Underground pipe cable laying quality on-line detection system based on machine vision

Abstract

The invention discloses an underground pipe cable laying quality online detection system based on machine vision, and belongs to the technical field of pipe cable quality detection. The system comprises a machine vision acquisition module, a data preprocessing module, a quality detection and analysis module, a result feedback module and a database, and all-directional and high-quality acquisition and standardized processing of pipe cable laying data in an underground complex environment are realized through cooperation of multiple modules in combination with a high-definition industrial camera, infrared light supplementing equipment and a self-adaptive position adjusting unit; interference of underground low-illumination, much dust and other environments on detection is effectively avoided, the integrity and accuracy of data acquisition are greatly improved, the labor cost and the operation safety risk are reduced, and through edge detection, gray level co-occurrence matrix and YOLO target detection, the detection accuracy is improved. Multi-dimensional accurate detection is carried out on core indexes such as pipe cable laying position accuracy, connection sealing performance and appearance integrity, accurate identification, positioning and quantitative evaluation of quality defects are achieved, and the problems of missing detection and false detection are effectively avoided.

Description

Technical Field

[0001] This invention relates to the field of pipeline and cable quality inspection technology, and in particular to an online inspection system for the quality of downhole pipeline and cable laying based on machine vision. Background Technology

[0002] Currently, quality inspection of underground cable laying relies heavily on manual inspection. Due to the harsh environment underground, including low lighting, high dust levels, and confined spaces, manual inspection is inefficient, subjective, and prone to missed or false detections. Furthermore, real-time monitoring of the laying process is difficult to achieve, making it impossible to promptly detect quality issues such as positional deviations, poor sealing, and surface damage, potentially leading to safety accidents such as cable wear, leaks, and breakage. Existing automated inspection technologies are mostly designed for surface or conventional environments, lacking adaptability to the complex underground environment. They suffer from unclear data acquisition, limited inspection dimensions, and insufficient defect identification accuracy. Moreover, the inspection results lack effective traceability and analysis mechanisms, making it difficult to form a closed-loop quality control system throughout the entire process. Summary of the Invention

[0003] The purpose of this invention is to provide an online inspection system for the quality of downhole cable laying based on machine vision, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: an online inspection system for the quality of underground cable laying based on machine vision, comprising a machine vision acquisition module, a data preprocessing module, a quality inspection and analysis module, a result feedback module, and a database;

[0005] The machine vision acquisition module is used to acquire image and video data in real time during the laying of underground pipelines and cables.

[0006] The data preprocessing module is used to preprocess the acquired image and video data;

[0007] The quality inspection and analysis module is used to perform multi-dimensional quality inspection and analysis on preprocessed image data and video data based on machine vision algorithms. The multi-dimensional quality inspection indicators include the accuracy of cable laying position, connection sealing and appearance integrity.

[0008] The result feedback module is used to feed back the test results to the control terminal in real time and generate a quality report;

[0009] The database is used to store collected data, preprocessed data, testing standards, and testing results.

[0010] Furthermore, the machine vision acquisition module includes several high-definition industrial cameras, infrared illumination devices, and a position adjustment unit;

[0011] The high-definition industrial cameras are distributed at key monitoring points along the underground cable laying path for the acquisition of underground environmental images and video data.

[0012] The infrared supplementary lighting device is used to provide a lighting environment and automatically adjust the supplementary lighting intensity according to the downhole environment illuminance;

[0013] The position adjustment unit is used to automatically adjust the angle and focal length of a high-definition industrial camera by driving a stepper motor based on the cable laying progress, real-time location information and preset monitoring path.

[0014] Furthermore, the data preprocessing module's processing steps include:

[0015] Gaussian filtering algorithm is used to filter noise in the collected data to remove random noise caused by dust and water vapor in the well; histogram equalization technology is used to enhance the dynamic range of gray levels in the image and improve the contrast between the pipeline and the background.

[0016] Based on preset camera intrinsic and extrinsic parameters and distortion coefficients, geometric calibration is performed on the images to correct perspective distortion and lens distortion; fragmented data collected from multiple perspectives are fused into complete panoramic data of cable laying through image stitching algorithms.

[0017] Furthermore, the method for detecting the cable laying position by the quality inspection and analysis module includes:

[0018] The contour features of the cable laying in the panoramic data were extracted by the Canny edge detection algorithm, and the center axis and position coordinates of the cable were obtained by combining the Hough linear transformation.

[0019] Retrieve preset standard laying path parameters from the database. The laying path parameters include the coordinates of the path centerline, the allowable offset threshold, and the curvature range.

[0020] The real-time detected cable position coordinates are compared point by point with the standard path coordinates to calculate the lateral offset, longitudinal offset, and path curvature deviation.

[0021] When any deviation value exceeds the preset threshold, it is marked as an abnormal position offset, and the mileage information and degree of offset of the abnormal position are recorded.

[0022] Furthermore, the point-by-point comparison and deviation calculation specifically include:

[0023] Establish a unified downhole spatial coordinate system with the starting point of cable laying as the origin, the laying direction as the X-axis, the horizontal direction perpendicular to the laying direction as the Y-axis, and the vertical direction perpendicular to the laying plane as the Z-axis. Map the standard path coordinates and the real-time detected cable position coordinates to this coordinate system.

[0024] The standard path is sampled uniformly according to a preset step size to obtain several standard sampling points. Each standard sampling point corresponds to a unique X-axis mileage coordinate and the corresponding Y and Z-axis standard coordinates.

[0025] Based on the real-time detection of the cable centerline, the corresponding real-time sampling points are extracted at the same X-axis mileage coordinates as the standard sampling points, and the real-time Y and Z-axis coordinates of each real-time sampling point are obtained.

[0026] The lateral offset is obtained by the difference in the Y-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate; the longitudinal offset is obtained by the difference in the Z-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate.

[0027] A standard curve segment is formed by selecting three consecutive adjacent standard sampling points, and the standard curvature of the curve segment is fitted by the least squares method; a real-time curve segment is formed by selecting three corresponding real-time sampling points, and the real-time curvature is obtained by fitting.

[0028] The path curvature deviation is obtained by calculating the absolute value of the difference between the standard curvature and the real-time curvature.

[0029] Furthermore, the method for the quality inspection and analysis module to detect the sealing performance of cable connections includes:

[0030] Using threshold segmentation and region growing algorithms, image regions of cable connection points are extracted from panoramic data of cable laying. The cable connection points include flanges, joints and clips.

[0031] By analyzing the gray-level co-occurrence matrix, the gray-level distribution characteristics of the image area of ​​the cable connection part are taken to identify abnormal features, including connection gaps, missing seals and sealant overflow.

[0032] Based on morphological operations, the abnormal features are dilated and eroded to eliminate interference noise and calculate the area, perimeter and depth of the abnormal region.

[0033] The calculation results are compared with the preset sealing standards in the database. When the area of ​​the abnormal area exceeds the threshold or the depth reaches the preset value, the connection sealing is deemed to be substandard.

[0034] Furthermore, the gray-level distribution characteristics of the cable connection area in the image are analyzed using the gray-level co-occurrence matrix, and abnormal features are identified, specifically including:

[0035] The extracted cable connection area image region is converted to grayscale, and the color image is converted to an 8-bit grayscale image with the grayscale level range set to 0-255.

[0036] Set the generation parameters for the gray-level co-occurrence matrix, taking distances d=1 and 3 pixels, and angles θ=0°, 45°, 90°, and 135°, and generate gray-level co-occurrence matrices at different distances and angles respectively;

[0037] Texture feature parameters are extracted from each gray-level co-occurrence matrix, including energy, entropy, contrast, correlation, and inverse difference moment.

[0038] The system retrieves the standard texture feature parameter threshold range corresponding to the qualified cable connection parts in the database and compares the feature parameters extracted in real time with the standard threshold.

[0039] When the energy is below the standard threshold, the entropy is above the standard threshold, the contrast exceeds the standard range, and the correlation and inverse moment deviate from the standard value by a preset proportion, it is determined that there is an abnormal grayscale distribution in the area.

[0040] By combining the grayscale distribution characteristics of abnormal areas, the type of sealing defect can be correlated;

[0041] If the entropy value of a local area changes abruptly and the contrast increases significantly, it corresponds to a connection gap; if the energy remains low and the inverse moment is abnormal, it corresponds to a missing seal; if the contrast of the edge area increases sharply and the correlation decreases, it corresponds to sealant overflow, thus completing the accurate identification and classification of abnormal features.

[0042] Furthermore, the method for detecting appearance defects in conduits and cables by the quality inspection and analysis module includes:

[0043] A cable appearance defect recognition model is constructed using the YOLO object detection algorithm. The cable appearance defect recognition model is trained and optimized using cable image samples containing defects such as breakage, scratches, deformation, corrosion, and bulging.

[0044] The cable laying panoramic data is scanned segment by segment using a cable appearance defect identification model to identify the defect parameters of various appearance defects, including type, location, quantity and size parameters.

[0045] Based on the defect size, the scope of impact, and the degree of harm to cable performance, a defect level classification standard is established, which includes minor, general, serious, and fatal defects.

[0046] The detected defect parameters are matched with the defect level standard, and the defect locations of severe and above defect levels are marked.

[0047] Furthermore, the result feedback module includes a real-time early warning unit and a report generation unit;

[0048] When the real-time early warning unit detects a quality anomaly, it alerts staff through audible and visual alarms, pop-up windows on the control terminal, and SMS notifications, and simultaneously pushes the location, type, level, and handling suggestions of the anomaly.

[0049] The report generation unit automatically summarizes the test data, anomaly information, handling records, and quality assessment results to generate a standardized quality test report that includes text descriptions, image evidence, and data statistics.

[0050] Furthermore, the database is a distributed storage architecture, divided into standard database, real-time database and historical database according to data type, supporting fast data retrieval, tracing and batch export.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] 1. This invention, through multi-module collaboration, combines a high-definition industrial camera, infrared supplementary lighting equipment, and an adaptive position adjustment unit to achieve comprehensive, high-quality acquisition and standardized processing of cable laying data in complex underground environments. It overcomes the limitations of manual inspection, effectively avoids interference from low-light and dusty environments underground, significantly improves the integrity and accuracy of data acquisition, and automates the inspection process, thereby significantly improving inspection efficiency and reducing labor costs and operational safety risks.

[0053] 2. This invention utilizes edge detection, grayscale co-occurrence matrix, and YOLO target detection to perform multi-dimensional and precise detection of core indicators such as cable laying location accuracy, connection sealing, and appearance integrity. It combines a unified coordinate system and point-by-point comparison method to quantify deviations, establish a defect level classification and anomaly classification mechanism, and achieve accurate identification, location, and quantitative assessment of quality defects. This effectively avoids missed and false detections, significantly improves detection accuracy, and can promptly identify various quality hazards during the laying process, providing data support for precise handling and ensuring the safety and standardization of cable laying.

[0054] 3. This invention enables rapid response and handling of quality anomalies through multi-mode real-time early warning and standardized report generation in the result feedback module, as well as the classification storage and traceability functions of the distributed database. It provides staff with intuitive and comprehensive quality data, facilitating the focus on key issues and efficient handling. It supports long-term data traceability and reuse, providing data support for quality review, standard optimization, and system algorithm iteration, forming a continuously optimized quality control system that adapts to the quality control needs of different downhole cable laying scenarios. Attached Figure Description

[0055] Figure 1 This is a schematic diagram of the online inspection process for the quality of underground cable laying according to the present invention. Detailed Implementation

[0056] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0057] Please see Figure 1 The present invention provides the following technical solutions:

[0058] The machine vision-based online inspection system for the quality of downhole cable laying includes a machine vision acquisition module, a data preprocessing module, a quality inspection and analysis module, a result feedback module, and a database.

[0059] The machine vision acquisition module is used to acquire image and video data in real time during the laying of underground cables;

[0060] The data preprocessing module is used to preprocess the acquired image and video data;

[0061] The quality inspection and analysis module is used to perform multi-dimensional quality inspection and analysis on preprocessed image and video data based on machine vision algorithms. The multi-dimensional quality inspection indicators include the accuracy of cable laying position, connection sealing and appearance integrity.

[0062] The results feedback module is used to feed back the test results to the control terminal in real time and generate a quality report;

[0063] The database is used to store collected data, preprocessed data, testing standards, and testing results, enabling data traceability and retrieval.

[0064] In the above embodiments, a complete online inspection system for the quality of underground cable laying is constructed through modular design. The machine vision acquisition module ensures the integrity and clarity of data acquisition in complex underground environments, the quality inspection and analysis module accurately covers core quality indicators, the result feedback module realizes real-time early warning and standardized report output, and the database supports data traceability and reuse. This system overcomes the limitations of low efficiency and strong subjectivity of manual inspection, and achieves automation, intelligence and real-time quality inspection through machine vision technology. This significantly improves inspection accuracy and efficiency, reduces safety risks in underground operations, and provides comprehensive and reliable technical support for the quality control of cable laying.

[0065] The machine vision acquisition module includes several high-definition industrial cameras, infrared illumination devices, and a position adjustment unit;

[0066] High-definition industrial cameras are distributed at key monitoring points along the underground cable laying path for the acquisition of underground environmental images and video data. They support image and video acquisition with resolutions of 1080P and above and frame rates of 30fps and above, and have the characteristics of being dustproof, waterproof and shockproof for underground use.

[0067] Infrared supplementary lighting equipment is used to provide lighting environment and automatically adjust the supplementary lighting intensity according to the downhole environment illumination to ensure the clarity of data acquisition under low illumination.

[0068] The position adjustment unit is used to automatically adjust the angle and focal length of a high-definition industrial camera driven by a stepper motor based on the progress of cable laying, real-time position information and preset monitoring path, to adapt to position changes during the cable laying process.

[0069] The data preprocessing module's processing steps include:

[0070] Gaussian filtering algorithm is used to filter noise in the collected data to remove random noise caused by dust and water vapor in the well; histogram equalization technology is used to enhance the dynamic range of gray levels in the image and improve the contrast between the pipeline and the background.

[0071] Based on preset camera intrinsic and extrinsic parameters and distortion coefficients, geometric calibration is performed on the images to correct perspective distortion and lens distortion; fragmented data collected from multiple perspectives are fused into complete panoramic data of cable laying through image stitching algorithms.

[0072] In the above embodiments, the combination of a high-definition industrial camera, an infrared supplementary lighting device, and a position adjustment unit specifically addresses the impact of harsh environments such as low illumination, high dust levels, and complex spaces underground on data acquisition. The high resolution and high frame rate of the high-definition industrial camera ensure the integrity of data acquisition details, while its dustproof, waterproof, and impact-resistant design adapts to the underground working environment. The adaptive supplementary lighting function of the infrared supplementary lighting device ensures the clarity of acquisition under different illumination levels. The automatic adaptation capability of the position adjustment unit dynamically optimizes the acquisition angle and focal length according to the progress of cable laying, achieving comprehensive and high-quality acquisition of underground cable laying data. Through multi-step technical means, the acquired data is optimized and processed to effectively eliminate noise interference and image distortion caused by the underground environment, providing accurate and comprehensive raw data support for subsequent quality inspection and analysis, avoiding detection errors caused by insufficient data acquisition, and improving the system's adaptability and stability to complex underground environments.

[0073] The quality inspection and analysis module implements the following methods for detecting the location of cable laying:

[0074] The contour features of the cable laying in the panoramic data were extracted by the Canny edge detection algorithm, and the center axis and position coordinates of the cable were obtained by combining the Hough linear transformation.

[0075] Retrieve preset standard laying path parameters from the database. The laying path parameters include the centerline coordinates of the path, the allowable offset threshold, and the curvature range.

[0076] The real-time detected cable position coordinates are compared point by point with the standard path coordinates to calculate the lateral offset, longitudinal offset, and path curvature deviation.

[0077] When any deviation value exceeds the preset threshold, it is marked as an abnormal position offset, and the mileage information and degree of offset of the abnormal position are recorded.

[0078] The point-by-point comparison and deviation calculation specifically include:

[0079] Establish a unified downhole spatial coordinate system with the starting point of cable laying as the origin, the laying direction as the X-axis, the horizontal direction perpendicular to the laying direction as the Y-axis, and the vertical direction perpendicular to the laying plane as the Z-axis. Map the standard path coordinates and the real-time detected cable position coordinates to this coordinate system.

[0080] The standard path is sampled uniformly according to a preset step size to obtain several standard sampling points. Each standard sampling point corresponds to a unique X-axis mileage coordinate and the corresponding Y and Z-axis standard coordinates.

[0081] Based on the real-time detection of the cable centerline, the corresponding real-time sampling points are extracted at the same X-axis mileage coordinates as the standard sampling points, and the real-time Y and Z-axis coordinates of each real-time sampling point are obtained.

[0082] The lateral offset is obtained by the difference in the Y-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate; the longitudinal offset is obtained by the difference in the Z-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate.

[0083] A standard curve segment is formed by selecting three consecutive adjacent standard sampling points, and the standard curvature of the curve segment is fitted by the least squares method; a real-time curve segment is formed by selecting three corresponding real-time sampling points, and the real-time curvature is obtained by fitting.

[0084] The path curvature deviation is obtained by calculating the absolute value of the difference between the standard curvature and the real-time curvature.

[0085] In the above embodiments, the detection of cable laying location achieves multi-dimensional accurate determination of position deviation through precise algorithm combination and standardized process. Edge detection and Hough line transformation ensure accurate extraction of cable outline and central axis. The establishment of a unified spatial coordinate system enables comparability between standard path and real-time position. Point-by-point comparison and deviation calculation methods comprehensively cover lateral and longitudinal offset and path curvature deviation, breaking through the limitations of single-dimensional position detection. Through multi-dimensional deviation analysis, the compliance of cable laying location is accurately reflected. Furthermore, the accurate location of abnormal positions is achieved through mileage information recording, ensuring real-time monitoring and accurate control of cable laying location, and effectively avoiding problems such as cable wear and connection failure caused by position deviation.

[0086] The quality inspection and analysis module implements the following methods for detecting the sealing performance of cable connections:

[0087] Using threshold segmentation and region growing algorithms, image regions of cable connection points are extracted from panoramic data of cable laying. Cable connection points include flanges, joints and clips.

[0088] By analyzing the gray-level co-occurrence matrix, the gray-level distribution characteristics of the image area of ​​the cable connection part are taken to identify abnormal features, including connection gaps, missing seals and sealant overflow.

[0089] Based on morphological operations, the abnormal features are dilated and eroded to eliminate interference noise and calculate the area, perimeter and depth of the abnormal region.

[0090] The calculation results are compared with the preset sealing standards in the database. When the area of ​​the abnormal area exceeds the threshold or the depth reaches the preset value, the connection sealing is deemed to be substandard.

[0091] Specifically, the gray-level distribution characteristics of the cable connection area image region are analyzed using the gray-level co-occurrence matrix to identify abnormal features, including:

[0092] The extracted cable connection area image region is converted to grayscale, and the color image is converted to an 8-bit grayscale image with the grayscale level range set to 0-255.

[0093] The generation parameters of the gray-level co-occurrence matrix are set, with distances d=1 and 3 pixels and angles θ=0°, 45°, 90° and 135°, respectively, to generate gray-level co-occurrence matrices at different distances and angles, comprehensively characterizing the spatial correlation of gray-level distribution in the region;

[0094] Texture feature parameters are extracted from each gray-level co-occurrence matrix. The texture feature parameters include energy, entropy, contrast, correlation, and inverse moment: energy reflects the uniformity of gray-level distribution, entropy reflects the disorder of gray-level information, contrast reflects the difference between gray levels, correlation reflects the correlation between adjacent gray levels, and inverse moment reflects the uniformity of texture.

[0095] The system retrieves the standard texture feature parameter threshold range corresponding to the qualified cable connection parts in the database and compares the feature parameters extracted in real time with the standard threshold.

[0096] When the energy is below the standard threshold, the entropy is above the standard threshold, the contrast exceeds the standard range, and the correlation and inverse moment deviate from the standard value by a preset proportion, it is determined that there is an abnormal grayscale distribution in the area.

[0097] By combining the grayscale distribution characteristics of abnormal areas, the type of sealing defect can be correlated;

[0098] If the entropy value of a local area changes abruptly and the contrast increases significantly, it corresponds to a connection gap; if the energy remains low and the inverse moment is abnormal, it corresponds to a missing seal; if the contrast of the edge area increases sharply and the correlation decreases, it corresponds to sealant overflow, thus completing the accurate identification and classification of abnormal features.

[0099] In the above embodiments, the quality inspection and analysis module detects the sealing performance of cable connections through a precise process of region extraction, feature analysis, and anomaly judgment. This enables accurate identification and classification of sealing defects. Threshold segmentation and region growing algorithms ensure accurate extraction of connection parts. The gray-level co-occurrence matrix characterizes gray-level distribution features from multiple dimensions. Morphological operations are combined to eliminate interference and quantify abnormal parameters. Finally, sealing performance is determined through standard comparison. For the types of sealing defects in cable connections, a precise correlation between features and defects is established, enabling accurate identification and quantification of anomalies such as connection gaps, missing seals, and sealant overflow. The design offers higher detection accuracy and greater targeting, effectively avoiding safety hazards caused by sealing defects.

[0100] The quality inspection and analysis module implements the following methods for detecting appearance defects in cables:

[0101] A cable appearance defect recognition model was constructed using the YOLO object detection algorithm. The model was trained and optimized using cable image samples containing defects such as breakage, scratches, deformation, corrosion, and bulging.

[0102] The cable laying panoramic data is scanned segment by segment using a cable appearance defect identification model to identify the defect parameters of various appearance defects, including type, location, quantity and size parameters.

[0103] Based on the defect size, the scope of impact, and the degree of harm to cable performance, a defect classification standard is established, which includes minor, general, serious, and fatal defects.

[0104] The detected defect parameters are matched with the defect level standard, and the defect locations of severe and above defect levels are marked.

[0105] In the above embodiments, the quality inspection and analysis module uses the YOLO target detection algorithm to build a dedicated recognition model for detecting appearance defects in conduits and cables. The model's recognition capability is improved through defect sample training and optimization. The model can quickly and accurately identify various appearance defects on the surface of conduits and cables, and extract key parameters such as defect type, location, and size. Combined with defect level classification standards, it realizes quality level determination and key defect marking, achieving automated and intelligent detection of appearance defects. Compared with manual visual inspection, it significantly improves the efficiency and accuracy of defect identification, avoiding the subjectivity and missed or false detection problems of manual inspection. At the same time, the defect level classification and key marking functions make it easy for staff to quickly focus on key issues, prioritize the handling of serious defects, improve the pertinence and efficiency of quality control, and ensure the overall quality of conduit and cable laying.

[0106] The results feedback module includes a real-time early warning unit and a report generation unit;

[0107] When the real-time early warning unit detects a quality anomaly, it alerts staff through audible and visual alarms, pop-up windows on the control terminal, and SMS notifications, and simultaneously pushes the location, type, level, and handling suggestions of the anomaly.

[0108] The report generation unit automatically summarizes the test data, abnormal information, handling records and quality assessment results, and generates a standardized quality test report that includes text descriptions, image evidence and data statistics. It supports filtering and exporting by laying section, time period, cable type and other dimensions.

[0109] The database uses a distributed storage architecture and is divided into standard database, real-time database, and historical database according to data type, supporting fast data retrieval, tracing, and batch export.

[0110] In the above embodiments, the result feedback module, through the collaborative design of the real-time early warning unit and the report generation unit, achieves rapid response to quality anomalies and standardized output of test results. The multi-mode reminder function of the real-time early warning unit ensures that staff are promptly informed of abnormal information, and the synchronously pushed anomaly details and handling suggestions provide support for rapid handling. The report generation unit integrates multi-dimensional data to generate standardized reports, supports multi-dimensional filtering and export, realizes real-time feedback and traceability management of test results, ensures timely handling of quality anomalies, avoids the expansion of defects, and provides comprehensive and standardized data support for subsequent evaluation and review of cable laying quality. The database adopts a distributed storage architecture and is classified and managed according to data type. The standard library stores test standards and parameter thresholds, the real-time library caches recent data to ensure fast access, and the historical library stores data for long-term traceability. The multi-mode early warning and flexible export functions improve the practicality and ease of use of the system and adapt to the quality control needs of different scenarios.

[0111] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A machine vision-based online inspection system for the quality of downhole cable laying, characterized in that, It includes a machine vision acquisition module, a data preprocessing module, a quality inspection and analysis module, a result feedback module, and a database; The machine vision acquisition module is used to acquire image and video data in real time during the laying of underground pipelines and cables. The data preprocessing module is used to preprocess the acquired image and video data; The quality inspection and analysis module is used to perform multi-dimensional quality inspection and analysis on preprocessed image data and video data based on machine vision algorithms. The multi-dimensional quality inspection indicators include the accuracy of cable laying position, connection sealing and appearance integrity. The result feedback module is used to feed back the test results to the control terminal in real time and generate a quality report; The database is used to store collected data, preprocessed data, testing standards, and testing results.

2. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The machine vision acquisition module includes several high-definition industrial cameras, infrared supplementary lighting equipment, and a position adjustment unit. The high-definition industrial cameras are distributed at key monitoring points along the underground cable laying path for the acquisition of underground environmental images and video data. The infrared supplementary lighting device is used to provide a lighting environment and automatically adjust the supplementary lighting intensity according to the downhole environment illuminance; The position adjustment unit is used to automatically adjust the angle and focal length of a high-definition industrial camera by driving a stepper motor based on the cable laying progress, real-time location information and preset monitoring path.

3. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The data preprocessing module's processing steps include: Gaussian filtering algorithm is used to filter noise in the collected data to remove random noise caused by dust and water vapor in the well; histogram equalization technology is used to enhance the dynamic range of gray levels in the image and improve the contrast between the pipeline and the background. Based on preset camera intrinsic and extrinsic parameters and distortion coefficients, geometric calibration is performed on the images to correct perspective distortion and lens distortion; fragmented data collected from multiple perspectives are fused into complete panoramic data of cable laying through image stitching algorithms.

4. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The quality inspection and analysis module implements the following methods for detecting the cable laying location: The contour features of the cable laying in the panoramic data were extracted by the Canny edge detection algorithm, and the center axis and position coordinates of the cable were obtained by combining the Hough linear transformation. Retrieve preset standard laying path parameters from the database. The laying path parameters include the coordinates of the path centerline, the allowable offset threshold, and the curvature range. The real-time detected cable position coordinates are compared point by point with the standard path coordinates to calculate the lateral offset, longitudinal offset, and path curvature deviation. When any deviation value exceeds the preset threshold, it is marked as an abnormal position offset, and the mileage information and degree of offset of the abnormal position are recorded.

5. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 4, characterized in that, Point-by-point comparison and deviation calculation specifically include: Establish a unified downhole spatial coordinate system with the starting point of cable laying as the origin, the laying direction as the X-axis, the horizontal direction perpendicular to the laying direction as the Y-axis, and the vertical direction perpendicular to the laying plane as the Z-axis. Map the standard path coordinates and the real-time detected cable position coordinates to this coordinate system. The standard path is sampled uniformly according to a preset step size to obtain several standard sampling points. Each standard sampling point corresponds to a unique X-axis mileage coordinate and the corresponding Y and Z-axis standard coordinates. Based on the real-time detection of the cable centerline, the corresponding real-time sampling points are extracted at the same X-axis mileage coordinates as the standard sampling points, and the real-time Y and Z-axis coordinates of each real-time sampling point are obtained. The lateral offset is obtained by the difference in the Y-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate; the longitudinal offset is obtained by the difference in the Z-axis coordinates between the real-time sampling point and the standard sampling point under the same X-axis mileage coordinate. A standard curve segment is formed by selecting three consecutive adjacent standard sampling points, and the standard curvature of the curve segment is fitted by the least squares method; a real-time curve segment is formed by selecting three corresponding real-time sampling points, and the real-time curvature is obtained by fitting. The path curvature deviation is obtained by calculating the absolute value of the difference between the standard curvature and the real-time curvature.

6. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The quality inspection and analysis module implements the following methods for detecting the sealing performance of cable connections: Using threshold segmentation and region growing algorithms, image regions of cable connection points are extracted from panoramic data of cable laying. The cable connection points include flanges, joints and clips. By analyzing the gray-level co-occurrence matrix of the image area of ​​the cable connection, abnormal features are identified, including connection gaps, missing seals, and sealant overflow. Based on morphological operations, the abnormal features are dilated and eroded to eliminate interference noise and calculate the area, perimeter and depth of the abnormal region. The calculation results are compared with the preset sealing standards in the database. When the area of ​​the abnormal area exceeds the threshold or the depth reaches the preset value, the connection sealing is deemed to be substandard.

7. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 6, characterized in that, The gray-level co-occurrence matrix analysis was used to analyze the gray-level distribution characteristics of the cable connection area in the image and identify abnormal features, specifically including: The extracted cable connection area image region is converted to grayscale, and the color image is converted to an 8-bit grayscale image with the grayscale level range set to 0-255. Set the generation parameters for the gray-level co-occurrence matrix, taking distances d=1 and 3 pixels, and angles θ=0°, 45°, 90°, and 135°, and generate gray-level co-occurrence matrices at different distances and angles respectively; Texture feature parameters are extracted from each gray-level co-occurrence matrix, including energy, entropy, contrast, correlation, and inverse difference moment. The system retrieves the standard texture feature parameter threshold range corresponding to the qualified cable connection parts in the database and compares the feature parameters extracted in real time with the standard threshold. When the energy is below the standard threshold, the entropy is above the standard threshold, the contrast exceeds the standard range, and the correlation and inverse moment deviate from the standard value by a preset proportion, it is determined that there is an abnormal grayscale distribution in the area. By combining the grayscale distribution characteristics of abnormal areas, the type of sealing defect can be correlated; If the entropy value of a local area changes abruptly and the contrast increases significantly, it corresponds to a connection gap; if the energy remains low and the inverse moment is abnormal, it corresponds to a missing seal; if the contrast of the edge area increases sharply and the correlation decreases, it corresponds to sealant overflow, thus completing the accurate identification and classification of abnormal features.

8. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The quality inspection and analysis module implements the following methods for detecting appearance defects in pipelines and cables: A cable appearance defect recognition model is constructed using the YOLO object detection algorithm. The cable appearance defect recognition model is trained and optimized using cable image samples containing defects such as breakage, scratches, deformation, corrosion, and bulging. The cable laying panoramic data is scanned segment by segment using a cable appearance defect identification model to identify the defect parameters of various appearance defects, including type, location, quantity and size parameters. Based on the defect size, the scope of impact, and the degree of harm to cable performance, a defect level classification standard is established, which includes minor, general, serious, and fatal defects. The detected defect parameters are matched with the defect level standard, and the defect locations of severe and above defect levels are marked.

9. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The result feedback module includes a real-time early warning unit and a report generation unit; When the real-time early warning unit detects a quality anomaly, it alerts staff through audible and visual alarms, pop-up windows on the control terminal, and SMS notifications, and simultaneously pushes the location, type, level, and handling suggestions of the anomaly. The report generation unit automatically summarizes the test data, anomaly information, handling records, and quality assessment results to generate a standardized quality test report that includes text descriptions, image evidence, and data statistics.

10. The machine vision-based online inspection system for the quality of downhole cable laying as described in claim 1, characterized in that, The database is a distributed storage architecture, divided into standard database, real-time database and historical database according to data type, and supports fast data retrieval, tracing and batch export.

Images