Plume model image data enhanced gas leakage visual detection method and system

Abstract

The invention belongs to the technical field of gas leakage monitoring, and particularly relates to a plume model image data enhanced gas leakage visual detection method and system, and the method comprises the steps: obtaining a background image of a to-be-detected site; establishing a pipeline three-dimensional physical model of a natural gas terminal on a to-be-detected site, setting initial conditions of gas leakage, and generating a gas flow concentration field image; superposing the gas flow concentration field image and the background image; performing data enhancement on the superposed image to obtain enhanced image data; and training the enhanced image data by adopting a deep learning method to obtain a gas leakage detection model, and automatically identifying a leakage area by adopting the gas leakage detection model. According to the method and the device, the enhanced image data can be trained by adopting a deep learning method under the condition of no actual leakage data, so that the gas leakage detection model is obtained, and more efficient and more accurate gas leakage detection and early warning are realized.

Description

Technical Field

[0001] The present application belongs to the technical field of gas leakage monitoring, and specifically relates to a gas leakage visual detection method and system using plume model image data enhancement. Background Art

[0002] Gas leaks are a common and dangerous safety hazard in industrial production, particularly in the oil, natural gas, and chemical industries. They can not only cause production halts and property losses, but can also trigger serious accidents such as fires and explosions, threatening human life and the environment. With the acceleration of global industrialization, the increasing frequency and serious consequences of gas leaks are driving a growing demand for gas leak detection technology across all industries. In environments such as oil and gas pipelines, chemical plants, and mining, timely detection and accurate location of gas leaks are crucial for preventing accidents and ensuring industrial safety.

[0003] Existing gas leak detection technologies primarily rely on sensor devices and traditional image recognition methods. Although these technologies are effective under certain conditions, they still have the following significant drawbacks: Lack of sufficient real-world leakage datasets: Existing deep learning-based gas leak detection methods typically rely on large, labeled datasets for training. However, due to the low frequency of actual leaks and the complex environments involved, obtaining large-scale, high-quality datasets of actual leaks is extremely difficult. This results in insufficient data for model training, impacting the accuracy and generalization of the detection models. The lack of sufficient, accurately labeled datasets of actual leaks leads to high false detection and false alarm rates in practical applications, impacting the accuracy and reliability of gas leak detection. Existing technologies are unable to address this issue, resulting in high false detection and missed detection rates in gas leak detection models.

[0004] Sensor layout is costly and limited by environmental conditions: Traditional gas leak detection technology relies on sensor networks, which not only requires the deployment of numerous sensors across a wide area, but also limits sensor placement in certain scenarios (such as underground pipelines, chemical plants, and confined spaces), resulting in blind spots and impacting overall monitoring effectiveness. Compared to traditional sensors, infrared cameras offer greater flexibility in covering a wide area without requiring extensive physical sensor installation. They are particularly suitable for hard-to-reach areas such as high altitudes, confined spaces, and remote pipelines.

[0005] Insufficient adaptability to complex gas leakage patterns: Existing gas leak detection technologies often rely on fixed rules or simple models, and are unable to adapt to complex gas flow and diffusion patterns in environments such as multiple gas leaks and shifting gas flow directions. Existing technologies suffer from low detection accuracy, particularly in scenarios where multiple gases are mixed or gas diffusion is complexly intertwined with the background environment. Summary of the Invention

[0006] An embodiment of the present application provides a gas leak visual detection method using plume model image data enhancement, which is at least used to solve the problem of high false detection and false alarm rates caused by the lack of sufficient and accurately labeled actual leakage data sets, thereby affecting the accuracy and reliability of gas leak detection.

[0007] An embodiment of the present application provides a gas leak visual detection method using plume model image data enhancement, the method comprising: Acquire a background image of the scene to be detected; Establish a 3D physical model of the pipeline of the natural gas terminal to be inspected, set the initial conditions for gas leakage, and generate a gas flow concentration field image; Overlaying the gas flow concentration field image with the background image; Performing data enhancement on the superimposed image to obtain enhanced image data; The enhanced image data is trained using a deep learning method to obtain a gas leak detection model, which is then used to automatically identify leakage areas.

[0008] Furthermore, the initial conditions include: gas type, leakage point location, pipeline pressure, air flow velocity and ambient temperature.

[0009] Furthermore, generating the gas flow concentration field image includes: simulating and calculating the gas flow through a plume model to obtain gas concentration distribution at different times and locations.

[0010] Furthermore, the simulation calculation of the gas flow by the plume model includes: Establish a diffusion equation that describes the gas concentration distribution and discretize it to obtain a discretized diffusion equation; A gas velocity field equation describing the gas flow behavior is established and discretized to obtain a discretized gas velocity field equation; Solve the discretized diffusion equation to obtain the value of the gas concentration field at each time step; Solve the discretized gas velocity field equation to obtain the value of the gas velocity field at each time step; The values of the gas concentration field at each time step and the values of the gas velocity field at each time step are converted into visual images.

[0011] Furthermore, converting the values of the gas concentration field at each time step and the values of the gas velocity field at each time step into a visual image includes: setting a threshold, converting values greater than the threshold into labels, and generating a bounding box of the gas leakage area according to the labels.

[0012] Furthermore, the gas flow concentration field image is superimposed with the background image, including: Align the gas concentration field image with the background image in spatial position; According to the weighting coefficient, the overlay ratio of the gas concentration field image and the background image is adjusted.

[0013] Furthermore, data enhancement is performed on the superimposed image to obtain enhanced image data, including: Randomly rotate the image at different angles to simulate leakage at different viewing angles; Flip the image horizontally or vertically to expand the leakage pattern at different angles; Zoom the image to simulate gas flow at different leak sizes; Adjust the brightness or contrast of the image to simulate leakage scenarios under different ambient light conditions.

[0014] Furthermore, when the gas leakage detection model is used to automatically identify the leakage area, the image information of the gas leakage is captured in real time by an infrared camera, matched with the gas flow concentration field image, and the leakage area is identified based on the matching results.

[0015] Another aspect of the present application provides a gas leak visual detection system with plume model image data enhancement, comprising: An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

[0016] A third aspect of an embodiment of the present application provides a gas leak visual detection system with plume model image data enhancement, comprising: Infrared camera, used to collect background images of the scene to be inspected; An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

[0017] Compared with the prior art, the embodiments of the present application have at least the following beneficial effects: By generating simulated gas flow data and combining it with background images collected on site, a diverse training data set is constructed. This method provides sufficiently rich data support for the deep learning model, significantly improving the accuracy of gas leak detection and reducing the occurrence of false detections and false alarms. Through this innovative data generation method, the present application can use deep learning methods to train enhanced image data in the absence of actual leakage data to obtain a gas leak detection model, thereby achieving more efficient and accurate gas leak detection and early warning. The technical support provided by this application not only provides an efficient solution for the gas leak monitoring system, but also has broad application prospects in the fields of industrial safety, environmental monitoring, etc., and can effectively reduce the risk of accidents caused by gas leaks and ensure personnel safety and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 A flow chart of a gas leak visual detection method using plume model image data enhancement provided in an embodiment of the present application; Figure 2 An image of a gas flow concentration field image generated in one state provided by an embodiment of the present application; Figure 3 The image provided in the embodiment of the present application is obtained by superimposing the gas flow concentration field image and the background image; Figure 4 This is a module block diagram of a gas leak visual detection system with plume model image data enhancement provided in an embodiment of the present application. DETAILED DESCRIPTION

[0019] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0020] This embodiment of the application combines plume models with deep learning technology to propose a gas leak visual detection method that uses plume model image data enhancement. By generating simulated images of gas flow, combining them with on-site environment images captured by a camera, and using deep learning algorithms to detect and predict gas leaks, it solves the problems of insufficient data sets, low detection accuracy, and difficult equipment deployment in existing technologies. Figure 1 A gas leak visual detection method using plume model image data enhancement is shown, comprising: S1 obtains the background image of the scene to be detected; An infrared camera can be used to capture the background image of the scene. The background image refers to the image of the scene other than the leaked gas. That is, the background image needs to be captured while ensuring that there is no leakage. In order to ensure the accuracy of gas leak detection, the natural gas terminal or the site to be monitored should be imaged first. The image acquisition device is preferably a cooled infrared focal plane array camera, and the wavelength range should be selected according to the type of gas to be detected, for example: The gas should be selected with a central wavelength of 3.3 Detector, detection The gas should be selected with a central wavelength of 10.6 The detector noise equivalent temperature difference (NETD) should be less than 30mK (to ensure low-noise imaging).

[0021] S2 establishes a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, and uses the three-dimensional physical model of the pipeline to establish a fluid dynamics numerical model, set the initial conditions for gas leakage, and generate a gas flow concentration field image. Figure 2 The gas flow concentration field image under one state is shown; First, at the natural gas terminal or site requiring monitoring, obtain the dimensions of the pipeline where leakage may occur, including the inner and outer wall dimensions, and simulate the size of the leak hole. Using tools such as SolidWorks 3D computer-aided design software, create a 3D physical model of the pipeline at the natural gas terminal. Import this 3D physical model into CFD software (fluid dynamics software) for meshing. Select an appropriate mesh type, such as structured, unstructured, or hybrid. For internal pipeline flow, a structured mesh is typically used. Increase mesh density in critical areas of the pipeline (such as elbows, tees, and valves) to improve simulation accuracy. Check mesh quality to avoid excessive distortion or oversized elements.

[0022] To generate a concentration field image of a gas leak, it is first necessary to set the initial conditions for the gas leak, such as gas type, leak location, pipeline pressure, airflow velocity, and ambient temperature. The plume model is used to simulate the gas flow and obtain the gas concentration distribution at different times and locations.

[0023] The Gaussian plume model here describes the gas concentration distribution of point source emissions under stable atmospheric conditions, assuming that the gas is Gaussian in the vertical and lateral directions.

[0024] S3 superimposes the gas flow concentration field image and the background image; Overlay the acquired background image with the generated gas flow concentration field image. Figure 3 The image shown is the result of superimposing the gas flow concentration field image and the background image. The purpose of this superposition process is to form a comprehensive training dataset by combining the simulated gas leakage image with the actual scene.

[0025] S4 performs data enhancement on the superimposed image to obtain enhanced image data; To increase the diversity of the training dataset and improve the robustness of the model, data augmentation operations can include: Rotation: Randomly rotate the image at different angles to simulate leakage from different viewing angles.

[0026] Flip: Flip the image horizontally or vertically to expand the leakage pattern at different azimuth angles.

[0027] Zoom: Zoom in or out on the image to simulate gas flow at different leak sizes.

[0028] Brightness Adjustment: Adjust the brightness or contrast of the image to simulate leakage scenarios under different ambient light conditions.

[0029] The data augmentation operation is to increase the diversity of the training dataset and enhance the model's generalization ability for different leakage scenarios.

[0030] S5 uses deep learning methods to train enhanced image data to obtain a gas leak detection model, and uses the gas leak detection model to automatically identify leakage areas.

[0031] After data preparation is complete, deep learning methods such as convolutional neural networks (CNNs) or image segmentation networks (UNets) are used to train the superimposed enhanced image data. Through labeled training data, deep learning can learn the characteristics of gas leaks and automatically identify leak areas.

[0032] The training process includes: Input data: superimposed image, including simulated gas flow concentration field image and background image.

[0033] Network structure: Use convolutional neural network (CNN) or image segmentation network (such as UNet) to perform feature extraction and region segmentation on the image.

[0034] Target output: Learn the gas leakage pattern through training and output the precise location of the leakage area.

[0035] The trained gas leak detection model can be used in actual gas leak detection tasks. It can detect gas leaks in real time in a field environment and provide accurate early warning information.

[0036] In one embodiment, simulating the flow of gas using a plume model includes: Establish a diffusion equation that describes the gas concentration distribution and discretize it to obtain a discretized diffusion equation; A gas velocity field equation describing the gas flow behavior is established and discretized to obtain a discretized gas velocity field equation; Solve the discretized diffusion equation to obtain the value of the gas concentration field at each time step; Solve the discretized gas velocity field equation to obtain the value of the gas velocity field at each time step; The values of the gas concentration field at each time step and the values of the gas velocity field at each time step are converted into visual images.

[0037] Specifically, in order to describe the gas concentration distribution, the diffusion equation is first introduced: , in, The gas is in position and time concentration; is the diffusion coefficient of the gas; is the source term, which represents the release rate of gas; is the Laplace operator of concentration, which indicates the change of concentration in space, specifically: , For gas leakage problems, the mass conservation equation can be applied to consider the impact of the gas leakage source in the pipeline. The mass conservation equation is: ,in, is the density of the gas; is the velocity vector of the gas.

[0038] In order to describe the gas flow in the pipeline during the leakage process, the Navier-Stokes equation is used to model the gas flow behavior of the gas velocity field equation: , in, is the pressure of the gas; is the dynamic viscosity of the gas; is an external force, such as gravity.

[0039] Leaked source items Describes the gas release rate from a leak, which is usually related to the pipeline pressure, leak size, and the physical properties of the leaking material. The leak rate can be calculated using the Hugoniot relationship or empirical formula as follows: , in, is the discharge coefficient of the leak, which depends on the shape of the leak and the characteristics of the fluid; is the area of the leak; It is the pressure difference between the inside and outside of the pipe.

[0040] In practical applications, the diffusion of gas leakage is not only affected by the pipeline pressure and the size of the leak, but also closely related to the ventilation mode of the surrounding environment. In order to simulate outdoor scenes, the wind speed field is introduced. , and taking into account the convection effect brought by wind. The diffusion equation can be written as: , In order to calculate the diffusion process of gas in a leakage scenario through numerical simulation, the above equations can be discretized using the finite difference method (FDM), finite element method (FEM) or finite volume method (FVM).

[0041] Using the finite difference method, for the diffusion equation in two dimensions, the discretized diffusion equation can be obtained after discretization: , in, Indicates that at time step Time, location The concentration at is the time step, and is the spatial step size. This equation describes how the gas concentration changes over time and space. At each time step, the gas concentration distribution is updated. Indicates the location The amount of gas released per unit time and per unit volume, Indicates that at time step Time, location The concentration at Indicates that at time step Time, location The concentration at Indicates that at time step Time, location The concentration at Represents that at time step Time, location The concentration at .

[0042] The discretized gas velocity field equation is obtained by using the discretized form of the Navier-Stokes equation: , in, Represents the time step Time, location The speed at Represents the time step Time, location The speed at which the

[0043] During the simulation, set boundary conditions and initial conditions that match the environment. Initial conditions: Set the initial values of gas concentration and flow rate. Boundary conditions: Such as pressure and air velocity at both ends of the pipe, temperature and gas concentration at the environmental boundary.

[0044] By solving the above discretized diffusion equation and discretized gas velocity field equation, the values of the gas concentration field at each time step and the values of the gas velocity field at each time step are obtained, and converted into visual images through graphical tools (such as Matplotlib, ParaView, etc.) to show the process of gas leakage and diffusion.

[0045] In one embodiment, converting the values of the gas concentration field at each time step and the values of the gas flow rate field at each time step into a visual image includes: setting a threshold, converting values greater than the threshold into labels, and generating a bounding box of the gas leakage area based on the labels.

[0046] In one embodiment, superimposing the gas flow concentration field image with the background image includes: aligning the gas concentration field image with the background image in space; and adjusting the superimposition ratio of the gas concentration field image and the background image according to a weighting coefficient.

[0047] Specifically include: Generating a bounding box for the gas leak area based on the label refers to automatically generating a bounding box for the gas leak area based on the pixel values of the concentration image. The generated gas concentration field image is matched with the background image so that the gas concentration field image and the background image are spatially aligned.

[0048] Adjust the transparency or perform weighted averaging on the two images to ensure that the gas leakage area is clearly visible while maintaining the details of the original background image.

[0049] Superimposed gas concentration image and background image It can be expressed as: , in, is the weighting coefficient, which controls the overlay ratio of the gas concentration field image and the background image.

[0050] In one embodiment, when automatically identifying a leak area using a gas leak detection model, an infrared camera is used to capture real-time image information of the gas leak, which is then matched with an image of the gas flow concentration field. The leak area is then identified based on the matching results. The infrared camera is preferably a cooled infrared focal plane array camera.

[0051] On the other hand, an embodiment of the present application provides a gas leak visual detection system with plume model image data enhancement, comprising: An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

[0052] On the other hand, the present invention also provides a gas leakage visual detection system with enhanced plume model image data, see Figure 4 The module block diagram of the gas leak visual detection system with plume model image data enhancement shown includes: an infrared camera for collecting background images of the scene to be detected; An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

[0053] The above-mentioned embodiments of the present application are not limited to traditional gas leak detection scenarios, but can also be widely used for leak detection in complex and dynamic environments, such as chemical plants, mines, oil and gas pipelines, confined spaces and other industrial and environmental scenarios. By generating simulation data and flexible infrared monitoring methods, the system can adapt to different gas types and leakage modes, has good scalability and flexibility, and can provide efficient solutions for gas leak monitoring in multiple industries. It has significant improvements in accuracy, efficiency, cost and other aspects compared to the existing technology. By generating rich data sets, using infrared cameras for flexible monitoring, and optimizing deep learning algorithms, gas leak detection is not only more accurate, but also more adaptable and real-time, solving the problems of false detection, missed detection, insufficient data and so on in the existing technology, and providing stronger technical support for industrial safety and environmental protection.

[0054] This application combines plume models with deep learning. By generating simulated gas flow images using plume models, this approach addresses the issue of insufficient gas leak datasets and provides sufficient training data. By leveraging infrared cameras and image overlay, this approach leverages the advantages of infrared cameras to achieve large-scale, non-contact, and flexible monitoring for gas leak detection. Image overlay enhances the authenticity and diversity of training data. Multiple image enhancement techniques increase the diversity of training data, improving the model's generalization and adaptability.

[0055] Existing technologies typically rely on physical sensors or traditional image recognition methods, which are prone to false detection and missed detection in complex environments. This is particularly true in situations where gas leaks are complex, gas types are diverse, and flow patterns vary, significantly reducing detection accuracy. The embodiments of the present application generate gas flow concentration field images and combine them with deep learning methods for training. This allows them to learn complex gas leakage patterns from large amounts of simulated data, thereby improving the accuracy and reliability of gas leak detection. This is particularly applicable to leaks involving multiple gas mixtures and highly variable patterns, effectively addressing the false detection and missed detection issues found in existing technologies.

[0056] Due to the low frequency of gas leakage incidents and the relative scarcity of actual leakage data sets, existing deep learning-based gas leakage detection systems have difficulty obtaining sufficient labeled data during training, which seriously restricts the performance of the model. By combining the plume model, this application can simulate the discretized diffusion equation and discretized gas velocity field equation under different environments to generate rich gas flow concentration field images. These simulated data can be used as training data sets to supplement existing data, greatly compensating for the problem of insufficient actual leakage data. In this way, not only the diversity of the training data set is improved, but also the generalization ability of deep learning is enhanced, enabling the detection system to show higher accuracy and adaptability in different scenarios.

[0057] Traditional gas leak detection technology relies on deploying a large number of sensor devices, requiring significant capital investment and making effective deployment difficult in complex environments (such as underground pipelines and confined spaces). In contrast, the embodiments of this application utilize only infrared cameras for gas leak monitoring. Infrared cameras detect gas leaks by detecting the infrared absorption characteristics of gases. Unrestricted by sensor placement, they offer flexible coverage over a wide area, making them particularly suitable for applications in difficult-to-reach or complex environments. This significantly reduces deployment costs, simplifies equipment installation and maintenance, and improves system deployment efficiency.

[0058] By combining infrared imaging with deep learning technology, this application can achieve rapid positioning and accurate identification of gas leakage areas. The infrared camera can capture the image information of the gas leakage in real time, and match it with the generated gas flow concentration field image, identify the leakage area based on the matching results, quickly identify the leakage area, and provide immediate detection results. This method not only improves the response speed of gas leak detection, but also provides accurate early warning information to staff in real time, enhances safety monitoring capabilities, and prevents potential accidents in a timely manner. This application is not only limited to traditional gas leak detection scenarios, but can also be widely used for leak detection in complex and dynamic environments, such as chemical plants, mines, oil and gas pipelines, confined spaces and other industrial and environmental scenarios. By generating simulation data and flexible infrared monitoring methods, the system can adapt to different gas types and leakage modes, has good scalability and flexibility, and can provide efficient solutions for gas leak monitoring in multiple industries.

[0059] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A gas leak visual detection method based on plume model image data enhancement, characterized in that: The method comprises: Acquire a background image of the scene to be detected; Establish a 3D physical model of the pipeline of the natural gas terminal to be inspected, set the initial conditions for gas leakage, and generate a gas flow concentration field image; Overlaying the gas flow concentration field image with the background image; Performing data enhancement on the superimposed image to obtain enhanced image data; The enhanced image data is trained using a deep learning method to obtain a gas leak detection model, which is then used to automatically identify leakage areas.

2. The gas leak visual detection method based on plume model image data enhancement according to claim 1, characterized in that: The initial conditions include: gas type, leak point location, pipeline pressure, air flow velocity and ambient temperature.

3. The gas leak visual detection method based on plume model image data enhancement according to claim 1, characterized in that: Generating a gas flow concentration field image includes: simulating and calculating the flow of gas through a plume model to obtain gas concentration distribution at different times and locations.

4. The gas leak visual detection method based on plume model image data enhancement according to claim 3, characterized in that: The simulation calculation of gas flow through plume model includes: Establish a diffusion equation that describes the gas concentration distribution and discretize it to obtain a discretized diffusion equation; A gas velocity field equation describing the gas flow behavior is established and discretized to obtain a discretized gas velocity field equation; Solve the discretized diffusion equation to obtain the value of the gas concentration field at each time step; Solve the discretized gas velocity field equation to obtain the value of the gas velocity field at each time step; The values of the gas concentration field at each time step and the values of the gas velocity field at each time step are converted into visual images.

5. The gas leak visual detection method based on plume model image data enhancement according to claim 4, characterized in that: Converting the values of the gas concentration field at each time step and the gas velocity field at each time step into a visual image includes: setting a threshold, converting values greater than the threshold into labels, and generating a bounding box of the gas leakage area according to the labels.

6. The gas leak visual detection method based on plume model image data enhancement according to claim 5, characterized in that: Overlay the gas flow concentration field image with the background image, including: Align the gas concentration field image with the background image in spatial position; According to the weighting coefficient, the overlay ratio of the gas concentration field image and the background image is adjusted.

7. The gas leak visual detection method based on plume model image data enhancement according to claim 1, characterized in that: The superimposed image is subjected to data enhancement to obtain enhanced image data, including: Randomly rotate the image at different angles to simulate leakage at different viewing angles; Flip the image horizontally or vertically to expand the leakage pattern at different angles; Zoom the image to simulate gas flow at different leak sizes; Adjust the brightness or contrast of the image to simulate leakage scenarios under different ambient light conditions.

8. The gas leak visual detection method based on plume model image data enhancement according to claim 1, characterized in that: When the gas leak detection model is used to automatically identify the leakage area, the image information of the gas leakage is captured in real time by an infrared camera, matched with the gas flow concentration field image, and the leakage area is identified based on the matching results.

9. A gas leak visual detection system enhanced with plume model image data, characterized in that: include: An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

10. A gas leak visual detection system with plume model image data enhancement, characterized in that: include: Infrared camera, used to collect background images of the scene to be inspected; An acquisition module, used to obtain a background image of the scene to be detected; The simulation module is used to establish a three-dimensional physical model of the pipeline of the natural gas terminal to be tested, set the initial conditions for gas leakage, and generate a gas flow concentration field image; A superposition module is used to superimpose the gas flow concentration field image with the background image; An enhancement module, used for performing data enhancement on the superimposed image to obtain enhanced image data; The learning module is used to train the enhanced image data using a deep learning method to obtain a gas leak detection model, and automatically identify the leakage area using the gas leak detection model.

Images