Pipeline micro-leakage fusion positioning method and system based on laser absorption spectrum and multi-modal imaging

Abstract

This invention relates to a method and system for fusion-based localization of pipeline micro-leaks based on laser absorption spectroscopy and multimodal imaging. It constructs a pixel-level spatial location-spectral concentration temporal coupling mapping relationship, breaking through the traditional discrete detection logic. It utilizes spectral concentration gradients to inversely constrain suspected areas in the image, while simultaneously using temporal changes in image texture to correct spectral drift noise. Addressing the unique characteristics of oil and gas micro-leaks—"low concentration, weak boundary, background-like appearance, and no obvious deformation"—a spectral threshold adaptive multi-scale weight decomposition mechanism is designed. The decomposition scale is adaptively adjusted based on the real-time methane concentration signal-to-noise ratio of the micro-leak. A spectral feature mask is introduced to eliminate interference from soil moisture, vegetation temperature difference, light changes, and salinity, achieving multi-scale adaptive enhancement of weak micro-leak features and removal of false anomalies. A methane concentration field-topographic image fusion visualization reconstruction graphic is constructed, intuitively presenting the correspondence between the leak source location, diffusion range, and surface anomalies.

Description

Technical Field

[0001] This invention relates to a method and system for fusion localization of pipeline microleakage based on laser absorption spectroscopy and multimodal imaging, belonging to the fields of optoelectronic technology and optical imaging technology. Background Technology

[0002] Long-distance onshore oil and gas pipelines have long service lifespans, and many pipelines suffer from problems such as pipe corrosion, uneven geological settlement, third-party mechanical damage, welding fatigue failure, and flange seal aging, making them highly susceptible to early micro-leakage that is not visible to the naked eye. Current mainstream leak monitoring solutions in the industry have three fundamental shortcomings:

[0003] First, technologies such as pressure wave method, flow balance method, negative pressure wave method, and pipeline inspection robot only produce significant hydraulic parameter disturbances for medium and large-scale leaks. For micropores with extremely low volatile oil seepage, the pipeline pressure, flow rate, and acoustic signals show almost no characteristic changes, and they are completely unable to detect early hidden dangers. They can only be dealt with passively after the seepage expands and forms a large area of ​​soil soaked in oil, thus losing the window of opportunity for early intervention.

[0004] Secondly, drones equipped with single infrared thermal imaging are currently the most common method for field inspections. They rely on the temperature difference caused by changes in the surface thermal conductivity due to leaking oil to make judgments. However, they are easily affected by the heat effect of direct sunlight, low temperature areas with water accumulation, temperature differences between vegetation growth and decay, dust and fog, and tree shadows. They can only delineate vague abnormal areas and cannot prove from a material level that the heat spot is caused by oil and gas volatilization. The false alarm rate is high in complex field conditions, the workload of maintenance personnel for verification is extremely large, and the recognition reliability is insufficient.

[0005] Third, while TDLAS laser absorption spectroscopy alone can utilize the infrared characteristic absorption of hydrocarbon molecules to achieve specific detection of trace gases in oil and gas, with a detection sensitivity of up to ppb level, it can only output a one-dimensional concentration distribution curve along the pipeline and lacks two-dimensional spatial morphology information. It can only determine whether "oil and gas volatilization exists in the pipeline" and cannot provide the precise coordinates of the leak point, the seepage range and diffusion pattern. Its positioning accuracy is poor and it cannot guide fixed-point maintenance operations.

[0006] Meanwhile, most existing multispectral image fusion algorithms only perform visual overlay processing of pixel grayscale and edge features, without introducing molecular spectral substance identification truth values ​​for strong constraint verification. Essentially, they remain pure visual image recognition algorithms, with limited anti-interference capabilities in complex field environments. In summary, the industry urgently needs a comprehensive microleakage precise localization method integrating molecular qualitative tracing, multimodal spatial imaging localization, and spectral-image cross-logic verification. Summary of the Invention

[0007] To overcome the shortcomings of existing oil pipeline micro-leakage monitoring, such as insufficient sensitivity, poor environmental adaptability of single detection systems, inaccurate spatial positioning due to spectral detection, and high false alarm rate of visual recognition, this invention provides a pipeline micro-leakage fusion positioning method and system based on laser absorption spectroscopy and multimodal imaging. It constructs a pixel-level spatial location-spectral concentration temporal coupling mapping relationship, breaking through the traditional discrete detection logic. It utilizes spectral concentration gradients to inversely constrain suspected areas in the image, while simultaneously using temporal changes in image texture to correct spectral drift noise. Targeting the unique characteristics of oil and gas micro-leakage—"low concentration, weak boundary, background-like appearance, and no obvious deformation"—an adaptive multi-scale weighted decomposition mechanism for spectral thresholds is designed. The decomposition scale is adaptively adjusted according to the real-time methane concentration signal-to-noise ratio of the micro-leakage. A spectral feature mask is introduced to remove interference from soil moisture, vegetation temperature difference, light changes, and salinity, achieving multi-scale adaptive enhancement and pseudo-anomaly removal of weak micro-leakage features. A methane concentration field-topographic image fusion visualization reconstruction graphic is constructed, intuitively presenting the relationship between the leak source location, diffusion range, and surface anomalies.

[0008] This invention enables qualitative identification of micro-leaking substances at the pipeline incipient stage, pixel-level location locking, and quantitative classification of leakage intensity. It improves the reliability of hazard identification in complex field scenarios from the underlying algorithm logic, and truly implements the pre-emptive prevention and control of oil and gas pipeline leakage risks.

[0009] The technical solution of the present invention is as follows:

[0010] The first aspect of the present invention provides a method for fusion localization of pipeline microleakage based on laser absorption spectroscopy and multimodal imaging, comprising:

[0011] Step 1: Construct a mobile observation platform to inspect oil pipelines and collect multimodal data of the oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images.

[0012] Step 2: Preprocess and register the acquired visible light image, infrared temperature field matrix, and laser active absorption imaging image;

[0013] Step 3: Perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and obtain suspected leakage candidate areas;

[0014] Step 4: Extract the pixel center coordinates of the suspected leak candidate area and map them to the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit. Based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory, determine whether the suspected leak candidate area is a real early micro-leak point of the pipeline.

[0015] Step 5: Calculate the path integral methane concentration corresponding to the actual early micro-leak point in the pipeline, obtain the oil and gas volatilization flux through the path integral methane concentration, divide the early warning into multiple levels according to the oil and gas volatilization flux, and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine the terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and mark the early warning information to complete the pipeline micro-leak location.

[0016] According to a preferred embodiment of the present invention, the specific implementation process of step 1 includes:

[0017] The mobile observation platform includes an unmanned aerial vehicle (UAV) inspection platform, equipped with a TDLAS laser absorption spectroscopy scanning unit, a visible light high-definition imaging unit, a mid- and long-wave infrared thermal imaging unit, and a near-infrared laser active illumination imaging unit.

[0018] The TDLAS laser absorption spectral scanning unit is used to acquire laser spectral absorption amplitude sequences, the visible light high-definition imaging unit is used to acquire visible light images, the mid- and long-wave infrared thermal imaging unit is used to acquire infrared temperature field matrices, and the near-infrared laser active illumination imaging unit is used to acquire laser active absorption imaging images.

[0019] The mobile observation platform inspects the oil pipeline. If the TDLAS laser absorption spectroscopy scanning unit detects an oil and gas volatilization cloud above the oil pipeline, it simultaneously collects multimodal data of the oil pipeline, including the laser spectral absorption amplitude sequence, visible light image, infrared temperature field matrix, and laser active absorption imaging image of the oil pipeline.

[0020] According to a preferred embodiment of the present invention, step 2 includes the following specific implementation process:

[0021] Preprocessing is performed in parallel on visible light images, infrared temperature field matrices, and laser active absorption imaging images, including bilateral filtering to remove Gaussian noise, dark channel prior algorithm to remove fog, and histogram equalization to improve details in low-contrast areas.

[0022] The Scale Invariant Feature Transform (SIFT) algorithm is used to perform coarse feature point matching on the preprocessed visible light image, infrared temperature field matrix, and active laser absorption imaging image. Then, geometric deviation is finely corrected by using a perspective transformation model. The visible light image, infrared temperature field matrix, and active laser absorption imaging image are uniformly projected onto the same geographic pixel coordinate system to complete pixel-level registration.

[0023] According to a preferred embodiment of the present invention, step 3 specifically includes the following steps:

[0024] For visible light images, infrared temperature field matrices, and laser active absorption imaging images that have completed pixel-level registration, wavelet basis is used to carry out multi-scale decomposition to obtain low-frequency background components and multi-scale high-frequency detail components, and to extract abnormal features, including edge contour abrupt regions, gray-level abnormal patches, infrared temperature distortion regions, and laser aerosol absorption high-brightness regions.

[0025] Among them, the edge contour abrupt change region includes: obtaining the local 8-neighbor gray-level gradient magnitude of each pixel in the visible light image. If the local gray-level change is greater than the change threshold and the number of connected pixels is greater than or equal to 50 pixels, it is regarded as the edge contour abrupt change region.

[0026] Gray-scale anomalous patches are defined as follows: the deviation between the average gray value of a pixel within a local window of each pixel in a visible light image and the average gray value of the surrounding background window is calculated. If the absolute value of the difference between the average gray value of a pixel in the local window and the average gray value of the background window is greater than 40, and the area of ​​the connected pixels is greater than 20 pixels, then it is considered a gray-scale anomalous patch.

[0027] The infrared temperature distortion region includes: the average temperature is calculated within a preset background window in the infrared temperature field matrix. If the absolute value of the temperature difference from the average temperature is greater than 1.5 degrees Celsius and the connected region is greater than 50 pixels, it is considered as the infrared temperature distortion region.

[0028] The high-brightness region of laser aerosol absorption includes: for each pixel in the laser active absorption imaging image, the absorption characteristics are compared with the surrounding neighborhood of the pixel, and the pixel with the largest absorption of laser by the oil and gas volatile gas cloud is taken as the high-brightness region of laser aerosol absorption.

[0029] The extracted abnormal features are fused, including spatially associating the visible light image, infrared temperature field matrix, and laser active absorption imaging image with the spatial location of the abnormal features; pixels with only one or two abnormal features at the same pixel location are removed, and static background interference is removed by background subtraction method to output suspected leakage candidate areas.

[0030] Further preferably, the range of the change threshold is 35-40.

[0031] According to a preferred embodiment of the present invention, step 4 includes the following specific implementation process:

[0032] Each suspected leakage candidate region includes N valid abnormal pixels, and the pixel center coordinates of the suspected leakage candidate region are (x, y, x). c , y c ) represents , ,in, , Represents the coordinates of the i-th valid abnormal pixel in the suspected leakage candidate region;

[0033] Read the pre-calibrated camera intrinsic parameter matrix and radial and tangential distortion parameters of the imaging unit corresponding to the suspected leakage candidate area, construct the pixel center coordinates into a homogeneous pixel vector, and use the inverse matrix of the intrinsic parameter matrix to obtain the normalized image plane coordinates carrying lens distortion; then use the distortion parameters to complete the distortion correction and obtain the normalized coordinates after eliminating lens distortion; take the camera optical center as the ray origin, combine the normalized coordinates with the unit z component to obtain the spatial ray direction vector in the camera coordinate system;

[0034] The installation extrinsic parameters of the imaging unit relative to the machine body, which are pre-calibrated by the mobile observation platform, are read, including the camera-machine body rotation matrix and translation vector. The spatial ray direction vector in the camera coordinate system is divided into the ray origin and the ray direction vector. The ray origin is taken as a three-dimensional spatial point and is transformed to the machine body coordinate system by performing rotation and translation transformations. The ray direction vector is only subjected to rotation transformation without translation operation to obtain the spatial ray in the machine body coordinate system.

[0035] The real-time attitude and GPS location information of the mobile observation platform are read to obtain the rotation matrix for transforming the body coordinate system to the geodetic ENU coordinate system and the geodetic origin position of the body. A rotation-translation transformation is performed on the ray origin in the body coordinate system, while the ray direction vector undergoes only a rotation transformation to obtain the spatial ray in the geodetic coordinate system. The spatial ray in the geodetic coordinate system is spatially intersected and matched with the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit to obtain the corresponding spatial points. The laser spectral absorption amplitude sequence of the spatial points is retrieved, and based on the laser spectral absorption amplitude sequence of the spatial points, it is determined whether the suspected leakage candidate area actually exists.

[0036] If there are suspected leakage candidate areas in the multimodal data, and the corresponding laser spectral absorption amplitude sequence detects characteristic absorption valleys of oil and gas hydrocarbons and the second harmonic demodulated oil and gas concentration is higher than the preset judgment threshold, then it is judged as a real early micro-leak point in the pipeline.

[0037] If there are only suspected leakage candidate areas, but there is no characteristic spectral response of oil and gas at the corresponding location, they are directly judged as environmental interference and eliminated.

[0038] More preferably, the preset judgment threshold is that the number of connected pixels is greater than 50 pixels and the concentration is greater than 5 ppm·m;

[0039] According to a preferred embodiment of the present invention, step 5 includes the following specific implementation process:

[0040] GPS geographic coordinates were bound to the actual early micro-leak points in the pipeline, and the laser spectral absorption amplitude sequence of the corresponding spatial points was retrieved. A wavelength modulation spectral second harmonic demodulation algorithm was used to perform digital phase-locked loop demodulation on the laser spectral absorption amplitude sequence to obtain the absorption peak depth. The calibration curve of the TDLAS laser absorption spectral scanning unit was read. The demodulated absorption peak depth was substituted into the calibration curve to invert and solve for the path integral methane concentration along the laser propagation path.

[0041] ;

[0042] in, The path integral is the methane concentration, H 2f Where is the absorption peak depth, k is the calibration slope, and b is the baseline offset;

[0043] If the calculated path integral methane concentration is lower than the preset detection limit, it is determined that there is no effective methane absorption at the actual early micro-leak point in the pipeline, and the actual early micro-leak point in the pipeline is removed; otherwise, the actual early micro-leak point in the pipeline is not removed.

[0044] Combining the flight altitude of the mobile observation platform, the pitch angle of the gimbal, and measured or estimated wind field parameters, a simplified Gaussian plume diffusion model is adopted to establish a mapping relationship between the path integral methane concentration and the actual early-stage micro-leakage flux in the pipeline. The mapping relationship is used to inversely deduce the oil and gas volatilization flux corresponding to the actual early-stage micro-leakage in the buried pipeline. According to the size of the oil and gas volatilization flux, three levels of early warning are defined: 0.5–2 L / d is the nascent warning level, 2–10 L / d is the slight leakage level, and greater than 10 L / d is the expansion risk level.

[0045] The infrared temperature field matrix output by the medium- and long-wave infrared thermal imaging unit is collected. Each element in the infrared temperature field matrix stores the measured surface temperature at the corresponding pixel location. Based on the boundary of the infrared temperature distortion zone, the temperature difference is judged pixel by pixel. When the pixel temperature difference decays to the background temperature noise threshold, the expansion stops. The boundary of the infrared temperature distortion zone to the position where the expansion stops is taken as the leakage temperature diffusion range.

[0046] Within the infrared temperature distortion region, a bicubic spline interpolation algorithm is used to spatially reconstruct discrete temperature sampling points within the leakage temperature diffusion range, resulting in a continuously distributed surface temperature field. This surface temperature field is then corrected by combining the spatial distribution of hydrocarbon concentrations corresponding to the laser spectral absorption amplitude sequence. This includes comparing the hydrocarbon concentration values ​​with the detection limit pixel-by-pixel in the surface temperature field; if the hydrocarbon concentration value and detection limit of the surface temperature field are lower than those of the hydrocarbon concentration value and detection limit of the spatial distribution, the corresponding position is set to 1; otherwise, it is set to 0, thus obtaining a hydrocarbon concentration confidence weight mask. The surface temperature field is then weighted and corrected using this mask, and the corrected temperature values ​​are superimposed on the surface temperature field to obtain the corrected surface temperature field. Finally, the temperature values ​​of the corrected surface temperature field are mapped to a preset color scale to generate a leakage diffusion heat map.

[0047] The GPS geographic coordinates, path integral methane concentration, infrared temperature field matrix, and terrain data of the actual early micro-leaking points in the pipeline are geospatially registered to construct a geographic grid, including a terrain elevation grid and a surface temperature grid. Each geographic grid stores geodetic coordinates, terrain elevation, and surface temperature information, and the geographic grid is resampled to the same spatial resolution as the infrared temperature field.

[0048] Based on the path integral methane concentration, geometric parameters observed by the TDLAS laser absorption spectroscopy scanning unit were collected, including the laser line-of-sight elevation angle and the effective vertical thickness of the leaking gas cloud. The effective optical path length of the laser passing through the methane gas cloud was calculated based on these geometric parameters. The path integral methane concentration was divided by the effective optical path length to obtain the equivalent average concentration of the methane gas cloud within the laser sampling area. Based on a simplified Gaussian plume diffusion model, combined with near-surface wind speed, atmospheric stability, and diffusion coefficient, the equivalent average concentration of the methane gas cloud was inversely extrapolated to solve for the equivalent near-surface initial gas concentration at the leak source outlet. The equivalent near-surface initial gas concentration at the leak source, environmental wind field parameters, and the distance passed through the gas cloud were then analyzed. The topographic elevation and surface temperature rasters registered in geospatial data are input using a Gaussian diffusion function. A topographic correction factor and a surface temperature convection correction factor are introduced. The entire geographic rasters in the inspection area are traversed, and the near-surface oil and gas concentration of each geographic raster is calculated point by point. The two-dimensional raster field of near-surface oil and gas concentration in the entire region is output. Among them, the topographic correction factor represents the blocking and confluence effect of topographic relief on oil and gas diffusion. The topographic correction factor is defined as 1 when the terrain is flat and 0 otherwise. The surface temperature convection correction factor represents the change in near-surface convection intensity caused by surface temperature difference. The surface temperature convection correction factor is defined as 1 when the surface temperature is equal to the background temperature and 0 otherwise.

[0049] Construct a visualization rendering sub-model, including mapping the oil and gas concentration values ​​of the two-dimensional raster field of near-surface oil and gas concentration in the whole region to gradient colors, overlaying topographic base map, pipeline vector information, leakage source location markers and leakage risk level indicators, and generating a leakage diffusion heat map with geographic information.

[0050] Mark the coordinates of the leak center point, the oil and gas diffusion boundary, and the warning level to form a standardized warning document. Push the warning document to the pipeline operation and maintenance management platform to complete the location of pipeline micro-leakage.

[0051] Further preferably, the calibration slope k is [0.001, 0.004]V / (ppm·m); the baseline offset b is [-0.02, 0.02]V; the preset detection limit is 1~5 ppm·m; and the background temperature noise threshold is 0.3 degrees Celsius or 0.5 degrees Celsius.

[0052] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging.

[0053] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging.

[0054] A second aspect of the present invention provides a pipeline micro-leakage fusion localization system based on laser absorption spectroscopy and multimodal imaging, comprising:

[0055] The multimodal data acquisition module is configured to: construct a mobile observation platform, conduct inspections of oil pipelines through the mobile observation platform, and collect multimodal data of oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images;

[0056] The preprocessing and registration module is configured to preprocess and register the acquired visible light images, infrared temperature field matrices, and laser active absorption imaging images.

[0057] The suspected leakage candidate area acquisition module is configured to: perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and acquire suspected leakage candidate areas.

[0058] The module for determining early-stage micro-leakage points in pipelines is configured to: extract the pixel center coordinates of suspected leak candidate areas and map them to the scanning trajectory of the TDLAS laser absorption spectral scanning unit; and determine whether the suspected leak candidate areas are early-stage micro-leakage points in pipelines based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory.

[0059] The early warning classification module is configured to: calculate the path integral methane concentration corresponding to the actual early micro-leakage point in the pipeline; obtain the oil and gas volatilization flux through the path integral methane concentration; classify the early warning into multiple levels according to the oil and gas volatilization flux; and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and label the early warning information to complete the pipeline micro-leakage location.

[0060] The beneficial effects of this invention are as follows:

[0061] 1. The detection sensitivity of this invention can reach the ppb level, which can capture trace oil seepage in the early stage of pipeline micropores and micro welds, filling the gap in early leakage detection of traditional hydraulic monitoring methods.

[0062] 2. Under complex field conditions, the false alarm rate is reduced by more than 70% compared to single infrared inspection, the reliability of hidden danger identification is greatly improved, and the cost of ineffective on-site verification by maintenance personnel is reduced.

[0063] 3. The leak point location accuracy reaches the sub-meter level, which can directly guide the fixed-point excavation and maintenance, reducing the manpower and material consumption of pipeline operation and maintenance.

[0064] 4. Multi-platform modular integration enables large-scale inspections of long-distance transmission lines and fixed-point monitoring of station equipment to be implemented.

[0065] 5. It can prevent major safety accidents such as soil / water pollution and explosion of flammable and explosive oil and gas caused by leakage in advance, and has significant environmental benefits and safety production value. Attached Figure Description

[0066] Figure 1 This is a flowchart of the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to the present invention;

[0067] Figure 2 This is a schematic diagram of the synchronous scanning optical path and on-site detection of the mobile observation platform of the present invention;

[0068] Figure 3 This is a flowchart illustrating the logic of multimodal feature fusion and laser spectrum cross-verification in this invention.

[0069] Figure 4 This is a schematic diagram of the typical characteristic absorption spectrum of oil and gas lasers. Detailed Implementation

[0070] The present invention will be further described below with reference to the embodiments and accompanying drawings, but is not limited thereto.

[0071] Example 1

[0072] Terminology Explanation:

[0073] 1. TDLAS (Tunable Semiconductor Laser Absorption Spectroscopy): Utilizes a tunable semiconductor laser to scan the characteristic absorption peaks of a gas, and quantitatively obtains the gas concentration based on the laser absorption signal, enabling long-distance, high-sensitivity, non-contact detection of hydrocarbon gases.

[0074] 2. Simplified Gaussian plume diffusion model: This is a simplified engineering version of the classic atmospheric diffusion model for continuous gas release from a continuous point source. It assumes that the gas concentration follows a Gaussian normal distribution on a cross section perpendicular to the average wind direction, ignores complex atmospheric turbulence parameters, and is suitable for UAV inspection of near-surface engineering applications.

[0075] 3. Gaussian diffusion function: It describes the distribution of gas concentration as it decreases with spatial distance after leaving the leak source due to turbulence. Essentially, it is a two-dimensional normal distribution function.

[0076] A method for fusion localization of pipeline microleakage based on laser absorption spectroscopy and multimodal imaging, such as Figure 1 As shown, it includes:

[0077] Step 1: Construct a mobile observation platform to inspect oil pipelines and collect multimodal data of the oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images.

[0078] Step 2: Preprocess and register the acquired visible light image, infrared temperature field matrix, and laser active absorption imaging image;

[0079] Step 3: Perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and obtain suspected leakage candidate areas;

[0080] Step 4: Extract the pixel center coordinates of the suspected leak candidate area and map them to the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit. Based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory, determine whether the suspected leak candidate area is a real early micro-leak point of the pipeline.

[0081] Step 5: Calculate the path integral methane concentration corresponding to the actual early micro-leak point in the pipeline, obtain the oil and gas volatilization flux through the path integral methane concentration, divide the early warning into multiple levels according to the oil and gas volatilization flux, and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine the terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and mark the early warning information to complete the pipeline micro-leak location.

[0082] Example 2

[0083] The difference between the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging described in Example 1 and the method described in Example 1 is as follows:

[0084] The specific implementation process of step 1 includes:

[0085] Mobile observation platforms include drone inspection platforms (such as DJI drones), equipped with TDLAS laser absorption spectroscopy scanning units, visible light high-definition imaging units, mid- and long-wave infrared thermal imaging units, and near-infrared laser active illumination imaging units;

[0086] The TDLAS laser absorption spectral scanning unit is used to acquire laser spectral absorption amplitude sequences, the visible light high-definition imaging unit is used to acquire visible light images, the mid- and long-wave infrared thermal imaging unit is used to acquire infrared temperature field matrices, and the near-infrared laser active illumination imaging unit is used to acquire laser active absorption imaging images.

[0087] The mobile observation platform inspects the oil pipeline. If the TDLAS laser absorption spectroscopy scanning unit detects an oil and gas volatilization cloud above the oil pipeline (characteristic absorption of the beam), it simultaneously collects multimodal data of the oil pipeline, including the laser spectral absorption amplitude sequence, visible light image, infrared temperature field matrix, and laser active absorption imaging image.

[0088] The TDLAS laser absorption spectroscopy scanning unit uses a mid-infrared distributed feedback laser with temperature stabilization control, a center wavelength locked at 1.65 µm, precisely matching the characteristic absorption peaks of the CH chemical bonds in alkane refined oils, and strictly controlling the wavelength coverage range within 1.5–1.7 μm. Figure 4 The optical path structure includes a laser emitting collimating lens, a fan-shaped scanning galvanometer, a surface echo receiving focusing lens, a photodetector, and a lock-in amplification and demodulation circuit. The laser beam is deflected by the galvanometer to form a fan-shaped scanning surface that is vertically projected downwards to the surface. After penetrating the oil and gas volatiles, selective light absorption occurs. The remaining transmitted light is scattered back to the receiver by the ground. The absorption amplitude is extracted by the second harmonic lock-in demodulation algorithm, and the volume fraction of oil and gas is inverted. The detection limit can reach the ppb level. The qualitative judgment is based on the spectral curve shown in Figure 4: when the detected waveform shows an obvious absorption peak at 1.65 μm and is higher than the threshold dotted line, it is determined that there are hydrocarbon oil and gas volatiles at this spatial location.

[0089] Visible light high-definition imaging unit: It adopts a global exposure industrial CMOS camera, paired with a low distortion fixed-focus lens, to collect two-dimensional morphological information such as surface oil stains, exposed pipe bodies, welds, flanges, and slope topography, providing visual morphological evidence for suspected leakage areas, with a resolution that meets sub-meter ground sampling distance requirements.

[0090] Mid- and long-wave infrared thermal imaging unit: working band 3-5μm, used to capture the surface low temperature anomaly area formed by the change in thermal conductivity after the seepage oil infiltrates the soil. It outputs a pixel-level temperature matrix, which can distinguish the thermal differences between dry soil, water accumulation, vegetation and oil-infiltrated soil. It is the core data source for coarse extraction of seepage areas at the image level.

[0091] Near-infrared laser active illumination imaging unit: Equipped with an 808nm or 915nm semiconductor laser filler and a matching imaging camera, it is mainly used in passive imaging failure conditions such as night, backlight, and tree canopy obstruction to excite oil and gas volatilization aerosols to generate scattered light spots, making up for the defects of pure visible light and infrared imaging in harsh environments, and realizing all-weather imaging acquisition.

[0092] The above four units are installed side by side on the gimbal, with the mechanical reference planes coplanar and the field of view covering downwards in the same direction, forming a common gimbal load structure; the UAV flight control system, gimbal stabilization system, edge computer-borne industrial control computer, wireless data transmission module, and GPS Beidou positioning module together constitute the complete airborne inspection system. The industrial control computer has the fusion positioning algorithm of this invention built into it to complete the entire process of data processing and cross-logic verification.

[0093] The specific implementation process of step 2 includes:

[0094] Preprocessing is performed in parallel on visible light images, infrared temperature field matrices, and laser active absorption imaging images, including bilateral filtering to remove Gaussian noise, dark channel prior algorithm to remove fog, and histogram equalization to improve details in low-contrast areas.

[0095] The Scale Invariant Feature Transform (SIFT) algorithm is used to perform coarse feature point matching on the preprocessed visible light image, infrared temperature field matrix, and active laser absorption imaging image. Then, a perspective transformation model (such as the Homography model, which solves the homography matrix using the Direct Linear Transform (DLT) algorithm, which can be implemented using the findHomography function in the OpenCV open-source library; this model is used to establish the pixel coordinate mapping relationship between images of different imaging units and complete the geometric registration between the visible light image, infrared temperature field, and active laser absorption imaging image) is used to perform fine geometric deviation correction. This unifies the projection of the visible light image, infrared temperature field matrix, and active laser absorption imaging image onto the same geographic pixel coordinate system, eliminates the abnormal area misalignment problem caused by field of view offset, and completes pixel-level registration.

[0096] The specific implementation process of step 3 includes:

[0097] For visible light images, infrared temperature field matrices, and laser active absorption imaging images that have completed pixel-level registration, wavelet basis is used to carry out multi-scale decomposition to obtain low-frequency background components and multi-scale high-frequency detail components, and to extract anomalous features, including edge contour abrupt change regions, gray-level anomalous patches, infrared temperature distortion regions, and laser aerosol (volatile gas cloud) absorption bright regions.

[0098] Among them, the edge contour abrupt change region includes: obtaining the local 8-neighbor gray-level gradient magnitude of each pixel in the visible light image. If the local gray-level change is greater than the change threshold and the number of connected pixels is greater than or equal to 50 pixels, it is regarded as the edge contour abrupt change region.

[0099] Gray-scale anomalous patches are defined as follows: the deviation between the average gray value of a pixel within a local window (15×15 pixels) of each pixel in a visible light image and the average gray value of the surrounding background window (e.g., 50-100 pixels in the surrounding area). If the absolute value of the difference between the average gray value of the local window and the average gray value of the background window is greater than 40, and the area of ​​the connected pixels is greater than 20 pixels, then it is considered a gray-scale anomalous patch.

[0100] The infrared temperature distortion region includes: the average temperature within a preset background window (31×31 pixels) in the infrared temperature field matrix. If the absolute value of the temperature difference from the average temperature is greater than 1.5 degrees Celsius and the connected region is greater than 50 pixels, it is considered as the infrared temperature distortion region.

[0101] The high-brightness area of ​​laser aerosol (volatile gas cloud) absorption includes: for each pixel in the laser active absorption imaging image, the absorption characteristics are compared with the surrounding neighborhood (within 40 pixels), and the pixel with the largest absorption of laser by the oil and gas volatile gas cloud is taken as the high-brightness area of ​​laser aerosol (volatile gas cloud) absorption.

[0102] The extracted abnormal features are fused, including spatially associating visible light images, infrared temperature field matrices, and laser active absorption imaging images with the spatial locations of the abnormal features. Pixels with only one or two abnormal features at the same pixel location are removed, and static background interference (pipeline fixing components, surface stones, weeds, etc.) is removed using background subtraction, outputting suspected leakage candidate areas. This effectively reduces information fragmentation and the number of false anomalies caused by independent processing of multi-source images, resulting in a set of all suspected leakage anomaly patches along the pipeline, which serves as a candidate target pool for subsequent verification.

[0103] The range of the change threshold is 35-40.

[0104] The specific implementation process of step 4 includes:

[0105] Each suspected leakage candidate region includes N valid abnormal pixels, and the pixel center coordinates of the suspected leakage candidate region are (x, y, x). c , yc ) represents , ,in, , Represents the coordinates of the i-th valid abnormal pixel in the suspected leakage candidate region;

[0106] Read the pre-calibrated camera intrinsic parameter matrix and radial and tangential distortion parameters of the imaging unit corresponding to the suspected leakage candidate area. Construct a homogeneous pixel vector from the pixel center coordinates. Use the inverse matrix of the intrinsic parameter matrix to obtain the normalized image plane coordinates carrying lens distortion. Then, use the distortion parameters to complete distortion correction and obtain the normalized coordinates after eliminating lens distortion. Take the camera optical center as the ray origin and combine the normalized coordinates with the unit z component to obtain the spatial ray direction vector in the camera coordinate system. All three-dimensional spatial points on this spatial ray that satisfy the depth scale factor greater than zero will be projected to the pixel center position. This step only obtains the spatial ray and does not solve for the unique three-dimensional spatial point. The depth scale factor is an unknown quantity.

[0107] The installation extrinsic parameters of the imaging unit relative to the machine body, which are pre-calibrated by the mobile observation platform, are read, including the camera-machine body rotation matrix and translation vector. The spatial ray direction vector in the camera coordinate system is divided into the ray origin and the ray direction vector. The ray origin is taken as a three-dimensional spatial point and is transformed to the machine body coordinate system by performing rotation and translation transformations. The ray direction vector is only subjected to rotation transformation without translation operation to obtain the spatial ray in the machine body coordinate system.

[0108] By reading the real-time attitude and GPS location information of the mobile observation platform, the rotation matrix for transforming the body coordinate system to the geodetic ENU coordinate system and the geodetic origin position of the body are obtained. A rotation-translation transformation is performed on the ray origin in the body coordinate system, while the ray direction vector undergoes only a rotation transformation, to obtain the spatial ray in the geodetic coordinate system. All three-dimensional points on this ray with a depth scale factor greater than zero are considered candidate geodetic spatial locations corresponding to the suspected seepage patch. Spatial intersection matching is performed between the spatial ray in the geodetic coordinate system and the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit to obtain the corresponding spatial points. The laser spectral absorption amplitude sequence of the spatial points is retrieved, and based on the laser spectral absorption amplitude sequence of the spatial points, it is determined whether the suspected seepage candidate area actually exists.

[0109] If a suspected leak candidate region exists in the multimodal data, and the corresponding laser spectral absorption amplitude sequence detects characteristic absorption valleys of oil and gas hydrocarbons, and the second harmonic demodulated oil and gas concentration is higher than the preset judgment threshold (connected pixels > 50 pixels and concentration > 5 ppm·m), then it is judged as a real early micro-leak point in the pipeline.

[0110] If there are only suspected leakage candidate areas, but there is no characteristic spectral response of oil and gas at the corresponding location, they are directly identified as environmental interference (water stains, vegetation temperature difference, shadows, dust spots, etc.) and eliminated.

[0111] The specific implementation process of step 5 includes:

[0112] GPS geographic coordinates were bound to the actual early micro-leak points in the pipeline, and the laser spectral absorption amplitude sequence of the corresponding spatial points was retrieved. A wavelength modulation spectral second harmonic demodulation algorithm (WMS-2f) was used to perform digital phase-locked loop (PLL) phase-sensitive demodulation on the laser spectral absorption amplitude sequence to obtain the absorption peak depth (a sawtooth scanning signal was superimposed on a high-frequency sinusoidal modulation signal, and the original spectral amplitude sequence was subjected to phase-sensitive detection and low-pass filtering with a reference signal at twice the modulation frequency to extract the second harmonic waveform; the absolute value of the peak value of the second harmonic waveform was taken as the absorption peak depth). The calibration curve of the TDLAS laser absorption spectral scanning unit was read (the calibration curve is the pre-calibrated relationship between the absorption peak depth and the path integral methane concentration; in this embodiment, the calibration curve uses a linear fitting model). H 2f To determine the absorption peak depth, C path (where k is the path integral methane concentration, k is the calibration slope, and b is the system baseline offset). Substitute the demodulated absorption peak depth into the calibration curve to invert and solve for the path integral methane concentration along the laser propagation path.

[0113] ;

[0114] in, The path integral is the methane concentration, H 2f Where is the absorption peak depth, k is the calibration slope, and b is the baseline offset;

[0115] If the calculated path integral methane concentration is lower than the preset detection limit, it is determined that there is no effective methane absorption at the actual early micro-leak point in the pipeline, and the actual early micro-leak point in the pipeline is removed; otherwise, the actual early micro-leak point in the pipeline is not removed.

[0116] Combining the flight altitude of the mobile observation platform, the pitch angle of the gimbal, and measured or estimated wind field parameters, a simplified Gaussian plume diffusion model is adopted to establish a mapping relationship between the path integral methane concentration and the actual early-stage micro-leakage flux in the pipeline. The mapping relationship is used to inversely deduce the oil and gas volatilization flux corresponding to the actual early-stage micro-leakage in the buried pipeline. According to the size of the oil and gas volatilization flux, three levels of early warning are defined: 0.5–2 L / d is the nascent warning level, 2–10 L / d is the slight leakage level, and greater than 10 L / d is the expansion risk level.

[0117] The infrared temperature field matrix output by the medium- and long-wave infrared thermal imaging unit is collected. Each element in the infrared temperature field matrix stores the measured surface temperature at the corresponding pixel location. Based on the boundary of the infrared temperature distortion zone, the temperature difference is judged pixel by pixel. When the pixel temperature difference decays to the background temperature noise threshold, the expansion stops. The boundary of the infrared temperature distortion zone to the position where the expansion stops is taken as the leakage temperature diffusion range.

[0118] Within the infrared temperature distortion region, a bicubic spline interpolation algorithm is used to spatially reconstruct discrete temperature sampling points within the leakage temperature diffusion range, resulting in a continuously distributed surface temperature field. This surface temperature field is then corrected by combining the spatial distribution of hydrocarbon concentrations corresponding to the laser spectral absorption amplitude sequence, eliminating temperature disturbances unrelated to leakage. This includes pixel-by-pixel comparison of the hydrocarbon concentration spatial distribution with the surface temperature field, checking the hydrocarbon concentration values ​​against the detection limit. If the hydrocarbon concentration value and detection limit of the surface temperature field are lower than those of the hydrocarbon concentration value and detection limit of the hydrocarbon concentration spatial distribution, the corresponding position is set to 1; otherwise, it is set to 0, thus obtaining the hydrocarbon concentration. Confidence weight mask; The surface temperature field is weighted and corrected using a hydrocarbon concentration confidence weight mask (by multiplying the calibration ratio, such as 8‰ ppm·m, by the surface temperature field). The corrected temperature values ​​are then superimposed on the surface temperature field to obtain the corrected surface temperature field. The temperature values ​​of the corrected surface temperature field are mapped to a preset color scale (the temperature values ​​of each pixel in the corrected surface temperature field are normalized to obtain the mapping coefficients, and linear interpolation is performed on the R, G, and B channels according to the mapping coefficients to solve for the RGB color components corresponding to each pixel), generating a seepage diffusion heat map.

[0119] The GPS geographic coordinates of the actual early micro-leakage points in the pipeline, the path integral methane concentration, the infrared temperature field matrix, and the terrain data are geospatially registered to construct a geographic grid, including a terrain elevation grid and a surface temperature grid. Each geographic grid stores geodetic coordinates, terrain elevation, and surface temperature information, and the geographic grid is resampled to the same spatial resolution as the infrared temperature field.

[0120] Based on the path integral methane concentration, geometric parameters observed by the TDLAS laser absorption spectroscopy scanning unit were collected, including the laser line-of-sight elevation angle and the effective vertical thickness of the leaking gas cloud. The effective optical path length of the laser passing through the methane gas cloud was calculated based on these geometric parameters. The path integral methane concentration was divided by the effective optical path length to obtain the equivalent average concentration of the methane gas cloud within the laser sampling area. Based on a simplified Gaussian plume diffusion model, combined with near-surface wind speed, atmospheric stability, and diffusion coefficient, the equivalent average concentration of the methane gas cloud was inversely extrapolated to solve for the equivalent near-surface initial gas concentration at the leak source outlet. The equivalent near-surface initial gas concentration at the leak source, environmental wind field parameters, and the distance passed through the gas cloud were then analyzed. The topographic elevation and surface temperature rasters registered in geospatial data are input using a Gaussian diffusion function. A topographic correction factor and a surface temperature convection correction factor are introduced. The entire geographic rasters in the inspection area are traversed, and the near-surface oil and gas concentration of each geographic raster is calculated point by point. The two-dimensional raster field of near-surface oil and gas concentration in the entire region is output. Among them, the topographic correction factor represents the blocking and confluence effect of topographic relief on oil and gas diffusion. The topographic correction factor is defined as 1 when the terrain is flat and 0 otherwise. The surface temperature convection correction factor represents the change in near-surface convection intensity caused by surface temperature difference. The surface temperature convection correction factor is defined as 1 when the surface temperature is equal to the background temperature and 0 otherwise.

[0121] A visualization rendering sub-model is constructed, which includes mapping the oil and gas concentration values ​​of the two-dimensional raster field of near-surface oil and gas concentration in the whole region to gradient colors (determining the range of oil and gas concentration mapping values, performing amplitude limiting and truncation processing on the oil and gas concentration values ​​of each raster in the two-dimensional raster field, normalizing the truncated concentration values ​​to obtain the mapping coefficients in the 0-1 range; performing linear interpolation operations on the R, G, and B channels according to the mapping coefficients, solving the RGB color components corresponding to each geographic raster, and generating a pseudo-color visualization image of oil and gas concentration gradient), and overlaying topographic base map, pipeline vector information, leakage source location markers, and leakage risk level indicators to generate a leakage diffusion heat map with geographic information.

[0122] Mark the coordinates of the leakage center point, the oil and gas diffusion boundary, and the warning level to form a standardized warning document. Push the warning document to the pipeline operation and maintenance management platform (such as the operation and maintenance platform adapted to the pipeline group or urban gas pipeline network, or a mobile app) to complete the pipeline micro-leakage location and form a closed-loop pre-control system of location, classification, and dispatch.

[0123] The calibration slope k is [0.001, 0.004]V / (ppm·m); the baseline offset b is [-0.02, 0.02]V; the preset detection limit is 1~5 ppm·m; and the background temperature noise threshold is 0.3 degrees Celsius or 0.5 degrees Celsius.

[0124] The core of this invention relies on the coupling verification logic of the specific absorption of infrared laser molecules in TDLAS and the multimodal pixel-level spatial positioning of visible light / infrared thermal imaging / laser active imaging to achieve the detection of micropore leakage at the initial stage. The overall execution flow corresponds to Figure 1, the hardware optical path structure corresponds to Figure 2, the core judgment logic corresponds to Figure 3, and the principle of qualitative absorption of oil and gas molecules is based on the spectral curve in Figure 4.

[0125] Unmanned aerial vehicle (UAV) mobile inspection of long-distance refined oil pipelines in the field:

[0126] The TDLAS probe, visible light camera, infrared thermal imager, and active laser illuminator are fixed to an integrated aluminum alloy base plate using locating pins and high-strength locking screws. The base plate undergoes aging stress relief treatment to prevent minor displacement of the components due to flight vibrations. The base plate is connected to the UAV pod via damping shock-absorbing pads to reduce the disturbance of propeller vibrations to the laser scanning optical path and imaging lens, ensuring that the laser fan-shaped scanning area in Figure 2 consistently coincides with the three imaging fields of view. After all components are installed, the flatness is checked using a dial indicator, and the flatness error is controlled within 0.02mm to ensure the stability of external parameter calibration. This allows for the calibration of the multi-source sensor.

[0127] The distortion coefficients of the three imaging cameras were solved using a checkerboard calibration plate to eliminate radial and tangential distortion of the lens, and the camera intrinsic parameter matrix and distortion vector were output to provide parameters for subsequent image correction. The TDLAS laser scanning galvanometer was used to complete the angle-spatial distance calibration, and a mathematical model of the galvanometer deflection angle and the coordinates of the ground scanning point was established to achieve accurate calculation of the laser spot position on the ground.

[0128] A high-precision calibration target board is placed in a flat and open outdoor area. The control gimbal is vertically aligned with the target board, and four devices are simultaneously triggered to collect a set of calibration data. The airborne industrial control computer uses the SIFT feature point matching algorithm to solve the rotation matrix and translation vector between visible light, infrared, and active laser imaging. Then, by using the coordinates of the laser scanning spot landing point on the target board, a mapping transformation matrix between the laser scanning coordinate system and the image pixel coordinate system is established. Finally, the unified geographic coordinate system binding of the four sets of devices is completed. The unified spatial coordinate calibration of step S1 in Figure 1 is realized from the hardware and algorithm levels, providing mathematical support for the subsequent mapping of pixel coordinates to spectral points (Figure 3 Logic).

[0129] The TDLAS system was calibrated using a standard concentration propane gas calibration cell to establish an absorption peak amplitude-gas concentration fitting curve and set a leakage detection threshold of 500 ppb. At the same time, fixed parameters such as image background difference threshold, wavelet fusion decomposition layer number, and minimum logarithm of SIFT feature point matching were preset in the algorithm layer and solidified into the airborne program to complete the preparation of the entire system before it leaves the factory.

[0130] Field operation methods:

[0131] Flight path planning and pre-flight equipment checks. For long-distance buried refined oil pipelines, the pipeline centerline KML coordinate file is imported into the ground station software. The UAV flight path is set to fly in a straight line directly above the pipeline, with a relative altitude of 10-120m and a horizontal flight speed of 3m / s. The gimbal stabilization function, laser emission status, real-time image transmission status, and BeiDou positioning signal strength are checked. After confirming that all modules are working properly, the takeoff command is executed. The UAV automatically travels along the pipeline at a constant speed according to the preset flight path and enters the synchronous acquisition mode to achieve synchronous acquisition of multiple source payloads by the gimbal.

[0132] Synchronous data real-time acquisition:

[0133] During the drone's flight, the onboard industrial control computer sends a synchronization trigger signal to achieve strict synchronous triggering of the laser scanning galvanometer and three imaging cameras: the TDLAS laser performs a fan-shaped reciprocating spot scan at a frequency of 10 Hz, and the laser beam penetrates vertically into the atmospheric space above the pipeline. If it passes through the oil and gas volatilization cloud above the pipeline micro-leak point, the beam undergoes characteristic absorption. The receiving end collects the modulated light intensity signal and demodulates and stores the absorption amplitude sequence; the three imaging devices simultaneously capture the original visible light image, infrared temperature field matrix, and laser active scattering image under the same field of view. All raw data is stored in real time to the onboard solid-state drive and simultaneously transmitted back to the ground monitoring backend with low latency via a communication link (4G or 5G). Figure 2 ).

[0134] Image preprocessing and pixel-level registration:

[0135] The airborne edge computing unit preprocesses the three sets of images in parallel, using bilateral filtering to remove Gaussian noise, dark channel prior algorithm to remove fog, and histogram equalization to enhance details in low-contrast areas. It calls the pre-calibrated intrinsic parameter matrix to correct lens distortion, and then completes SIFT feature point matching and fine registration through the perspective transformation matrix obtained by extrinsic parameter calibration. This unifies the projection of the three images with different imaging principles and different field sizes onto the same pixel grid, completely eliminating the problem of abnormal area misalignment caused by the field of view offset of the imaging device, and ensuring the uniqueness of the coordinates of the leaking patches extracted subsequently.

[0136] Multi-scale feature fusion and extraction of suspected leakage candidate regions:

[0137] Three-layer wavelet multi-scale decomposition feature fusion is performed on the three registered images to extract three types of features: low-temperature distortion region of infrared image, edge contour of oil stain in visible light, and high-brightness region of aerosol scattering in laser active imaging. The adaptive background difference method is used to perform difference operation between the current frame image and the pipeline reference background image to remove static fixed interference such as pipeline cement slope protection, stones, weeds, etc. After superposition and fusion, several closed irregular patches are output, all of which are marked as suspected leakage candidate areas and stored in the candidate target pool, and then enter the verification process shown in Figure 3.

[0138] Spectral truth cross-logic determination:

[0139] The geometric center pixel coordinates of each candidate leakage patch are read one by one. A pre-stored coordinate mapping transformation matrix is ​​called to reverse-calculate the two-dimensional pixel coordinates into ground geographic coordinates, which are then further mapped to the corresponding laser sampling point on the TDLAS laser fan-shaped scanning trajectory. The laser spectral data stored at that point is retrieved, and the judgment rule in Figure 4 is used to perform a double-condition "AND logic" judgment:

[0140] Condition 1: The image contains closed abnormal patches (suspected leakage at the visual level);

[0141] Condition 2: The corresponding spectral curve shows an absorption peak at 1.65 μm, and the demodulated oil and gas concentration is greater than the 500 ppb threshold.

[0142] When both conditions are met simultaneously, Figure 3 indicates the branch direction to be the "true early micro-leakage point"; if only condition 1 is met and there is no spectral absorption signal, it is directly identified as environmental interference such as water accumulation at low temperatures, vegetation temperature differences, tree shade, or dust spots, and is therefore eliminated. This step fundamentally solves the industry pain point of excessively high false alarm rates in single infrared inspections, and is also the core creative contribution of this invention that distinguishes it from existing image fusion technologies.

[0143] Output of leakage quantification, classification, and location results:

[0144] For real leak points that pass dual verification, BeiDou high-precision geographic coordinates are bound to them. Oil and gas volatilization flux is retrieved based on spectral absorption peak depth fitting, and classified into three warning levels according to flux size: 0.5–2 L / d volatilization is the nascent warning level, 2–10 L / d is the minor leak level, and greater than 10 L / d is the expanding risk level. Infrared temperature field pixel temperature values ​​are interpolated to generate a thermal cloud map of leaked oil and gas diffusion, visually displaying the leak's impact range. Finally, the location coordinates, risk level, thermal map, and on-site captured images are packaged and uploaded to the pipeline operation and maintenance platform. The platform automatically pushes different handling work orders according to the level: nascent level is reviewed quarterly, minor level requires on-site investigation within 72 hours, and expanding risk level requires immediate on-site excavation and repair, completing the closed-loop management of early micro-leaks.

[0145] Fixed-point, 24 / 7 monitoring of oil depots and stations:

[0146] Fixed integrated monitoring pan-tilt units are installed above the oil depot flange assembly, valve stuffing boxes, and storage tank discharge pipes. These units perform 24 / 7 small-area fan-shaped laser scanning and multispectral imaging. The algorithm continuously executes fusion verification logic; if stable trace oil and gas absorption signals are detected multiple times consecutively, and the corresponding location in the infrared image shows abnormal oil seepage temperature at the gasket, an audible and visual alarm is immediately triggered, and a warning is pushed to the backend. This enables unattended, proactive prevention of minor leaks at the site, preventing the long-term accumulation of trace oil seepage from static seals and the resulting safety hazards.

[0147] This embodiment is applied in the tank area, pump area and dense pipe gallery area of ​​finished oil depot. It focuses on monitoring long-term minor leakage of static equipment such as flange gasket aging leakage, valve stuffing box slight oil leakage, and short pipe weld micro-crack oil leakage. The hardware also uses the same gimbal load architecture as shown in Figure 2, but the mounting form is changed from UAV airborne to steel structure column fixed erection.

[0148] The core of the monitoring host is still an integrated four-module rigid pan-tilt unit. The internal component arrangement, optical path form, laser band, and imaging type are completely consistent with those of Example 1: the TDLAS unit maintains the 3.2-3.5μm mid-infrared alkane characteristic absorption band and completes the qualitative identification of oil and gas by relying on the spectral curve; the visible light camera is used to photograph the oil stains on the flange surface; the infrared thermal imaging captures the low-temperature area of ​​the gasket oil seepage; and the 808nm active laser illumination is used for aerosol scattering imaging under no-light conditions at night.

[0149] The supporting auxiliary hardware includes: stainless steel anti-corrosion column, horizontal adjustment flange, electric servo motor for pan / tilt angle, outdoor industrial-grade control box, 24V DC power supply, surge protection module, fiber optic Ethernet communication module, and local audible and visual alarm. The pan / tilt unit achieves small-angle fan-shaped reciprocating scanning via electric servo motor, covering a 15m × 12m area of ​​the site equipment in front, capable of covering 3-5 sets of parallel pipe flanges and valve groups. The hardware spatial scanning trajectory and imaging field of view overlap logic still follow the optical path diagram in Appendix 2. The monitoring steps are as follows:

[0150] Periodic synchronous data acquisition:

[0151] After the equipment is powered on, it enters the automatic monitoring mode. The electric servo drives the gimbal to swing back and forth at a small angle. The TDLAS laser beam forms a fan-shaped scanning surface covering the flange, valve, and pipe weld area below. During each round of scanning, the visible light, infrared, and active laser cameras are triggered simultaneously to capture the current field of view image. When the laser beam sweeps over the oil and gas cloud that seeps out and evaporates upward, an absorption peak is generated at the 3.31μm position. The detector collects the modulation signal and sends it to the phase-locked loop circuit to calculate the concentration value. All spectral sequences and image data are stored in the local database in real time, completing one round of raw data acquisition.

[0152] Image registration and suspected region extraction:

[0153] The industrial control computer automatically preprocesses the three sets of images in this round by denoising and dehazing, and uses fixed top-view calibration parameters to complete perspective transformation and pixel-level registration. It then uses wavelet multi-scale feature fusion to overlay infrared low-temperature patches, visible light oil stains, and laser scattering spot features, and performs background subtraction with the long-term collected site baseline image to remove fixed static targets such as valve bodies, pipe insulation layers, and supports. This process identifies closed regions in the images with abrupt changes in grayscale, temperature, and scattering characteristics as a set of suspected leak candidate patches. The site environment contains strong interference sources such as temperature differences in equipment insulation layers, water accumulation pits, and shade from sunshades. Relying solely on infrared imaging would generate numerous false alarms; therefore, the spectral truth rejection mechanism shown in Figure 3 must be used in the next step.

[0154] Spectral and image cross-validation and logical verification:

[0155] like Figure 3 As shown, the system traverses each candidate leakage patch, maps the pixel center point to the corresponding spatial sampling point of the current laser scan, retrieves the spectral data of that point and compares it with the absorption curve to determine whether there is a valid oil and gas signal, and strictly performs dual condition judgment:

[0156] 1) Abnormal infiltration / low-temperature plaques were detected in the image;

[0157] 2) A characteristic absorption peak of 3.31 μm was detected in the corresponding spatial location spectrum, and the concentration was higher than the 800 ppb field threshold.

[0158] If both conditions are met, it is determined to be a genuine micro-leakage of the flange or valve; if only the image shows anomalies and the spectrum shows no hydrocarbon absorption signal, it is automatically determined to be insulation layer temperature difference, ground water accumulation, or shadow interference, and is directly filtered without triggering any alarm. If three consecutive scanning cycles determine it as a valid leak point, the system locks that coordinate as a stable potential hazard point to avoid occasional misjudgments caused by instantaneous airflow disturbances.

[0159] Quantitative grading and multi-level early warning output:

[0160] For genuine micro-leakage hazards that have passed multiple rounds of verification, the system classifies them into three risk levels based on spectral inversion of volatile flux: the initial warning level only records daily alerts in the background log; the minor leakage level pushes maintenance reminders to the mobile app; and the expanding risk level simultaneously triggers on-site red audible and visual alarms, a pop-up alarm in the oil depot control room, and SMS notifications to maintenance personnel. Simultaneously, the system automatically captures infrared thermal diffusion maps, visible light oil stain photos, and spectral absorption curve screenshots of the leakage point as attachments for archiving, forming a traceable monitoring record. Maintenance personnel, based on the precise locations marked by the system, disassemble and replace the corresponding flange packing glands and gaskets, eliminating the risk of flammable and explosive oil and gas accumulation caused by long-term micro-leakage from static seals.

[0161] Example 3

[0162] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging as described in Embodiment 1 or 2.

[0163] Example 4

[0164] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging as described in Embodiment 1 or 2.

[0165] Example 5

[0166] A pipeline micro-leakage fusion localization system based on laser absorption spectroscopy and multimodal imaging includes:

[0167] The multimodal data acquisition module is configured to: construct a mobile observation platform, conduct inspections of oil pipelines through the mobile observation platform, and collect multimodal data of oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images;

[0168] The preprocessing and registration module is configured to preprocess and register the acquired visible light images, infrared temperature field matrices, and laser active absorption imaging images.

[0169] The suspected leakage candidate area acquisition module is configured to: perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and acquire suspected leakage candidate areas.

[0170] The module for determining early-stage micro-leakage points in pipelines is configured to: extract the pixel center coordinates of suspected leak candidate areas and map them to the scanning trajectory of the TDLAS laser absorption spectral scanning unit; and determine whether the suspected leak candidate areas are early-stage micro-leakage points in pipelines based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory.

[0171] The early warning classification module is configured to: calculate the path integral methane concentration corresponding to the actual early micro-leakage point in the pipeline; obtain the oil and gas volatilization flux through the path integral methane concentration; classify the early warning into multiple levels according to the oil and gas volatilization flux; and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and label the early warning information to complete the pipeline micro-leakage location.

Claims

1. A method for fusion localization of micro-leakage in pipelines based on laser absorption spectroscopy and multimodal imaging, characterized in that, include: Step 1: Construct a mobile observation platform to inspect oil pipelines and collect multimodal data of the oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images. Step 2: Preprocess and register the acquired visible light image, infrared temperature field matrix, and laser active absorption imaging image; Step 3: Perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and obtain suspected leakage candidate areas; Step 4: Extract the pixel center coordinates of the suspected leak candidate area and map them to the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit. Based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory, determine whether the suspected leak candidate area is a real early micro-leak point of the pipeline. Step 5: Calculate the path integral methane concentration corresponding to the actual early micro-leak point in the pipeline, obtain the oil and gas volatilization flux through the path integral methane concentration, divide the early warning into multiple levels according to the oil and gas volatilization flux, and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine the terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and mark the early warning information to complete the pipeline micro-leak location.

2. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 1, characterized in that, The specific implementation process of step 1 includes: The mobile observation platform includes an unmanned aerial vehicle (UAV) inspection platform, equipped with a TDLAS laser absorption spectroscopy scanning unit, a visible light high-definition imaging unit, a mid- and long-wave infrared thermal imaging unit, and a near-infrared laser active illumination imaging unit. The TDLAS laser absorption spectral scanning unit is used to acquire laser spectral absorption amplitude sequences, the visible light high-definition imaging unit is used to acquire visible light images, the mid- and long-wave infrared thermal imaging unit is used to acquire infrared temperature field matrices, and the near-infrared laser active illumination imaging unit is used to acquire laser active absorption imaging images. The mobile observation platform inspects the oil pipeline. If the TDLAS laser absorption spectroscopy scanning unit detects an oil and gas volatilization cloud above the oil pipeline, it simultaneously collects multimodal data of the oil pipeline, including the laser spectral absorption amplitude sequence, visible light image, infrared temperature field matrix, and laser active absorption imaging image of the oil pipeline.

3. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 2, characterized in that, The specific implementation process of step 2 includes: Preprocessing is performed in parallel on visible light images, infrared temperature field matrices, and laser active absorption imaging images, including bilateral filtering to remove Gaussian noise, dark channel prior algorithm to remove fog, and histogram equalization to improve details in low-contrast areas. The Scale Invariant Feature Transform (SIFT) algorithm is used to perform coarse feature point matching on the preprocessed visible light image, infrared temperature field matrix, and active laser absorption imaging image. Then, geometric deviation is finely corrected by using a perspective transformation model. The visible light image, infrared temperature field matrix, and active laser absorption imaging image are uniformly projected onto the same geographic pixel coordinate system to complete pixel-level registration.

4. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 3, characterized in that, The specific implementation process of step 3 includes: For visible light images, infrared temperature field matrices, and laser active absorption imaging images that have completed pixel-level registration, wavelet basis is used to carry out multi-scale decomposition to obtain low-frequency background components and multi-scale high-frequency detail components, and to extract abnormal features, including edge contour abrupt regions, gray-level abnormal patches, infrared temperature distortion regions, and laser aerosol absorption high-brightness regions. Among them, the edge contour abrupt change region includes: obtaining the local 8-neighbor gray-level gradient magnitude of each pixel in the visible light image. If the local gray-level change is greater than the change threshold and the number of connected pixels is greater than or equal to 50 pixels, it is regarded as the edge contour abrupt change region. Gray-scale anomalous patches are defined as follows: the deviation between the average gray value of a pixel within a local window of each pixel in a visible light image and the average gray value of the surrounding background window is calculated. If the absolute value of the difference between the average gray value of a pixel in the local window and the average gray value of the background window is greater than 40, and the area of ​​the connected pixels is greater than 20 pixels, then it is considered a gray-scale anomalous patch. The infrared temperature distortion region includes: the average temperature is calculated within a preset background window in the infrared temperature field matrix. If the absolute value of the temperature difference from the average temperature is greater than 1.5 degrees Celsius and the connected region is greater than 50 pixels, it is considered as the infrared temperature distortion region. The high-brightness region of laser aerosol absorption includes: for each pixel in the laser active absorption imaging image, the absorption characteristics are compared with the surrounding neighborhood of the pixel, and the pixel with the largest absorption of laser by the oil and gas volatile gas cloud is taken as the high-brightness region of laser aerosol absorption. The extracted abnormal features are fused, including spatially associating the visible light image, infrared temperature field matrix, and laser active absorption imaging image, and aligning them with the spatial positions of the abnormal features; pixels with only one or two abnormal features at the same pixel location are removed, and static background interference is removed by background subtraction method to output suspected leakage candidate areas. The range of the change threshold is 35-40.

5. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 4, characterized in that, The specific implementation process of step 4 includes: Each suspected leakage candidate region includes N valid abnormal pixels, and the pixel center coordinates of the suspected leakage candidate region are (x, y, x). c ,y c ) represents , ,in, , Represents the coordinates of the i-th valid abnormal pixel in the suspected leakage candidate region; Read the pre-calibrated camera intrinsic parameter matrix and radial and tangential distortion parameters of the imaging unit corresponding to the suspected leakage candidate area, construct the pixel center coordinates into a homogeneous pixel vector, and use the inverse matrix of the intrinsic parameter matrix to obtain the normalized image plane coordinates carrying lens distortion; then use the distortion parameters to complete the distortion correction and obtain the normalized coordinates after eliminating lens distortion; take the camera optical center as the ray origin, combine the normalized coordinates with the unit z component to obtain the spatial ray direction vector in the camera coordinate system; The installation extrinsic parameters of the imaging unit relative to the machine body, which are pre-calibrated by the mobile observation platform, are read, including the camera-machine body rotation matrix and translation vector. The spatial ray direction vector in the camera coordinate system is divided into the ray origin and the ray direction vector. The ray origin is taken as a three-dimensional spatial point and is transformed to the machine body coordinate system by performing rotation and translation transformations. The ray direction vector is only subjected to rotation transformation without translation operation to obtain the spatial ray in the machine body coordinate system. The real-time attitude and GPS location information of the mobile observation platform are read to obtain the rotation matrix for transforming the body coordinate system to the geodetic ENU coordinate system and the geodetic origin position of the body. A rotation-translation transformation is performed on the ray origin in the body coordinate system, while the ray direction vector undergoes only a rotation transformation to obtain the spatial ray in the geodetic coordinate system. The spatial ray in the geodetic coordinate system is spatially intersected and matched with the scanning trajectory of the TDLAS laser absorption spectroscopy scanning unit to obtain the corresponding spatial points. The laser spectral absorption amplitude sequence of the spatial points is retrieved, and based on the laser spectral absorption amplitude sequence of the spatial points, it is determined whether the suspected leakage candidate area actually exists. If there are suspected leakage candidate areas in the multimodal data, and the corresponding laser spectral absorption amplitude sequence detects characteristic absorption valleys of oil and gas hydrocarbons and the second harmonic demodulated oil and gas concentration is higher than the preset judgment threshold, then it is judged as a real early micro-leak point in the pipeline. If there are only suspected leakage candidate areas, but there is no characteristic spectral response of oil and gas at the corresponding location, they are directly judged as environmental interference and eliminated. The preset threshold is that the number of connected pixels is greater than 50 and the concentration is greater than 5 ppm·m.

6. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 5, characterized in that, The specific implementation process of step 5 includes: GPS geographic coordinates were bound to the actual early micro-leak points in the pipeline, and the laser spectral absorption amplitude sequence of the corresponding spatial points was retrieved. A wavelength modulation spectral second harmonic demodulation algorithm was used to perform digital phase-locked loop demodulation on the laser spectral absorption amplitude sequence to obtain the absorption peak depth. The calibration curve of the TDLAS laser absorption spectral scanning unit was read. The demodulated absorption peak depth was substituted into the calibration curve to invert and solve for the path integral methane concentration along the laser propagation path. ； in, The path integral is the methane concentration, H 2f Where is the absorption peak depth, k is the calibration slope, and b is the baseline offset; If the calculated path integral methane concentration is lower than the preset detection limit, it is determined that there is no effective methane absorption at the actual early micro-leak point in the pipeline, and the actual early micro-leak point in the pipeline is removed; otherwise, the actual early micro-leak point in the pipeline is not removed. Combining the flight altitude of the mobile observation platform, the pitch angle of the gimbal, and measured or estimated wind field parameters, a simplified Gaussian plume diffusion model is adopted to establish a mapping relationship between the path integral methane concentration and the actual early-stage micro-leakage flux in the pipeline. The mapping relationship is used to inversely deduce the oil and gas volatilization flux corresponding to the actual early-stage micro-leakage in the buried pipeline. According to the size of the oil and gas volatilization flux, three levels of early warning are defined: 0.5–2 L / d is the nascent warning level, 2–10 L / d is the slight leakage level, and greater than 10 L / d is the expansion risk level. The infrared temperature field matrix output by the medium- and long-wave infrared thermal imaging unit is collected. Each element in the infrared temperature field matrix stores the measured surface temperature at the corresponding pixel location. Based on the boundary of the infrared temperature distortion zone, the temperature difference is judged pixel by pixel. When the pixel temperature difference decays to the background temperature noise threshold, the expansion stops. The boundary of the infrared temperature distortion zone to the position where the expansion stops is taken as the leakage temperature diffusion range. Within the infrared temperature distortion region, a bicubic spline interpolation algorithm is used to spatially reconstruct discrete temperature sampling points within the leakage temperature diffusion range, resulting in a continuously distributed surface temperature field. This surface temperature field is then corrected by combining the spatial distribution of hydrocarbon concentrations corresponding to the laser spectral absorption amplitude sequence. This includes pixel-by-pixel comparison of hydrocarbon concentration values ​​with the detection limit in the surface temperature field; if the hydrocarbon concentration value and detection limit in the surface temperature field are lower than those in the spatial distribution, the corresponding position is set to 1; otherwise, it is set to 0, thus obtaining a hydrocarbon concentration confidence weight mask. The surface temperature field is then weighted and corrected using this mask, and the corrected temperature value is superimposed on the surface temperature field to obtain the corrected surface temperature field. Finally, the temperature values ​​of the corrected surface temperature field are mapped to a preset color scale to generate a leakage diffusion thermal map. The GPS geographic coordinates of the actual early micro-leakage points in the pipeline, the path integral methane concentration, the infrared temperature field matrix, and the terrain data are geospatially registered to construct a geographic grid, including a terrain elevation grid and a surface temperature grid. Each geographic grid stores geodetic coordinates, terrain elevation, and surface temperature information, and the geographic grid is resampled to the same spatial resolution as the infrared temperature field. Based on the path integral methane concentration, geometric parameters observed by the TDLAS laser absorption spectroscopy scanning unit were collected, including the laser line-of-sight elevation angle and the effective vertical thickness of the leaking gas cloud. The effective optical path length of the laser passing through the methane gas cloud was calculated based on these geometric parameters. The path integral methane concentration was divided by the effective optical path length to obtain the equivalent average concentration of the methane gas cloud within the laser sampling area. Based on a simplified Gaussian plume diffusion model, combined with near-surface wind speed, atmospheric stability, and diffusion coefficient, the equivalent average concentration of the methane gas cloud was inversely extrapolated to solve for the equivalent near-surface initial gas concentration at the leak source outlet. The equivalent near-surface initial gas concentration at the leak source, environmental wind field parameters, and the distance passed through the gas cloud were then analyzed. The topographic elevation and surface temperature rasters registered in geospatial data are input using a Gaussian diffusion function. A topographic correction factor and a surface temperature convection correction factor are introduced. The entire geographic rasters in the inspection area are traversed, and the near-surface oil and gas concentration of each geographic raster is calculated point by point. The two-dimensional raster field of near-surface oil and gas concentration in the entire region is output. Among them, the topographic correction factor represents the blocking and confluence effect of topographic relief on oil and gas diffusion. The topographic correction factor is defined as 1 when the terrain is flat and 0 otherwise. The surface temperature convection correction factor represents the change in near-surface convection intensity caused by surface temperature difference. The surface temperature convection correction factor is defined as 1 when the surface temperature is equal to the background temperature and 0 otherwise. Construct a visualization rendering sub-model, including mapping the oil and gas concentration values ​​of the two-dimensional raster field of near-surface oil and gas concentration in the whole region to gradient colors, overlaying topographic base map, pipeline vector information, leakage source location markers and leakage risk level indicators, and generating a leakage diffusion heat map with geographic information. Mark the coordinates of the leak center point, the oil and gas diffusion boundary, and the warning level to form a standardized warning document. Push the warning document to the pipeline operation and maintenance management platform to complete the location of pipeline micro-leakage.

7. The pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging according to claim 6, characterized in that, The calibration slope k is [0.001, 0.004]V / (ppm·m); the baseline offset b is [-0.02, 0.02]V; the preset detection limit is 1~5 ppm·m; and the background temperature noise threshold is 0.3 degrees Celsius or 0.5 degrees Celsius.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging as described in any one of claims 1-7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the pipeline micro-leakage fusion localization method based on laser absorption spectroscopy and multimodal imaging as described in any one of claims 1-7.

10. A pipeline micro-leakage fusion positioning system based on laser absorption spectroscopy and multimodal imaging, characterized in that, include: The multimodal data acquisition module is configured to: construct a mobile observation platform, conduct inspections of oil pipelines through the mobile observation platform, and collect multimodal data of oil pipelines, including laser spectral absorption amplitude sequences, visible light images, infrared temperature field matrices, and laser active absorption imaging images; The preprocessing and registration module is configured to preprocess and register the acquired visible light images, infrared temperature field matrices, and laser active absorption imaging images. The suspected leakage candidate area acquisition module is configured to: perform multi-scale decomposition and extract abnormal features from the registered visible light image, infrared temperature field matrix and laser active absorption imaging image, perform feature fusion on the abnormal features, and acquire suspected leakage candidate areas. The module for determining early-stage micro-leakage points in pipelines is configured to: extract the pixel center coordinates of suspected leak candidate areas and map them to the scanning trajectory of the TDLAS laser absorption spectral scanning unit; and determine whether the suspected leak candidate areas are early-stage micro-leakage points in pipelines based on the laser spectral absorption amplitude sequence of the spatial points corresponding to the scanning trajectory. The early warning classification module is configured to: calculate the path integral methane concentration corresponding to the actual early micro-leakage point in the pipeline; obtain the oil and gas volatilization flux through the path integral methane concentration; classify the early warning into multiple levels according to the oil and gas volatilization flux; and generate a leakage diffusion thermal map using the infrared temperature field matrix; combine terrain elevation information and surface temperature information to construct a visualization rendering sub-model, and label the early warning information to complete the pipeline micro-leakage location.

Images