A smart gas leak monitoring system based on infrared thermal imaging

Abstract

This invention relates to the field of image analysis technology, and more particularly to an intelligent gas leak monitoring system based on infrared thermal imaging, comprising: a pipeline robot; a feature region extraction module for determining feature regions in an infrared image based on local temperature difference and thermal steepness reference values; an analysis selection module for determining whether to perform cloud feature analysis or test analysis based on the number of cloud feature images when feature regions exist; a feature analysis module for determining whether a leak exists in cloud feature analysis based on gas absorption depth and peak interval variation coefficient; a test analysis module for applying low-frequency vibration to a target pipeline area and determining whether it is under suspected leak conditions based on multi-band absorption ratio; and a suspected leak analysis module for determining the region category of the target pipeline area based on the spatial convergence of the leak source and determining a processing strategy based on the region category. This invention can improve the positioning accuracy of leak sources.

Description

Technical Field

[0001] This invention relates to the field of image analysis technology, and in particular to an intelligent gas leak monitoring system based on infrared thermal imaging. Background Technology

[0002] Gas leaks, especially those involving hazardous gases (such as methane, carbon monoxide, and hydrogen sulfide) in industries like petrochemicals and coking, are a major cause of serious safety accidents, including explosions, poisonings, and environmental pollution. The raw coal gas transported in coking plant pipelines reaches temperatures as high as 700–800℃ and is rich in flammable, explosive, and toxic components such as tar, benzene, and naphthalene. Furthermore, a coking layer easily forms on the inner wall of these pipelines. The presence of this coking layer not only reduces pipeline transport efficiency but also produces infrared thermal imaging characteristics similar to actual gas leaks due to its shedding or breathing effect, leading to frequent false alarms using traditional detection methods. Therefore, how to accurately locate leak sources using infrared thermal imaging technology has become a pressing technical problem for those skilled in the art.

[0003] Chinese Patent Publication No. CN115331108A discloses a method for detecting hazardous gas leaks based on infrared images, including the following steps: 1) using an infrared thermal imager to photograph the production environment; 2) constructing a real dataset Dreal and a synthetic dataset Dsynthesis using the acquired images; 3) constructing a leak gas semantic segmentation network LGSNet; 4) adding skip connections containing SE modules to the backbone network; 5) adding a PPM module between the encoder and decoder; 6) adding two Dropout layers after the PPM module, and adding a convolutional kernel between the two Dropout layers; 7) training the training set Dsynthesis-train to obtain training weights w; 8) inputting the image to perform leak gas detection. However, the above scheme has the following problems: relying solely on deep learning networks for gas leak semantic segmentation detection, it cannot adaptively select the analysis method according to the actual state of the image, resulting in poor accuracy and environmental adaptability in gas leak monitoring. Summary of the Invention

[0004] To address this issue, the present invention provides an intelligent gas leak monitoring system based on infrared thermal imaging, which overcomes the problem in existing technologies that cannot adaptively select analysis methods according to the actual state of the image, resulting in poor monitoring accuracy and environmental adaptability of gas leaks.

[0005] To achieve the above objectives, the present invention provides a gas leak intelligent monitoring system based on infrared thermal imaging, comprising:

[0006] A pipeline robot, used to acquire infrared images of a target pipeline area within a first preset time period;

[0007] A feature region extraction module, which is connected to the pipeline robot, is used to determine the feature regions in the infrared image based on local temperature difference and thermal steepness reference values.

[0008] An analysis and selection module, which is connected to the pipeline robot and the feature region extraction module respectively, is used to determine whether to perform cloud feature analysis or test analysis based on the number of cloud feature images when feature regions exist; wherein, cloud feature images are determined based on inter-frame optical flow motion coefficients and regional texture entropy values;

[0009] The feature analysis module, which is connected to the analysis selection module, is used to determine whether there is a leak in the target pipeline area based on the gas absorption depth and the peak interval variation coefficient in cloud feature analysis.

[0010] The test analysis module is connected to the pipeline robot and the analysis selection module respectively. It is used to apply low-frequency vibration to the target pipeline area during the test analysis, and determine whether the target pipeline area is under suspected leakage conditions based on the multi-band absorption ratio of the infrared image collected in the second preset time period after the application.

[0011] The suspected leakage analysis module, which is connected to the test analysis module, is used to determine the region category of the target pipeline area based on the spatial convergence of the leakage source under suspected leakage conditions, and to determine the processing strategy based on the region category.

[0012] The processing strategy includes determining whether a leak exists based on the leak plume spread ratio, determining whether a leak exists based on the edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient, and determining whether to adjust the monitoring frequency based on a comprehensive evaluation value.

[0013] Furthermore, the feature region extraction module determines the sub-regions whose local temperature difference is greater than or equal to a preset local temperature difference and whose thermal steepness reference value is greater than or equal to a preset thermal steepness reference value as feature regions.

[0014] Furthermore, the analysis selection module determines to perform cloud feature analysis when the number of cloud feature images is greater than or equal to the preset number of cloud feature images.

[0015] The analysis and selection module determines to conduct test analysis when the number of cloud feature images is less than the preset number of cloud feature images.

[0016] Furthermore, the analysis and selection module determines that an infrared image with an inter-frame optical flow motion coefficient greater than a preset inter-frame optical flow motion coefficient and a regional texture entropy value greater than a preset regional texture entropy value is a cloud feature image.

[0017] Furthermore, the feature analysis module determines that there is a leak in the target pipeline area when the gas absorption depth is greater than the preset gas absorption depth and the peak interval variation coefficient is greater than the preset peak interval variation coefficient.

[0018] Furthermore, the test analysis module determines that the target pipeline area is under suspected leakage conditions when the multi-band absorption ratio is greater than or equal to the preset multi-band absorption ratio.

[0019] Furthermore, the suspected analysis module determines the region category of the target pipeline area based on the spatial convergence of the leakage source, including:

[0020] If the spatial convergence of the leakage source is greater than or equal to the first preset spatial convergence of the leakage source, it is determined to be a Class I region.

[0021] If the spatial convergence of the leakage source is less than the first preset spatial convergence of the leakage source and greater than or equal to the second preset spatial convergence of the leakage source, it is determined to be a type II region.

[0022] If the spatial convergence of the leakage source is less than the second preset spatial convergence of the leakage source, it is determined to be a Class III region;

[0023] Among them, the spatial convergence of the first preset leakage source is greater than that of the spatial convergence of the second preset leakage source.

[0024] Furthermore, the suspected analysis module, for a certain type of area, determines whether a leak exists based on the leak plume spread ratio;

[0025] If the leakage plume extension ratio is greater than the preset leakage plume extension ratio, it is determined that there is a leak in the target pipeline area.

[0026] Furthermore, for the second type of region, the suspected analysis module uses a decision-making strategy based on the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient to determine whether a leak exists.

[0027] If the edge sharpness of the thermal anomaly is greater than the preset edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient is greater than the preset vibration-induced modulation coefficient, then it is determined that there is a leak in the target pipeline area.

[0028] Furthermore, the suspected analysis module determines whether to adjust the monitoring frequency based on a comprehensive evaluation value for the three types of areas.

[0029] If the comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value, the monitoring frequency will be adjusted.

[0030] Compared with the prior art, the beneficial effects of the present invention are as follows: In the technical solution of the present invention, the degree of thermal anomaly of the pipeline wall and the steepness of the temperature field distribution are effectively reflected by the local temperature difference and thermal steepness reference value. Then, based on the local temperature difference and thermal steepness reference value, the feature area in the infrared image is determined, which is conducive to accurately screening out the suspected thermal anomaly parts of leakage in the complex background of the raw coal gas pipeline. Then, when there is a feature area, cloud feature analysis or test analysis is determined based on the number of cloud feature images. This allows the detection path to be adaptively switched according to the significance of the actual leakage characterization. When the cloud features are sufficient, passive analysis is used to achieve rapid judgment. When the cloud features are insufficient, active vibration excitation test is switched to achieve enhanced detection of small leaks, thereby achieving a dynamic balance between detection efficiency and sensitivity.

[0031] Furthermore, in the cloud feature analysis of this invention, the gas absorption depth and peak interval variation coefficient effectively reflect the degree of absorption of infrared radiation by the leaking cloud and the random fluctuation characteristics of the cloud's appearance time. Thus, the presence of a leak is determined based on the gas absorption depth and peak interval variation coefficient, which is beneficial for accurately identifying real leaks under complex operating conditions of raw coal gas pipelines. Only when the gas absorption depth is sufficient to characterize a significant gas absorption effect and the peak interval variation coefficient is large enough to reflect the randomness of the leak occurrence is a leak determined, thereby significantly reducing the false alarm rate caused by periodic false leak signals and improving the reliability of leak determination.

[0032] Furthermore, this invention effectively reflects the difference in radiation attenuation characteristics between the mid-wave infrared and long-wave infrared bands of the target pipeline area through the multi-band absorption ratio. Based on the multi-band absorption ratio of the infrared image acquired during the second preset time period after application, it determines whether the area is under suspected leakage conditions. This helps distinguish between actual gas leaks and non-gaseous interference factors. Only when the multi-band absorption ratio reaches or exceeds a preset threshold is it determined to be under suspected leakage conditions. Thus, the difference in absorption between the two bands effectively eliminates interference from non-gaseous factors on the detection results. Furthermore, under suspected leakage conditions, the regional category of the target pipeline area is determined based on the spatial convergence of the leakage source, achieving accurate location and classification of the leakage source and providing a reliable basis for the formulation of subsequent differentiated processing strategies.

[0033] Furthermore, this invention adaptively selects different processing strategies based on region categories, which is beneficial for adopting differentiated judgment criteria for the spatial characteristics and signal strength of different types of leak sources, achieving optimized allocation of detection resources and accurate adaptation of leak judgment. For Class I regions, determining the presence of a leak based on the leak plume extension ratio is beneficial for utilizing the geometric characteristics of the slender plume shape of the actual leak cloud under high-speed jet conditions, and quickly identifying fixed-point source leaks by using the ratio of the major axis to the minor axis of the cloud's minimum circumscribed rectangle, achieving efficient judgment. For Class II regions, determining the presence of a leak based on the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient is beneficial for achieving high-sensitivity joint detection and capturing weak leak signals when the spatial convergence is moderate, by comprehensively considering the severity of temperature change at the leak point edge and the modulation enhancement effect of vibration on gas absorption. For Class III regions, determining whether to adjust the monitoring frequency based on the comprehensive evaluation value is beneficial for assessing the severity of pipeline defects by weighted fusion of the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient when the leak characteristics are not yet clear but there are potential hazards, dynamically adjusting the inspection cycle, and achieving a shift from passive response to proactive preventive maintenance. Attached Figure Description

[0034] Figure 1 This is a module connection diagram of the intelligent gas leak monitoring system based on infrared thermal imaging according to the present invention;

[0035] Figure 2 This is a flowchart illustrating the present invention for cloud feature analysis or test analysis based on the determination of the number of cloud feature images.

[0036] Figure 3 This is a flowchart illustrating the process of determining whether a suspected leakage condition is present based on multi-band absorption ratio according to the present invention.

[0037] Figure 4 This is a flowchart illustrating how the present invention determines the processing strategy based on region category. Detailed Implementation

[0038] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0039] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0040] Please see Figures 1 to 4 As shown, the present invention provides a gas leak intelligent monitoring system based on infrared thermal imaging, comprising:

[0041] A pipeline robot, used to acquire infrared images of a target pipeline area within a first preset time period;

[0042] A feature region extraction module, which is connected to the pipeline robot, is used to determine the feature regions in the infrared image based on local temperature difference and thermal steepness reference values.

[0043] An analysis and selection module, which is connected to the pipeline robot and the feature region extraction module respectively, is used to determine whether to perform cloud feature analysis or test analysis based on the number of cloud feature images when feature regions exist; wherein, cloud feature images are determined based on inter-frame optical flow motion coefficients and regional texture entropy values;

[0044] The feature analysis module, which is connected to the analysis selection module, is used to determine whether there is a leak in the target pipeline area based on the gas absorption depth and the peak interval variation coefficient in cloud feature analysis.

[0045] The test analysis module is connected to the pipeline robot and the analysis selection module respectively. It is used to apply low-frequency vibration to the target pipeline area during the test analysis, and determine whether the target pipeline area is under suspected leakage conditions based on the multi-band absorption ratio of the infrared image collected in the second preset time period after the application.

[0046] The suspected leakage analysis module, which is connected to the test analysis module, is used to determine the region category of the target pipeline area based on the spatial convergence of the leakage source under suspected leakage conditions, and to determine the processing strategy based on the region category.

[0047] The processing strategy includes determining whether a leak exists based on the leak plume spread ratio, determining whether a leak exists based on the edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient, and determining whether to adjust the monitoring frequency based on a comprehensive evaluation value.

[0048] The application scenario of this invention is the infrared thermal imaging gas leak detection of raw coal gas pipelines using pipeline robots. Given the actual working conditions of raw coal gas pipelines—high temperatures, high dust concentrations, strong thermal interference from coking layers, and the susceptibility to false alarms from single infrared features—traditional detection methods struggle to distinguish between genuine gas leaks and interference such as coking hot spots and localized overheating of the pipe wall. Therefore, this invention employs a multi-feature hierarchical judgment combined with active vibration excitation to progressively eliminate interference and improve the reliability of leak identification.

[0049] The pipeline robot carries a dual-band infrared thermal imager, moves along the inner wall of the raw coal gas pipeline, and continuously collects infrared image sequences of the target pipeline area within a preset first preset time period to identify whether there is a raw coal gas leak and the extent of the leak.

[0050] In this embodiment, the first preset time period is 2 minutes long. During the first time period, only infrared images of the target pipeline area in the mid-wave infrared band are captured; the frame rate of the infrared thermal imager is 1Hz, and the number of frames acquired is 120.

[0051] The target pipeline area refers to the section of pipeline that needs to be monitored for gas leaks during the pipeline robot's inspection of raw coal gas pipelines.

[0052] Specifically, the feature region extraction module determines the sub-regions whose local temperature difference is greater than or equal to a preset local temperature difference and whose thermal steepness reference value is greater than or equal to a preset thermal steepness reference value as feature regions.

[0053] The method for confirming the sub-region is as follows:

[0054] The infrared image is evenly divided into several rectangular sub-regions of equal area along the horizontal and vertical directions. The area of ​​a single rectangular sub-region is S. S is preset to A×B pixels according to the imaging resolution, detection distance and target pipe wall size of the pipeline robot. The values ​​of A and B range from 10 to 30. In this embodiment, A and B are both set to 20.

[0055] The method for determining the local temperature difference is as follows:

[0056] For a single sub-region, calculate the average radiation temperature T1 corresponding to all pixels within that sub-region, and calculate the average radiation temperature T2 corresponding to all pixels in the surrounding neighborhood (the neighborhood being a ring of rectangular sub-regions adjacent to the sub-region). The local temperature difference is then calculated. ;

[0057] The radiation temperature corresponding to a single pixel is obtained by converting the calibration parameters obtained from the previous radiation calibration by the infrared thermal imager carried by the pipeline robot;

[0058] The specific steps for the preliminary radiation calibration are as follows:

[0059] Several blackbody or surface source calibration heat sources with known standard temperatures are selected, covering the temperature range that may occur during normal operation and leakage of raw coal gas pipelines, preferably from -20℃ to 450℃. Each calibration heat source is placed sequentially within the field of view of the pipeline robot, keeping the shooting distance and imaging angle consistent with the actual monitoring. Infrared images of each calibration heat source at different set temperatures are acquired by the infrared thermal imager, and the grayscale value or digital quantization value of each pixel in the calibration heat source area in each frame of infrared image is recorded. A mapping relationship between grayscale values ​​and known standard temperatures is established to generate a calibration data table. The calibration data table also includes the corresponding parameters of infrared image pixel grayscale values ​​and actual temperatures under different ambient temperatures and shooting distances.

[0060] The thermal steepness reference value for a single sub-region is the average thermal steepness of each pixel in that sub-region; the Sobel operator is used to calculate the horizontal gradient of each pixel in that region. and vertical gradient Thus, the thermal steepness G of that pixel is obtained: The Sobel operator is used to calculate the horizontal gradient Gx and vertical gradient Gy of each pixel, which is a common technique used by those skilled in the art and will not be elaborated on in detail.

[0061] The preset local temperature difference and preset thermal steepness reference values ​​are determined through experimental calibration of the on-site working conditions, temperature interference characteristics, and leak samples of the raw coal gas pipeline. They can also be adaptively adjusted according to the detection sensitivity requirements. The local temperature difference and thermal steepness reference values ​​effectively reflect the degree of thermal anomaly of the pipeline wall. The calibration rule is that the greater the detection requirement for minor leaks, the smaller the preset local temperature difference and preset thermal steepness reference values ​​should be. In this embodiment, the preset local temperature difference is 0.2℃ and the preset thermal steepness reference value is 3.0.

[0062] It should be noted that if no characteristic area is found, it is determined that there is no leak in the target pipeline area.

[0063] Specifically, the analysis and selection module determines to perform cloud feature analysis when the number of cloud feature images is greater than or equal to the preset number of cloud feature images.

[0064] The analysis and selection module determines to conduct test analysis when the number of cloud feature images is less than the preset number of cloud feature images.

[0065] Specifically, the number of cloud feature images is the number of cloud feature images in the infrared image;

[0066] The user can determine the preset number of cloud feature images based on the actual application scenario. This number is used to characterize the frequency of cloud feature images appearing in the total infrared images within a preset time period, thereby reflecting whether there are continuous signs of gas leakage in the pipeline area. The higher the user's sensitivity requirement for detecting minor leaks and early leaks, the smaller the preset number of cloud feature images should be. In this embodiment, the preset number of cloud feature images is set to 70% of the total number of infrared images collected.

[0067] Specifically, the analysis and selection module determines that an infrared image with an inter-frame optical flow motion coefficient greater than a preset inter-frame optical flow motion coefficient and a regional texture entropy value greater than a preset regional texture entropy value is a cloud feature image.

[0068] For a single infrared image, the formula for calculating the inter-frame optical flow motion coefficient F is:

[0069]

[0070] Where N is the number of pixels in the infrared image, and T0 is the average radiation temperature corresponding to each pixel in the infrared image. Let be the radiation temperature corresponding to the i-th pixel in the infrared image. Let i be the radiation temperature of the i-th pixel in the infrared image that is adjacent to and sequentially follows the current infrared image, where i = 1, 2, ..., N. The movement of the leaking gas cloud causes significant temperature changes in the corresponding pixels of adjacent frames, while the temperature changes in a fixed background or static interference are relatively small. Therefore, the sum of the temperature differences between corresponding pixels of adjacent frames is used to characterize the degree of dynamic change in the temperature field, effectively reflecting the intensity of cloud movement.

[0071] The formula for calculating the region texture entropy value H corresponding to a single infrared image is:

[0072]

[0073] Where K is the number of different pixel values. is the ratio of the number of pixels with a pixel value of m to N, where m is 1, 2, ..., K; the edges of the leaking gas cloud are irregular and the distribution is messy, with poor uniformity of pixel value distribution, corresponding to a high texture entropy value; the background areas such as the pipe wall and coking layer have relatively uniform and regular textures and low texture entropy values, so they can be used to distinguish gas clouds from stable backgrounds.

[0074] The preset values ​​of the inter-frame optical flow motion coefficient and the preset regional texture entropy value are determined through experimental calibration using on-site operating conditions, temperature interference characteristics, and leak samples of raw coal gas pipelines. They can also be adaptively adjusted according to detection sensitivity requirements. The inter-frame optical flow motion coefficient characterizes the dynamic changes in the temperature field between image sequences, while the regional texture entropy value characterizes the disorder of the texture in the image region. The combination of these two values ​​effectively reflects whether candidate regions in infrared images exhibit the dynamic movement and disordered texture characteristics unique to leaking gas clouds. The calibration rule is that the higher the detection requirement for minute leaks or early leaks, the smaller the preset values ​​of the inter-frame optical flow motion coefficient and the preset regional texture entropy value should be. In this embodiment, the preset inter-frame optical flow motion coefficient is 0.3, and the preset regional texture entropy value is 2.5.

[0075] Specifically, the feature analysis module determines that there is a leak in the target pipeline area when the gas absorption depth is greater than the preset gas absorption depth and the peak interval variation coefficient is greater than the preset peak interval variation coefficient.

[0076] Specifically, the feature analysis module determines that there is no leakage in the target pipeline area when the gas absorption depth is less than or equal to a preset gas absorption depth or the peak interval variation coefficient is less than or equal to a preset peak interval variation coefficient.

[0077] The cloud regions in each cloud feature image are identified by a pre-trained semantic segmentation model and denoted as the first region. Based on the first region, a ring-shaped region is extracted by extending outward by 10 pixels, which does not overlap with the first region and is denoted as the second region. The pre-trained semantic segmentation model adopts a deep learning-based semantic segmentation network and is pre-trained with a large number of cloud feature image samples. It can accurately identify cloud regions in cloud feature images. The specific training process of the pre-trained model is a common technique used by those skilled in the art and will not be described in detail.

[0078] The gas absorption depth is the average value of the sub-gas absorption depths corresponding to the feature images of each cloud cluster in the first preset time period. The sub-gas absorption depth corresponding to a single cloud cluster feature image = (average value of radiation temperature corresponding to each pixel in the entire second region - average value of radiation temperature corresponding to each pixel in the entire first region) / average value of radiation temperature corresponding to each pixel in the entire second region.

[0079] The cloud feature images in the infrared images captured in each first preset time period are identified, and the time between two adjacent cloud feature images is recorded as a time interval. The peak interval variation coefficient is the ratio of the standard deviation of each time interval to the mean of each time interval. Interferences such as coking breathing and thermal expansion and contraction are periodic, and the time interval of cloud appearance is relatively stable, with a small standard deviation and low variation coefficient. Real leakage occurs randomly, and the time interval of cloud appearance is highly discrete, so the variation coefficient is higher.

[0080] The values ​​of the preset gas absorption depth and the preset peak interval variation coefficient are determined through experimental calibration using on-site operating conditions, temperature interference characteristics, and leak samples of raw coal gas pipelines. They can also be adaptively adjusted according to detection sensitivity requirements. The gas absorption depth effectively reflects the degree of temperature difference between the cloud region and the background region, thus characterizing the absorption capacity of the leaked gas for infrared radiation. The greater the difference and the stronger the absorption capacity, the higher the probability of a leak. The peak interval variation coefficient effectively reflects the regularity of the time interval between cloud appearances. The stronger the periodicity (the smaller the variation coefficient), the more likely it is to be due to interference such as coking layer breathing; the stronger the randomness (the larger the variation coefficient), the more likely it is to be a real leak. For small leaks and early leaks, the higher the detection sensitivity requirement, the smaller the values ​​of the preset gas absorption depth and the preset peak interval variation coefficient. In this embodiment, the preset gas absorption depth is 0.08, and the preset peak interval variation coefficient is 0.5.

[0081] The principle is as follows: When raw coal gas leaks, the leaking gas forms a diffuse cloud with infrared absorption characteristics inside the pipeline. This cloud selectively absorbs the infrared energy radiated outward from the pipe wall, resulting in a significantly lower infrared radiation temperature in the cloud region compared to the background region without gas coverage. This temperature difference is quantified by the gas absorption depth; the greater the absorption depth, the stronger the attenuation effect of the gas on infrared radiation, and the higher the credibility of the leak. Furthermore, real gas leaks typically exhibit random, non-periodic diffusion characteristics, with a highly discrete time interval between cloud appearances. In contrast, interferences such as coking and shedding of the pipeline wall, and thermal expansion and contraction, often show periodic or quasi-periodic changes. The peak interval variation coefficient effectively characterizes the degree of time interval dispersion. A larger variation coefficient indicates that the cloud appearance pattern is closer to a random distribution, better eliminating periodic interference and thus more reliably confirming a real leak.

[0082] Specifically, the test analysis module determines that the target pipeline area is under suspected leakage conditions when the multi-band absorption ratio is greater than or equal to the preset multi-band absorption ratio.

[0083] The test and analysis module responds to the condition that the multi-band absorption ratio is less than the preset multi-band absorption ratio, determines that the target pipeline area is not under suspected leakage conditions, and determines that there is no leakage in the target pipeline area.

[0084] The test analysis aims to enhance the distinguishability between real leaks and interference sources through active excitation. This embodiment uses low-frequency vibration-assisted multi-band absorption ratio detection, and the specific implementation steps are as follows:

[0085] The pipeline robot is controlled to move to the vicinity of the target pipeline area, and the gimbal is adjusted so that the dual-band infrared thermal imager is pointing directly at the area. The pipeline robot is equipped with a controllable low-frequency vibration generator to apply mechanical vibration of a known frequency to the pipe wall. Those skilled in the art can select the specific type of the controllable low-frequency vibration generator according to the actual working conditions, including but not limited to piezoelectric ceramic actuators, high-temperature special motors, or standard vibration motors equipped with thermal protection devices.

[0086] The vibration parameters are set as follows: vibration frequency f = 100Hz; vibration amplitude (peak-to-peak displacement) is controlled at... Within the range; the vibration duration is 30s.

[0087] Simultaneously with the vibration start command issued by the control system, the dual-band infrared thermal imager is triggered to begin acquiring images. During the vibration period, L frames of infrared images are continuously acquired, which is recorded as the second preset time period. In this embodiment, the frame rate of the infrared thermal imager is 1Hz, and 30 frames are acquired, corresponding to a time period of 30 seconds.

[0088] The dual-band infrared thermal imager mounted on the pipeline robot simultaneously images the target pipeline area in the mid-wave infrared and long-wave infrared bands. The mid-wave infrared response band is 3.2–3.5 μm, covering the strong absorption peaks of target gases such as methane and benzene in raw coal gas. The long-wave infrared response band is 8–14 μm; in this band, the absorption of target gases is extremely weak or negligible, and it is mainly used to obtain background radiation attenuation information under conditions of no gas absorption characteristics.

[0089] Multi-band absorption ratio = average of sub-gas absorption depth of each cloud feature image in the mid-wave infrared band within the second preset time period / average of sub-gas absorption depth of each cloud feature image in the long-wave infrared band within the second preset time period.

[0090] The preset multi-band absorption ratio is determined through experimental calibration of the on-site working conditions, temperature interference characteristics, and leak samples of the raw coal gas pipeline. It can also be adaptively adjusted according to the detection sensitivity requirements. The multi-band absorption ratio effectively reflects the difference in radiation attenuation between the mid-wave infrared and long-wave infrared bands of the cloud region, thereby distinguishing between real gas leaks and non-gas interference. The greater the demand for high-sensitivity detection, the smaller the preset multi-band absorption ratio should be. In this embodiment, the preset multi-band absorption ratio is 1.5.

[0091] The principle is as follows: Components such as methane and benzene in raw coal gas have characteristic absorption peaks in the mid-wave infrared band, and leaked gas clouds will significantly attenuate the infrared radiation in this band. However, in the long-wave infrared band, the target gas has almost no infrared absorption; the radiation attenuation mainly comes from background factors such as pipe walls and dust. Through simultaneous dual-band imaging, the selective absorption of the gas can be separated from non-specific interference such as background thermal radiation and dust scattering. The multi-band absorption ratio quantifies the ratio of the attenuation caused by gas absorption in the mid-wave band to the background attenuation in the long-wave band. The larger this ratio, the higher the credibility of the thermal anomaly being caused by gas infrared absorption.

[0092] Simultaneously, applying low-frequency controllable vibration to the pipe wall causes the dispersion state and cloud morphology of the actual leaking gas to change systematically with the vibration, resulting in significant fluctuations in the gas absorption depth; while interference sources such as coking hot spots and localized overheating of the pipe wall show no obvious response under vibration excitation. By combining vibration excitation with dual-band infrared detection, leakage characteristics can be further enhanced and fixed interference suppressed, enabling reliable identification of suspected leak areas.

[0093] Specifically, the suspected analysis module determines the region category of the target pipeline area based on the spatial convergence of the leak source, including:

[0094] If the spatial convergence of the leakage source is greater than or equal to the first preset spatial convergence of the leakage source, it is determined to be a Class I region.

[0095] If the spatial convergence of the leakage source is less than the first preset spatial convergence of the leakage source and greater than or equal to the second preset spatial convergence of the leakage source, it is determined to be a type II region.

[0096] If the spatial convergence of the leakage source is less than the second preset spatial convergence of the leakage source, it is determined to be a Class III region;

[0097] Among them, the spatial convergence of the first preset leakage source is greater than that of the spatial convergence of the second preset leakage source.

[0098] The spatial convergence of the leakage source is the maximum value among the subspace convergences corresponding to each cloud region in the mid-wave infrared band feature images within the second preset time period. The formula for calculating the subspace convergence S corresponding to a single cloud region is:

[0099]

[0100] in, For the standard deviation of the circle, R is the average length. |, Let w be the azimuth angle of the edge point of each cloud cluster relative to the centroid. Coordinates of the centroid ( ,y0) = (the average x-coordinate value of all pixels in the cloud region, the average y-coordinate value of all pixels in the cloud region); N0 is the total number of edge pixels in the cloud region, and edge pixels are pixels located on the boundary of the cloud region; The standard deviation of a circle when the direction angles are uniformly distributed. The theoretical maximum value of S is [0,1]. The larger the value, the more concentrated the direction angle of the edge points is, that is, the cloud spreads outward from a small angle sector near the centroid, which conforms to the geometric characteristics of fixed point source leakage; the smaller the value, the more the edge points are scattered throughout the circumference, and the cloud presents a circular expansion or a diffuse shape without a fixed source point.

[0101] The values ​​of the first and second preset leakage source spatial convergence are determined through experimental calibration of the on-site working conditions, temperature interference characteristics, and leakage samples of the raw coal gas pipeline. They can also be adaptively adjusted according to the detection sensitivity requirements. The leakage source spatial convergence effectively reflects the probability that the cloud originates from a fixed point source. The greater the reliability requirement for leakage determination, the larger the values ​​of the first and second preset leakage source spatial convergence. In this embodiment, the first preset leakage source spatial convergence is 0.8 and the second preset leakage source spatial convergence is 0.5.

[0102] The principle is that: real pipeline leaks are mostly point source jet leaks caused by defects at fixed locations. The leaking gas cloud will diffuse outward in a directional manner from a fixed leak point on the pipe wall. The azimuth distribution of the edge points of the cloud relative to the centroid is highly concentrated, showing obvious directionality and convergence. On the other hand, non-leakage thermal anomalies such as coking and shedding, local hot spots, and dust interference usually do not have fixed leak sources. The cloud is mostly irregularly diffused and omnidirectionally expanding, with the azimuth distribution of the edge points being uniform and discrete.

[0103] The spatial convergence of the leakage source is quantitatively characterized by the spatial concentration and directional characteristics of the cloud by statistically calculating the azimuth angle of the edge pixels of the cloud. The larger the value, the more concentrated the cloud diffusion direction is, and the more it conforms to the geometric morphological characteristics of fixed point source leakage. The smaller the value, the more uniform and diffuse the cloud distribution is, and the more it tends to be non-leakage thermal interference.

[0104] Based on this physical characteristic, the present invention divides the target pipeline area into different categories by the spatial convergence of the leakage source, realizes the morphological characteristics classification of the suspected leakage area, and then adopts a differentiated judgment strategy for different convergence characteristics, which not only ensures the accurate identification of point source type real leakage, but also avoids misjudgment of diffuse interference, thereby improving the reliability and pertinence of leakage detection.

[0105] Specifically, the suspected analysis module targets a type of area and uses a processing strategy based on the leakage plume spread ratio to determine whether a leak exists.

[0106] If the leakage plume extension ratio is greater than the preset leakage plume extension ratio, it is determined that there is a leak in the target pipeline area.

[0107] If the leakage plume extension ratio is less than or equal to the preset leakage plume extension ratio, it is determined that there is no leakage in the target pipeline area.

[0108] The leakage plume spread ratio is the maximum value among the sub-spread ratios corresponding to cloud regions with subspace convergence in the [0.5, 0.8] range in the mid-infrared band feature images of each cloud cluster within the second preset time period.

[0109] The sub-spread ratio corresponding to a single cloud region is the ratio of the major axis length to the minor axis length of that cloud region;

[0110] The major axis length is the length of the longer side of the smallest bounding rectangle of the cloud region, and the minor axis length is the length of the shorter side of the smallest bounding rectangle of the cloud region.

[0111] The preset leak plume extension ratio is determined by experimental calibration of the on-site working conditions, temperature interference characteristics, and leak samples of the raw coal gas pipeline. It can also be adaptively adjusted according to the detection sensitivity requirements. The leak plume extension ratio effectively reflects whether the cloud area presents a slender plume shape formed by high-speed jetting. The greater the requirement for the anti-false alarm capability of leak detection (i.e., to avoid misjudging circular or near-circular interference as a leak), the larger the value of the preset leak plume extension ratio. In this embodiment, the preset leak plume extension ratio is 2.5.

[0112] The principle is as follows: a certain type of region corresponds to an area with high spatial convergence of the leakage source and obvious point source diffusion characteristics. Such thermal anomalies are more likely to originate from gas jet leakage at fixed defects in the pipe wall. In infrared images, real pipeline gas jet leakage usually appears as a slender feather-like shape extending outward from the leakage point. The cloud outline shows significant anisotropy, with the major axis dimension being much larger than the minor axis dimension. Even if interferences such as coke deposits and local overheating spots show a certain degree of spatial convergence, their shapes are still mostly close to circular or elliptical, with smaller differences between the major and minor axes.

[0113] The leak plume extension ratio, measured by the ratio of the major axis to the minor axis of the cloud's minimum circumscribed rectangle, quantifies the degree of stretching and extension in a thermal anomaly region. A larger ratio indicates that the cloud morphology closely resembles a typical jet plume, increasing the probability of a genuine leak; a smaller ratio suggests a more rounded morphology, making it more likely to be a non-leakage interference. Using this parameter for secondary assessment of a type of region allows for the elimination of interference from areas whose morphology does not meet the characteristics of a jet leak, while retaining the features of point source leaks, significantly improving the leak detection's resistance to false alarms.

[0114] Specifically, the suspected analysis module targets the second type of area, and the judgment and processing strategy is to determine whether there is a leak based on the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient.

[0115] If the edge sharpness of the thermal anomaly is greater than the preset edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient is greater than the preset vibration-induced modulation coefficient, then it is determined that there is a leak in the target pipeline area.

[0116] If the edge sharpness of the thermal anomaly is less than or equal to the preset edge sharpness of the thermal anomaly, or the vibration-induced modulation coefficient is less than or equal to the preset vibration-induced modulation coefficient, then it is determined that there is no leakage in the target pipeline area.

[0117] Thermal anomaly edge sharpness = average value of temperature difference reference value corresponding to each edge pixel in the feature image of each cloud cluster in the mid-wave infrared band within the second preset time period / Δd;

[0118] For a single edge pixel, starting from the edge point (xi, yi), move d pixels outward (towards the background) along the normal vector direction (nx, ny) to obtain the outer point (xout, yout); move d pixels inward along the opposite normal direction (−nx, −ny) to obtain the inner point (xin, yin). In this embodiment, the moving distance d is 2 pixels. Record the radiation temperature Tin of the inner point and the radiation temperature Tout of the outer point. , ;

[0119] Temperature difference reference value for a single edge pixel The physical distance Δd (unit: mm) between the inner and outer points is calculated. This distance is obtained by multiplying the number of moving pixels 2d by the spatial resolution of the infrared thermal imager (the actual number of millimeters per pixel): Δd = 2d × Rspatial, where Rspatial is the spatial resolution, which is 0.2 mm / pixel in this embodiment.

[0120] Vibration-induced modulation coefficient = the difference between the maximum and minimum values ​​of the sub-gas absorption depth of each cloud feature image in the mid-wave infrared band during the second preset time period / the difference between the maximum and minimum values ​​of the sub-gas absorption depth of each cloud feature image during the first preset time period.

[0121] The values ​​of the preset thermal anomaly edge sharpness and the preset vibration-induced modulation coefficient are determined through experimental calibration of the on-site working conditions, temperature interference characteristics, and leak samples of the raw coal gas pipeline. They can also be adaptively adjusted according to the detection sensitivity requirements. The thermal anomaly edge sharpness and the vibration-induced modulation coefficient effectively reflect the degree of temperature change at the hot spot boundary and the modulation enhancement effect of vibration on the gas absorption depth. The greater the requirement for detection sensitivity of micro-leakage and early-stage leakage, the smaller the values ​​of the preset thermal anomaly edge sharpness and the preset vibration-induced modulation coefficient. In this embodiment, the preset thermal anomaly edge sharpness is 5.0℃ / mm and the preset vibration-induced modulation coefficient is 3.0.

[0122] The principle is that the spatial convergence of the leakage source in the second type of region is in the medium range, and it is difficult to reliably distinguish between real leakage and interference by relying solely on morphological characteristics. Therefore, the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient are combined for dual judgment to improve the ability to identify weak and early leakage.

[0123] The gas cloud formed by a real leak is caused by the selective absorption of infrared radiation by the leaking gas. There is a clear and steep temperature boundary between the cloud and the background pipe wall. The temperature changes drastically along the normal of the edge, so the edge of the thermal anomaly is sharp. In contrast, interferences such as coking hot spots, local overheating, and dust adhesion are mostly gradual heat distributions with gentle temperature transitions at the edges, resulting in a lower edge sharpness of the thermal anomaly.

[0124] Meanwhile, under low-frequency vibration excitation, the dispersion state and distribution position of the leaked gas cloud will change significantly with the vibration of the pipe wall, resulting in significant fluctuations in the gas absorption depth. This makes the characteristic change amplitude during vibration much greater than before vibration, i.e., the vibration-induced modulation coefficient is high. On the other hand, interference sources such as fixed hot spots and coking are not affected by vibration, and the gas absorption depth remains basically stable, with a low vibration-induced modulation coefficient.

[0125] By simultaneously using edge sharpness and vibration modulation coefficient for joint judgment, it is possible to effectively distinguish between real leaks and various types of interference, and to accurately identify small leaks and early leaks in areas with moderate convergence.

[0126] Specifically, the suspected analysis module targets three types of areas, and the judgment and processing strategy is to determine whether to adjust the monitoring frequency based on the comprehensive evaluation value;

[0127] If the comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value, the monitoring frequency will be adjusted.

[0128] The comprehensive evaluation value is calculated as follows: (Sharpness of thermal anomaly edge / Preset sharpness of thermal anomaly edge × 0.5) + (Vibration-induced modulation coefficient / Preset vibration-induced modulation coefficient × 0.5).

[0129] The user can determine the value of the preset comprehensive evaluation value according to the actual application scenario. The comprehensive evaluation value effectively reflects the severity of the comprehensive defects in the target pipeline area (combining edge steepness and vibration response characteristics). The greater the user's demand for rapid response and fine detection, the smaller the value of the preset comprehensive evaluation value. In this embodiment, the preset comprehensive evaluation value is 0.5.

[0130] When adjusting the monitoring frequency, the initial monitoring frequency for the target pipeline area was changed from once every 30 days to once every 7 days.

[0131] The principle is that the spatial convergence of the leakage source in the three types of areas is low, and the cloud is distributed in a diffuse manner. They do not meet the morphological characteristics for direct judgment as leakage. However, these areas often correspond to early and potential anomalies such as micro-cracks and loose defects in the pipe wall. Although they have not formed obvious jet leakage, they have the risk of gradually developing into actual leakage.

[0132] The comprehensive evaluation value combines the sharpness of the thermal anomaly edge with the vibration-induced modulation coefficient using normalized weighting. This comprehensive evaluation value reflects both the degree of thermal boundary abrupt change and the degree of vibration response in the region, quantifying the severity of potential defects. A higher comprehensive evaluation value indicates that the thermal anomaly characteristics and dynamic response characteristics of the region are closer to early-stage leakage, and the higher the potential leakage risk.

[0133] To this end, this invention increases the monitoring frequency and shortens the inspection cycle for areas that reach the risk threshold, enabling key monitoring and trend tracking of low-convergence, high-risk areas. Without causing a large number of misjudgments, it can identify potential leakage hazards in advance, prevent early anomalies from going unnoticed and gradually developing into serious leakage accidents, and ensure the safe operation of raw coal gas pipelines.

[0134] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A gas leak intelligent monitoring system based on infrared thermal imaging, characterized in that, include: A pipeline robot, used to acquire infrared images of a target pipeline area within a first preset time period; A feature region extraction module, which is connected to the pipeline robot, is used to determine the feature regions in the infrared image based on local temperature difference and thermal steepness reference values. An analysis and selection module, which is connected to the pipeline robot and the feature region extraction module respectively, is used to determine whether to perform cloud feature analysis or test analysis based on the number of cloud feature images when feature regions exist; wherein, cloud feature images are determined based on inter-frame optical flow motion coefficients and regional texture entropy values; The feature analysis module, which is connected to the analysis selection module, is used to determine whether there is a leak in the target pipeline area based on the gas absorption depth and the peak interval variation coefficient in cloud feature analysis. The test analysis module is connected to the pipeline robot and the analysis selection module respectively. It is used to apply low-frequency vibration to the target pipeline area during the test analysis, and determine whether the target pipeline area is under suspected leakage conditions based on the multi-band absorption ratio of the infrared image collected in the second preset time period after the application. The suspected leakage analysis module, which is connected to the test analysis module, is used to determine the region category of the target pipeline area based on the spatial convergence of the leakage source under suspected leakage conditions, and to determine the processing strategy based on the region category. The processing strategy includes determining whether a leak exists based on the leak plume spread ratio, determining whether a leak exists based on the edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient, and determining whether to adjust the monitoring frequency based on a comprehensive evaluation value.

2. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 1, characterized in that, The feature region extraction module determines the sub-regions whose local temperature difference is greater than or equal to a preset local temperature difference and whose thermal steepness reference value is greater than or equal to a preset thermal steepness reference value as feature regions.

3. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 2, characterized in that, The analysis and selection module determines to perform cloud feature analysis when the number of cloud feature images is greater than or equal to the preset number of cloud feature images. The analysis and selection module determines to conduct test analysis when the number of cloud feature images is less than the preset number of cloud feature images.

4. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 3, characterized in that, The analysis and selection module determines that infrared images with inter-frame optical flow motion coefficients greater than preset inter-frame optical flow motion coefficients and regional texture entropy values ​​greater than preset regional texture entropy values ​​are cloud feature images.

5. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 4, characterized in that, The feature analysis module determines that there is a leak in the target pipeline area when the gas absorption depth is greater than the preset gas absorption depth and the peak interval variation coefficient is greater than the preset peak interval variation coefficient.

6. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 3, characterized in that, The test and analysis module determines that the target pipeline area is under suspected leakage conditions if the multi-band absorption ratio is greater than or equal to the preset multi-band absorption ratio.

7. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 6, characterized in that, The suspected analysis module determines the region category of the target pipeline area based on the spatial convergence of the leak source, including: If the spatial convergence of the leakage source is greater than or equal to the first preset spatial convergence of the leakage source, it is determined to be a Class I region. If the spatial convergence of the leakage source is less than the first preset spatial convergence of the leakage source and greater than or equal to the second preset spatial convergence of the leakage source, it is determined to be a type II region. If the spatial convergence of the leakage source is less than the second preset spatial convergence of the leakage source, it is determined to be a Class III region; Among them, the spatial convergence of the first preset leakage source is greater than that of the spatial convergence of the second preset leakage source.

8. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 7, characterized in that, The suspected analysis module targets a type of area and uses a processing strategy based on the leakage plume extension ratio to determine whether a leak exists. If the leakage plume extension ratio is greater than the preset leakage plume extension ratio, it is determined that there is a leak in the target pipeline area.

9. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 8, characterized in that, The suspected analysis module targets two types of areas, and the judgment and processing strategy is to determine whether there is a leak based on the sharpness of the thermal anomaly edge and the vibration-induced modulation coefficient. If the edge sharpness of the thermal anomaly is greater than the preset edge sharpness of the thermal anomaly and the vibration-induced modulation coefficient is greater than the preset vibration-induced modulation coefficient, then it is determined that there is a leak in the target pipeline area.

10. The intelligent gas leak monitoring system based on infrared thermal imaging according to claim 9, characterized in that, The suspected analysis module targets three types of areas, and the judgment and processing strategy is to determine whether to adjust the monitoring frequency based on the comprehensive evaluation value. If the comprehensive evaluation value is greater than or equal to the preset comprehensive evaluation value, the monitoring frequency will be adjusted.

Images