Natural gas in-service station three-dimensional modeling method and system based on laser point cloud scanning

Abstract

The invention relates to the field of three-dimensional modeling, and discloses a natural gas in-service station three-dimensional modeling method and system based on laser point cloud scanning, which are used for providing support for fine management and safe operation and maintenance of a natural gas in-service station. Comprising the following steps: establishing a field station facility feature rule base, performing geometric feature recognition and semantic segmentation on structured point cloud to obtain a semantic point cloud data set, extracting key geometric parameters from the semantic point cloud data set, generating optimized modeling parameters by combining standard facility specifications, constructing and verifying a facility component model, and establishing a spatial topological relation according to the model. And space assembly and topology reconstruction are completed, and a three-dimensional operation and maintenance model is generated through integration. According to the method, high-precision three-dimensional reconstruction of the station facilities is realized, the problems of insufficient precision, low efficiency, update lag and the like of a traditional modeling method are solved, and the digital management level of the station is remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling, and in particular to a method and system for 3D modeling of in-service natural gas stations based on laser point cloud scanning. Background Technology

[0002] In the natural gas industry, in-service natural gas stations serve as crucial hubs for natural gas transmission, storage, and distribution, and their safe and efficient operation is of paramount importance. With the continuous growth in natural gas demand and the increasing complexity of station facilities, more stringent requirements are being placed on the precise management of these facilities, operational and maintenance decision support, and safety risk prevention and control. Traditional station management methods primarily rely on two-dimensional drawings and manual inspections. These methods have significant limitations in terms of the intuitiveness, comprehensiveness, and real-time nature of information expression, making it difficult to meet the demands of modern, refined station management.

[0003] In recent years, 3D modeling technology has been increasingly widely applied in the industrial field, bringing new ideas to site management. By constructing 3D models of sites, the layout of facilities, spatial relationships, and equipment status can be displayed intuitively and comprehensively, providing maintenance personnel with more convenient and efficient management tools. However, most existing 3D modeling methods for sites are based on manual measurement or traditional photogrammetry techniques, which have many shortcomings in practical applications. Manual measurement is not only labor-intensive and inefficient, but its accuracy is also greatly affected by the skill level of personnel and the measurement environment, making it difficult to guarantee the accuracy and consistency of data. Although traditional photogrammetry techniques have improved data acquisition efficiency to some extent, their ability to capture facility details in complex scenes is limited, especially in poor lighting conditions or with occlusion, making it difficult to obtain high-quality point cloud data, thus affecting the accuracy and completeness of the 3D model.

[0004] Furthermore, existing 3D modeling methods often lack the effective extraction and utilization of semantic information about facilities. The generated models only possess geometric shape information and cannot accurately identify and label the type, function, and attributes of the facilities, thus limiting the application value of the models. During operation and maintenance management, maintenance personnel need to spend a significant amount of time manually interpreting and analyzing the models, making it difficult to quickly obtain key information and affecting the timeliness and accuracy of operation and maintenance decisions. At the same time, existing 3D models lack dynamic update mechanisms and cannot reflect changes in site facilities in a timely manner, resulting in discrepancies between the model and the actual scene, reducing the model's credibility and practicality.

[0005] Therefore, we propose a method and system for 3D modeling of in-service natural gas stations based on laser point cloud scanning to solve the above problems. Summary of the Invention

[0006] This invention provides a method and system for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning, which can support the refined management and safe operation and maintenance of in-service natural gas stations.

[0007] The first aspect of this invention provides a method for 3D modeling of in-service natural gas stations based on laser point cloud scanning. The method includes: performing multi-level scanning of the in-service station to acquire multi-source point cloud data, including macroscopic point clouds of the overall station layout and fine point clouds of key facilities; processing the multi-source point cloud data to generate structured point cloud data in a unified coordinate system; performing geometric feature recognition and semantic segmentation on the structured point cloud data based on a preset station facility feature rule library to generate a semantic point cloud dataset; generating facility component models based on the semantic point cloud dataset using a parametric modeling method, wherein the parametric modeling combines standard facility specifications and the actual dimensions extracted from the point cloud; and performing spatial assembly and topological relationship reconstruction based on the facility component models to integrate and obtain a 3D operable and maintainable model.

[0008] Optionally, in a first implementation of the first aspect of the present invention, the method includes: formulating a multi-level scanning scheme based on the site layout plan and facility distribution plan, including mobile scanning path planning and fixed scanning station layout; performing continuous scanning along a predetermined path according to the multi-level scanning scheme to obtain macroscopic point cloud data of the overall site layout; identifying the location and range of key facility areas based on the macroscopic point cloud data, performing multi-angle fine scanning of the key facilities to obtain fine point cloud data of the key facilities; processing the macroscopic point cloud data and fine point cloud data to generate a preliminary registered point cloud dataset; and performing point cloud density equalization and overlapping area optimization processing on the preliminary registered point cloud dataset to generate multi-source point cloud data.

[0009] Optionally, in the second implementation of the first aspect of the present invention, the method includes: establishing a station facility feature rule library based on station facility engineering standards and a geometric feature library; extracting geometric features from the structured point cloud data based on the station facility feature rule library to generate a feature parameter set, including normal vectors, curvature, and size parameters; identifying pipe components using a cylinder fitting algorithm, identifying valve components using bounding box detection, and identifying tank components using surface fitting based on the feature parameter set, generating preliminary classification results for facility components; generating verified facility classification data based on the preliminary classification results for facility components; and assigning corresponding semantic labels to each facility component based on the verified facility classification data to generate a semantic point cloud dataset.

[0010] Optionally, in the third implementation of the first aspect of the present invention, the following steps are taken: First, station design drawings, equipment specifications, and engineering acceptance data are collected. Standard geometric parameters and spatial relationship constraints of the facilities are extracted to generate a standard feature library for station facilities. Second, based on historical point cloud data processing experience, geometric feature recognition rules and parameter extraction methods for different types of facilities are summarized to generate a set of facility recognition experience rules. Third, the standard feature library for station facilities is fused and optimized with the set of facility recognition experience rules to generate an initial station facility feature rule library. Fourth, based on typical station sample point cloud data, the initial station facility feature rule library is tested for recognition effectiveness and its parameters are adjusted to generate an optimized station facility feature rule library. Fifth, a version management and update mechanism for the station facility feature rule library is established to continuously improve the rule content based on actual recognition results, generating the final station facility feature rule library system.

[0011] Optionally, in the fourth implementation of the first aspect of the present invention, the method includes: extracting key geometric parameters of various facility components based on the semantic point cloud dataset, generating a measured size dataset of the facility components, including the diameter and direction of pipes, the connection dimensions of valves, and the diameter and height parameters of storage tanks; matching the corresponding standard facility model according to the measured size dataset to generate a standard facility parameter set; generating optimized facility modeling parameters based on the measured size dataset and the standard facility parameter set; generating a three-dimensional geometric model of the facility components by calling a parametric modeling engine according to the optimized facility modeling parameters, generating a preliminary facility component model; and comparing the spatial position and verifying the geometric accuracy of the preliminary facility component model with the semantic point cloud dataset to generate a final facility component model.

[0012] Optionally, in the fifth implementation of the first aspect of the present invention, the method includes: establishing spatial topological relationship data between facilities based on the spatial location and engineering connection relationship of the facility component model; performing spatial assembly and connection relationship reconstruction on the facility component model according to the spatial topological relationship data to generate a preliminary integrated three-dimensional assembly model of the station; performing collision detection and clearance analysis between facilities based on the three-dimensional assembly model of the station to generate a safety space analysis report; integrating the safety space analysis report with the three-dimensional assembly model of the station to generate a three-dimensional operable model; and performing spatial feasibility verification of the renovation plan and simulation analysis of the maintenance path based on the three-dimensional operable model to generate station operation and maintenance decision support data.

[0013] Optionally, in a sixth implementation of the first aspect of the present invention, the safety space analysis report is integrated with the three-dimensional assembly model of the site to generate a three-dimensional operable and maintainable model, wherein the safety performance index is SEI: The weighting coefficients are assigned values ​​according to the risk level.

[0014] Optionally, in the seventh implementation of the first aspect of the present invention, the method further includes: scanning the changed areas of the in-service facility to obtain incremental point cloud data of the changed areas; spatially registering and comparing the incremental point cloud data with an existing three-dimensional operable model to generate a facility change analysis report; based on the facility change analysis report, performing partial updates and version management on the three-dimensional operable model to generate an updated version of the three-dimensional operable model; associating and storing the three-dimensional operable model with the corresponding facility operation and maintenance records to generate a facility full lifecycle three-dimensional model database; and based on the facility full lifecycle three-dimensional model database, performing facility evolution analysis and renovation effect evaluation to generate a facility evolution analysis report.

[0015] Optionally, in the eighth implementation of the first aspect of the present invention, based on the spatial distribution characteristics of the incremental point cloud data, feature matching is performed with the corresponding region of the three-dimensional operable model to generate registered incremental point cloud data; point-by-point distance calculation and geometric feature comparison are performed on the registered incremental point cloud data and the three-dimensional operable model to generate preliminary change detection results; based on a preset facility change type rule base, the preliminary change detection results are identified and classified for change type to generate facility change classification results; the facility change classification results are verified for engineering rationality and manually confirmed to generate facility change data; based on the facility change data, combined with facility engineering attributes and maintenance records, a site facility change analysis report is generated.

[0016] A second aspect of this invention provides a 3D modeling system for in-service natural gas stations based on laser point cloud scanning. The system comprises: a scanning module for performing multi-level scanning of the in-service station to acquire multi-source point cloud data, including macroscopic point clouds of the overall station layout and fine point clouds of key facilities; a processing module for processing the multi-source point cloud data to generate structured point cloud data in a unified coordinate system; a segmentation module for performing geometric feature recognition and semantic segmentation on the structured point cloud data based on a preset station facility feature rule library to generate a semantic point cloud dataset; a modeling module for generating facility component models based on the semantic point cloud dataset using a parametric modeling method, wherein the parametric modeling combines standard facility specifications and the actual dimensions extracted from the point cloud; and an integration module for performing spatial assembly and topological relationship reconstruction based on the facility component models to obtain a 3D operable and maintainable model.

[0017] The mechanism of this invention is as follows: a maintainable three-dimensional model is formed through spatial assembly and topology reconstruction, and a dynamic update mechanism is introduced to realize the continuous synchronization between the model and the actual state of the site, which breaks through the limitations of traditional modeling such as incomplete data, manual segmentation, and lagging model updates. Beneficial effects: The multi-level scanning and optimized data processing methods ensure high-quality data acquisition and processing, the parametric modeling and rigorous verification process guarantee the high accuracy of the model, and the multi-level scanning scheme improves the overall scanning efficiency and shortens the modeling cycle. The three-dimensional operational model integrates multiple functions such as spatial assembly, topology relationship, safety analysis and operation and maintenance decision support, which can provide comprehensive and intuitive information support for the daily operation and maintenance, safety management and transformation decision of the site, and improve the efficiency of operation and maintenance management. Through incremental point cloud data processing and model update mechanisms, changes in site facilities can be reflected in a timely manner, ensuring the timeliness and accuracy of the model. The full lifecycle 3D model database provides historical data support for long-term site planning, facility evolution analysis, and evaluation of renovation effects, which helps to optimize site operation and management strategies. The introduction of safety space analysis reports and safety performance indices makes site safety management more scientific and quantifiable. Verification of modification schemes and simulation analysis of maintenance paths based on three-dimensional operational models can identify potential problems in advance, optimize schemes and paths, reduce safety risks, and improve the accuracy and reliability of decision-making. Accurate 3D models and comprehensive operation and maintenance information support help reduce the workload and errors of manual inspections, detect facility failures and safety hazards in advance, carry out timely maintenance and repairs, reduce operation and maintenance costs and safety risks, and ensure the safe and stable operation of the site. Attached Figure Description

[0018] Figure 1 This is a schematic diagram of an embodiment of the three-dimensional modeling method for in-service natural gas stations based on laser point cloud scanning in this invention. Figure 2 A schematic diagram illustrating multi-level scanning of a natural gas station using a mobile laser scanning vehicle and a fixed scanner; Figure 3 This is a schematic diagram of another embodiment of the three-dimensional modeling method for in-service natural gas stations based on laser point cloud scanning in this invention. Figure 4 This is a schematic diagram of an embodiment of the three-dimensional modeling system for in-service natural gas stations based on laser point cloud scanning in this invention. Figure 5 This is a schematic diagram of an embodiment of a 3D modeling device for in-service natural gas stations based on laser point cloud scanning, as described in this invention. Detailed Implementation

[0019] This invention provides a method and system for 3D modeling of in-service natural gas stations based on laser point cloud scanning, supporting refined management and safe operation and maintenance of in-service natural gas stations. The terms "first," "second," "third," "fourth," etc. (if present) in the specification, claims, and accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments described herein can be implemented in a sequence other than that illustrated or described herein. Furthermore, the terms "comprising" or "having" and any variations thereof are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or device that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0020] For ease of understanding, the specific process of the embodiments of the present invention is described below. Please refer to [link / reference]. Figure 1 One embodiment of the three-dimensional modeling method for in-service natural gas stations based on laser point cloud scanning in this invention includes: 101. Multi-level scanning of in-service sites is carried out using mobile laser scanning systems and fixed scanning equipment to obtain multi-source point cloud data containing complete geometric features of the facilities. The multi-source point cloud data includes macro point clouds of the overall layout of the site and fine point clouds of key facilities. It is understood that the executing entity of this invention can be a 3D modeling system for in-service natural gas stations based on laser point cloud scanning, or it can be a terminal or a server; the specific implementation is not limited here. This embodiment of the invention will be described using a server as an example.

[0021] It should be noted that, in order to achieve 3D modeling of an in-service natural gas gate station covering an area of ​​approximately 0.4 square kilometers, the project team adopted a collaborative operation scheme that integrates mobile laser scanning systems and fixed scanning equipment, aiming to efficiently acquire complete geometric information from the overall layout to key facilities.

[0022] Three types of scanning equipment are configured to meet the acquisition needs of different scales and precisions: A vehicle-mounted mobile laser scanning system: used to quickly acquire macroscopic point clouds of the entire site. This system integrates a high-precision laser scanner and an IMU / GNSS combined navigation system. It is planned to travel along a preset route at a speed not exceeding 20 km / h, and can complete data acquisition of the main roads in the site within 2 hours. The point cloud density is expected to reach 200 points per square meter, meeting the measurement needs of the overall layout and macroscopic dimensions. A drone-based LiDAR system: for areas within the site that are difficult for personnel to access or pose safety risks, such as the tops of tall storage tanks and complex overhead pipelines, aerial surveying is conducted using airborne LiDAR such as the DJI Zenmuse L1. The drone will fly along a preset flight path at an altitude of 50 meters and a scanning frequency of 160 kHz to ensure sufficient point cloud density to identify pipeline routes and the outlines of large equipment, effectively compensating for blind spots in ground scanning. A ground-based fixed 3D laser scanner: using high-performance equipment such as the Faro or Leica RTC360, to perform detailed scanning of facilities in key process areas such as valve groups, metering skids, and compressors. The scanner will be deployed at various locations around key facilities to ensure that the overlap between adjacent stations exceeds 60%, and target spheres will be placed on critical equipment to achieve high-precision automatic stitching of multi-site point cloud data. A single-station scan will take approximately 3 minutes, and the overall plan is to deploy about 80 stations to ensure the integrity and millimeter-level accuracy of the point cloud data for critical facilities.

[0023] A site survey was conducted to identify scanning boundaries and key areas. Next, the vehicle-mounted system traveled along the planned route, rapidly acquiring large-scale point cloud and panoramic images of the site. Simultaneously, a drone was launched to scan the superstructure, including the tank area and pipe corridors. Finally, under stationary conditions, a ground-based fixed scanner performed multi-angle, detailed scans of core facilities such as the pressure regulating skid and valve area, and targets were deployed around the equipment to assist in subsequent data registration.

[0024] Through the aforementioned collaborative work, the following multi-source point cloud datasets were ultimately obtained: Macro-level point cloud: Collected by vehicle-mounted and UAV systems, forming a comprehensive point cloud covering the entire site with absolute coordinates, clearly reflecting the spatial distribution of buildings, roads, main pipelines, and large storage tanks within the site. Fine-grained point cloud: Acquired by a ground-based fixed scanner, achieving millimeter-level precision dense point cloud coverage for key facilities, clearly revealing the minute corrosion conditions on valve flanges, bolts, and even pipe surfaces, providing a solid foundation for subsequent refined modeling.

[0025] 102. Perform registration, fusion, and noise filtering on multi-source point cloud data to generate structured point cloud data in a unified coordinate system, where the structured point cloud data serves as input data for facility identification and segmentation. It should be noted that after completing data collection, the project team obtained three types of core point cloud data: the overall site point cloud (approximately 800 million points, accuracy ±5cm) acquired by a vehicle-mounted mobile laser scanning system; the superstructure point cloud (approximately 300 million points, mainly covering tank tops and pipe corridors) acquired by UAV LiDAR; and the detailed point cloud of key facilities (approximately 120 stations, each with an accuracy of ±2mm) acquired by a ground-based fixed scanner. These data suffer from inconsistencies in coordinate systems, uneven density, and various types of noise.

[0026] The core of registration is to unify all data into the same world coordinate system. The project adopts a coarse-to-fine strategy: coarse registration: using control point targets deployed within the site (15 high-precision targets were evenly deployed throughout the site) and the stable geometric features of the facilities themselves (the center of the tank, the corner points of the valve group), the vehicle-mounted point cloud and the UAV point cloud are initially aligned to the absolute coordinate system. After the initial registration, the average deviation of the corresponding points is controlled within 10 centimeters.

[0027] Fine registration: For fine point clouds scanned by ground fixed stations, due to their low overlap and large accuracy differences with macro point clouds, a feature-based registration method is adopted. Specifically, significant features such as the circular outline of valve flanges and the straight axis of pipelines are automatically extracted from the fine point cloud and matched with corresponding features in the macro point cloud. Based on this, an iterative nearest-point algorithm is applied for fine optimization, so that the point cloud registration error of the critical facility area is ultimately better than 5 mm.

[0028] To address the noise in massive point clouds, the project team adopted a layered processing strategy: Large-scale outlier removal: Statistical filtering algorithms were used to analyze the average distance between each point and its 50 nearest neighbors. A Gaussian model was established for the distance distribution, and points with a mean distance exceeding 3 standard deviations (isolated points caused by birds or temporary personnel movement) were identified as outliers and removed. Small-scale noise smoothing: For high-frequency noise in the point cloud data caused by dust or slight equipment vibration, a bilateral filtering algorithm was used. This algorithm can smooth the data while preserving the edge and detail information of critical parts such as pipe welds and valve threads. Data simplification: To balance data volume and detail, voxel downsampling was performed on the point cloud in non-critical areas (open ground). A 2cm cube grid was used to resample the point cloud, retaining only one representative point within each grid, significantly improving data processing efficiency.

[0029] After the above processing, the originally scattered, multi-source, and noisy point clouds were merged into a structured point cloud dataset in a unified coordinate system. The dataset contains approximately 2.5 billion valid points, and its overall accuracy meets the requirements of engineering applications.

[0030] 103. Based on the preset site facility feature rule base, perform geometric feature recognition and semantic segmentation on the structured point cloud data to generate a semantic point cloud dataset with facility type labels. The semantic point cloud dataset includes classification point clouds of facilities such as pipelines, valves, and storage tanks. It should be noted that, in order to achieve 3D modeling of a natural gas gate station covering an area of ​​approximately 0.4 square kilometers, after completing point cloud registration and denoising, facility identification and semantic segmentation need to be performed on the structured point cloud (approximately 2.5 billion points) in a unified coordinate system. Using a geometric feature rule base and deep learning algorithms, the point cloud data is transformed into a semantic dataset with facility type labels. The specific implementation is as follows: A feature rule base for facility construction was established. The rule base includes the following predefined rules based on industry standards and typical facility geometric features: Pipelines: Cylindrical model, diameter range 0.1–1.5 meters, aspect ratio >10, continuous point cloud curvature features. Valves: Composite geometry (flange is a circular plane, valve body is a cube or sphere), diameter matches the pipeline, high-density point cloud exists at the connection. Storage tanks: Regular curved surface (cylindrical or spherical), diameter 5–30 meters, height 3–15 meters, point cloud density at the top is significantly higher than on the sides. Feature parameters: A total of 12 facility types and over 200 geometric rules (curvature threshold, dimensional tolerance, spatial topology relationships) were defined.

[0031] The multi-level semantic segmentation process uses the PointNet++ network framework to perform coarse segmentation of the entire point cloud. The model is pre-trained on a dataset containing 5000 labeled samples (94% accuracy), directly extracting basic geometric features (planes, curved surfaces, cylinders) from the point cloud, and initially classifying the point cloud into major categories such as "pipes", "equipment", and "buildings".

[0032] Based on the preliminary classification results, secondary optimization was performed using a feature rule base: Pipeline identification: A cylinder was fitted using the RANSAC algorithm to extract point clouds of pipelines with a diameter of 0.3 meters and a length of 50 meters. Points with abnormal curvature (bending sections) were removed and labeled as "DN300 gas transmission pipeline". Valve identification: The valve area point cloud (approximately 2 million points per station) was segmented using Euclidean clustering to generate independent point clusters. The normal vector and geometric center of each point cluster were calculated, and valve templates in the rule base (flange roundness > 0.9, valve body size error < 5%) were matched, labeling them as "ball valve" or "gate valve". Tank segmentation: Based on a global context awareness algorithm, the dome surface of the tank top and the cylindrical surface of the side walls were identified. Combined with the point cloud density distribution (top point spacing < 2 cm, side wall point spacing < 5 cm), it was labeled as "5000m". 3 Liquefied gas storage tank.

[0033] Semantic label verification and fusion conflict resolution: For facilities with overlapping segmentation boundaries (pipe and valve connections), a dual attention mechanism module is used; point cloud labels are weighted and allocated to ensure clear boundaries.

[0034] Global consistency check: Verify topological relationships (valves must be connected to pipelines, and storage tanks must have supporting foundations), correct incorrect annotations, and if an isolated "valve" point cloud has no pipeline connection, it is reclassified as "other equipment".

[0035] The final generated semantic point cloud dataset includes: facility classification point clouds: pipelines (12 km), valves (86), storage tanks (4), compressors (3), etc., with each point labeled (type, size, material). Data accuracy: the average intersection-over-union (mIoU) ratio for facility classification reaches 91.5%, and the segmentation error of key components is <3 cm.

[0036] 104. Based on semantic point cloud datasets, a facility component model with engineering dimensional accuracy is generated through parametric modeling methods, where parametric modeling combines standard facility specifications and actual dimensions extracted from point clouds; It should be noted that, in order to achieve 3D modeling of an in-service natural gas gate station covering an area of ​​approximately 0.4 square kilometers, after completing the semantic segmentation of the point cloud (obtaining a classified point cloud containing 12 types of facilities such as pipelines, valves, and storage tanks), it is necessary to convert the semantic point cloud into a facility component model with engineering accuracy using parametric modeling methods. The core of this step is to combine the actual dimensions extracted from the point cloud with a standard facility specification library to achieve rapid and accurate facility modeling. Standard specification library construction: Based on industry-standard drawing sets (GB / T standards), a parametric template library for site facilities is established. Valves are classified by type (ball valve, gate valve), pressure rating (PN1.6MPa, PN4.0MPa), and size (DN50-DN300). Storage tanks are classified by volume (1000m³). 3 -5000m 3 The system is classified by structure (vertical cylindrical, spherical) and predefined key parameters (diameter, height, connection flange size) for each facility.

[0037] Point cloud size extraction: Automatically calculate the actual dimensions of facilities from semantic point clouds: Pipes: Fit a cylindrical point cloud using the RANSAC algorithm to extract the diameter (DN200 pipe, actual measured diameter in the point cloud is 219mm) and length (point cloud length of one pipe segment is 12.5m). Valves: Identify characteristic dimensions such as flange spacing and valve body height (point cloud of a ball valve shows a flange center distance of 350mm and a valve body height of 480mm). Storage tanks: Fit a curved point cloud surface to calculate the tank diameter (tank diameter extracted from the point cloud is 12.8m) and cylinder height (8.2m).

[0038] The parameter-driven modeling process matches the dimensions extracted from the point cloud with the standard specification library and verifies their rationality. For example, the flange center distance of a valve point cloud is 350mm, which deviates from the standard PN4.0MPa, DN200 ball valve specification (standard center distance is 355mm) by only 1.4%, which is within the allowable tolerance (±5%). Therefore, the standard model is directly adopted. If the deviation exceeds the limit (the point cloud shows a pipe wall thickness of 12mm, but the standard DN200 pipe wall thickness is 7.1mm), it is marked as abnormal and requires manual review.

[0039] Using PML or similar parametric tools (Revit family library, SolidWorks driver), input the matched parameters into the preset model template to automatically generate facility components. (Based on a 5000m...) 3 Taking a vertical storage tank as an example: Template call: Select the storage tank template, input the diameter (12.8m) and height (8.2m) extracted from the point cloud, and the program will automatically calculate parameters such as the tank wall thickness (standard is 12mm) and the curvature of the end caps. Component generation: The model is composed of basic bodies (cylindrical body, dished end caps, and supports), and the dimensions of each component are linked through parameters (the support height is adjusted synchronously with the tank height).

[0040] Based on the point cloud to supplement non-standard details, the point cloud of a certain valve shows that its handwheel has local rust and dents. After generating a standard handwheel model, the program adds dent features (about 2mm deep) to the model surface through point cloud mesh deviation analysis to ensure the geometric consistency between the model and the actual object.

[0041] The following table 1 shows examples of parametric modeling for some facilities, illustrating the integration process of point cloud data with standard specifications: Table 1 The generated facility component models have engineering-grade precision (dimensional errors are generally less than 3mm) and retain parametric attributes (modifying the diameter can automatically update the associated structure). The ladder platform model of the storage tank is generated based on the actual spacing of the point cloud, rather than relying entirely on the standard spacing, thus avoiding conflicts with the site.

[0042] All models are associated with semantic tags (valve model, pressure rating), supporting subsequent topology assembly and operation and maintenance analysis.

[0043] 105. Spatial assembly and topological relationship reconstruction of facility component models are performed to generate a three-dimensional operable model containing complete geometric information of the facilities and safety space analysis results. The three-dimensional operable model supports site renovation planning and maintenance scheme verification.

[0044] It should be noted that, in order to integrate the 3D operational model of the natural gas gate station, which covers an area of ​​approximately 0.4 square kilometers, this step requires spatial assembly and topological reconstruction of the facility components (pipelines, valves, storage tanks, etc.) generated by parametric modeling, and embedding safety analysis capabilities. Spatial assembly and topology reconstruction, spatial positioning assembly: All facility component models (12 km pipeline, 86 valves, 4 storage tanks) are equipped with semantic tags and engineering coordinates (accuracy ±3mm). Through a spatial matching algorithm, components are automatically positioned in the global coordinate system: pipelines are precisely aligned with valve flange faces using endpoint coordinates (flange center coordinates extracted from the point cloud: X=1023.5m, Y=88.2m, Z=1.05m). The tank foundations are aligned with the center of the foundation extracted from the point cloud, with verticality error controlled within ±5mm.

[0045] Topology Reconstruction: A topology network is constructed based on the connection logic between facilities: Pipelines and valves are automatically associated through interface relationships, generating pipeline-valve-pipeline connection paths. Process System Hierarchy: The main pipeline network (pressure level 4.0MPa) and auxiliary pipelines (sewage pipes, fire water pipes) are managed in layers to ensure the correctness of fluid simulation logic.

[0046] Integrated safety space analysis and safety distance checks: Employing bounding box detection technology (AABB / OBB algorithm), high-risk areas are automatically analyzed. A 15m radius firebreak is established around the storage tank. If an illegally placed object is detected (only 8m from the tank), the system automatically marks it as a red alert. Valve operation passage width is required to be ≥1.2m. If the model detects a passageway of only 0.9m in a certain area, a maintenance reminder is triggered.

[0047] Maintenance space simulation: Virtual inspection robot path planning verifies whether the hoisting space meets the maintenance needs of large equipment (compressor), simulates the crane entering the compressor room, dynamically detects the turning radius and pipe spacing (≥2m), finds a pipeline interference (spacing 1.3m), and generates a modification plan.

[0048] Embedded Operation and Maintenance Functions: The model integrates a real-time monitoring interface, supporting the following applications: Retrofit Planning: When adding new pipeline routes, the model automatically performs collision detection to avoid conflicts with existing structures (avoiding underground cable corridors). Emergency Drills: Simulates natural gas leak scenarios, dynamically displaying the gas diffusion range (a 50m radius danger zone) and planning evacuation routes. Maintenance Verification: Virtually disassembles valves to verify tool operating space (0.6m radius required for wrench rotation), identifying operational obstacles in advance.

[0049] Table 2 shows a comparison of key analytical indicators. The following are examples of core indicators for safe space analysis: Table 2 The generated 3D operable model covers full-field geometric information (2.5 billion triangular facets) and safety analysis results, and supports: dynamic management: associating with IoT sensors (pressure, temperature) to update equipment status in real time.

[0050] In this embodiment of the invention, a mobile laser scanning system, an unmanned aerial vehicle (UAV) LiDAR system, and a ground-based fixed 3D laser scanner are integrated to collect data for different scales and precision requirements. The vehicle-mounted system rapidly acquires the overall macroscopic point cloud of the site, the UAV fills in the blind spots of the ground scanning, and the ground-based fixed scanner achieves millimeter-level precision scanning of key facilities, comprehensively and efficiently acquiring complete geometric information from the overall layout to key facilities. A coarse-to-fine registration strategy is adopted, using control point targets and facility geometric features for coarse registration, followed by fine registration based on feature extraction and iterative nearest-point algorithms. This effectively solves the problems of inconsistent coordinate systems, low overlap, and large precision differences in multi-source point clouds, ensuring that the point cloud registration error for key facility areas is better than 5 millimeters. For noise in the massive point cloud, a layered processing strategy is adopted. Statistical filtering algorithms are used to filter out large-scale outliers, and bilateral filtering algorithms are used to smooth small-scale noise. Simultaneously, data simplification is performed, preserving detailed information of key parts while improving data processing efficiency. The final result is a structured point cloud dataset with approximately 2.5 billion valid points and precision meeting engineering application requirements. All facility component models carry semantic tags and engineering coordinates, and are automatically located in the global coordinate system through spatial matching algorithms. A topology network is constructed based on the connection logic between facilities to achieve precise spatial assembly of facilities and reconstruction of correct topological relationships. The system integrates safety space analysis functions, uses bounding box detection technology to check safety distances, simulates maintenance spaces, embeds operation and maintenance functions, supports applications such as renovation planning, emergency drills, and maintenance verification, and generates a three-dimensional operable model that covers the entire site's geometric information and safety analysis results, providing strong support for the dynamic management of the site.

[0051] Please see Figures 2-3 Another embodiment of the three-dimensional modeling method for in-service natural gas stations based on laser point cloud scanning in this invention includes: 201. Multi-level scanning of in-service sites is carried out using mobile laser scanning systems and fixed scanning equipment to obtain multi-source point cloud data containing complete geometric features of the facilities. The multi-source point cloud data includes macro point clouds of the overall layout of the site and fine point clouds of key facilities. Specifically, based on the site layout plan and facility distribution map, a multi-level scanning scheme is formulated, which includes mobile scanning path planning and fixed scanning station layout. The multi-level scanning scheme serves as the basis for scanning execution. According to the multi-level scanning scheme, a vehicle-mounted mobile laser scanning system continuously scans along a predetermined path to obtain macroscopic point cloud data of the overall site layout. The macroscopic point cloud data contains the spatial location and contour information of the site facilities. Based on the macroscopic point cloud data, the location and range of key facility areas are identified, and multi-angle fine scanning is performed on the key facilities using fixed scanning equipment to obtain fine point cloud data of the key facilities. The fine point cloud data contains surface details and geometric features of connection parts of the facilities. The macroscopic point cloud data and fine point cloud data are time-stamped and initially aligned with coordinate systems to generate a pre-registered point cloud dataset, which is used for subsequent data fusion. The pre-registered point cloud dataset is then subjected to point cloud density equalization and overlapping area optimization to generate multi-source point cloud data containing complete geometric features of the facilities. The multi-source point cloud data serves as input data for subsequent point cloud processing steps.

[0052] It should be noted that this comprehensive scanning project for a natural gas processing plant covers an area of ​​approximately 5 hectares, and its core facilities include a separation zone, compressor units, storage tank area, and a complex network of process pipelines. The project employs a multi-level scanning strategy: based on the plant layout plan, a closed-loop vehicle-mounted mobile scanning path with a total length of approximately 2.5 kilometers was planned to ensure coverage of all main process channels. Thirty-two fixed scanning stations were pre-selected around 15 key facilities, including compressors and large storage tanks, as fine-tuning scanning points.

[0053] Macroscopic point cloud acquisition: An off-road vehicle equipped with a Riegl VUX-240 laser scanning system travels along a predetermined path at a speed of 15 kilometers per hour. The system emits 1 million laser points per second, and the points are calculated in real time using a SLAM (Simultaneous Localization and Mapping) algorithm. This results in the acquisition of a macroscopic point cloud dataset covering the entire plant, totaling approximately 1.5 billion points with a point density of 500-800 points per square meter. This data clearly reflects the overall outline and relative position of each facility, the length and width dimensions of the compressor unit plant (30m × 15m), and its spatial relationship with the main pipelines.

[0054] Fine-scale point cloud scanning: Based on macroscopic point clouds, 12 key areas requiring fine-scale modeling were identified, including compressor inlet and outlet valve assemblies and pipe flange connections. Subsequently, these areas were scanned at multiple stations using a FaroFocusPremium fixed scanner. Eight scanning stations were set up around a single compressor, with each station scanning for approximately 6 minutes, ultimately yielding a fine-scale point cloud with a point density of up to 20,000 points per square meter, clearly capturing the texture of valve handwheels and the geometric features of flange bolts.

[0055] Both macroscopic and fine-grained point clouds recorded GPS timestamps. In the preprocessing stage, initial synchronization was performed based on GPS time. This was achieved by extracting seven common feature points from both types of point clouds (top of the storage tank escalator, corner of the factory wall, etc.) for initial coordinate system alignment. Due to the low density of the moving scan point cloud, density equalization was applied to overlapping areas to stabilize the overall density at approximately 2000 points per square meter, generating a multi-source point cloud dataset of approximately 1.8 billion points.

[0056] 202. Perform registration, fusion, and noise filtering on multi-source point cloud data to generate structured point cloud data in a unified coordinate system, where the structured point cloud data serves as input data for facility identification and segmentation. Specifically, feature point extraction and matching are performed on multi-source point cloud data to generate preliminary registered point cloud data, which is used for subsequent precise registration. Based on the preliminary registered point cloud data, precise registration is optimized through an iterative nearest-point algorithm to generate precisely registered point cloud data, which is used for noise filtering. Statistical outlier filtering is performed on the precisely registered point cloud data to remove noise and outliers, generating denoised point cloud data, which is used for coordinate system unification. Coordinate system transformation and spatial index construction are performed on the denoised point cloud data to generate structured point cloud data in a unified coordinate system, which serves as input data for facility identification and segmentation.

[0057] It should be noted that the macroscopic point cloud (approximately 1.5 billion points, with an average density of 500 points / square meter) obtained by vehicle-mounted mobile scanning is fused with the fine point cloud (with a maximum density of 20,000 points / square meter) of 12 key facilities obtained by fixed scanning.

[0058] A multi-feature fusion algorithm was employed to extract feature points from two types of point clouds. For the compressor region, the algorithm identified 1245 feature points in the macroscopic point cloud and 28670 feature points in the fine point cloud, based on local depth and normal angle features. Through feature descriptor matching, 538 reliable feature point pairs were initially generated, and a preliminary transformation matrix was calculated. The fine point cloud was then preliminarily aligned to the macroscopic coordinate system, with the registration error initially controlled within approximately 15 cm.

[0059] Precise registration optimization: Based on the preliminary registration results, the iterative nearest point algorithm is applied for precise registration. This algorithm iteratively calculates the nearest point pairs and optimizes the transformation matrix. After 50 iterations, the registration error between the macro point cloud and the fine point cloud is reduced to an average of less than 3 mm, and the registration accuracy of key connection parts (flanges) reaches 1.5 mm, meeting the accuracy requirements of engineering modeling.

[0060] Noise filtering was performed on the precisely registered and merged point cloud (approximately 1.8 billion points). A statistical outlier filtering algorithm was used to analyze the 20 nearest neighbors of each point and calculate the average distance. A standard deviation factor of 2.0 was set as a threshold to filter out outliers with abnormal distance distributions. This step removed approximately 0.8% of the total point cloud, including noise and outliers (approximately 14.4 million points), effectively eliminating interference points caused by personnel movement, vehicle dust, and equipment reflections during scanning.

[0061] The denoised point cloud was uniformly transformed to a station-independent coordinate system (with the station's central control point as the origin), and an octree spatial index was constructed. This index divides the point cloud space into cubic voxels with a side length of 0.5 meters, enabling efficient management of massive point clouds and supporting rapid spatial queries for subsequent facility identification algorithms.

[0062] 203. Based on the preset site facility feature rule base, perform geometric feature recognition and semantic segmentation on the structured point cloud data to generate a semantic point cloud dataset with facility type labels, wherein the semantic point cloud dataset contains classification point clouds of facilities such as pipelines, valves, and storage tanks; Specifically, based on the engineering standards and geometric feature library of the facility, a feature rule library for the facility is established. This library contains geometric feature descriptions and recognition rules for various types of facilities, serving as the basis for geometric feature recognition. Based on this feature rule library, geometric features are extracted from structured point cloud data to generate a feature parameter set containing normal vectors, curvature, and size parameters. This feature parameter set is used for facility component recognition. Based on the feature parameter set, pipe components are identified using a cylinder fitting algorithm, valve components are identified using bounding box detection, and storage tank components are identified using surface fitting, generating preliminary classification results for facility components. The preliminary classification results are then subjected to topological relationship verification and spatial location correction to generate verified facility classification data, which is used for semantic label assignment. Based on the verified facility classification data, a corresponding semantic label is assigned to each facility component, generating a semantic point cloud dataset with facility type labels. This semantic point cloud dataset serves as input data for parametric modeling.

[0063] Furthermore, based on the engineering standards and geometric feature library of the facility, a feature rule library for the facility is established. This includes: collecting facility design drawings, equipment specifications, and engineering acceptance data; extracting standard geometric parameters and spatial relationship constraints of the facilities; generating a standard feature library for the facility, which contains the theoretical dimensions and installation specifications of various facilities; summarizing geometric feature recognition rules and parameter extraction methods for different types of facilities based on historical point cloud data processing experience; generating a set of facility recognition experience rules, which includes feature extraction thresholds and recognition priority settings; merging and optimizing the standard feature library and the set of facility recognition experience rules to generate an initial feature rule library for the facility, which is used for rule verification; testing the recognition effect and optimizing the parameters of the initial feature rule library based on typical facility sample point cloud data to generate an optimized feature rule library for the facility, which is used for geometric feature recognition; and establishing a version management and update mechanism for the feature rule library to continuously improve the rule content based on actual recognition results, generating a self-improving feature rule library system for the facility.

[0064] It should be noted that, for the structured point cloud data (approximately 1.8 billion points, already registered and denoised) of a certain natural gas processing plant, automatic classification and labeling of facilities such as pipelines, valves, and storage tanks were achieved through geometric feature recognition and semantic segmentation: Based on the design drawings of this site (nominal pipe diameter DN200-DN500, valve model Z41H-16C, storage tank volume 1000m³) 3 Based on historical point cloud processing experience, a feature rule base was established. The rule base clearly defines the geometric features and recognition thresholds for various facilities: pipeline curvature threshold <0.05, valve bounding box aspect ratio range 1.2-1.8, and storage tank surface fitting residual <2cm. Key rules are shown in Table 3 below: Table 3 Geometric feature extraction and component recognition: Feature extraction: Calculate the normal vector (neighborhood radius 5cm) and curvature features of each point in the point cloud, and extract the size parameters (the estimated diameter of the pipe segment point cloud is DN300, with an error of ±1.5cm).

[0065] Component Identification: Pipelines: Using a random sampling consensus algorithm to fit cylinders, 85 pipelines were identified, with a total length of approximately 2km and a diameter identification accuracy of 98%. Valves: 42 valve components were identified through bounding box detection, and verified using flange hole spacing (standard spacing 15cm) to eliminate false targets. Storage Tanks: 6 storage tanks were identified based on surface fitting, with the top spherical crown curvature radius deviating from the design value (12m) by less than 1%.

[0066] Topology verification and semantic annotation: Verify the connection relationship between pipes and valves (a distance of less than 10cm between the pipe end and the valve center is considered a valid connection), and correct 3 topology errors caused by point cloud occlusion.

[0067] Semantic labels “process pipeline-DN300” and “ball valve-Z41H-16C” were assigned to each component to generate a semantic point cloud dataset, with an overall classification accuracy of 95.2% (based on manual sampling verification).

[0068] 204. Based on semantic point cloud datasets, a facility component model with engineering dimensional accuracy is generated through parametric modeling methods, where parametric modeling combines standard facility specifications and actual dimensions extracted from point clouds; Specifically, based on the semantic point cloud dataset, key geometric parameters of various facility components are extracted to generate a measured dimension dataset of the facility components. This dataset includes parameters such as pipe diameter and orientation, valve connection dimensions, and tank diameter and height. Based on the measured dimension dataset, corresponding standard facility models are matched from a standard facility specification library to generate a standard facility parameter set, which contains standard dimensions and engineering specifications. Based on the measured dimension dataset and the standard facility parameter set, parameters are compared and specifications are adjusted to generate optimized facility modeling parameters, which are used for parametric modeling. Based on the optimized facility modeling parameters, a parametric modeling engine is invoked to generate a 3D geometric model of the facility components, creating a preliminary facility component model, which is used for accuracy verification. The preliminary facility component model is then compared spatially with the semantic point cloud dataset for geometric accuracy verification, generating a final facility component model with engineering dimensional accuracy. This final facility component model is used for the integration and assembly of the 3D operable model.

[0069] Furthermore, based on the measured size dataset and the standard facility parameter set, parameter comparison and specification adjustment are performed to generate optimized facility modeling parameters. This includes: statistical analysis of the measured size dataset to identify abnormal size data exceeding the standard tolerance range and generating a size anomaly report, which is used for parameter adjustment decisions; based on the size anomaly report and in conjunction with the facility installation allowable deviation specifications, a parameter adjustment plan is formulated, which includes size correction values ​​and adjustment priorities; according to the parameter adjustment plan, the measured dimensions exceeding the allowable deviation are progressively adjusted to generate a preliminary optimized modeling parameter set, which is used for model generation; the preliminary optimized modeling parameter set is input into the parametric modeling engine for trial modeling, generating a trial model and extracting its key dimensions to generate trial model size data; the trial model size data is compared a second time with the measured size dataset to verify the parameter optimization effect and generate validated optimized facility modeling parameters.

[0070] It should be noted that, based on the semantic point cloud dataset of a certain natural gas processing plant (which has been generated through step 203 and includes classified point clouds and labels for facilities such as pipelines, valves, and storage tanks), a high-precision 3D model is generated using parametric modeling methods: Key geometric parameter extraction: Measured dimensions of various facilities were extracted from the semantic point cloud: Pipelines: Diameter and direction were extracted using a cylinder fitting algorithm. The average diameter of a process pipeline after point cloud fitting was measured to be 308 mm (standard nominal diameter DN300), with a direction angle deviation of <0.5°. Valves: External dimensions were obtained through bounding box detection. The flange spacing of a gate valve was measured to be 155 mm (standard value 150 mm) from the point cloud, with a bolt hole center distance deviation of ±2 mm. Storage tanks: Diameter and height were calculated based on surface fitting. The point cloud fitting of a liquefied natural gas storage tank yielded a diameter of 12.02 meters and a height of 20.15 meters (design value 12.00 meters × 20.00 meters). After generating the measured dimension dataset, abnormal data was identified: A section of pipeline with a measured diameter of 290 mm (exceeding the DN300 tolerance range of ±5 mm) was marked as a dimensional anomaly.

[0071] Standard specification matching and parameter optimization: Match measured dimensions with the standard facility specification library (the piping standard library includes ANSI / ASME B36.10 specifications). When adjusting parameters, use the tolerance verification formula: The tolerance rate is set according to the facility type (1.5% for pipelines and 2% for valves). Adjustment plans are developed for abnormal data, as shown in Table 4 below: Table 4 The optimized parameters were verified through trial modeling: the pipe diameter of 302 mm was input into the parametric modeling engine (the parametric module of FreeCAD), and after generating the trial model, its diameter was extracted to be 301.5 mm. The error between the experimental point cloud and the actual diameter was less than 1 mm, confirming the effectiveness of the parameters.

[0072] The parametric modeling engine (based on the 3DMAX template library) is invoked, and the optimized parameters (pipe diameter 302 mm, valve flange spacing 155 mm) are input to automatically generate a preliminary facility component model.

[0073] Spatial comparison was performed between the model and the semantic point cloud: the average distance error between the tank model and the point cloud was 3.2 mm, which is lower than the allowable error of 5 mm; the point cloud fitting residual for valve connection parts was <1 mm. Local parameter fine-tuning was performed on areas (pipe bends) that did not meet the accuracy requirements to generate facility component models that meet engineering accuracy (error <0.1%).

[0074] 205. Spatial assembly and topological relationship reconstruction of facility component models are performed to generate a three-dimensional operable model containing complete geometric information of the facilities and safety space analysis results. The three-dimensional operable model supports site renovation planning and maintenance scheme verification.

[0075] Specifically, based on the spatial location and engineering connection relationships of the facility component models, spatial topological relationship data between facilities is established. This data includes pipeline connection relationships, equipment adjacency relationships, and spatial constraint information. Based on this data, the facility component models are spatially assembled and their connection relationships reconstructed, generating a preliminary integrated 3D assembly model of the site. This model is used for safety space analysis. Based on this model, collision detection and clearance analysis are performed between facilities, generating a safety space analysis report. This report includes information on collision points and areas with insufficient clearance. The safety space analysis report is integrated with the 3D assembly model to generate a 3D operable and maintainable model containing safety warning information. This model is used for retrofit planning analysis. Based on this model, spatial feasibility verification of the retrofit plan and simulation analysis of maintenance paths are conducted, generating site operation and maintenance decision support data containing retrofit suggestions and maintenance paths.

[0076] It should be noted that, for a certain natural gas processing plant's facility components (including 85 pipelines, 42 valves, and 6 storage tanks) that have already undergone parametric modeling, a three-dimensional operable model supporting operation and maintenance decisions is generated through spatial assembly, topology reconstruction, and safety analysis. Based on the engineering connections of facility components (flange connections between pipes and valves, pipeline connections between storage tanks and pumps), spatial coordinates and direction vectors of the components are extracted to establish a topological relationship dataset: Pipeline connection relationships: DN300 main pipe is connected to 8 ball valves, with interface coordinate accuracy error <2cm. Equipment adjacency constraints: The standard distance between the compressor and the buffer tank is 3 meters, while the actual point cloud measurement value is 2.95 meters. Spatial constraints: The width of the maintenance passage is required to be ≥1.5 meters, and the minimum passage width calibrated in the model is 1.45 meters.

[0077] The component models are imported into the 3D engine (Unity3D), and connection points are automatically aligned based on the topology data. Through coordinate system transformation (rigid body transformation matrix), the center of the pipe flange is matched with the valve interface, with the maximum assembly deviation controlled within 5mm. The initially integrated assembly model contains 1,247 component units, covering five functional areas including the process area and storage area.

[0078] Collision Detection: Bounding box algorithm was used to detect interference between facilities. Three collision points were found, including interference between a pipe and a support (overlap depth 12mm), and two collisions between cable trays and ventilation ducts. Clearance Analysis: The minimum clearance height for maintenance work was defined as 2.2 meters. Analysis revealed a pipeline above the compressor unit only 2.1 meters above the ground, generating a high-risk warning. The safety analysis results are summarized in Table 5 below: Table 5 Integrate the safety analysis report with the assembly model to generate a maintainable model with early warning information. Define the Safety Performance Index (SEI) to evaluate the renovation plan: The weighting coefficients are assigned according to risk level (high risk = 0.6, medium risk = 0.3, low risk = 0.1). The initial SEI of this model is 82%, which is improved to 94% after optimization.

[0079] Based on this model, a retrofit plan for adding a new filter was simulated: the space requirements for pipe welding were verified (requiring an operating radius of 0.8 meters), and the maintenance path was simulated to ensure the safe passage of hoisting equipment (minimum turning radius of 4 meters). The optimized maintenance plan reduced the estimated construction time by 15%.

[0080] 206. Regularly scan the changed areas of in-service facilities using a mobile laser scanning system to obtain incremental point cloud data of the changed areas. This incremental point cloud data is used for model updates. Spatial registration and difference comparison are performed between the incremental point cloud data and the existing 3D operable maintenance model to generate a facility change analysis report. This report includes information on newly added, demolished, and modified facilities. Based on the facility change analysis report, the 3D operable maintenance model is partially updated and version managed to generate an updated version. This updated version is used for historical version tracing. Each version of the 3D operable maintenance model is associated and stored with the corresponding facility operation and maintenance records to generate a 3D model database for the entire lifecycle of the facility. This database supports historical status backtracking. Based on the 3D model database, facility evolution analysis and modification effect evaluation are performed, generating a facility evolution analysis report.

[0081] Specifically, the incremental point cloud data is spatially registered and compared with the existing 3D operable model to generate a site facility change analysis report. This includes: based on the spatial distribution characteristics of the incremental point cloud data, feature matching is performed between it and the corresponding area of ​​the 3D operable model to generate registered incremental point cloud data, which is used for precise comparison; point-by-point distance calculation and geometric feature comparison are performed between the registered incremental point cloud data and the 3D operable model to generate preliminary change detection results, which include the location and range of suspected change areas; based on a preset facility change type rule base, the preliminary change detection results are identified and classified into change types to generate facility change classification results, which include type identifiers for newly added facilities, demolished facilities, and renovated facilities; the facility change classification results are verified for engineering rationality and manually confirmed to generate verified facility change data, which is used for report generation; based on the verified facility change data, combined with facility engineering attributes and maintenance records, a site facility change analysis report containing change details and impact analysis is generated.

[0082] It should be noted that dynamic model maintenance is achieved through periodic scanning and incremental updates: Scanning Execution: A mobile laser scanning system (RieglVUX-240) is used quarterly to scan the changed areas of the site. This scan covered the western process modification area of ​​the plant (approximately 0.8 hectares), traveling along a predetermined route at a speed of 10 km / h, acquiring approximately 320 million incremental point cloud data points, with a point density of 800 points / square meter. The scan simultaneously recorded GPS timestamps and control point coordinates (reference points T01 and T02) to ensure initial alignment with the original model coordinate system (CGCS2000).

[0083] Registration Process: Feature matching is performed between the incremental point cloud and the V2.1 model. First, six common feature points (top of the tank ladder and center of the compressor flange) are extracted from the two types of data. The transformation matrix is ​​calculated using the singular value decomposition algorithm. The average error after registration is ≤2 mm.

[0084] Difference detection: Point-by-point distance calculation and geometric feature comparison are used: Set distance threshold: 5 cm (above the threshold is considered a change).

[0085] Three types of changes were detected: New additions: a DN200 pipe section (30 meters long) and two ball valves (model Z41H-16C). Removal: an old filter and its connecting short pipe. Modification: relocation of the compressor inlet pipe (offset of 1.2 meters). Preliminary results were generated containing 12 suspected change areas.

[0086] Based on a pre-defined facility change rule base (pipeline connection logic, equipment installation specifications), the preliminary results are verified through engineering procedures: Manual confirmation: By comparing maintenance records, it is confirmed that the removed filter was scrapped last month, and the newly added pipeline is a process upgrade project. The generated report is shown in Table 6 below: Table 6 Model partial updates and version management: Partial updates: Based on the report, only the changed model components are updated: New pipes and valves are generated using the parametric modeling engine (diameter error <1mm). Filter models are deleted and their topology links are removed. Pipeline routing is adjusted, and collision detection is performed again (no conflicts). Version management: A new version V2.2 is generated and associated with the maintenance record (maintenance work order 2024-09-101). Historical versions (V2.1, V2.0) are stored in the full lifecycle database, supporting backtracking by timestamp.

[0087] Evolution analysis and assessment, full life cycle database query: Comparing V2.2 with the version V1.5 from a year ago, it was found that the station has added a total of 280 meters of pipeline and upgraded 8 key equipment.

[0088] The Facility Evolution Activity Index (AEI) is proposed to assess the intensity of change. The weighting coefficients (new additions = 0.6, renovations = 0.3, demolitions = 0.1) result in an AEI of 15.3 (a 20% increase from the previous year), reflecting the period of intensive renovations.

[0089] Evaluation of the transformation effect: Based on the model simulation of the flow of the new pipeline, the pass rate was verified to be 18%, and optimization suggestions were generated.

[0090] In this embodiment of the invention, a multi-level scanning scheme is formulated based on the site layout plan and facility distribution map, including mobile scanning path planning and fixed scanning station layout. This makes the scanning work more systematic and planned, avoids blind scanning, greatly improves scanning efficiency, and reduces scanning time and cost. In the point cloud data processing, timestamp synchronization and preliminary coordinate system alignment are performed on macro point cloud data and fine point cloud data, and point cloud density balancing and overlapping area optimization are performed on the preliminarily registered point cloud dataset. This effectively solves the problems of coordinate inconsistency and density unevenness that may occur in the process of multi-source point cloud data fusion, improves the quality and usability of point cloud data, and provides more accurate data support for subsequent facility identification and modeling. Based on the engineering standards and geometric feature library of station facilities, a station facility feature rule library was established, which includes descriptions and identification rules of geometric features of various facilities. Through continuous integration, optimization, and testing, it can accurately identify different types of facilities, providing a scientific basis for geometric feature identification and improving the accuracy and reliability of identification. Through geometric feature extraction and component identification, combined with topological relationship verification and spatial position correction, automatic classification and labeling of facilities such as pipelines, valves, and storage tanks were achieved, generating semantic point cloud datasets with facility type labels, which greatly improved data processing efficiency, reduced manual intervention, and reduced human error. By comparing the preliminary facility component model with the semantic point cloud dataset for spatial location and verifying geometric accuracy, and by fine-tuning the local parameters of areas that do not meet the accuracy requirements, the final facility component model is ensured to have high-precision engineering dimensions, providing accurate component model support for the three-dimensional operability model of the site. Based on the three-dimensional operability model, the spatial feasibility of the renovation plan and the simulation analysis of the maintenance path can be carried out, which can identify potential problems in the plan in advance, make timely adjustments and optimizations, reduce rework and modifications in actual construction, improve construction efficiency, and reduce construction costs. By associating and storing the various versions of the 3D operable and maintainable models with their corresponding site operation and maintenance records, a 3D model database of the site's entire lifecycle is generated, supporting historical status retrospection. Based on this database, site facility evolution analysis and renovation effect evaluation can be performed, providing a deeper understanding of the site's change history and renovation effects, offering crucial reference for future site development and decision-making.

[0091] The above describes the 3D modeling method for in-service natural gas stations based on laser point cloud scanning in embodiments of the present invention. The following describes the 3D modeling system for in-service natural gas stations based on laser point cloud scanning in embodiments of the present invention. Please refer to [link / reference]. Figure 4 An embodiment of the 3D modeling system for in-service natural gas stations based on laser point cloud scanning in this invention includes: a scanning module 301, used to perform multi-level scanning of the in-service station to acquire multi-source point cloud data, including macro-level point clouds of the overall station layout and fine point clouds of key facilities; a processing module 302, used to process the multi-source point cloud data to generate structured point cloud data in a unified coordinate system; a segmentation module 303, used to perform geometric feature recognition and semantic segmentation on the structured point cloud data based on a preset station facility feature rule library to generate a semantic point cloud dataset; a modeling module 304, used to generate facility component models based on the semantic point cloud dataset using a parametric modeling method, wherein the parametric modeling combines standard facility specifications and the actual dimensions extracted from the point cloud; and an integration module 305, used to perform spatial assembly and topological relationship reconstruction based on the facility component models to integrate and obtain a 3D operable model.

[0092] above Figure 4 The three-dimensional modeling system for in-service natural gas stations based on laser point cloud scanning in this embodiment of the invention will be described in detail from the perspective of modular functional entities. The three-dimensional modeling equipment for in-service natural gas stations based on laser point cloud scanning in this embodiment of the invention will be described in detail from the perspective of hardware processing.

[0093] Figure 5 This is a schematic diagram of the structure of a 3D modeling device for in-service natural gas stations based on laser point cloud scanning, provided in an embodiment of the present invention. The 3D modeling device 400 for in-service natural gas stations based on laser point cloud scanning may include a processor 401 and a memory 402. The memory 402 is used to store program instructions and / or data, and the processor 401 is used to execute the program instructions stored in the memory 402, thereby implementing the method described in the above-described method embodiment.

[0094] Optionally, the memory 402 and the processor 401 are coupled. The coupling is an indirect coupling or communication connection between devices, units, or modules, and can be electrical, mechanical, or other forms, for information interaction between devices, units, or modules.

[0095] Optionally, the 3D modeling equipment 400 for in-service natural gas stations based on laser point cloud scanning may further include a communication interface 403. The communication interface 403 is used to communicate with other devices via a transmission medium, for example, transmitting received signals from other communication devices to the processor 401, or transmitting signals from the processor 401 to other communication devices. The communication interface 403 may be a transceiver or an interface circuit, such as a transceiver circuit or transceiver chip.

[0096] This application embodiment does not limit the specific connection medium between the processor 401, memory 402, and communication interface 403. This application embodiment... Figure 5 The processor 401, memory 402, and communication interface 403 are connected via a bus 404. Figure 5 The connections between other components are shown in bold and are for illustrative purposes only, not as limiting information. The bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, Figure 5 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0097] The present invention also provides a 3D modeling device for in-service natural gas stations based on laser point cloud scanning. The 3D modeling device for in-service natural gas stations based on laser point cloud scanning includes a memory and a processor. The memory stores computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor performs the steps of the 3D modeling method for in-service natural gas stations based on laser point cloud scanning in the above embodiments.

[0098] The present invention also provides a computer-readable storage medium, which can be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium, wherein the computer-readable storage medium stores instructions that, when the instructions are executed on a computer, cause the computer to perform the steps of the method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning.

[0099] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0100] 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, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0101] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention 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 the present invention.

Claims

1. A method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning, characterized in that, include: Multi-level scanning of in-service stations was conducted to obtain multi-source point cloud data, including macro point clouds of the overall station layout and fine point clouds of key facilities. The multi-source point cloud data is processed to generate structured point cloud data in a unified coordinate system; Based on a pre-defined rule base for site facilities features, geometric feature recognition and semantic segmentation are performed on the structured point cloud data to generate a semantic point cloud dataset. Based on the semantic point cloud dataset, facility component models are generated using a parametric modeling method, which combines standard facility specifications with the actual dimensions extracted from the point cloud. Based on the facility component model, spatial assembly and topological relationship reconstruction are performed to obtain a three-dimensional operable model.

2. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 1, characterized in that, include: Based on the site layout plan and facility distribution plan, a multi-level scanning scheme was developed, including mobile scanning path planning and fixed scanning station layout. According to the multi-level scanning scheme, continuous scanning is performed along a predetermined path to obtain macroscopic point cloud data of the overall layout of the site; Based on the macroscopic point cloud data, the location and extent of key facility areas are identified, and multi-angle fine scanning is performed on the key facilities to obtain fine point cloud data of the key facilities. The macroscopic point cloud data and the fine point cloud data are processed to generate a preliminary registered point cloud dataset; The point cloud dataset that has been initially registered is subjected to point cloud density equalization and overlapping region optimization processing to generate multi-source point cloud data.

3. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 2, characterized in that, include: Based on the engineering standards and geometric feature library of station facilities, establish a feature rule library for station facilities; Based on the site facility feature rule base, geometric features are extracted from the structured point cloud data to generate a feature parameter set, including normal vector, curvature, and size parameters; Based on the feature parameter set, pipeline components are identified by cylinder fitting algorithm, valve components are identified by bounding box detection, and storage tank components are identified by surface fitting, generating preliminary classification results for facility components. Based on the preliminary classification of facility components, validated facility classification data is generated; Based on the verified facility classification data, a corresponding semantic label is assigned to each facility component to generate a semantic point cloud dataset.

4. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 3, characterized in that, Collect station design drawings, equipment specifications and engineering acceptance data, extract standard geometric parameters and spatial relationship constraints of facilities, and generate a standard feature library of station facilities; Based on historical point cloud data processing experience, we summarize the geometric feature recognition rules and parameter extraction methods for different types of facilities, and generate a set of facility recognition experience rules. The standard feature library of the station facilities is integrated and optimized with the experience rule set for facility identification to generate an initial feature rule library for station facilities. Based on typical station sample point cloud data, the initial station facility feature rule base is tested for recognition effect and the parameters are optimized to generate an optimized station facility feature rule base. Establish a version management and update mechanism for the station facility feature rule base, continuously improve the rule content based on the actual recognition effect, and generate the final station facility feature rule base system.

5. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 3, characterized in that, include: Based on the semantic point cloud dataset, key geometric parameters of various facility components are extracted to generate a dataset of measured dimensions of facility components, including the diameter and direction of pipes, the connection dimensions of valves, and the diameter and height parameters of storage tanks. Based on the measured size dataset, match the corresponding standard facility model to generate a standard facility parameter set; Optimized facility modeling parameters are generated based on the measured size dataset and the standard facility parameter set. Based on the optimized facility modeling parameters, the parametric modeling engine is invoked to generate a 3D geometric model of the facility components, thus generating a preliminary facility component model. The preliminary facility component model is compared with the semantic point cloud dataset in terms of spatial location and geometric accuracy to generate the final facility component model.

6. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 5, characterized in that, include: Based on the spatial location and engineering connection relationship of the facility component model, establish spatial topological relationship data between facilities; Based on the spatial topology data, the facility component model is spatially assembled and the connection relationship is reconstructed to generate a preliminary integrated three-dimensional assembly model of the station. Based on the three-dimensional assembly model of the site, collision detection and clearance analysis are performed between facilities to generate a safety space analysis report. The safety space analysis report is integrated with the three-dimensional assembly model of the site to generate a three-dimensional operable and maintainable model; Based on the aforementioned three-dimensional operability model, spatial feasibility verification of the modification scheme and simulation analysis of the maintenance path are performed to generate site operation and maintenance decision support data.

7. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 6, integrates the safety space analysis report with the three-dimensional assembly model of the station to generate a three-dimensional operable and maintainable model, wherein the safety performance index is SEI: in, Weighting coefficients are assigned based on risk level.

8. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 1, characterized in that, Also includes: Perform a change area scan on in-service stations to obtain incremental point cloud data of the change areas; The incremental point cloud data is spatially registered and the differences are compared with the existing three-dimensional operable model to generate a site facility change analysis report. Based on the site facility change analysis report, the three-dimensional operability model is partially updated and version managed to generate an updated version of the three-dimensional operability model. The three-dimensional operational model is associated and stored with the corresponding site operation and maintenance records to generate a three-dimensional model database of the entire life cycle of the site; Based on the three-dimensional model database of the entire life cycle of the station, the evolution analysis of the station facilities and the evaluation of the transformation effect are carried out, and a station facility evolution analysis report is generated.

9. The method for three-dimensional modeling of in-service natural gas stations based on laser point cloud scanning according to claim 8, characterized in that, Based on the spatial distribution characteristics of the incremental point cloud data, feature matching is performed with the corresponding region of the three-dimensional operable model to generate registered incremental point cloud data. The registered incremental point cloud data is compared with the three-dimensional operable model by point-by-point distance calculation and geometric feature comparison to generate preliminary change detection results. Based on a pre-defined rule base for facility change types, the preliminary change detection results are identified and classified according to change type, and facility change classification results are generated. The engineering rationality verification and manual confirmation of the facility change classification results are performed to generate facility change data. Based on facility change data, combined with facility engineering attributes and maintenance records, a site facility change analysis report is generated.

10. A 3D modeling system for in-service natural gas stations based on laser point cloud scanning, characterized in that, The 3D modeling system for in-service natural gas stations based on laser point cloud scanning includes: The scanning module is used to perform multi-level scanning of in-service stations to acquire multi-source point cloud data, including macro point clouds of the overall station layout and fine point clouds of key facilities. The processing module is used to process the multi-source point cloud data to generate structured point cloud data in a unified coordinate system. The segmentation module is used to perform geometric feature recognition and semantic segmentation on the structured point cloud data based on a preset site facility feature rule base, and generate a semantic point cloud dataset. The modeling module is used to generate facility component models based on the semantic point cloud dataset using a parametric modeling method, wherein the parametric modeling combines standard facility specifications and the actual dimensions extracted from the point cloud; An integration module is used to perform spatial assembly and topological relationship reconstruction based on the facility component model, and integrate them to obtain a three-dimensional operable model.

Images