metrological detection process.

[0048] According to an embodiment of the present invention, the step of inputting the multimodal data anomaly features into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and generating an automatic early warning prompt when an abnormal threshold is triggered, specifically includes: The multimodal data anomaly features and historical error correction samples are input into the deep learning error correction model to complete the compensation calculation for indication error and environmental drift. Obtain the standard metrological testing and verification results generated by the deep learning error correction model, and bind and package the verification results, the test object ID, the operator information and the timestamp on the blockchain; When the abnormal characteristics exceed the preset safety threshold, an alarm signal is automatically sent to trigger a manual review process or restart the process.

[0049] It should be noted that, in this embodiment, the deep learning error correction model is used to learn the error variation patterns and environmental drift characteristics in historical metrological testing data. By jointly analyzing the multimodal data anomaly features identified during the current testing process with historical error correction samples, it achieves indication error compensation and environmental drift compensation, thereby improving the accuracy and stability of metrological testing results. The blockchain adopts a consortium blockchain or permissioned blockchain architecture. After hashing key data such as standard metrological testing verification results, testing object ID, operator information, and timestamps, it generates corresponding data digests and writes them into the blockchain ledger, realizing the immutability, traceability, and reliable evidence storage of testing data. When the detected anomaly level exceeds a preset security threshold, an anomaly alarm mechanism is automatically activated, and the anomaly information is pushed to the testing terminal or management platform to ensure the reliability and security of the testing results.

[0050] Specifically, in this embodiment, the multimodal data anomaly features identified in the previous stage, such as image reading jumps, human transcription errors, and sudden changes in environmental parameters, are input together with historical error correction samples into a deep learning error correction model. The model learns the distribution characteristics of historical errors and the influence of environmental changes on measurement values, calculates the indication compensation amount and environmental drift compensation amount corresponding to the current detection data, and outputs the corrected standard metrological testing verification result. Subsequently, the standard metrological testing verification result is acquired, associated and encapsulated with the detection object ID, operator identity information, detection timestamp, and detection equipment identifier. After generating a corresponding data digest through hash operation, it is packaged into block data and uploaded to the blockchain network for on-chain notarization, ensuring the authenticity, integrity, and traceability of the detection data during storage and circulation.

[0051] Furthermore, in this embodiment, simultaneously, the corresponding abnormal risk level is calculated based on the abnormal characteristics of multimodal data. When the abnormal risk level exceeds a preset safety threshold, an alarm message is automatically generated and sent through the detection terminal, management platform, or message push module to trigger the manual review process or re-execute the detection process. When the abnormal risk level does not exceed the preset safety threshold, the standard metrological testing verification result is output and the current metrological testing task is completed.

[0052] According to an embodiment of the present invention, before fusing the image recognition dataset, environmental parameter sensor data, and text semantic dataset, the method further includes: Obtain data quality evaluation indicators for each modality of data, wherein the data quality evaluation indicators include at least image sharpness, character recognition confidence, text parsing confidence, sensor stability, and data integrity; Calculate the corresponding modal reliability coefficient based on the data quality evaluation index; The fusion weights of each modality in the multi-source fusion process are dynamically adjusted based on the aforementioned credibility coefficient. Multimodal data fusion is completed based on the adjusted fusion weights.

[0053] It should be noted that in this embodiment, the same weights are not always used for each data source in the fusion calculation. Instead, the reliability coefficient is dynamically calculated based on the quality of each modality of data. When the data quality of a certain data source deteriorates due to the influence of the on-site environment, its weight in the fusion calculation is automatically reduced, while the weights of other reliable data sources are increased accordingly, thereby ensuring the stability and robustness of the fusion result.

[0054] Specifically, in this embodiment, the image sharpness evaluation value, average confidence score of OCR character recognition, confidence score of text entity recognition, environmental sensor stability coefficient, and data integrity rate are calculated respectively. These evaluation indicators are then input into the credibility assessment module to generate corresponding credibility coefficients. Subsequently, the contribution ratios of image data, text data, and environmental data in the fusion model are dynamically allocated based on the credibility coefficients. When blurring, reflection, or occlusion is detected in the image, the weight of image data is automatically reduced, while the fusion ratio of text semantics and environmental parameters is increased. When abnormal drift occurs in the environmental sensor, the weight of environmental parameters is reduced, thereby ensuring that the fusion result has high reliability.

[0055] According to an embodiment of the present invention, the method further includes: Establish a detection trend prediction model based on historical detection records of the same detection object; Calculate the degree of deviation of the current detection result based on the prediction result; The degree of deviation is used as the input for anomaly detection.

[0056] It should be noted that, in this embodiment, a historical testing file is established for each testing object, and a continuous trend is formed based on the results of each test to reflect the long-term performance change pattern of the measuring equipment.

[0057] Specifically, in this embodiment, historical detection records are retrieved according to the detection object ID, a sequence of indicated value changes is established based on the detection time order, and the reasonable range of the current detection result is predicted in conjunction with the trend of environmental parameter changes. When the current detection indicated value exceeds the predicted range, the prediction deviation is calculated, and the prediction deviation is used as a new abnormal feature input into the multi-source anomaly detection model to achieve early identification of abnormal situations caused by long-term aging of equipment, performance degradation, or sudden failure, thereby improving the accuracy of anomaly detection.

[0058] According to an embodiment of the present invention, after generating detection results and outputting alarm information based on the anomaly features of the multimodal data, the method further includes: Acquire the multimodal data anomaly features, image recognition dataset, text semantic dataset, and environmental parameter sensor data; Based on a pre-built anomaly association rule base, the anomaly features of the multimodal data are correlated and inferred with each data source to determine the corresponding anomaly causes; Based on the cause of the anomaly, corresponding anomaly explanation information is generated, wherein the anomaly explanation information includes at least one of the following: anomaly source, influencing factors, and handling suggestions; The anomaly explanation information is associated with the standard metrological testing and verification results and stored, and simultaneously uploaded to the blockchain for trusted evidence preservation.

[0059] It should be noted that, in this embodiment, the anomaly association rule base is used to establish the correlation between image recognition results, text semantic information, environmental parameters, and anomaly detection results. This anomaly association rule base can be constructed based on metrological testing standards, historical testing experience, and typical anomaly cases. After completing multimodal anomaly detection, it not only outputs anomaly judgment results but also further combines the correlation between multi-source data to automatically analyze the causes of anomalies, generating anomaly explanation information that reflects the source and influencing factors of the anomalies, thereby improving the understandability of the detection results and the efficiency of manual review.

[0060] Specifically, in this embodiment, image recognition datasets, text semantic datasets, environmental parameter sensor data, and multimodal data anomaly features are acquired, and all detection information in the corresponding detection process is retrieved based on the detection object ID. Subsequently, image recognition confidence, character recognition results, text entity information, ambient temperature, ambient humidity, environmental vibration parameters, and anomaly detection results are input into an anomaly association rule base for association analysis, and the source of the anomaly is determined based on the association relationships between various anomaly features.

[0061] For example, when the image recognition confidence level is lower than a preset threshold and the ambient humidity is consistently higher than a preset humidity threshold, the cause of the anomaly is determined to be that the high humidity at the detection site leads to a decrease in the quality of the acquired image; when the device number is inconsistent with the device number in the historical detection record, the system determines that the cause of the anomaly is that the detection object was entered incorrectly or manually; when the ambient vibration amplitude exceeds a preset vibration threshold and the instrument reading fluctuates continuously, the cause of the anomaly is determined to be that external mechanical vibration affects the detection stability; when there is a significant conflict between the allowable error in the text semantic parsing result and the image recognition reading, the cause of the anomaly is determined to be that the detection procedure was referenced incorrectly or the detection parameters were configured abnormally.

[0062] Furthermore, in this embodiment, anomaly explanation information is automatically generated based on the determined cause of the anomaly. The anomaly explanation information includes at least the source of the anomaly, influencing factors, anomaly level, and corresponding handling suggestions. The anomaly explanation information and the standard metrological testing and verification results are written into the test report. At the same time, it is uploaded to the blockchain for trusted storage after being associated with the test object ID, operator information, and timestamp, so as to facilitate subsequent test traceability, manual review, and anomaly analysis.

[0063] According to an embodiment of the present invention, the method further includes: Obtain information on the equipment type, specifications, and testing task of the object to be tested; Match the corresponding detection strategy template based on the category to which the detection object belongs; The detection parameter configuration for this metrological testing task is automatically generated based on the detection strategy template. Complete subsequent image acquisition, text parsing, and anomaly detection according to the detection parameters configured.

[0064] It should be noted that, in this embodiment, a detection strategy template library is pre-established, with different detection processes and parameters corresponding to different types of measuring instruments. The detection strategy template includes at least the target detection area, key identification objects, environmental parameter collection items, anomaly judgment rules, and error correction parameters. When the object being detected changes, the corresponding template can be automatically matched, eliminating the need for manual reconfiguration of the detection process and improving the automation level and efficiency of the detection process for different measuring instruments.

[0065] Specifically, in this embodiment, based on the device model and category obtained from OCR recognition, combined with the testing task information in the testing order, the type of measuring instrument to which the current testing object belongs is automatically determined, and the corresponding template is retrieved from the testing strategy template library. For example, when the testing object is a pressure gauge, the stamp recognition, range recognition, and allowable error verification process are automatically configured; when the testing object is an electronic balance, zero-point stability analysis and environmental vibration monitoring are automatically added; when the testing object is an electronic scale, load indication consistency analysis and temperature and humidity compensation parameters are added. Subsequently, based on the generated testing strategy, the image acquisition parameters, OCR recognition area, environmental sensor sampling frequency, and anomaly judgment threshold are dynamically adjusted to achieve adaptive configuration of the testing process for different measuring devices.

[0066] According to an embodiment of the present invention, the method further includes: The detection reliability score is calculated based on the image recognition results, text parsing results, environmental parameters, and anomaly detection results. The reliability level of the detection results is determined based on the aforementioned detection reliability score; The system will automatically determine whether to proceed to the manual review process based on the stated trust level.

[0067] It should be noted that, in this embodiment, the detection credibility score is used to comprehensively evaluate the credibility of the measurement and testing results. The credibility score comprehensively considers factors such as image recognition quality, text parsing accuracy, environmental stability, consistency of multi-source data, and abnormal detection results to form a unified credibility evaluation index, providing a basis for automatic review and manual verification of the test results.

[0068] Specifically, in this embodiment, the average confidence score of OCR character recognition, the confidence score of text entity recognition, the stability of environmental parameters, the spatiotemporal consistency metric, and the residual error after error correction are obtained and normalized respectively. Then, a comprehensive credibility score is calculated based on a preset evaluation model, and credibility levels are divided according to the score range. When the comprehensive credibility score is higher than a first preset threshold, the standard metrological verification result is directly output; when the comprehensive credibility score is between the first and second preset thresholds, a manual review prompt is generated; when the comprehensive credibility score is lower than the second preset threshold, the image acquisition, anomaly detection, or error correction process is automatically retried to ensure the reliability of the final detection result. Finally, the credibility score and the standard metrological verification result are jointly written into the detection report and simultaneously uploaded to the blockchain for trusted storage.

[0069] Figure 5 A block diagram of a metrology and testing data verification system based on the fusion of OCR and NLP is shown.

[0070] like Figure 5 As shown, this invention discloses a metrological testing data verification system based on the fusion of OCR and NLP, including a memory and a processor. The memory includes a metrological testing data verification method program based on the fusion of OCR and NLP. When the processor executes the metrological testing data verification method program based on the fusion of OCR and NLP, it performs the following steps: Acquire raw image data and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process. The raw image data includes at least the metrological equipment identification mark and real-time reading. The original image data is subjected to target detection and character localization using a preset recognition model. Key information from the equipment nameplate and the readings of pointer / digital displays are extracted to obtain an image recognition dataset. The detection document text is subjected to entity extraction and semantic parsing using a natural language processing parsing engine. Key indicators of the verification procedure and historical detection parameters are extracted to obtain a text semantic dataset. The image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features. The abnormal features of the multimodal data are input into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered.

[0071] It should be noted that when the metrological testing data verification system based on the fusion of OCR and NLP disclosed in this invention is applied, the specific process corresponds to the metrological testing data verification method based on the fusion of OCR and NLP described in the above embodiments. Since the specific implementation details of the system application are consistent with the content of the metrological testing data verification method based on the fusion of OCR and NLP described above, no further details will be provided in this embodiment.

[0072] A third aspect of the present invention provides a computer-readable storage medium comprising a program for a measurement and testing data verification method based on the fusion of OCR and NLP. When the program for the measurement and testing data verification method based on the fusion of OCR and NLP is executed by a processor, it implements the steps of the measurement and testing data verification method based on the fusion of OCR and NLP as described in any of the preceding claims.

[0073] The fourth aspect of the present invention provides a computer program product comprising: computer program code, which, when run on a computer, causes the computer to execute any of the methods described in the embodiments of the metrological testing data verification method based on the fusion of OCR and NLP.

[0074] The fifth aspect of the present invention provides an electronic device, the electronic device comprising: a processor and a memory; wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory, so that the electronic device performs the steps of a metrological testing data verification method based on the fusion of OCR and NLP as described in any of the preceding claims.

[0075] The terms “component,” “module,” “system,” etc., used in this specification are used to refer to computer-related entities, hardware, firmware, combinations of hardware and software, software, or software in execution. For example, a component can be, but is not limited to, a process running on a processor, a processor, an object, an executable file, an execution thread, a program, and / or a computer. As illustrated, applications running on computing devices and computing devices can both be components. One or more components may reside in a process and / or an execution thread, and components may be located on a single computer and / or distributed among two or more computers. Furthermore, these components can be executed from various computer-readable media on which various data structures are stored. Components can communicate, for example, via local and / or remote processes based on signals having one or more data packets (e.g., data from two components interacting with another component between a local system, a distributed system, and / or a network, such as the Internet interacting with other systems via signals).

[0076] Those skilled in the art will recognize that the various illustrative logical blocks and steps 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.

[0077] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0078] In the embodiments provided by this invention, it should be understood that the disclosed systems, 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 system, 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 apparatuses or units may be electrical, mechanical, or other forms.

[0079] 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.

[0080] 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.

[0081] In the above embodiments, the functions of each functional unit can be implemented entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. A computer program product includes one or more computer instructions (programs). When the computer program instructions (programs) are loaded and 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., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (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 integrates one or more available media. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., high-density digital video discs (DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0082] The present invention discloses a measurement and testing data verification method and system based on the integration of OCR and NLP, which can significantly shorten the verification cycle, reduce human input errors, break down the barriers of multi-source heterogeneous data, and achieve the beneficial effects of fully unmanned intelligent management and control based on blockchain and RPA.

[0083] The above embodiments are merely illustrative of the principles and effects of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of the present invention. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in the present invention should still be covered by the claims of the present invention.

Claims

1. A method for verifying metrological testing data based on the fusion of OCR and NLP, characterized in that, Includes the following steps: Acquire raw image data and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process. The raw image data includes at least the metrological equipment identification mark and real-time reading. The original image data is subjected to target detection and character localization using a preset recognition model. Key information from the equipment nameplate and the readings of pointer / digital displays are extracted to obtain an image recognition dataset. The detection document text is subjected to entity extraction and semantic parsing using a natural language processing parsing engine. Key indicators of the verification procedure and historical detection parameters are extracted to obtain a text semantic dataset. The image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features. The abnormal features of the multimodal data are input into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered.

2. The measurement and testing data verification method based on the fusion of OCR and NLP according to claim 1, characterized in that, The acquisition of raw image data, and the simultaneous collection of environmental parameter sensor data and detection document text during the metrological testing process, specifically includes: The raw image data, environmental parameter sensor data, and detection document text are acquired in parallel through distributed mobile acquisition terminals or sensor network interfaces. The environmental parameter sensor data includes temperature, humidity, and vibration parameters; The testing documentation includes the testing order, verification procedures, and historical testing record sheet; and When the network is offline, all types of data are encrypted and cached offline, and the cached data is automatically uploaded to the cloud server for two-way synchronization after the network is restored.

3. The measurement and testing data verification method based on the fusion of OCR and NLP according to claim 2, characterized in that, The process of using a preset recognition model to perform target detection and character localization on the original image data, extracting key information from the equipment nameplate and the readings of pointer / digital displays, and obtaining an image recognition dataset specifically includes: The recognition model includes an improved YOLOv8 recognition model, wherein an attention mechanism is introduced into the feature extraction network of the recognition model; The improved YOLOv8 recognition model is used to perform localization detection on the original image data, wherein the localization area includes at least the equipment nameplate, laser engraving, stamped markings, and instrument pointers / digital displays; The positioning area is segmented and recognized by an optical character recognition algorithm to extract key information of the device nameplate and the indicated value data in real time. The image recognition dataset is obtained by verifying the range rationality and format based on a preset metrology standard database.

4. The measurement and testing data verification method based on the fusion of OCR and NLP according to claim 3, characterized in that, The process involves using a natural language processing parsing engine to perform entity extraction and semantic parsing on the detected document text, extracting key indicators of the verification procedure and historical detection parameters to obtain a text semantic dataset, specifically including: Acquire a dictionary of specialized terminology and a named entity recognition model for a specific metrology field; The natural language processing parsing engine parses the unstructured paragraphs in the detected document text and extracts structured triplet information including at least the verification items, measurement upper limit, allowable error and equipment number; A time-series linked knowledge base is constructed based on the structured triple information and the detection parameters in historical data to obtain the text semantic dataset.

5. The measurement and testing data verification method based on the fusion of OCR and NLP according to claim 4, characterized in that, The step of inputting the image recognition dataset, the environmental parameter sensor data, and the text semantic dataset into a multi-source data fusion and anomaly detection model to calculate a spatiotemporal consistency metric to identify multimodal data anomaly features specifically includes: The image recognition dataset, environmental parameter sensor data, and text semantic dataset are preprocessed, wherein the preprocessing mechanism includes at least filtering and noise reduction, missing value imputation, and standardization and alignment of multi-source heterogeneous data. Using preset statistical principles, the environmental parameter sensor data and the indicated data are initially screened for extreme value anomalies, and the first type of suspicious anomalies are marked. The aligned multi-source heterogeneous data is input into a deep learning neural network based on an autoencoder, and the reconstruction error and spatiotemporal correlation are calculated to obtain the spatiotemporal consistency metric. The multimodal data anomaly features are comprehensively identified based on the spatiotemporal consistency metric, wherein the multimodal data anomaly features include at least one of image reading jumps, human transcription errors, and environmental parameter mutations.

6. The measurement and testing data verification method based on the fusion of OCR and NLP according to claim 5, characterized in that, The step of inputting the multimodal data anomaly features into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and generating an automatic early warning prompt when an abnormal threshold is triggered, specifically includes: The multimodal data anomaly features and historical error correction samples are input into the deep learning error correction model to complete the compensation calculation for indication error and environmental drift. Obtain the standard metrological testing and verification results generated by the deep learning error correction model, and bind and package the verification results, the test object ID, the operator information and the timestamp on the blockchain; When the abnormal characteristics exceed the preset safety threshold, an alarm signal is automatically sent to trigger a manual review process or restart the process.

7. A metrological testing data verification system based on the fusion of OCR and NLP, characterized in that, The system includes a memory and a processor. The memory contains a program for a measurement and testing data verification method based on the fusion of OCR and NLP. When the processor executes the program, the measurement and testing data verification method based on the fusion of OCR and NLP performs the following steps: Acquire raw image data and simultaneously collect environmental parameter sensor data and test document text during the metrological testing process. The raw image data includes at least the metrological equipment identification mark and real-time reading. The original image data is subjected to target detection and character localization using a preset recognition model. Key information from the equipment nameplate and the readings of pointer / digital displays are extracted to obtain an image recognition dataset. The detection document text is subjected to entity extraction and semantic parsing using a natural language processing parsing engine. Key indicators of the verification procedure and historical detection parameters are extracted to obtain a text semantic dataset. The image recognition dataset, the environmental parameter sensor data, and the text semantic dataset are input into the multi-source data fusion and anomaly detection model to calculate the spatiotemporal consistency metric to identify multimodal data anomaly features. The abnormal features of the multimodal data are input into a deep learning error correction model to obtain the corrected standard metrological detection and verification results, and an automatic early warning prompt is generated when the threshold is triggered.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a program for a measurement and testing data verification method based on the fusion of OCR and NLP. When the program is executed by a processor, it implements the steps of a measurement and testing data verification method based on the fusion of OCR and NLP as described in any one of claims 1 to 6.

9. A computer program product, characterized in that, The computer program product includes computer program code, which, when run on a computer, enables the computer to implement the steps of a metrological testing data verification method based on the fusion of OCR and NLP as described in any one of claims 1 to 6.

10. An electronic device, characterized in that, The electronic device includes a processor and a memory; wherein the memory is used to store a computer program, and the processor is used to execute the computer program stored in the memory to enable the electronic device to perform the steps of the measurement and testing data verification method based on the fusion of OCR and NLP as described in any one of claims 1 to 6.

Images