Gas pipe network hidden danger identification method and system based on unmanned detection vehicle

Abstract

The invention relates to the technical field of gas pipe network detection, and discloses a gas pipe network hidden danger identification method and system based on an unmanned detection vehicle, and the method comprises the steps: generating an inspection path and a detection mode of the unmanned detection vehicle based on the GIS coordinates and historical hidden danger data of a gas pipe network; acquiring multi-source detection data through a multi-sensor system configured on the unmanned detection vehicle during inspection; abnormal feature extraction is carried out on the multi-source detection data, and preliminary hidden danger identification is carried out based on abnormal features to determine preliminarily judged hidden danger points and corresponding grading classification; surrounding detection and multi-azimuth measurement are carried out on the primary judgment hidden danger point, whether the primary judgment hidden danger point is a leakage source or not is determined, and a surrounding environment risk value is evaluated; and starting an emergency response based on the surrounding environment risk value, and updating the inspection path. The inspection resource utilization efficiency and the hidden danger discovery rate are improved.

Description

Technical Field

[0001] This invention relates to the field of gas pipeline network inspection technology, and in particular to a method and system for identifying potential hazards in gas pipeline networks based on an unmanned inspection vehicle. Background Technology

[0002] With the continuous expansion of urban gas pipeline networks and their increasing service life, potential hazards such as gas leaks, damage from third-party construction, and land subsidence are on the rise, posing a serious threat to public safety. Traditional manual inspection methods suffer from low efficiency, incomplete coverage, delayed response, and susceptibility to environmental and human factors, making it difficult to meet the real-time and accuracy requirements for safe operation of gas pipeline networks in the high-density and complex environments of modern cities. In recent years, unmanned inspection vehicles, as an intelligent and automated inspection platform, have gradually become an important technical means for identifying potential hazards in gas pipeline networks due to their flexible deployment, all-weather operation, and multi-sensor fusion capabilities.

[0003] However, existing unmanned vehicle-based gas pipeline network inspection methods generally suffer from the following shortcomings: First, the inspection path planning lacks deep integration of historical hazard data and geographic information systems, resulting in unreasonable resource allocation and insufficient coverage of key areas; second, multi-source heterogeneous sensor data (such as gas concentration, images, acoustics, spatial location, etc.) lacks an effective spatiotemporal alignment and fusion mechanism, making it difficult to achieve high-precision extraction of abnormal features; third, the hazard identification process relies heavily on single sensor judgment, lacking a closed-loop logic of initial judgment—verification—grading—response, making it impossible to effectively distinguish between real leak sources and interference signals, and the emergency response strategy is disconnected from risk assessment, making it difficult to support dynamic scheduling and intelligent decision-making. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles, aiming to solve one or more of the above-mentioned problems.

[0005] This invention provides a method for identifying potential hazards in gas pipeline networks based on an unmanned inspection vehicle, comprising: Receive GIS coordinates and historical hazard data of the gas pipeline network, generate inspection path and detection mode of unmanned inspection vehicle based on the GIS coordinates and historical hazard data, and control unmanned inspection vehicle to carry out inspection according to the inspection path and detection mode; During inspections, the unmanned inspection vehicle acquires multi-source detection data through a multi-sensor system configured on the vehicle. The multi-source detection data includes geospatial data, visible light and infrared image data, gas concentration gradient data, and acoustic feature data. The multi-source detection data is spatiotemporally aligned and fused, and abnormal features are extracted from the multi-source detection data. Based on the abnormal features, the potential hazards are initially identified, and the initial hazard points and their corresponding hierarchical classifications are determined. Surround detection and multi-directional measurement are carried out on the initially identified potential hazard points to determine whether the initially identified potential hazard points are leakage sources. If they are leakage sources, the risk value of the surrounding environment is assessed. An emergency response is initiated based on the risk value of the surrounding environment, and the inspection route is updated based on the source of the leak.

[0006] Preferably, the inspection path of the unmanned inspection vehicle is generated based on the GIS coordinates and historical hazard data, including: The historical hazard data includes the coordinates of areas with a high incidence of hazards; Based on the GIS coordinates, the gas pipeline network is divided into several regions. Each region includes only the coordinates of a high-risk area, and the regions do not overlap. Obtain the starting coordinates of the unmanned inspection vehicle, and generate the shortest inspection path to traverse each area based on the starting coordinates; Based on the coordinates of high-risk areas in each region, an intra-regional inspection path is generated, and the initial coordinates of the intra-regional inspection path are the coordinates of the high-risk areas. The inspection path of the unmanned inspection vehicle is obtained by merging the shortest inspection path and the inspection path within the area.

[0007] Preferably, the detection mode of the unmanned detection vehicle is generated based on the GIS coordinates and historical hazard data, including: The detection mode includes inspection frequency and inspection method; If the GIS coordinates are the coordinates of a high-risk area, then the inspection frequency is set to the first inspection frequency; otherwise, it is set to the second inspection frequency, with the first inspection frequency being higher than the second inspection frequency. Obtain meteorological environmental data of the gas pipeline network, and determine the inspection method based on the meteorological environmental data. The inspection method includes daytime inspection method and nighttime inspection method.

[0008] Preferably, the multi-sensor system includes: lidar, multispectral camera, inertial navigation system, gas concentration sensor array, and acoustic sensor.

[0009] Preferably, anomaly feature extraction is performed on multi-source detection data, including: Image feature extraction is performed on the geospatial data, visible light and infrared image data to obtain surface subsidence features and third-party construction features; Gas feature extraction is performed on the gas concentration gradient data to obtain gas concentration abrupt change characteristics and gas continuous leakage trend characteristics; Acoustic features are extracted from the acoustic feature data to obtain the gas leak spectrum features and abnormal sound source features.

[0010] Preferably, the preliminary identification of potential hazards based on the aforementioned abnormal features, determining the initial hazard points and their corresponding classifications, includes: Construct a mapping database between abnormal features and hazard types, where hazard types include pipeline leakage, third-party construction interference, surface subsidence threat, and abnormal sound source interference; The extracted surface subsidence features, third-party construction features, gas concentration abrupt change features, gas continuous leakage trend features, gas leakage spectrum features, and abnormal sound source features are input into the mapping relationship library to match and obtain the corresponding preliminary hazard type; A hierarchical index system is constructed based on the confidence value and feature intensity of each abnormal feature. The comprehensive risk score of the initially identified hidden danger points is calculated according to the hierarchical index system. The initially identified hidden danger points are divided into three risk levels: high, medium and low, according to the comprehensive risk score, and associated with the corresponding hidden danger classification labels.

[0011] Preferably, the preliminary identified potential hazard point is subjected to surround detection and multi-directional measurement to determine whether the preliminary identified potential hazard point is a leakage source, including: The system automatically performs a circular multi-directional moving detection with a radius of 2-5 meters centered on the initially identified potential hazard point; gas concentration and acoustic data are repeatedly collected at at least 8 equally divided points along the circular path; Based on multi-directional measurement data, a triangulation algorithm is used to calculate the three-dimensional coordinates of the initially identified potential hazard point. A leak source probability model is constructed by combining the gas concentration gradient change curve and the acoustic signal attenuation law. When the output value of the leak source probability model is greater than the preset probability threshold, the initially identified potential hazard point is determined to be a leak source. At the same time, the precise three-dimensional coordinates, gas concentration peak value, and diffusion direction of the leak source are recorded. If the output value is less than or equal to the preset probability threshold, it is determined to be a non-leak source.

[0012] Preferably, when the initially identified potential hazard point is determined to be a leak source, the risk value of the surrounding environment is assessed, including: analyzing the population density, building distribution and traffic flow within a preset range around the leak source, assessing the leak diffusion trend based on real-time wind speed and direction, and determining the risk value of the surrounding environment.

[0013] Preferably, the emergency response is initiated based on the surrounding environmental risk value, and the inspection route is updated based on the leak source, including: When the risk value of the surrounding environment exceeds the first emergency threshold, a Level 1 emergency response is automatically triggered, including uploading the coordinates of the leak source and the risk assessment report to the monitoring center in real time, activating the audible and visual alarm device simultaneously, and pushing evacuation warning information to smart terminals within a 500-meter radius through the vehicle communication module. When the risk value of the surrounding environment is between the second emergency threshold and the first emergency threshold, a level-two emergency response is initiated, continuous monitoring of the leak source is carried out and dynamic data is uploaded, and the nearest emergency repair team is dispatched to the site. When the risk value of the surrounding environment is lower than the second emergency threshold, a level 3 emergency response is initiated, a hazard handling work order is generated, and it is included in the routine maintenance plan. When a leak source is identified, it is marked as a key re-inspection node, prioritized for insertion into the current inspection sequence, the inspection order of subsequent paths is adjusted, and the inspection frequency of the leak source is increased.

[0014] This invention also discloses a gas pipeline network hazard identification system based on an unmanned inspection vehicle, used to apply the above-mentioned gas pipeline network hazard identification method based on an unmanned inspection vehicle. The system includes: The inspection control module is configured to receive GIS coordinates and historical hazard data of the gas pipeline network, generate inspection paths and detection modes for unmanned inspection vehicles based on the GIS coordinates and historical hazard data, and control the unmanned inspection vehicles to perform inspections according to the inspection paths and detection modes. The data acquisition module is configured to acquire multi-source detection data through the multi-sensor system configured on the unmanned inspection vehicle during inspection. The multi-source detection data includes geospatial data, visible light and infrared image data, gas concentration gradient data and acoustic feature data. The hazard identification module is configured to perform spatiotemporal alignment and fusion processing on the multi-source detection data, extract abnormal features from the multi-source detection data, and perform preliminary identification of hazards based on the abnormal features to determine the initial hazard points and their corresponding hierarchical classifications. The hazard verification and response module is configured to perform surround detection and multi-directional measurement on the initially identified hazard point to determine whether the initially identified hazard point is a leak source. If it is a leak source, the module assesses the risk value of the surrounding environment, initiates an emergency response based on the risk value of the surrounding environment, and updates the inspection path based on the leak source.

[0015] Compared with existing technologies, the beneficial effects of this invention are as follows: By integrating GIS coordinates and historical hazard data (especially in high-risk areas), the system can automatically generate the shortest inspection path covering the entire network and prioritizing high-risk areas. It can also dynamically configure inspection frequency and day / night inspection methods based on regional attributes, significantly improving the efficiency of inspection resource utilization and the hazard detection rate. Utilizing the lidar, multispectral camera, gas concentration sensor array, acoustic sensor, and inertial navigation system integrated on the unmanned inspection vehicle, multimodal data is collected simultaneously. Through spatiotemporal alignment and fusion processing, the limitations of single sensors are effectively overcome, providing a highly consistent and high-dimensional data foundation for subsequent anomaly feature extraction. By constructing a mapping relationship library between anomaly features and hazard types, and establishing a hierarchical index system based on confidence level and feature intensity, the risk level of initially identified hazard points can be classified. Furthermore, through a ring-shaped multi-directional detection and triangulation algorithm centered on the hazard point, combined with gas gradient and acoustic attenuation laws, a leak source probability model is constructed, effectively eliminating false alarms and accurately confirming the location and diffusion characteristics of the leak source. Once the leak source is identified, the system can calculate the surrounding environmental risk value based on environmental parameters such as real-time population density, building distribution, traffic flow, and wind direction and speed, and automatically trigger a level one to three emergency response (including alarm, evacuation warning, emergency repair dispatch or work order generation) according to preset thresholds; at the same time, the leak source is marked as a key re-inspection node, and subsequent inspection paths are dynamically inserted and optimized to achieve a collaborative closed loop of hazard handling and preventive inspection. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0017] Figure 1 This is a flowchart illustrating a method for identifying potential hazards in gas pipeline networks based on an unmanned inspection vehicle, according to the present invention.

[0018] Figure 2 This is a functional block diagram of a gas pipeline hidden danger identification system based on an unmanned inspection vehicle according to the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] like Figure 1 As shown, this invention provides a method for identifying potential hazards in gas pipeline networks based on an unmanned inspection vehicle, including: The system receives GIS coordinates and historical hazard data of the gas pipeline network, generates inspection paths and detection modes for unmanned inspection vehicles based on the GIS coordinates and historical hazard data, and controls the unmanned inspection vehicles to carry out inspections according to the inspection paths and detection modes.

[0021] Specifically, this application utilizes structured spatial information (i.e., GIS coordinates of gas pipeline networks) and empirical risk information (i.e., historical hazard data) to generate targeted inspection paths and detection modes for unmanned inspection vehicles through intelligent analysis.

[0022] Specifically, GIS coordinates refer to the precise spatial location (usually expressed in latitude and longitude or projected coordinates) of key facilities such as pipe sections, valves, pressure regulating stations, and joints in a gas pipeline network within a geographic information system, constituting the spatial topology of the pipeline network. Historical hazard data includes information such as the location, occurrence time, hazard type, and severity of leak points, areas affected by third-party construction interference, surface subsidence areas, and high-corrosion sections discovered in previous inspections or accident records.

[0023] For example, suppose the total length of the gas pipeline network in area A of a city is 200 kilometers. Section B has experienced five pipeline damage incidents due to third-party construction in the past three years. Area C has experienced two minor leaks due to groundwater erosion. Area D is a newly built residential area with a relatively new pipeline network but frequent construction in the surrounding area. The system first receives a complete GIS layer of the gas pipeline network for the area (containing the coordinates of all pipeline nodes) and a historical hazard database provided by the maintenance department (containing hazard records for areas B, C, and D). Then, the system divides the pipeline network into several logical areas, each centered on a historically high-risk hazard point; for example, a high-risk zone with a radius of 300 meters is defined with the center point of section B as the center. Next, based on the current parking location of the unmanned inspection vehicle (e.g., the coordinates of a maintenance station), the system calculates the shortest path traversing all high-risk areas (e.g., using the Traveling Salesman Problem (TSP) optimization algorithm). Upon entering each high-risk area, the system automatically switches to a "high-density surround scan" sub-path (e.g., a spiral or grid-like path) instead of a simple straight line. Meanwhile, for areas B and C, the system sets the inspection frequency to "once a day" (first inspection frequency), while for area D it sets it to "twice a week" (second inspection frequency). If the weather forecast indicates strong winds that night, the system may adjust the originally planned nighttime infrared detection to a daytime visible light + gas concentration combined detection to avoid wind interference with the infrared imaging effect. Finally, the unmanned inspection vehicle automatically sets off according to the generated composite path and dynamic detection mode, and sequentially completes the detailed inspection tasks of each area.

[0024] During inspections, the unmanned inspection vehicle acquires multi-source detection data through a multi-sensor system, including geospatial data, visible light and infrared image data, gas concentration gradient data, and acoustic feature data.

[0025] Specifically, this step is the data perception stage in the hazard identification process. Its core lies in relying on the multi-sensor fusion system carried by the unmanned inspection vehicle to synchronously and continuously collect raw inspection data of multiple physical dimensions during the inspection process, providing a high-dimensional and complementary information foundation for subsequent spatiotemporal alignment, abnormal feature extraction and hazard judgment.

[0026] Specifically, the multi-sensor system refers to a variety of sensing devices integrated on the unmanned inspection vehicle, each performing its own function while working collaboratively to jointly construct a three-dimensional perception capability of the surrounding environment of the gas pipeline network. Geospatial data, obtained in real-time by a high-precision positioning module (such as RTK-GNSS combined with an inertial navigation system), includes information such as the vehicle's position, attitude, and driving trajectory, used to accurately annotate the spatial coordinates of other sensor data. Visible light and infrared image data: Visible light cameras are used to capture visible anomalies on the ground (such as construction marks, abnormal vegetation withering, and ground cracks); infrared thermal imagers are used to detect localized temperature anomalies caused by gas leaks (such as low-temperature patches caused by the endothermic effect of methane leaks). Gas concentration gradient data: Multiple high-sensitivity gas sensors (such as laser methane telemetry devices and electrochemical sensor arrays) deployed at different heights or directions on the vehicle body measure the concentration of gas components in the air and their spatial rate of change in real time, forming a concentration gradient field, which helps determine the direction and intensity of the leak. Acoustic feature data is collected by using high-sensitivity microphones or ultrasonic sensor arrays to acquire sound wave signals around the pipeline, especially the specific frequency noise (usually above 20 kHz) generated when gas leaks under high pressure, to help identify minor leaks.

[0027] For example: Suppose an unmanned inspection vehicle is performing a nighttime patrol in an old urban area of ​​a city. This area has a dense network of underground gas pipelines and recent road renovations. The vehicle's onboard RTK-GNSS and inertial navigation system record the vehicle's precise location at a frequency of 10 times per second (with an error of less than 5 cm), forming a high-density trajectory point. A visible light camera, under streetlight illumination, photographs the road surface and discovers a newly excavated but not fully backfilled trench, suspected to be due to third-party construction. An infrared thermal imager detects a localized low-temperature area of ​​approximately 0.8°C at the same location, with a linearly extending shape consistent with the underground pipeline's direction. As the gas concentration sensor array passes through this area, it detects a rapid increase in methane concentration from a background value of 1 ppm to 15 ppm, forming a significant concentration attenuation gradient behind the vehicle. Simultaneously, the vehicle's acoustic sensors capture a high-frequency whistling sound lasting 3 seconds with a center frequency of 28 kHz, which, after preliminary filtering analysis, matches the acoustic spectrum characteristics of a high-pressure gas leak.

[0028] The multi-source detection data is spatiotemporally aligned and fused, and abnormal features are extracted from the multi-source detection data. Based on the abnormal features, the potential hazards are initially identified, and the initial hazard points and their corresponding classifications are determined.

[0029] Specifically, this step involves aligning and fusing heterogeneous detection data from different sensors, timestamps, and spatial reference systems through a unified spatiotemporal benchmark to form structured and highly correlated comprehensive perception information. On this basis, it further extracts discriminative abnormal features and makes a preliminary judgment on potential hazards based on preset rules or models, outputting "preliminary hazard points" with risk level and type labels.

[0030] Specifically, spatiotemporal alignment refers to using geospatial data as a reference coordinate system to precisely map various data, such as visible light / infrared images, gas concentration readings, and acoustic signals, onto a unified spatiotemporal grid according to the acquisition time and location, ensuring that multimodal data from the same physical location are comparable and fusionable. Fusion processing refers to performing complementary enhancement or cross-validation on the aligned multi-source data. For example, using the location of surface cracks identified by image recognition to constrain the spatial range of gas concentration anomalies, or using acoustic leakage spectra to corroborate the authenticity of infrared low-temperature regions, thereby improving data credibility.

[0031] Surround detection and multi-directional measurement are carried out on the initially identified potential hazard points to determine whether the initially identified potential hazard points are leakage sources. If they are leakage sources, the risk value of the surrounding environment is assessed. Specifically, the purpose of this step is to conduct a refined and proactive secondary detection of the initially identified potential hazards in order to eliminate false alarms, improve the accuracy of judgment, and immediately initiate a quantitative assessment of the potential hazards to the surrounding environment after confirming that it is a real leak source.

[0032] An emergency response is initiated based on the risk value of the surrounding environment, and the inspection route is updated based on the source of the leak.

[0033] Specifically, the core of this step is to transform the leak sources and their quantified risk values ​​identified in the previous steps into specific emergency response actions, and simultaneously feed them back to the inspection and dispatch system to achieve dynamic adjustment of subsequent task paths, thereby forming a complete intelligent closed loop of "detection-identification-verification-response-optimization".

[0034] In some embodiments of this application, generating an inspection path for an unmanned inspection vehicle based on the GIS coordinates and historical hazard data includes: the historical hazard data including coordinates of high-risk hazard areas; dividing the gas pipeline network into several regions based on the GIS coordinates, each region including only one high-risk hazard area coordinate, and each region not overlapping; obtaining the starting coordinates of the unmanned inspection vehicle, and generating the shortest inspection path traversing each region based on the starting coordinates; generating an intra-regional inspection path based on the high-risk hazard area coordinates in each region, wherein the initial coordinates of the intra-regional inspection path are the high-risk hazard area coordinates; and merging the shortest inspection path and the intra-regional inspection path to obtain the inspection path of the unmanned inspection vehicle.

[0035] Specifically, this solution scientifically and efficiently generates a complete inspection path for unmanned inspection vehicles, taking into account both overall coverage of the gas pipeline network and detailed inspection of key potential hazard areas, avoiding duplication or omissions, and improving the efficiency of inspection resource utilization and the ability to discover potential hazards.

[0036] Specifically, the system receives GIS coordinate data and historical hazard data of the gas pipeline network. The historical hazard data explicitly includes coordinates of several "high-risk area areas," representing locations where leaks, construction damage, or other abnormal events have frequently occurred in the past. Next, based on the overall spatial distribution of the gas pipeline network (i.e., GIS coordinates), the entire network area is divided into several non-overlapping sub-regions. Each sub-region contains exactly one high-risk area coordinate, ensuring that each high-risk point is independently focused on, while avoiding path conflicts or duplicate inspections caused by overlapping areas. Then, the system obtains the current starting position of the unmanned inspection vehicle (e.g., garage or repair station coordinates). Using this as a starting point, a global path that sequentially visits all sub-regions is calculated. This path aims to minimize the total travel distance and is typically generated using path optimization algorithms (e.g., solving the traveling salesman problem), and is called the "shortest inspection path." For each sub-region, using the coordinates of its internal high-risk area as a starting point, a local path (e.g., a grid-like, spiral, or circular route) is planned specifically for fine-tuning the pipeline network within that region. This is called the "intra-region inspection path," ensuring high-density, multi-angle inspection of the areas surrounding high-risk points. Finally, the shortest global inspection path is spliced ​​and merged with the regional inspection paths of each sub-region to form a continuous and executable complete inspection path for the unmanned inspection vehicle to execute in sequence.

[0037] For example: Suppose a city's gas pipeline network covers three areas, A, B, and C. Historical data shows three high-risk areas, located at P1, P2, and P3 respectively. The system first divides the entire pipeline network into three non-overlapping areas, R1, R2, and R3, where R1 contains P1, R2 contains P2, and R3 contains P3. An unmanned inspection vehicle departs from maintenance station S. The system calculates the shortest global path to access R1→R2→R3 as S→P1→P2→P3. Upon entering R1, the vehicle does not pass directly through but instead scans the pipelines within a 50-meter radius of R1 along a pre-defined "U-shaped" path, starting from P1. After completion, it moves to R2, starting from P2 and executing a refined path within R2, and so on. The final complete inspection path is: S → (enter R1) → P1 → refined path within R1 → (leave R1) → P2 → refined path within R2 → P3 → refined path within R3 → return. This approach ensures both overall efficiency and targeted coverage of high-risk areas.

[0038] In some embodiments of this application, a detection mode for an unmanned inspection vehicle is generated based on the GIS coordinates and historical hazard data, including: the detection mode includes an inspection frequency and an inspection method; if the GIS coordinates are coordinates of a high-risk area, the inspection frequency is set to a first inspection frequency; otherwise, it is set to a second inspection frequency, wherein the first inspection frequency is higher than the second inspection frequency; meteorological environmental data of the gas pipeline network is acquired, and the inspection method is determined based on the meteorological environmental data, wherein the inspection method includes a daytime inspection method and a nighttime inspection method.

[0039] Specifically, based on the risk level and external environmental conditions of different areas of the gas pipeline network, the detection mode of the unmanned inspection vehicle is dynamically configured to achieve differentiated investment of inspection resources and environmentally adaptable adjustment of detection strategies, thereby improving inspection efficiency and feasibility while ensuring the effectiveness of hazard identification.

[0040] Specifically, the inspection frequency refers to the time interval for repeated inspections of a certain location or area. The system first determines whether the currently processed GIS coordinates belong to the "high-risk area coordinates" marked in historical hazard data. If so, the inspection frequency for that area is set as the first inspection frequency (e.g., once a day or once every 12 hours); if not, it is set as the second inspection frequency (e.g., once a week or once every 72 hours), with the first inspection frequency being higher than the second to reflect the focus on monitoring high-risk areas. The inspection method refers to the specific operational mode of the inspection, mainly divided into daytime inspection and nighttime inspection. The system acquires current or forecast meteorological environmental data (including light intensity, visibility, wind speed, precipitation, temperature, and humidity), and selects a more suitable inspection period accordingly. For example, under conditions of heavy rainfall, dense fog, or low light, visible light imaging is poor, so daytime inspections may be prioritized; while when high temperatures and strong sunlight cause severe surface heat interference, infrared detection is more suitable for nighttime, and nighttime inspection can be selected. Furthermore, some acoustic or gas sensors are more sensitive at night when background noise is low, thus supporting nighttime operations. By combining meteorological data with dynamic decision-making, the sensors are ensured to operate under optimal environmental conditions, improving detection reliability.

[0041] For example, point G on a certain road section was marked by historical data as an area with frequent third-party construction, indicating a high risk of potential hazards. Therefore, the system set its inspection frequency to the highest level – once a day. Meanwhile, weather forecasts indicate sustained high temperatures (surface temperatures exceeding 45°C) during the day for the next three days, potentially interfering with infrared thermal imaging's identification of low-temperature leak patches. Based on this, the system decided to adopt a "nighttime inspection mode" for this point, scheduling unmanned inspection vehicles to perform tasks between 10:00 PM and 2:00 AM the following day. During this time, the ambient temperature is stable, and background noise is low, which is conducive to the optimal performance of infrared and acoustic sensors. On another road section, point H has no historical hazard records, so the system set its inspection frequency to the second highest level – twice a week. Due to recent sunny weather and ample sunlight, a "daytime inspection mode" was adopted, fully utilizing visible light cameras to identify abnormal excavation or subsidence signs on the surface. Through this mechanism, the detection mode can reflect regional risk differences and adapt to environmental changes, achieving accurate, efficient, and reliable gas pipeline network inspections.

[0042] In some embodiments of this application, the multi-sensor system includes: lidar, multispectral camera, inertial navigation system, gas concentration sensor array, and acoustic sensor.

[0043] Specifically, lidar is used to scan the vehicle's surroundings in real time, build a 3D point cloud map with centimeter-level accuracy, identify terrain and obstacles, and detect abnormal surface uplift or subsidence.

[0044] The multispectral cameras include: a high-resolution visible light camera: used to capture high-definition images of the ground surface to identify third-party construction, illegal encroachment, vegetation status, and the integrity of marker posts; an infrared thermal imager: used to detect abnormal surface temperature distribution and locate soil temperature changes caused by underground pipeline leaks or abnormal temperatures in the pipeline itself; and an ultraviolet imager: used in specific modes to directly observe the "ultraviolet plume" generated by minute leaks in high-pressure natural gas pipelines.

[0045] Inertial navigation systems provide high-frequency vehicle attitude, acceleration, and angular velocity information, complementing GNSS, ensuring positioning continuity in signal-blocked areas, and assisting in sensor data alignment.

[0046] The gas concentration sensor array includes a high-sensitivity, wide-range methane sensor: the core detection element, typically employing tunable diode laser absorption spectroscopy or catalytic combustion principles to accurately measure methane concentration. An auxiliary gas concentration gradient sensor array consists of multiple low-cost, fast-response gas sensors positioned at the front, rear, left, and right of the vehicle body to initially determine the approximate direction of the leak source (through concentration gradient comparison).

[0047] Acoustic sensors include ultrasonic microphones / sensors: used to capture high-frequency ultrasonic waves (typically 40-100 kHz) generated when gas leaks occur in pipelines, a frequency band less affected by ambient noise. Acoustic imagers or microphone arrays: used to spatially locate and visualize the leak sound source, forming an "acoustic camera" image that visually displays the leak point. Ground vibration sensors: used to monitor micro-vibrations in the ground caused by leaks or third-party construction.

[0048] In some embodiments of this application, anomaly feature extraction is performed on multi-source detection data, including: image feature extraction from the geospatial data, visible light and infrared image data to obtain surface subsidence features and third-party construction features; gas feature extraction from the gas concentration gradient data to obtain gas concentration abrupt change features and gas continuous leakage trend features; and acoustic feature extraction from the acoustic feature data to obtain gas leakage spectrum features and abnormal sound source features.

[0049] Specifically, the system extracts discriminative abnormal features from multi-source detection data, transforming the original perceived information into quantifiable and comparable hazard characterization indicators, providing a structured basis for accurate matching of hazard types and scientific determination of risk levels.

[0050] Specifically, the spatiotemporally aligned and fused multi-source detection data is classified and processed, with targeted feature extraction methods implemented according to data type. First, for geospatial data and visible and infrared image data, image analysis techniques are used for feature extraction. By comparing the current landform with historical benchmark models or adjacent area conditions, abnormal changes in ground elevation are identified, thus obtaining surface subsidence characteristics. Simultaneously, through target detection or texture analysis methods, typical markers such as construction machinery, fencing, and newly excavated trenches are identified in visible light images, or abnormal heat distribution caused by human disturbance is detected in infrared images, thereby extracting third-party construction features. Second, for gas concentration gradient data, its spatial and temporal variation patterns are analyzed. If the gas concentration increases sharply over a short distance (e.g., an increase exceeding a set threshold per meter), a sudden gas concentration change is identified; if a gas concentration consistently higher than the background value is detected at a fixed location without a significant attenuation trend, a continuous gas leakage trend is identified. These two types of features together reflect the intensity and stability of the leakage. Finally, for acoustic feature data, spectral analysis is performed on the collected audio signals. When gas leaks under high pressure, it will produce a high-frequency whistling sound in a specific frequency range (usually 20–50 kHz). The system extracts the spectral characteristics of gas leakage by identifying the energy concentration phenomenon in this frequency band. In addition, if sudden or periodic abnormal sounds (such as knocking or drilling sounds) that are not in the environmental background are detected, they are marked as abnormal sound source characteristics to help determine human interference or pipeline structural damage.

[0051] For example, during an inspection, when an unmanned inspection vehicle passed through a certain road section, visible light images showed newly piled building materials and yellow construction barriers on the roadside. The image analysis module extracted "third-party construction features" based on this. Simultaneously, point cloud data generated by the lidar showed that the local ground in this area had subsided by about 8 centimeters compared to the surrounding area, forming "surface subsidence features." The gas sensor array recorded a sudden increase in methane concentration from 2 ppm to 35 ppm within a 3-meter distance, indicating the presence of "gas concentration abrupt change features." Subsequently, after staying in place for 10 seconds, the concentration remained above 30 ppm, further confirming "gas continuous leakage trend features." The acoustic sensor simultaneously captured a high-frequency noise with a center frequency of 28 kHz lasting for 5 seconds. Spectral analysis confirmed that it matched the acoustic model of high-pressure gas leakage, extracting "gas leakage spectral features." In addition, intermittent metallic knocking sounds were detected and marked as "abnormal sound source features."

[0052] In some embodiments of this application, the preliminary identification of potential hazards based on the aforementioned abnormal features and the determination of initial hazard points and their corresponding classifications include: constructing a mapping relationship library between abnormal features and hazard types, wherein hazard types include pipeline leakage, third-party construction interference, surface subsidence threats, and abnormal sound source interference; inputting the extracted surface subsidence features, third-party construction features, gas concentration mutation features, gas continuous leakage trend features, gas leakage spectrum features, and abnormal sound source features into the mapping relationship library to match and obtain the corresponding preliminary hazard types; constructing a classification index system based on the confidence value and feature intensity of each abnormal feature, wherein the confidence value is determined by the combined acquisition accuracy of the multi-sensor system and the accuracy of the feature extraction algorithm, and the feature intensity is quantified by the deviation of the feature parameters from a preset threshold; calculating the comprehensive risk score of the initial hazard point according to the classification index system; classifying the initial hazard point into three risk levels—high, medium, and low—according to the comprehensive risk score, and associating it with the corresponding hazard classification label.

[0053] Specifically, the extracted multi-dimensional abnormal features are transformed into structured hazard assessment results. By matching hazard types and quantifying risk levels, "preliminary hazard points" with clear classification labels and risk levels are formed, providing a reliable basis for subsequent verification, detection, and emergency response.

[0054] Specifically, the system first constructs a mapping database between abnormal features and hazard types. This database predefines logical association rules between typical hazard types (including pipeline leaks, third-party construction interference, land subsidence threats, and abnormal sound source interference) and various abnormal features. For example, "gas concentration mutation feature + gas leak spectrum feature" primarily points to "pipeline leak"; "third-party construction feature + abnormal sound source feature" is more likely to correspond to "third-party construction interference"; and "land subsidence feature" appearing alone is associated with "land subsidence threat". After obtaining the six types of abnormal features extracted (land subsidence feature, third-party construction feature, gas concentration mutation feature, gas continuous leakage trend feature, gas leak spectrum feature, and abnormal sound source feature), the system inputs them into the mapping database and uses rule matching or pattern recognition methods to determine the most likely initial hazard type. Subsequently, the system quantifies the risk of the initially identified hazard points based on a hierarchical indicator system. This system comprehensively considers two key factors: first, the confidence value of each anomalous feature, which is jointly determined by the acquisition accuracy of the corresponding sensor (such as the resolution of a gas sensor and the clarity of an image sensor) and the accuracy of the feature extraction algorithm (such as the recall rate of a target detection model); second, the feature intensity of each anomalous feature, which is quantified by comparing the actual feature parameters (such as sedimentation, concentration change rate, and sound energy amplitude) with preset thresholds and calculating their deviation. The system performs weighted fusion based on the confidence and intensity of each feature to calculate the comprehensive risk score of the initially identified hazard point. According to the preset scoring range, the hazard point is divided into three risk levels: high, medium, and low, and this level is bound to the aforementioned preliminary hazard type to form a complete preliminary judgment result, including the hazard location, type label, and risk level.

[0055] For example: The following features were extracted from a certain inspection point: third-party construction features (90% confidence, high feature strength, and the excavator was clearly identified in the image); abnormal sound source features (85% confidence, regular drilling sounds were detected); gas concentration abrupt change features (80% confidence, the concentration increased from 1 ppm to 25 ppm within 3 seconds); gas leak spectrum features (75% confidence, the energy in the 28 kHz frequency band was significantly enhanced).

[0056] The mapping database matching results show that both construction interference and leakage signs exist simultaneously, with the most likely hazard type being "pipeline leakage caused by third-party construction." The grading index system weights the four valid features (high-confidence features have higher weights), resulting in a comprehensive risk score of 86 (out of 100). Based on the classification criteria (≥80 for high risk, 50–79 for medium risk, <50 for low risk), this point is classified as high risk, and the initial hazard information is output as: "Hazard type: Pipeline leakage (third-party construction interference); Risk level: High."

[0057] In some embodiments of this application, a surround detection and multi-directional measurement are performed on the initially identified potential hazard point to determine whether the initially identified potential hazard point is a leakage source. This includes: automatically performing a ring-shaped multi-directional moving detection around the initially identified potential hazard point with a radius of 2-5 meters; repeatedly collecting gas concentration and acoustic data at at least 8 equally divided points along the ring path; calculating the three-dimensional coordinates of the initially identified potential hazard point using a triangulation algorithm based on the multi-directional measurement data, and constructing a leakage source probability model by combining the gas concentration gradient change curve and the acoustic signal attenuation law; when the output value of the leakage source probability model is greater than a preset probability threshold, the initially identified potential hazard point is determined to be a leakage source, and the precise three-dimensional coordinates, gas concentration peak value, and diffusion direction of the leakage source are recorded; if the output value is less than or equal to the preset probability threshold, it is determined to be a non-leakage source.

[0058] Specifically, proactive and refined secondary verification is conducted on initially identified potential hazards. Through comprehensive, multi-directional data collection and fusion analysis, it is scientifically determined whether the hazard is a real gas leak source, thereby effectively eliminating false alarms and improving the accuracy and reliability of hazard confirmation.

[0059] Specifically, once a preliminary hazard point is identified, the unmanned detection vehicle automatically suspends its routine inspection tasks, drives to the vicinity of the point, and plans a circular detection path with a radius of 2 to 5 meters centered on the preliminary hazard point. This radius can be dynamically adjusted according to the site environment (such as road width and obstacle distribution) or the preliminary risk level to ensure coverage of the potential leak impact area. Along this circular path, the system sets at least eight equally spaced sampling points (i.e., one sampling point every 45 degrees), stopping at each point and simultaneously and repeatedly collecting gas concentration and acoustic data. Multiple, multi-angle measurements help capture the spatial distribution characteristics of the leak signal, avoiding misjudgments due to accidental interference from a single point. Subsequently, a comprehensive analysis is performed based on the collected multi-directional data: firstly, using a triangulation algorithm, combined with the peak gas concentration or sound source intensity at each measuring point, the three-dimensional spatial coordinates of the leak source are deduced; simultaneously, the trend of gas concentration variation along the circular path is analyzed to form a concentration gradient curve, and the attenuation law of acoustic signals with increasing distance is examined. These physical characteristics are input into a pre-built leak source probability model. This model, by integrating indicators such as concentration gradient steepness, sound source direction consistency, and location convergence, outputs a probability value indicating that "this point is a real leak source." If this probability value is greater than a preset probability threshold (e.g., 85%), the initially identified potential hazard point is determined to be a leak source, and its precise three-dimensional coordinates, the detected gas concentration peak, and the gas diffusion direction inferred from the wind direction and concentration distribution are recorded. If the probability value is less than or equal to the threshold, the current anomaly is considered to be caused by interfering factors (such as residual odor, environmental noise, or image misidentification), and it is determined to be a non-leak source, without triggering subsequent emergency procedures.

[0060] For example, a potential hazard point is initially identified as being located next to a city sidewalk. The system controls an unmanned detection vehicle to generate a circular path with a radius of 3 meters centered on this point, and collects data sequentially at 8 points. The measurement results show that the gas concentration at points 3, 4, and 5 is significantly higher (42 ppm, 58 ppm, and 51 ppm, respectively), while the concentration at the other points is below 5 ppm. The acoustic sensor detects concentrated energy in the 25–30 kHz frequency band at the same three points. Based on this, the triangulation algorithm calculates that the coordinates of the leak source deviate from the initial identification point by only 0.3 meters; the gas concentration gradient shows a clear single-peak distribution, and the acoustic signal attenuation conforms to the point source diffusion law. The leak source probability model, combining the above information, outputs a probability value of 91%, exceeding the 85% threshold. Therefore, the system confirms that this point is the actual leak source and records its precise location, peak concentration of 58 ppm, and diffusion direction as northwest (combined with real-time wind direction).

[0061] In some embodiments of this application, when the initial suspected hazard point is determined to be a leakage source, the risk value of the surrounding environment is assessed, including: analyzing the population density, building distribution and traffic flow within a preset range around the leakage source, assessing the leakage diffusion trend based on real-time wind speed and direction, and determining the risk value of the surrounding environment.

[0062] Specifically, after confirming that the initial hazard point is a real leak source, a comprehensive assessment is conducted on the actual degree of harm it may cause to the surrounding public safety and environment, generating a quantifiable "surrounding environment risk value" to provide a scientific and objective basis for subsequent graded emergency response.

[0063] Specifically, once a preliminary hazard point is identified as a leak source, the surrounding environmental risk assessment process is immediately initiated. First, a pre-defined assessment area centered on the leak source (e.g., a radius of 300 or 500 meters) is defined. This area can be adjusted based on pipeline pressure levels, gas type, or urban area attributes. Within this area, the system accesses multi-source urban basic data and real-time environmental information: Population density: By accessing urban population heat maps or mobile signaling data, the degree of population concentration in the area during the current time period is obtained; Building distribution: Using GIS building layers, the number, type, and airtightness of sensitive locations such as residences, schools, hospitals, and shopping malls are identified (e.g., basements and underground garages are prone to gas accumulation); Traffic flow: Combining traffic monitoring or navigation platform data, road traffic density is determined, and the difficulty of evacuation and the risk of potential ignition sources (e.g., vehicle engines) are assessed; Real-time wind speed and direction: Current wind conditions are obtained through a meteorological interface or a vehicle-mounted micro-weather station to simulate the main direction and speed of gas diffusion. Based on the above factors, a comprehensive risk assessment model is constructed: population exposure, sensitive building weight, and traffic activity are used as static risk factors, while wind direction and wind speed (which affect dilution rate) are used as dynamic correction factors. A normalized "surrounding environment risk value" (e.g., 0–100 points) is obtained through weighted calculation. The higher the value, the more serious the public safety consequences that a leak event may cause.

[0064] In some embodiments of this application, an emergency response is initiated based on the surrounding environmental risk value, and the inspection route is updated based on the leak source. This includes: when the surrounding environmental risk value is higher than a first emergency threshold, a Level 1 emergency response is automatically triggered, including uploading the leak source coordinates and risk assessment report to the monitoring center in real time, simultaneously activating the audible and visual alarm device, and pushing evacuation warning information to smart terminals within a 500-meter radius via the vehicle communication module; when the surrounding environmental risk value is between a second emergency threshold and a first emergency threshold, a Level 2 emergency response is initiated, continuously monitoring the leak source and uploading dynamic data, while simultaneously dispatching the nearest emergency repair team to the site; when the surrounding environmental risk value is lower than the second emergency threshold, a Level 3 emergency response is initiated, generating a hazard handling work order and incorporating it into the regular maintenance plan; when a leak source is identified, the leak source is marked as a key re-inspection node, prioritized for insertion into the current inspection sequence, the inspection order of subsequent routes is adjusted, and the inspection frequency of the leak source is increased.

[0065] Specifically, based on the level of risk to the surrounding environment caused by the leak source, a matching graded emergency response measure is automatically triggered, and the subsequent inspection tasks of the unmanned detection vehicle are optimized simultaneously to achieve a closed-loop management of "risk-response-tracking" that not only ensures public safety but also improves operation and maintenance efficiency.

[0066] Specifically, after confirming the leak source and calculating the risk value of the surrounding environment, the system compares this risk value with two preset thresholds and executes the corresponding level of emergency response: Level 1 Emergency Response: When the risk value is higher than the first emergency threshold (e.g., ≥80 points), it indicates a significant public safety threat. The system immediately and automatically executes several high-priority actions: uploading the precise coordinates of the leak source and the risk assessment report (including population, buildings, wind direction, etc.) to the city's gas monitoring center in real time; simultaneously activating the audible and visual alarm device on the unmanned detection vehicle, emitting warning sounds and flashing lights to alert on-site personnel; and simultaneously pushing evacuation warning information to smart terminals (such as community apps, property management systems, and residents' mobile phones) within a 500-meter radius of the leak source via the vehicle's communication module (such as 5G, NB-IoT, or a private network), prompting users to avoid open flames and evacuate quickly. Level 2 Emergency Response: When the risk value is between the second and first emergency thresholds (e.g., 50–79 points), it indicates a moderate risk, requiring manual intervention but not large-scale evacuation. The system continuously monitors the leak source in situ or within a small area, uploading dynamic data such as gas concentration and acoustic signals in real time. Simultaneously, it automatically dispatches the nearest repair team, pushing task instructions and on-site information to ensure timely response. Level 3 Emergency Response: When the risk value is below the second emergency threshold (e.g., <50 points), it indicates a weak leak or a low-sensitivity area, with manageable risk. The system automatically generates a structured hazard handling work order, including location, type, and risk level information, and incorporates it into the routine maintenance plan for sequential processing by the operations and maintenance department. Furthermore, regardless of the risk level, once a real leak source is confirmed, the system marks that location as a "key re-inspection node" in the background. In current or subsequent inspection tasks, this node is prioritized in the inspection sequence, adjusting the original path order to ensure early re-inspection, and increasing its inspection frequency (e.g., from once a week to multiple times a day) to verify the effectiveness of repairs or monitor for recurrence.

[0067] For example, if a leak source is assessed as having a risk score of 85, exceeding the first emergency threshold (80 points), the system immediately triggers a Level 1 response: an alarm window pops up on the monitoring center's large screen, displaying a map of the leak point and risk details; the unmanned detection vehicle activates its red and blue flashing lights and plays a voice message saying "Gas leak, do not approach"; over 200 households within a 500-meter radius receive an app notification: "Gas leak detected on XX Road. Please close the gas valve, open windows for ventilation, avoid using electrical appliances, and follow community instructions." Simultaneously, the leak point is marked as a key re-inspection node. The unmanned detection vehicle, originally scheduled to continue patrolling the eastern area, automatically plans its return route after receiving the alarm and returns to the point two hours later for re-testing; within the next 48 hours, the system schedules inspections every six hours until the concentration is confirmed to be zero and repairs completed. Through this process, precise grading of emergency response and dynamic optimization of inspection strategies are achieved, effectively supporting the closed-loop safety management of the urban gas pipeline network.

[0068] like Figure 2 As shown, this invention discloses a gas pipeline network hidden danger identification system based on an unmanned inspection vehicle, used to apply the aforementioned gas pipeline network hidden danger identification method based on an unmanned inspection vehicle. The system includes: The inspection control module is configured to receive GIS coordinates and historical hazard data of the gas pipeline network, generate inspection paths and detection modes for unmanned inspection vehicles based on the GIS coordinates and historical hazard data, and control the unmanned inspection vehicles to perform inspections according to the inspection paths and detection modes. The data acquisition module is configured to acquire multi-source detection data through the multi-sensor system configured on the unmanned inspection vehicle during inspection. The multi-source detection data includes geospatial data, visible light and infrared image data, gas concentration gradient data and acoustic feature data. The hazard identification module is configured to perform spatiotemporal alignment and fusion processing on the multi-source detection data, extract abnormal features from the multi-source detection data, and perform preliminary identification of hazards based on the abnormal features to determine the initial hazard points and their corresponding hierarchical classifications. The hazard verification and response module is configured to perform surround detection and multi-directional measurement on the initially identified hazard point to determine whether the initially identified hazard point is a leak source. If it is a leak source, the module assesses the risk value of the surrounding environment, initiates an emergency response based on the risk value of the surrounding environment, and updates the inspection path based on the leak source.

[0069] This invention achieves full automation and intelligence in the identification of potential hazards in gas pipeline networks through multi-module collaboration. The inspection control module generates personalized inspection paths and detection modes based on GIS coordinates and historical hazard data, ensuring that unmanned inspection vehicles focus on high-risk areas, improving the targeting and efficiency of inspections. The data acquisition module utilizes multi-sensor systems to collect geospatial, image, gas, and acoustic data, providing a comprehensive and rich information foundation for hazard identification and avoiding the limitations of single data types. The hazard identification module, through spatiotemporal alignment fusion and abnormal feature extraction, combined with a hierarchical classification mechanism, can accurately locate preliminary hazard points, significantly improving the accuracy of hazard identification and early warning capabilities. The hazard verification and response module's linkage of surround detection, leak source confirmation, environmental risk assessment, and emergency response not only effectively avoids misjudgments but also rapidly activates appropriate emergency measures after leak confirmation and dynamically adjusts the inspection path, forming a complete closed-loop management system from hazard discovery, identification, verification, to emergency response. This significantly improves the safety of gas pipeline network operation and the timeliness of emergency response, reducing the probability of accidents and potential losses.

[0070] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program goods. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program goods embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0071] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program goods according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0072] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0073] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles, characterized in that, include: Receive GIS coordinates and historical hazard data of the gas pipeline network, generate inspection path and detection mode of unmanned inspection vehicle based on the GIS coordinates and historical hazard data, and control unmanned inspection vehicle to carry out inspection according to the inspection path and detection mode; During inspections, the unmanned inspection vehicle acquires multi-source detection data through a multi-sensor system configured on the vehicle. The multi-source detection data includes geospatial data, visible light and infrared image data, gas concentration gradient data, and acoustic feature data. The multi-source detection data is spatiotemporally aligned and fused, and abnormal features are extracted from the multi-source detection data. Based on the abnormal features, the potential hazards are initially identified, and the initial hazard points and their corresponding hierarchical classifications are determined. Surround detection and multi-directional measurement are carried out on the initially identified potential hazard points to determine whether the initially identified potential hazard points are leakage sources. If they are leakage sources, the risk value of the surrounding environment is assessed. An emergency response is initiated based on the risk value of the surrounding environment, and the inspection route is updated based on the source of the leak.

2. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, Based on the GIS coordinates and historical hazard data, an inspection path for the unmanned inspection vehicle is generated, including: The historical hazard data includes the coordinates of areas with a high incidence of hazards; Based on the GIS coordinates, the gas pipeline network is divided into several regions. Each region includes only the coordinates of a high-risk area, and the regions do not overlap. Obtain the starting coordinates of the unmanned inspection vehicle, and generate the shortest inspection path to traverse each area based on the starting coordinates; Based on the coordinates of high-risk areas in each region, an intra-regional inspection path is generated, and the initial coordinates of the intra-regional inspection path are the coordinates of the high-risk areas. The inspection path of the unmanned inspection vehicle is obtained by merging the shortest inspection path and the inspection path within the area.

3. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 2, characterized in that, Based on the aforementioned GIS coordinates and historical hazard data, an unmanned detection vehicle's detection mode is generated, including: The detection mode includes inspection frequency and inspection method; If the GIS coordinates are the coordinates of a high-risk area, then the inspection frequency is set to the first inspection frequency; otherwise, it is set to the second inspection frequency, with the first inspection frequency being higher than the second inspection frequency. Obtain meteorological environmental data of the gas pipeline network, and determine the inspection method based on the meteorological environmental data. The inspection method includes daytime inspection method and nighttime inspection method.

4. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, The multi-sensor system includes: lidar, multispectral camera, inertial navigation system, gas concentration sensor array, and acoustic sensor.

5. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, Anomaly feature extraction is performed on multi-source detection data, including: Image feature extraction is performed on the geospatial data, visible light and infrared image data to obtain surface subsidence features and third-party construction features; Gas feature extraction is performed on the gas concentration gradient data to obtain gas concentration abrupt change characteristics and gas continuous leakage trend characteristics; Acoustic features are extracted from the acoustic feature data to obtain the gas leak spectrum features and abnormal sound source features.

6. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 5, characterized in that, Based on the aforementioned abnormal characteristics, preliminary identification of potential hazards is performed to determine the initial hazard points and their corresponding classifications, including: Construct a mapping database between abnormal features and hazard types, where hazard types include pipeline leakage, third-party construction interference, surface subsidence threat, and abnormal sound source interference; The extracted surface subsidence features, third-party construction features, gas concentration abrupt change features, gas continuous leakage trend features, gas leakage spectrum features, and abnormal sound source features are input into the mapping relationship library to match and obtain the corresponding preliminary hazard type; A hierarchical index system is constructed based on the confidence value and feature intensity of each abnormal feature. The comprehensive risk score of the initially identified hidden danger points is calculated according to the hierarchical index system. The initially identified hidden danger points are divided into three risk levels: high, medium and low, according to the comprehensive risk score, and associated with the corresponding hidden danger classification labels.

7. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, Surround detection and multi-directional measurement are performed on the initially identified potential hazard points to determine whether the initially identified hazard points are leakage sources, including: The system automatically performs a circular multi-directional moving detection with a radius of 2-5 meters centered on the initially identified potential hazard point; gas concentration and acoustic data are repeatedly collected at at least 8 equally divided points along the circular path; Based on multi-directional measurement data, a triangulation algorithm is used to calculate the three-dimensional coordinates of the initially identified potential hazard point. A leak source probability model is constructed by combining the gas concentration gradient change curve and the acoustic signal attenuation law. When the output value of the leak source probability model is greater than the preset probability threshold, the initially identified potential hazard point is determined to be a leak source. At the same time, the precise three-dimensional coordinates, gas concentration peak value, and diffusion direction of the leak source are recorded. If the output value is less than or equal to the preset probability threshold, it is determined to be a non-leak source.

8. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, When the initial suspected hazard point is determined to be the leakage source, the risk value of the surrounding environment is assessed, including: analyzing the population density, building distribution and traffic flow within a preset range around the leakage source, assessing the leakage diffusion trend based on real-time wind speed and direction, and determining the risk value of the surrounding environment.

9. The method for identifying potential hazards in gas pipeline networks based on unmanned inspection vehicles according to claim 1, characterized in that, An emergency response is initiated based on the surrounding environmental risk value, and the inspection route is updated based on the leak source, including: When the risk value of the surrounding environment exceeds the first emergency threshold, a Level 1 emergency response is automatically triggered, including uploading the coordinates of the leak source and the risk assessment report to the monitoring center in real time, activating the audible and visual alarm device simultaneously, and pushing evacuation warning information to smart terminals within a 500-meter radius through the vehicle communication module. When the risk value of the surrounding environment is between the second emergency threshold and the first emergency threshold, a level-two emergency response is initiated, continuous monitoring of the leak source is carried out and dynamic data is uploaded, and the nearest emergency repair team is dispatched to the site. When the risk value of the surrounding environment is lower than the second emergency threshold, a level 3 emergency response is initiated, a hazard handling work order is generated, and it is included in the routine maintenance plan. When a leak source is identified, it is marked as a key re-inspection node, prioritized for insertion into the current inspection sequence, the inspection order of subsequent paths is adjusted, and the inspection frequency of the leak source is increased.

10. A gas pipeline network hazard identification system based on an unmanned inspection vehicle, used to apply the gas pipeline network hazard identification method based on an unmanned inspection vehicle as described in any one of claims 1-9, characterized in that, The system includes: The inspection control module is configured to receive GIS coordinates and historical hazard data of the gas pipeline network, generate inspection paths and detection modes for unmanned inspection vehicles based on the GIS coordinates and historical hazard data, and control the unmanned inspection vehicles to perform inspections according to the inspection paths and detection modes. The data acquisition module is configured to acquire multi-source detection data through the multi-sensor system configured on the unmanned inspection vehicle during inspection. The multi-source detection data includes geospatial data, visible light and infrared image data, gas concentration gradient data and acoustic feature data. The hazard identification module is configured to perform spatiotemporal alignment and fusion processing on the multi-source detection data, extract abnormal features from the multi-source detection data, and perform preliminary identification of hazards based on the abnormal features to determine the initial hazard points and their corresponding hierarchical classifications. The hazard verification and response module is configured to perform surround detection and multi-directional measurement on the initially identified hazard point to determine whether the initially identified hazard point is a leak source. If it is a leak source, the module assesses the risk value of the surrounding environment, initiates an emergency response based on the risk value of the surrounding environment, and updates the inspection path based on the leak source.

Images