A pipeline detection engineering data quality evaluation method, system and device

Abstract

This application relates to the field of pipeline inspection technology, and in particular to a method, system, and equipment for quality assessment of pipeline inspection engineering data. The method includes: acquiring a pipeline inspection engineering dataset to be evaluated; performing multi-dimensional quality analysis on the pipeline inspection engineering dataset to obtain authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and crawling speed standardization verification results, replacing subjective and inefficient manual experience judgment with objective and consistent algorithmic judgment; and generating a comprehensive quality evaluation report of the pipeline inspection engineering dataset based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and crawling speed standardization verification results, thus completing a closed loop from data quality inspection to quality control and value enhancement.

Description

Technical Field

[0001] This application belongs to the field of pipeline inspection technology, and in particular relates to a method, system and equipment for quality assessment of pipeline inspection engineering data. Background Technology

[0002] Pipeline inspection is a core component in assessing the health of underground pipe networks and ensuring their safe operation. Internal detection technologies, such as closed-circuit television (CCTV) inspection, collect continuous images and synchronous spatiotemporal data (such as distance and timestamps) inside the pipeline to form a structured pipeline inspection engineering dataset. This dataset serves as the sole original basis for subsequent defect identification, condition assessment, and maintenance decisions, and its quality directly determines the accuracy and reliability of all derived analysis results.

[0003] Current technologies for controlling the quality of pipeline inspection project data mainly rely on manual review and experience-based judgment by inspection personnel after the fact. This approach has inherent drawbacks such as low efficiency, strong subjectivity, and difficulty in standardizing. More importantly, current technologies lack a systematic and automated analysis method, making it impossible to conduct a synchronous, quantitative, and traceable comprehensive assessment of the authenticity of the data, the standardization of the collection process, and the completeness of the content logic in the project dataset.

[0004] Due to the aforementioned technological gaps, low-quality or invalid test data can easily be mixed into subsequent processes, leading to inaccurate analysis foundations and resulting in misjudgments, missed judgments, or duplicate tests. This severely restricts the improvement of the standardization and intelligence level of pipeline inspection operations and brings additional operation and maintenance costs and safety hazards. Summary of the Invention

[0005] This application provides a method, system, and equipment for quality assessment of pipeline inspection engineering data, which can solve the problems of low assessment efficiency, large subjective bias, and inconsistent quality standards caused by the reliance on manual experience for retrospective review in the existing pipeline inspection engineering data quality assessment process.

[0006] In a first aspect, embodiments of this application provide a method for quality assessment of pipeline inspection engineering data, including: Obtain the dataset of pipeline inspection projects to be evaluated; A multi-dimensional quality analysis was performed on the pipeline inspection project dataset to obtain the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and creep speed standardization inspection results. Based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and creep speed standardization verification results, a comprehensive quality evaluation report of the pipeline inspection engineering dataset is generated.

[0007] The technical solutions described in this application embodiment have at least the following technical effects: The quality assessment method for pipeline inspection engineering data provided in this application provides a unified and complete structured data input foundation for the subsequent systematic and algorithmic quality assessment process by acquiring the pipeline inspection engineering dataset to be evaluated, ensuring the originality and processability of the assessment object. Multi-dimensional quality analysis of the pipeline inspection engineering dataset yields results for authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and crawling speed standardization verification. This enables synchronous, automated, and quantitative review of core quality attributes such as data source credibility, operational process standardization, and content logical completeness, replacing subjective judgment with objective and consistent algorithmic analysis. This approach fundamentally addresses the systemic flaws of traditional quality inspection models, such as inefficiency, inconsistent standards, and susceptibility to oversights, by eliminating the need for inefficient manual judgment. Based on the results of authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and climbing speed standardization inspection, a comprehensive quality evaluation report for the pipeline inspection engineering dataset is generated. This integrates multi-dimensional and fragmented algorithmic analysis conclusions into a structured output containing overall grade, sub-item evidence, precise problem location, and corrective suggestions. This facilitates reliable data quality certification, problem tracing, targeted rectification, and the flow of high-quality data to downstream analysis applications, thus completing a closed loop from data quality inspection to quality control and value enhancement.

[0008] Secondly, embodiments of this application provide a quality assessment system for pipeline inspection engineering data, including: The acquisition unit is used to acquire the pipeline inspection project dataset to be evaluated. The analysis unit is used to perform multi-dimensional quality analysis on the pipeline inspection project dataset to obtain the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results. The evaluation unit is used to generate a comprehensive quality evaluation report for the pipeline inspection engineering dataset based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results.

[0009] Thirdly, embodiments of this application provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method as described in any of the first aspects above.

[0010] Fourthly, embodiments of this application provide a computer program product that, when run on an electronic device, causes the electronic device to perform the method described in any of the first aspects above.

[0011] It is understood that the beneficial effects of the second to fourth aspects mentioned above can be found in the relevant descriptions in the above aspects, and will not be repeated here. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a schematic flowchart of a method for quality assessment of pipeline inspection engineering data provided in an embodiment of this application; Figure 2 This is a schematic diagram illustrating the operation of a method for quality assessment of pipeline inspection engineering data provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of a quality assessment system for pipeline inspection engineering data provided in one embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0014] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0015] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0016] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0017] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determination" or "if the described condition or event is detected" may be interpreted, depending on the context, as "once determination," "in response to determination," "once the described condition or event is detected," or "in response to the detection of the described condition or event."

[0018] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0019] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0020] Pipeline inspection is a core component in assessing the health of underground pipe networks and ensuring their safe operation. Internal detection technologies, such as closed-circuit television (CCTV) inspection, collect continuous images and synchronous spatiotemporal data (such as distance and timestamps) inside the pipeline to form a structured pipeline inspection engineering dataset. This dataset serves as the sole original basis for subsequent defect identification, condition assessment, and maintenance decisions, and its quality directly determines the accuracy and reliability of all derived analysis results.

[0021] Current technologies for controlling the quality of pipeline inspection project data mainly rely on manual review and experience-based judgment by inspection personnel after the fact. This approach has inherent drawbacks such as low efficiency, strong subjectivity, and difficulty in standardizing. More importantly, current technologies lack a systematic and automated analysis method, making it impossible to conduct a synchronous, quantitative, and traceable comprehensive assessment of the authenticity of the data, the standardization of the collection process, and the completeness of the content logic in the project dataset.

[0022] Due to the aforementioned technological gaps, low-quality or invalid test data can easily be mixed into subsequent processes, leading to inaccurate analysis foundations and resulting in misjudgments, missed judgments, or duplicate tests. This severely restricts the improvement of the standardization and intelligence level of pipeline inspection operations and brings additional operation and maintenance costs and safety hazards.

[0023] To address the aforementioned issues, this application provides a method, system, and equipment for quality assessment of pipeline inspection engineering data. This method acquires a dataset of pipeline inspection engineering data to be evaluated, providing a unified and complete structured data input foundation for subsequent systematic and algorithmic quality assessment processes, ensuring the originality and processability of the assessment object. Multi-dimensional quality analysis is performed on the pipeline inspection engineering dataset to obtain results for authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and crawling speed standardization verification. This enables synchronous, automated, and quantitative review of core quality attributes such as data source credibility, operational process standardization, and content logical completeness, replacing subjective and inefficient manual experience-based judgment with objective and consistent algorithmic judgment. This approach fundamentally solves the systemic defects of traditional quality inspection models, such as low efficiency, inconsistent standards, and susceptibility to oversights. Based on the results of authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and climbing speed standardization inspection, a comprehensive quality evaluation report for pipeline inspection engineering datasets is generated. This integrates multi-dimensional and scattered algorithm analysis conclusions into a structured output containing overall level, sub-item evidence, precise problem location, and correction suggestions. This facilitates credible data quality certification, problem tracing, targeted rectification, and the flow of high-quality data to downstream analysis and applications, thus completing a closed loop from data quality inspection to quality control and value enhancement.

[0024] The quality assessment method for pipeline inspection engineering data provided in this application embodiment can be applied to electronic devices. In this case, the electronic device is the executing entity of the quality assessment method for pipeline inspection engineering data provided in this application embodiment. This application embodiment does not impose any restrictions on the specific type of electronic device.

[0025] For example, electronic devices can be ultra-mobile personal computers (UMPCs), netbooks, desktop computers, computers, laptops, communication equipment, computing devices, satellite wireless equipment, etc.

[0026] To better understand the quality assessment method for pipeline inspection engineering data provided in the embodiments of this application, the specific implementation process of the quality assessment method for pipeline inspection engineering data provided in the embodiments of this application will be described by way of example below.

[0027] Figure 1 This illustration shows a schematic flowchart of a method for quality assessment of pipeline inspection engineering data provided in an embodiment of this application. The method for quality assessment of pipeline inspection engineering data includes: S100: Obtain the pipeline inspection project dataset to be evaluated.

[0028] As you can understand, the pipeline inspection project dataset is a structured collection of data generated synchronously during pipeline closed-circuit television (CCTV) inspection operations. Please refer to [link / reference needed]. Figure 2 The dataset not only includes video stream data recording images of the pipeline's interior, but also embeds metadata (such as equipment number and timestamp) associated with each frame or specific time point, as well as equipment runtime sequence data (such as cumulative distance recorded by the encoder and travel speed markers). The pipeline inspection project dataset can be obtained by directly reading from the inspection equipment's storage medium, downloading from the inspection project management platform, or transmitting via a dedicated data interface. The purpose of obtaining this dataset is to provide a complete and tamper-proof raw data foundation for subsequent automated quality analysis.

[0029] S200 performs multi-dimensional quality analysis on the pipeline inspection engineering dataset to obtain the results of authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and climbing speed standardization inspection.

[0030] It is understandable that the design of multi-dimensional quality analysis of pipeline inspection engineering datasets aims to fundamentally solve the systemic defects inherent in traditional methods that rely on manual review and experience-based judgment, such as inefficiency, strong subjectivity, inconsistent standards, and biased evaluation. Traditional methods require operators to process multiple non-standardized rules in their minds simultaneously, including source credibility, operational standardization, and content completeness, leading to high cognitive load, inconsistent judgments from different individuals and at different times, and significant efficiency bottlenecks in linear review models. Therefore, this invention explicitly defines and decouples quality assessment into five core dimensions: authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and crawling speed standardization verification, and constructs corresponding algorithms to implement them. By encoding the evaluation rules for each dimension (such as regular expression templates, speed thresholds, and statistical thresholds) into executable, unified algorithmic logic, this method achieves objectification and consistency of evaluation standards, freeing evaluation conclusions from personal experience. Simultaneously, it can automatically, in parallel, or systematically schedule these algorithm modules to perform high-speed scanning of the dataset, liberating quality inspection work from time-consuming manual review and achieving an order-of-magnitude improvement in efficiency. More importantly, each algorithm module outputs structured and quantitative intermediate results, rather than simple yes / no judgments. This not only generates detailed data evidence to support the conclusions, but also provides a direct basis for accurate problem tracing, targeted rectification, and subsequent data correction, thereby providing technical support for the standardization and intelligentization of pipeline inspection operations.

[0031] This method constructs a multi-dimensional quality assessment system. The definition and design principles of its five dimensions follow a complete logical chain from data source, operational starting point, process benchmark, to behavioral norms, aiming to systematically replace and surpass human experience-based judgment. Multi-dimensional quality analysis of pipeline inspection engineering datasets does not involve isolated verification of single-dimensional quality assessments, but rather consistent reasoning regarding the completeness and verifiability of the video evidence chain in the pipeline inspection engineering dataset: observable information in the videos (such as headers, mileage, and frame continuity) is structured into computable evidence, jointly verified with business databases, ledgers, GIS records, and regulatory constraints, ultimately outputting sub-item conclusions and auditable evidence. To maintain semantic and structural consistency among evaluation items, this method uses pipe segments as the smallest evaluation object, evaluating any pipe segment... Let the video being detected be a frame sequence. The frame rate is Business records (ledger / GIS / project database) are as follows: The defect identification result is (Including category, mileage, confidence level, etc.). Based on the specifications for header, continuous metering, zeroing compensation, speed control, etc., video quality evaluation is formalized into a complete mapping process of evidence extraction, constraint testing, and decision fusion: ,in For a structured collection of evidence, Threshold and tolerance parameters (such as field consistency threshold, length tolerance, speed classification threshold, etc.). This auditable evidence package provides the basis for judgment and verifiable positioning information such as timestamps, mileage segments, and keyframes. For each of the five quality evaluation items (authenticity, coverage integrity, starting point compliance, starting point zeroing, and crawling speed standardization), this application requires verification. Output triples: ,in To determine whether or not it passes, For continuous scoring (used in regular scoring). The criteria for judgment (such as keyframes, OCR fields, ledger matching items, speeding intervals, and reset times) are summarized as follows: Collection of evidence The organization is: ,in For structured fields in the header (pipe segment number, starting and ending wells, pipe diameter, date, etc.); It is a mileage sequence; As a time reference; This is a sequence of scene stages (surface / wellhead / inside pipe, etc.). Alignment mapping between defects and video timelines supports traceable auditing.

[0032] Authenticity verification aims to confirm the credibility and logical consistency of data sources. Its design follows the principles of multi-source information cross-validation and algorithmic fault-tolerant matching, ensuring data authenticity through the comparison of textual metadata and physical measurement data. Coverage integrity verification focuses on the continuity and completeness of the detection process. Its core is the principle of anomaly pattern recognition based on spatiotemporal data sequences, automatically locating interruptions and jumps by analyzing the mathematical continuity of distance-time series. Starting point compliance verification verifies whether the operation starts from the designated manhole, employing a multi-level analysis principle from scene understanding to object recognition, integrating color features and target detection to form a three-dimensional evidence chain. Starting point zeroing verification confirms the correctness of the distance measurement benchmark and follows the principles of physical process-driven and compensation-oriented approaches. It not only judges zeroing stability through statistical analysis but also calculates precise compensation values ​​based on the equipment's physical model in case of anomalies. Climbing speed compliance verification continuously monitors the travel process. Its design is based on the principles of dynamic threshold matching and precise event aggregation, achieving quantitative identification and statistics of non-compliant sections by calculating instantaneous speed and comparing it with the standard threshold associated with the pipe diameter. These five dimensions together constitute a closed-loop evaluation logic from data identity verification, operation start point verification, measurement benchmark calibration to process behavior monitoring. Their design principles all point to the core goals of algorithmization, quantification, and traceability, transforming the traditional fuzzy, subjective, and discrete quality judgment into an objective, consistent, and standardized technical process that can directly support problem location and correction.

[0033] In one possible implementation, in step S200, a multi-dimensional quality analysis is performed on the pipeline inspection engineering dataset to obtain the authenticity verification results, including: S211, sample the pipeline inspection engineering dataset to determine the corresponding key data units.

[0034] It is understandable that representative data segments containing key verification information, i.e., key data units, can be efficiently selected from the video stream data of pipeline inspection engineering datasets. In practice, this is reflected in a frame sampling strategy. For example, several frames (i.e., keyframes) can be extracted as key data units from the beginning (e.g., the first 30 seconds), the middle stable segment, and the end of the dataset at fixed intervals or based on content change detection. Key data units typically contain project identification text (e.g., manhole cover number, construction unit name) and distance information that can be used for length verification. The sampling strategy is designed to balance computational efficiency with the comprehensiveness of information coverage.

[0035] S212, determine the text information fields and the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset based on all key data units.

[0036] Understandably, information can be extracted and aggregated from the key data units obtained from sampling to generate two types of core evidence data for authenticity comparison. The text information field is semantic information extracted from visual data, and the distance difference parameter between the first and last data units is geometric consistency evidence derived from physical measurement data. The combination of these two types of evidence achieves dual verification of the authenticity and logical consistency of the data content.

[0037] Optionally, S212, based on all key data units, determine the text information fields and the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset, including: S2121, perform information extraction operations on all key data units to obtain the corresponding text recognition results.

[0038] It can be understood that information extraction specifically refers to the application of Optical Character Recognition (OCR) technology. Specifically, each key data unit (keyframe image) undergoes preprocessing (such as grayscale conversion, binarization, and denoising) to enhance the contrast of text regions. Then, a pre-trained OCR engine (e.g., a deep learning model supporting mixed Chinese and English recognition, such as an optimized version of PaddleOCR or Tesseract) is invoked to scan and recognize the image, converting the text regions in the image into machine-readable text strings. The recognition result for each key data unit can contain multiple text blocks and their positional information within the image.

[0039] For example, the OCR engine can be customized based on the industrial-grade deep learning model PP-OCRv4. Before training, a mixed Chinese and English text dataset adapted to keyframe images is constructed, integrating the ICDAR2015 and MSRA-TD500 general datasets with a self-made keyframe annotation dataset. All samples are then subjected to enhancement operations including random ±15° rotation, 0.5-2x scaling, Gaussian blur (kernel size 3×3 / 5×5), salt-and-pepper noise (noise density 0.01-0.03), and random cropping. Simultaneously, the training, validation, and test sets are divided in an 8:1:1 ratio, and the text is then processed. Character annotation coordinates are normalized. In terms of network architecture, MobileNetV3_large_x0_5 is selected as the backbone network to balance accuracy and inference efficiency. The Neck layer uses a DBHead detection head paired with an RNNNEncoder feature fusion layer. The Head layer combines CRNN and Attention mechanisms to achieve end-to-end text detection and recognition. Training hyperparameters are set to a batch size of 32 on a single GPU and an initial learning rate of 0.001. Cosine annealing is used to decay the learning rate from 0.001 to 1e-5. The model was trained using AdamW with a weight decay coefficient of 0.0005 and a batch normalized momentum of 0.9. The training consisted of 200 epochs, with model checkpoints validated and saved every 10 epochs. For optimization, a joint training mode for detection and recognition was adopted. The detection branch incorporated FocalLoss loss to alleviate the imbalance between positive and negative samples in text regions, while the recognition branch used a CTC+Attention hybrid loss to improve the recognition of low-contrast and blurred text. L2 regularization was also added to suppress overfitting. Furthermore, based on the image quality characteristics of keyframe images, similar low-contrast and locally blurred text were incorporated into the training process. Real samples are used for targeted reinforcement training. After 150 rounds of training, a learning rate fine-tuning strategy is initiated, reducing the learning rate to 1 / 10 of the original and continuing training for another 50 rounds. After the validation set metrics stabilize, channel pruning and INT8 quantization optimization are performed on the model at a ratio of 0.2 to remove redundant network nodes and reduce computational load. Finally, the training completion criteria are character recognition accuracy (CR) ≥ 99.0%, text detection recall (R) ≥ 98.5%, and detection precision (P) ≥ 98.0%. After evaluation on the test set, the pruned and quantized model is solidified, forming a customized OCR engine adapted to key data units.

[0040] S2122, extracts the corresponding text information fields from the text recognition results through predefined pattern matching rules; the text information fields include unit name, detection location, pipeline number, date and time.

[0041] It's understandable that pattern matching rules consist of a series of regular expressions. Since text identification at the inspection site typically follows certain formats (e.g., pipeline numbers may follow a region-road-type-serial number format, and dates are YYYY-MM-DD format), the system pre-stores regular expression templates for fields such as unit name, inspection location, pipeline number, and date / time. By traversing the text recognition results of all key data units, these regular expressions can be used for matching and extraction, structurally extracting the target field content from different frames of potentially incomplete recognized text. For example, from identified fragments such as "XX Municipal Engineering Company," "2023-10-26," and "Yan'an Road Sewage Pipe W-012," the corresponding fields can be categorized and filled.

[0042] S2123, based on the analysis of all key data units, determine the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset.

[0043] Understandably, we can find the timestamps corresponding to the first key data unit (first frame or first pipeline number frame) and the last key data unit (tail frame or last pipeline number frame) from the device runtime sequence data associated with the dataset, and then obtain the encoder distance readings corresponding to these two timestamps. Calculating the difference between these two distance values ​​yields the first-tail data unit distance difference parameter. This parameter represents the pipeline length estimated from the video content for this detection.

[0044] S213, compare the text information field and the distance difference parameter between the first and last data units with the corresponding benchmark data in the preset engineering database to obtain the authenticity verification result.

[0045] It is understandable that the authenticity verification employs dual evidence gating, namely metadata consistency and detection length consistency, aiming to simultaneously confirm the video. With database records Consistent in identity and scale. Text information fields verify whether the video corresponds to the specified task / segment, and the distance difference parameter between the first and last data units verifies whether the detection length is within the design tolerance range. Header / overlay subtitle fields are extracted into text information fields using OCR. However, its reliability is affected by non-standard information entry and recognition errors caused by visual degradation; if only the distance difference parameter between the first and last data units is relied upon to detect the length, it may be affected by counter anomalies, starting point compensation, or video truncation interference. Therefore, authenticity is determined when both pieces of evidence pass, and auditable evidence is output. The pre-set engineering database stores standard information for planned inspection tasks, including the contracted construction unit (unit name), the exact location and number of the pipeline to be inspected (inspection location, pipeline number), the planned inspection date, and the design / known length of the pipeline. Text fields such as the unit name extracted in S212 can be matched with database records using string similarity (e.g., using an edit distance algorithm). Simultaneously, the calculated distance difference between the first and last data units is compared with the known pipeline length in the database, allowing a reasonable measurement error range (e.g., ±2%). Finally, considering both text matching and length consistency, a comprehensive authenticity verification result is given (e.g., authentic, 95% confidence, or questionable, excessive length deviation).

[0046] For example, during metadata consistency checks, for and China participated in the verification Each field (such as task / unit, location, date, start and end hash numbers, etc., obtained by OCR recognition) is calculated for fuzzy similarity and then weighted and summarized: ,in From The parsed result of the first Each field value This is the standard value for the database; It can include features such as abbreviation merging, number format normalization, and character noise tolerance. This represents the importance weight of the field. When checking length consistency (order-of-magnitude alignment), it is determined from the mileage sequence. The detection length is calculated from the first and last valid readings and compared with the design length in the ledger / GIS. The relative error is obtained by comparison: ,in Used to absorb starting point compensation, reading jitter, and normal measurement errors; authenticity is defined by the following conditions: Continuous scoring is available in the normal mode. ,For example In a stringent mode, it can be used as a veto. (Audit evidence) Including a list of conflicting fields, similarity, And related keyframes / timestamps to support verification and location.

[0047] In one possible implementation, in step S200, a multi-dimensional quality analysis is performed on the pipeline inspection engineering dataset to obtain the coverage integrity verification results, including: S221, perform equal-interval sampling on the pipeline inspection engineering dataset to obtain the corresponding equal-interval data sequence.

[0048] Understandably, to analyze the continuity of the detection process, uniform observations are needed throughout the entire detection journey. This step samples from the entire dataset at fixed time intervals (e.g., 1 frame per second) or fixed distance intervals (e.g., every 0.1 meters) to generate a temporally or spatially uniformly distributed, equally spaced data sequence. This equally spaced data sequence provides dense observation points for coverage integrity analysis.

[0049] S222, extract the corresponding distance values ​​from all equally spaced data sequences to determine the corresponding distance dataset.

[0050] It is understandable that the encoder distance value can be extracted from the metadata associated with each data unit (frame) in the above equally spaced data sequence. All the distance values ​​extracted in chronological order are arranged to form a distance dataset synchronized with the detection process. The distance dataset intuitively reflects the change of the crawler's position in the pipeline over time.

[0051] S223, determine the distance change trend data based on the distance dataset, and identify the corresponding acquisition behavior characteristics through the distance change trend data; wherein, the acquisition behavior characteristics are used to reflect the acquisition interruption, abnormal jump in distance value, and reverse acquisition behavior that exist when acquiring pipeline inspection engineering dataset.

[0052] It is understandable that distance change trend data can be obtained by calculating the first-order difference (instantaneous velocity) and second-order difference (acceleration) of the distance dataset. The analysis logic includes: 1) Acquisition interruption: Identifying segments where the distance value remains unchanged or changes very little (below the minimum displacement resolution of the device) across multiple consecutive sampling points. This means that the device is paused but video recording continues, or the video was edited into static frames later. 2) Abnormal jumps in distance values: Detecting discontinuous and large increases or decreases in distance values ​​(e.g., an increase of tens of meters between two adjacent frames, far exceeding the maximum crawling speed of the device), indicating data transmission errors or later human tampering with the data. 3) Reverse acquisition behavior: Detecting whether the distance dataset shows a continuous decreasing sequence, which indicates that the detection device is moving from the other end of the pipeline towards the starting point, belonging to supplementary shooting or reverse detection. These patterns of acquisition interruption, abnormal jumps in distance values, and reverse acquisition behavior that exist when acquiring pipeline inspection engineering datasets are collectively referred to as acquisition behavior characteristics.

[0053] S224, Obtain the coverage integrity verification result based on the characteristics of the acquisition behavior.

[0054] Understandably, it can summarize all identified data collection behavior characteristics, such as determining whether there is an interruption (yes / no), the duration and location of the interruption; whether there are abnormal jump points; whether there is a reverse shooting segment and whether it is completely connected to the forward segment. Based on a set of predefined integrity rules (e.g., allowing one interruption of less than 10 seconds, but the location must be marked; allowing complete bidirectional shooting, and the sum of the coverage intervals should be greater than 95% of the pipe length), it comprehensively judges whether the detection completely covers the target pipe segment and generates a coverage integrity verification result, which includes the conclusion of whether the overall coverage is complete and the specific problem segment location information.

[0055] For example, the coverage integrity verification result can be obtained by first analyzing the mileage-time series. Sparse sampling is performed to obtain the set of observation points. ,in Time (seconds) Mileage reading: By changing point detection Obtain the set of variable points Divide the video into A fragment .

[0056] For each segment Use the minimum / maximum mileage within that segment. Constructing the coverage area The length of overlap between adjacent segments: This indicates the possibility of rewinding, repeated shooting, or overlapping reshoots. Let the union of the coverage areas of all segments be... Its total length With design length The relative gap ratio is defined as: The larger the value, the more significant the gap. If Even if the video is segmented, it can be considered "segmented but fully covered," avoiding misjudgment of compliant reshooting situations. To further improve interpretability and pinpoint the cause of segmentation, this application introduces obstruction / stagnation clues and dual-end reshooting clues. Obstruction / stagnation is indicated by a long period of low change in the image: [The text abruptly ends here, likely due to an incomplete translation or source material.] For the sampling time used for this discrimination, If the frame difference interval is used, then... Measuring the difference between two frames ( (counting non-zero pixels), and using a threshold With minimum continuous running length Determine if a "long stationary segment" exists: The double-end reshoot / reverse advancement is indicated by a significant decrease in mileage, defined as: ,in The minimum decrease threshold for suppressing reading noise; when At this time, it indicates that there may be a process of reshooting from the other end or reversing the shooting process, which needs to be combined with... Further verification is performed against the gap interval to check for any remaining omissions. Finally, the coverage integrity module outputs a triplet. :in Give by the tag ( (for configurable notch tolerance) As a continuous score in the regular mode; evidence package The fragment structure is also given. Timestamp / mileage boundaries, overlap The mileage range corresponding to the gap, the blocked / stagnant range and The tags are used to support review and reshoot decisions.

[0057] In one possible implementation, in step S200, a multi-dimensional quality analysis is performed on the pipeline inspection project dataset to obtain the starting point compliance verification results, including: S231, based on the pipeline inspection engineering dataset, perform content parsing to extract the initial data segment.

[0058] Understandably, the compliance verification at the starting point ensures that the inspection video begins in the environment surrounding the ground inspection manhole, recording necessary on-site information. Data from a time window (e.g., the first 30 seconds or the first 100 frames) after the start of the pipeline inspection project dataset can be used as the initial data segment. This initial data segment corresponds to the initial stage from when the inspection equipment is lowered from the ground inspection manhole until it is fully inside the pipeline and begins stable inspection.

[0059] S232, the initial data segment is transformed into the specified feature space for threshold segmentation to determine the color distribution feature data of different environmental elements.

[0060] Understandably, to perform scene-level macro-environmental analysis, the video frames in the initial data segment can be converted from the RGB color space to the HSV color space. This space is less sensitive to changes in lighting and is more conducive to color segmentation. Subsequently, threshold ranges are set in the HSV space for typical outdoor and inspection well environmental elements such as the sky (high brightness, blue / cyan hues), ground / vegetation (green and brown hues), and concrete well walls (gray hues). For each frame, the proportion of pixels falling within each threshold range is statistically analyzed to obtain a set of color distribution characteristic data that changes over time, which is used to quantitatively analyze whether the scene possesses the characteristics of an open outdoor environment.

[0061] S233, Based on the initial data segment, identify the target entity and determine the manhole cover detection data corresponding to the initial data segment.

[0062] It's understandable that target entity recognition specifically refers to the detection of manhole covers. A deep learning-based object detection model (e.g., YOLO or Faster R-CNN architecture) can be used, trained on a large number of images labeled with manhole covers. Inputting video frames from the initial data segment into the model, the model outputs whether a manhole cover was detected, the bounding box location of the cover, and the confidence score. Simultaneously, the traditional Hough circle transform algorithm can be used to detect circular outlines in the image, serving as supplementary verification of the manhole cover's shape. Combining the results of both methods yields the manhole cover detection data, indicating which frames detected the manhole cover and their corresponding confidence scores.

[0063] For example, the manhole cover target entity recognition model can be customized and trained based on the YOLOv8n lightweight architecture. Before training, a dedicated manhole cover detection dataset is constructed, which integrates publicly available road facility datasets with self-made manhole cover labeled samples. This dataset covers round / square manhole covers, different lighting conditions (strong light / backlight / weak light), different wear levels (intact / damaged / missing), and complex backgrounds (road surface / green belt / water accumulation). All samples are subjected to image enhancement operations including random ±20° rotation, 0.3-1.2x scaling, random folding, Gaussian blur (kernel size 3×3), and random brightness and contrast adjustment (±0.2). The training set, validation set, and test set are divided into a 7:2:1 ratio. The bounding box annotation format was normalized (converted to YOLO format) and pixel normalization was performed. In terms of network configuration, the original C2f backbone feature extraction layer and SPPF spatial pyramid pooling layer of YOLOv8n were retained. For the small target detection characteristics of manhole covers, a feature fusion branch with a 1×1 convolutional kernel was added to the Neck layer to enhance shallow detail feature extraction. The Head layer retained the decoupled detection head and optimized the anchor box size to adapt to the actual pixel scale of the manhole cover (three sets of preset anchor box sizes were used to adapt to different sizes of manhole covers). The training hyperparameters were set to single-GPU training, batch size of 16, and initial learning rate of 0.001, using a cosine annealing learning rate strategy, starting the learning rate from 0.0... The weight decay was gradually reduced from 01 to 1e-6. The optimizer used was AdamW with a weight decay coefficient of 0.0005 and momentum of 0.937. The total number of training rounds was 300. Every 20 rounds, the optimal model checkpoint was evaluated on the validation set and saved. An early stopping mechanism was also set (training was terminated if there was no improvement in mAP50 on the validation set for 30 consecutive rounds). In terms of optimization logic, CIoULoss was used as the loss function to improve the accuracy of bounding box regression. FocalLoss (γ=2, α=0.25) was introduced to alleviate the class imbalance problem between manhole cover samples and background samples. L2 regularization and random mosaic data augmentation were added to suppress model overfitting. For the actual scenario of manhole cover detection, small targets were added during training. The training weights for occluded manhole cover samples were adjusted to enhance the model's ability to recognize complex scenes. After 200 training iterations, a fine-tuning strategy was initiated, reducing the learning rate to 5e-5 and decreasing data augmentation intensity to focus on model accuracy convergence. After model training, ONNX format conversion and NMS (non-maximum suppression, IoU threshold set to 0.45) optimization were performed to remove redundant detection boxes. A supplementary validation was performed using the traditional Hough circle transform algorithm (setting the circle radius threshold to adapt to the manhole cover size, and setting the Hough gradient accumulator threshold to 80). The final validation set showed mAP50 ≥ 99.2%, manhole cover detection precision ≥ 98.8%, recall ≥ 98.5%, and single-frame inference confidence threshold ≥ 0.7 is the criterion for determining training completion. After evaluating the model across the entire scene on the test set, the model is solidified. The detection results of the model are then fused with the shape verification results of the Hough circle transform for a final determination (if both are detected, the confidence is weighted higher; if only one is detected, a second verification is performed). The final output includes the manhole cover detection frame number, bounding box position, and fused confidence score.

[0064] S234, the starting point compliance verification result is obtained based on manhole cover detection data and color distribution feature data.

[0065] To improve the robustness and accuracy of the algorithm, this method employs a multi-frame voting mechanism, for example, analyzing the first N frames of the initial data segment. For each frame, it combines its color distribution feature data (whether outdoor features are displayed) and manhole cover detection data (whether a manhole cover is detected) to give a preliminary judgment of whether the frame is compliant or non-compliant. Finally, it calculates the proportion of compliant frames in the N frames. If the proportion exceeds a preset threshold (e.g., 70%), the compliance verification of the entire detection starting point is considered successful. The final result will include the conclusion of compliance, as well as key features used to support the conclusion (e.g., the highest confidence level of manhole cover detection, the average proportion of sky pixels). In specific implementation, the manhole cover target entity recognition model performs each frame... Output the binary classification confidence score: (Inside the pipe) and (External environment). To suppress single-frame fluctuations, continuous... The moment when a frame exceeds the threshold is determined: ,in The built-in confidence threshold is used for the tube. For stable length. The compliance of the starting point is determined accordingly: The first is the confidence threshold for the external scene; the second ensures that the external scene actually existed before entry. Evidence package. Output The keyframes and timestamps before and after (corresponding to the peak of external confidence and the stable interval within the system) are used for verification and tracing.

[0066] In one possible implementation, in step S200, a multi-dimensional quality analysis is performed on the pipeline inspection project dataset to obtain the starting point zeroing verification result, including: S241, based on the initial data segment, extract the distance value of its initial part of the data sequence to obtain the initial distance sequence.

[0067] It is understandable that zeroing the starting point ensures that the distance recording of the detection device starts from zero, avoiding systematic measurement errors. Therefore, it is necessary to determine the distance reading when the detection device is first started. From the defined initial data segment, the first few seconds (e.g., the first 10 seconds) of video frames can be extracted, and the corresponding encoder distance values ​​can be extracted from the metadata of these frames. These values ​​are then arranged in chronological order to form an initial distance sequence. The initial distance sequence is used to analyze the initial state of the distance counter when the device is started.

[0068] S242, perform statistical and trend change analysis on the initial distance sequence to determine the corresponding zeroing stability analysis results and reference state transition point; where the reference state transition point is the reference point for the detection equipment to enter the pipeline network from the ground.

[0069] Understandably, multi-dimensional statistical calculations and time-series trend analysis can be performed on the initial distance sequence. First, the mean, median, standard deviation, range, and fluctuation coefficient of the distance difference between consecutive frames are calculated. The standard deviation ≤ a preset threshold and the absolute value of the distance difference between consecutive frames ≤ a fixed step size are used as criteria to obtain the zeroing stability analysis results. If the indicators meet the requirements, the distance record zeroing state is determined to be stable at device startup, with no initial offset error. If the indicators exceed the threshold, the zeroing state is determined to be abnormal, with systematic measurement bias. Simultaneously, time-series trend fitting and abrupt change detection are performed on the initial distance sequence. The method calculates the mean distance and slope of change within a sliding window (the window size adapts to the frame interval). Combined with the second-order difference method, it identifies key nodes in the sequence that transition from a stable, unchanging state (slope approaching 0, corresponding to the equipment being stationary on the ground) to a continuously linearly increasing state (slope stabilizing at a fixed value, corresponding to the equipment starting to move into the pipeline network). The frame number and distance value corresponding to this node are marked as reference state transition points, serving as the time and distance benchmarks for the detection equipment entering the pipeline network from the ground. The distance change characteristics before and after the transition point are recorded synchronously, providing a basis for distance calibration of subsequent pipeline network detection data.

[0070] Optionally, S242, statistical and trend change analysis is performed on the initial distance sequence to determine the corresponding zeroing stability analysis results and reference state transition point, including: S2421, perform statistical and trend analysis on the initial distance sequence to determine the corresponding variance and coefficient of variation.

[0071] It is understandable that variance measures the dispersion of data, while the coefficient of variation eliminates the influence of dimensions. In an ideal zero-valued state, with the equipment stationary, the distance value should fluctuate slightly around a very small value (close to 0), therefore both variance and coefficient of variation should be very small.

[0072] S2422, stability analysis based on variance and coefficient of variation, yields the results of zeroing stability analysis.

[0073] Understandably, the calculated variance and coefficient of variation can be compared with a preset stability threshold. If both are below the threshold, the zeroing operation is considered stable; if either is above the threshold, the zeroing operation is considered unstable or suspected of not being zeroed. The zeroing stability analysis result is the primary indicator for determining whether the zeroing operation was successfully executed.

[0074] S2423, perform second-order difference calculation on the initial distance sequence, identify the inflection point of distance change in the initial distance sequence, and determine the inflection point as the reference state transition point.

[0075] It can be understood that the second-order difference is essentially the calculation of the rate of change of distance (velocity) (acceleration). When the equipment goes from being stationary on the ground (velocity ≈ 0) to being lowered into the well at a constant positive velocity, its acceleration will have a pulse change from 0 to a positive value. By analyzing the second-order difference of the initial distance sequence, the inflection point of this significant change can be found. The moment corresponding to this inflection point is identified as the reference state transition point, marking the beginning of the equipment's meaningful movement into the pipeline.

[0076] S243, Based on the zeroing stability analysis results, the reference state transition point, and the first distance value of the initial distance sequence, determine the zeroing operation status of the detection device's starting point.

[0077] It is understandable that the final logical judgment of the zeroing operation can be made by combining the zeroing stability analysis results, the reference state transition point, and the first distance value of the initial distance sequence. The decision logic is as follows: If the zeroing stability analysis result is stable, and the first distance value of the initial distance sequence is close to 0 (within the error range), then the zeroing operation is directly judged as successful. If the stability result is unstable, or the first distance value is significantly greater than 0, then it is judged as abnormal (i.e., the zeroing was unsuccessful). S244, when the zeroing operation status of the detection equipment starting point is abnormal, perform distance compensation on the pipeline inspection project dataset and determine the compensation distance value.

[0078] It is understandable that when a zeroing anomaly is determined, a systematic distance offset needs to be calculated so that all distance readings can be corrected in subsequent analyses such as defect location to ensure the accuracy of spatial position.

[0079] Optionally, in S244, if the zeroing operation status of the detection equipment's starting point is abnormal, distance compensation is performed on the pipeline inspection project dataset to determine the compensation distance value, including: S2441, when the zeroing operation status of the detection equipment is abnormal, acquire the physical parameters of the equipment; among which, the physical parameters of the equipment include the equipment length parameter and the lens focal length parameter.

[0080] It's understandable that compensation calculations require the physical dimensions of the detection device itself, i.e., its physical parameters. These parameters are stored in a device parameter knowledge base and can be retrieved by detecting the device model. The device length parameter refers to the length of the crawler body, while the lens focal length parameter affects the physical offset between the lens focus and the distance measurement reference point.

[0081] S2442 performs distance compensation calculation based on the device length parameter, lens focal length parameter, and the first distance value of the initial distance sequence to obtain the compensated distance value.

[0082] Understandably, based on the actual physical principles of pipeline inspection, a compensation model can be constructed. For example, the compensated distance value C = the first distance value D0 of the initial distance sequence + the equipment length parameter L + the lens focal length compensation value F. Here, D0 is the measured initial systematic error, and L and F are the inherent physical offsets of the equipment. A total compensation value C is calculated through the compensation model, and this value is subsequently added to all the original distance readings in the dataset to correct for systematic bias.

[0083] S245, based on the compensation distance value and the starting point zeroing operation status, obtain the starting point zeroing verification result.

[0084] It is understandable that it can integrate and output the starting point zeroing verification results, and can include the zeroing operation status (success / abnormal). When the status is abnormal, it will also give the calculated compensation distance value, providing data users with a clear basis for correction.

[0085] For example, the starting point zeroing verification can specifically take the mileage observation within the initial short window as... (Obtained by sparse sampling within the window), let the first valid reading be Stability is assessed by constructing a monotonic confidence mapping using variance. And together with the origin tolerance, it determines the zeroing determination: ,in This is the origin tolerance (0.01 in implementation). This is the stability threshold (0.9 in the implementation). A smaller value indicates a more stable initial reading, thus... The closer to 1. When and When this occurs, it is defined as "positive zero-point offset caused by starting point offset", and the compensation distance is output. Perform a uniform translation correction on the entire mileage sequence: In the formula This indicates whether a clue to "entering the effective metering section within the pipe" has been detected (this can be determined by the scenario stage). (or mileage point support). For equipment geometric length compensation, Fixed offset compensation for viewpoint / focal length. Corrected. This will serve as a unified mileage benchmark for subsequent coverage, crawl rate, and defect location, outputting a triplet. Consecutive scores are acceptable. Evidence package supply , , And its triggering basis (keyframe / timestamp and entry clue) to meet the requirements of verification and auditability.

[0086] In one possible implementation, in step S200, a multi-dimensional quality analysis is performed on the pipeline inspection engineering dataset to obtain the climbing speed compliance inspection results, including: S251, based on the timestamp sequence and distance sequence in the pipeline inspection project dataset, calculate the travel speed of the inspection equipment in the pipeline.

[0087] It is understandable that the speed compliance inspection verifies whether the travel speed of the testing equipment meets the specifications based on the speed standard corresponding to the pipe diameter. Therefore, high-precision timestamps and corresponding encoder distance values ​​are extracted from the dataset to form a synchronized timestamp sequence and distance sequence. To reduce the impact of instantaneous measurement noise, the central difference method can be used to calculate the instantaneous speed: v_i=(d_{i+1}-d_{i-1}) / (t_{i+1}-t_{i-1}). By traversing the sequence, the instantaneous speed sequence for the entire testing process is obtained.

[0088] S252 compares the travel speed with a preset speed threshold based on pipeline attributes to identify speeding violation sections.

[0089] It is understandable that different pipe diameters and materials have corresponding recommended or specified testing speeds (for example, for pipes with a diameter less than 300mm, the recommended speed should not exceed 0.1 m / s). Based on the attributes of the pipe being tested, the corresponding standard speed threshold can be retrieved from the rule base. A sliding window (e.g., a 5-second window) is used to traverse the instantaneous speed sequence. If the average speed within the window consistently exceeds the threshold, the window is marked as a potential speeding violation zone.

[0090] S253 aggregates and statistically analyzes the identified violation sections to generate a climbing speed compliance test result that includes the violation location, duration, and maximum speed.

[0091] It is understandable that sporadic and brief overspeeding flags may occur due to equipment vibration or computational noise. Therefore, based on the principle of spatiotemporal continuity, temporally adjacent or nearby violation points can be aggregated into independent violation events. For each aggregated violation event, its start and end distances (thus obtaining the violation location and duration) are calculated, and the peak speed within the event is recorded. A list of crawl speed compliance verification results is generated, clearly listing the specific parameters of all violation events, as well as the overall speed statistics (such as average speed and overspeed percentage).

[0092] For example, the creepage rate compliance test can be performed by pipe diameter. (from) or header field Define the upper limit function for speed: The observation points are obtained by sampling the video at a fixed step size. ,in The frame number, For mileage readings, In seconds ( (For frame rate). Speed ​​is estimated using first-order difference: To reduce the impact of stalled / obstructed sections, a minimum effective speed threshold is set. Filter near-zero samples and calculate robust mean and peak values: Define the overspeed point set Merge adjacent overspeeding points into a continuous event set. For any event Its duration is: The compliance determination adopts a triple constraint of mean-peak-event: the mean does not exceed the speed limit, and the peak is subject to a coefficient. Restrictions are in place, and sustained speeding incidents are not permitted. Output triples :in Used for a veto in strict mode; continuous scoring in normal mode. can be and Normalized construction. Evidence package. supply , and the start and end times of each speeding incident. Duration With the corresponding mileage segment.

[0093] S300 generates a comprehensive quality evaluation report for pipeline inspection engineering datasets based on the results of authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and creep speed standardization inspection.

[0094] It is understandable that intermediate results generated from five dimensions (which may include structured data, text descriptions, confidence levels, numerical values, lists, etc.) can be summarized. Following a pre-defined report template, all results are organized into a structured comprehensive quality evaluation report. This report may include: 1) an overall quality level (e.g., Excellent / Good / Average / Poor), based on weighted scores for each dimension; 2) detailed evaluation of each item, listing the inspection results, evidence, and specific problems found for each dimension; and 3) problem identification and recommendations, clearly indicating the specific time points or distance segments where quality problems exist in the dataset, and possibly providing operational suggestions (e.g., recommending uniform compensation of Z meters for distance data between X meters and Y meters). The report can be exported in PDF, JSON, or XML formats for easy archiving, sharing, or integration into a larger pipeline asset management system. This enables automated and standardized evaluation of pipeline inspection project datasets and inspection video quality, fundamentally solving the problem of unverifiable evidence in traditional manual judgment and laying the technical foundation for quality auditing.

[0095] To adapt to different scenarios, this application defines two decision-making modes. Zero tolerance mode (strict mode): Zero tolerance is applicable to acceptance, dispute review or high-risk sections. If any verification fails, the entire process fails. ,in This is the indicator function. The zero-tolerance mode reflects the rigid constraints of the standard on continuous measurement, process integrity, and traceability. The quantitative grading mode (standard mode) is suitable for routine sampling and maintenance decisions, using weighted scoring. The overall quality score is defined as: It also allows setting a minimum pass threshold. Key thresholds (e.g., authenticity must be verified): Weight It can be configured according to application scenarios. For example, during acceptance testing, the focus can be on traceability and coverage integrity; during disposal and location testing, the focus can be on coverage and speed compliance; and during operational spot checks, the focus can be on starting point and speed specifications. To achieve efficient processing, a sampling-driven strategy is adopted, performing evidence extraction only on keyframes and sparse sampling points, thus reducing complexity to a minimum. ( (For sampling intervals), to avoid the computational burden of full-frame processing.

[0096] For example, this application divides the output into a decision layer and an evidence layer, with the decision layer outputting five verification results. and the overall score under the normal mode The evidence layer outputs verifiable evidence corresponding to each conclusion. (Such as field conflict list, length deviation, segment / gap interval, starting keyframe, zeroing stability and compensation amount, overspeed events, etc.), and summarized as: The overall decision-making process remains consistent with the dual-mode approach: the stringent mode employs a threshold-based system. The standard model uses weighted scoring. And take equal weight under the default configuration .

[0097] Corresponding to the pipeline inspection engineering data quality assessment method in the above embodiments, this application also provides a pipeline inspection engineering data quality assessment system, wherein each unit of the system can implement each step of the pipeline inspection engineering data quality assessment method. Figure 3 This paper presents a structural block diagram of a pipeline inspection engineering data quality assessment system provided in an embodiment of this application. For ease of explanation, only the parts related to the embodiments of this application are shown.

[0098] Reference Figure 3 The quality assessment system for the pipeline inspection project data includes: The acquisition unit is used to acquire the pipeline inspection project dataset to be evaluated. The analysis unit is used to perform multi-dimensional quality analysis on the pipeline inspection project dataset to obtain the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results. The evaluation unit is used to generate a comprehensive quality evaluation report for the pipeline inspection engineering dataset based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results.

[0099] It should be noted that the information interaction and execution process between the above systems / units are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, and they will not be repeated here.

[0100] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the system can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit module can exist physically separately, or two or more unit modules can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0101] This application also provides an electronic device. Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 4 As shown, the electronic device 6 of this embodiment includes: at least one processor 60 ( Figure 4 Only one is shown in the image), at least one memory 61 ( Figure 4 (Only one is shown in the image) and a computer program 62 stored in the at least one memory 61 and executable on the at least one processor 60, wherein when the processor 60 executes the computer program 62, it causes the electronic device 6 to perform the steps in any of the above embodiments of the quality assessment method for pipeline inspection engineering data, or causes the electronic device 6 to perform the functions of each unit in the above embodiments of the systems.

[0102] For example, the computer program 62 may be divided into one or more units, which are stored in the memory 61 and executed by the processor 60 to complete this application. The one or more units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 62 in the quality assessment method for pipeline inspection engineering data.

[0103] Electronic device 6 can be a computing device or terminal device such as a desktop computer, laptop, handheld computer, or cloud server. This electronic device may include, but is not limited to, a processor 60 and a memory 61. Those skilled in the art will understand that... Figure 4 This is merely an example of electronic device 6 and does not constitute a limitation on electronic device 6. It may include more or fewer components than shown, or combine certain components, or different components, such as input / output devices, network access devices, buses, etc.

[0104] The processor 60 can be a Central Processing Unit (CPU), or it can be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor.

[0105] In some embodiments, the memory 61 may be an internal storage unit of the electronic device 6, such as a hard disk or memory of the electronic device 6. In other embodiments, the memory 61 may be an external storage device of the electronic device 6, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the electronic device 6. Furthermore, the memory 61 may include both internal and external storage units of the electronic device 6. The memory 61 is used to store the operating system, applications, bootloader, data, and other programs, such as the program code of the computer program. The memory 61 can also be used to temporarily store data that has been output or will be output.

[0106] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the above method embodiments.

[0107] This application provides a computer program product that, when run on an electronic device, causes the electronic device to perform the steps in any of the above method embodiments.

[0108] If the integrated unit 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, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to an electronic device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0109] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0110] 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 implementation should not be considered beyond the scope of this application.

[0111] In the embodiments provided in this application, it should be understood that the disclosed pipeline inspection engineering data quality assessment system / electronic device and method can be implemented in other ways. For example, the pipeline inspection engineering data quality assessment system / electronic device 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 displayed or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0112] 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.

[0113] The above-described embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. A method of quality assessment of pipeline inspection engineering data, characterized by, include: Obtain the dataset of pipeline inspection projects to be evaluated; A multi-dimensional quality analysis was performed on the pipeline inspection project dataset to obtain the results of authenticity verification, coverage integrity verification, starting point compliance verification, starting point zeroing verification, and creep speed standardization inspection. Based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and creep speed standardization verification results, a comprehensive quality evaluation report of the pipeline inspection engineering dataset is generated.

2. The method of claim 1, wherein, The multi-dimensional quality analysis of the pipeline inspection project dataset to obtain the authenticity verification results includes: Data sampling was performed on the pipeline inspection project dataset to identify several key data units. Based on all the key data units, determine the text information fields and the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset; The text information field and the distance difference parameter between the first and last data units are compared with the corresponding benchmark data in the preset engineering database to obtain the authenticity verification result.

3. The method as described in claim 2, characterized in that, The step of determining the text information fields and the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset based on all the key data units includes: Perform information extraction operations on all the key data units to obtain the corresponding text recognition results; The corresponding text information fields are extracted from the text recognition results using predefined pattern matching rules; the text information fields include unit name, detection location, pipeline number, date and time; Based on the analysis of all the key data units, the distance difference parameter between the first and last data units corresponding to the pipeline inspection project dataset is determined.

4. The method of claim 1, wherein, A multi-dimensional quality analysis was performed on the pipeline inspection project dataset to obtain the coverage integrity verification results, including: The pipeline inspection project dataset is sampled at equal intervals to obtain the corresponding equally interval data sequence; Extract the corresponding distance values ​​from all the equally spaced data sequences to determine the corresponding distance dataset; Based on the distance dataset, distance change trend data is determined, and corresponding acquisition behavior characteristics are identified through the distance change trend data; wherein, the acquisition behavior characteristics are used to reflect acquisition interruptions, abnormal jumps in distance values, and reverse acquisition behaviors that exist when acquiring the pipeline inspection project dataset; The coverage integrity verification result is obtained based on the characteristics of the acquisition behavior.

5. The method of claim 1, wherein, The multi-dimensional quality analysis of the pipeline inspection project dataset yields the starting point compliance verification results, including: Based on the pipeline inspection project dataset, content parsing was performed to extract the initial data segment; The initial data segment is converted to a specified feature space for threshold segmentation to determine the color distribution feature data of different environmental elements; Target entity identification is performed based on the initial data segment to determine the manhole cover detection data corresponding to the initial data segment; The starting point compliance verification result is obtained based on the manhole cover detection data and the color distribution feature data.

6. The method as described in claim 5, characterized in that, The multi-dimensional quality analysis of the pipeline inspection project dataset, to obtain the starting point zeroing verification results, includes: Based on the initial data segment, the distance value of its initial part data sequence is extracted to obtain the initial distance sequence; Statistical and trend change analysis is performed on the initial distance sequence to determine the corresponding zeroing stability analysis results and reference state transition points; wherein, the reference state transition point is the reference point for the detection equipment to enter the pipeline network from the ground; Based on the zeroing stability analysis results, the reference state transition point, and the first distance value of the initial distance sequence, the zeroing operation state of the detection device is determined. If the zeroing operation of the detection equipment is abnormal, distance compensation is performed on the pipeline inspection project dataset to determine the compensation distance value. Based on the compensation distance value and the starting point zeroing operation status, the starting point zeroing verification result is obtained.

7. The method as described in claim 6, characterized in that, The step of performing statistical and trend change analysis on the initial distance sequence to determine the corresponding zeroing stability analysis results and reference state transition points includes: Statistical and trend analysis was performed on the initial distance sequence to determine the corresponding variance and coefficient of variation; Stability analysis was performed based on the variance and coefficient of variation to obtain the zeroing stability analysis results; The initial distance sequence is subjected to second-order difference calculation to identify the inflection point of distance change in the initial distance sequence, and the inflection point is determined as the reference state transition point.

8. The method as described in claim 6, characterized in that, When the zeroing operation at the starting point of the detection equipment is abnormal, distance compensation is performed on the pipeline inspection project dataset to determine the compensation distance value, including: When the zeroing operation of the detection device's starting point is abnormal, the device's physical parameters are acquired; wherein, the device's physical parameters include the device's length parameter and the lens's focal length parameter; Based on the device length parameter, the lens focal length parameter, and the first distance value of the initial distance sequence, distance compensation calculation is performed to obtain the compensated distance value.

9. A quality evaluation system of pipeline inspection engineering data, characterized by, include: The acquisition unit is used to acquire the pipeline inspection project dataset to be evaluated. The analysis unit is used to perform multi-dimensional quality analysis on the pipeline inspection project dataset to obtain the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results. The evaluation unit is used to generate a comprehensive quality evaluation report for the pipeline inspection engineering dataset based on the authenticity verification results, coverage integrity verification results, starting point compliance verification results, starting point zeroing verification results, and climbing speed standardization inspection results. 10.A pipeline inspection engineering data quality evaluation device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized by, When the processor executes the computer program, it implements the method as described in any one of claims 1 to 8.

Images