Steam pipeline leakage monitoring system based on image recognition
By applying image recognition technology and deep learning algorithms in the steam pipeline leakage monitoring system, real-time analysis of steam pipeline images is solved, and the problem that traditional monitoring methods are difficult to achieve real-time and accurate monitoring is significantly improved, which is significantly improved the efficiency and accuracy of leakage detection.
Patent Information
- Application Number
- CN202510039536.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Priority Date
- 2024-07-18
- Filing Date
- 2025-01-10
- Publication Date
- 2025-05-30
AI Technical Summary
Traditional steam pipeline leakage monitoring methods rely on manual inspection and simple sensors, making it difficult to achieve comprehensive and real-time monitoring of the pipeline system, and there are safety risks and efficiency problems.
The steam pipeline leakage monitoring system based on image recognition is adopted, and high-resolution cameras and deep learning algorithms are used to capture and analyze the images around the steam pipeline in real time, judge the existence and severity of the leakage, and issue an alarm through the data display and alarm feedback unit.
It significantly improves the accuracy and efficiency of steam pipeline leakage detection, reduces the risk of safety accidents and maintenance costs, and realizes real-time and accurate monitoring of steam pipelines.
Smart Images
Figure CN120070328A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the technical field of industrial pipeline monitoring and safety, and particularly relates to a steam pipeline leakage monitoring system based on image recognition. Background Art
[0002] With the continuous improvement of the global industrialization level and the widespread application of automation technology, the demand for the monitoring and maintenance of key industrial facilities is increasing day by day, especially in fields with high safety importance, such as petrochemical, nuclear power generation, and other heavy industrial production lines. In these fields, as a key component for transporting energy and raw materials, the safe operation of steam pipelines is directly related to the stability and safety of the entire production process. Therefore, timely and accurately monitoring whether there is a leakage in the steam pipeline has become an important link to ensure production efficiency and environmental safety.
[0003] However, traditional steam pipeline leakage monitoring methods mainly rely on regular manual inspections or monitoring systems based on simple sensors. These methods not only consume time and effort, but also are difficult to achieve comprehensive and real-time monitoring of the pipeline system. In a complex industrial environment, such as high temperature, high pressure, and the presence of toxic chemicals, this monitoring method has huge safety risks and efficiency problems. In addition, due to the subjectivity and non-continuity of manual inspections, the ability to identify minor leaks is extremely limited, which may lead to missed or false detections, increasing potential safety hazards and economic losses.
[0004] With the rapid development of artificial intelligence technology, especially image recognition and deep learning technology, new technical means have been provided to solve the above problems. These technologies can automatically identify signs of steam leakage by analyzing monitoring videos or image data, and achieve real-time and accurate monitoring of steam pipeline leakage. Compared with traditional methods, the monitoring system based on image recognition can significantly improve the accuracy and efficiency of detection, reduce the dependence on manual inspections, and at the same time reduce the safety risks of working in complex and dangerous environments.
[0005] Therefore, developing an efficient and reliable steam pipeline leakage monitoring system based on image recognition has important practical significance and application value for improving the safety monitoring level of industrial pipelines, reducing safety risks, and promoting the development of industrial automation and intelligence. Summary of the Invention
[0006] The purpose of the present invention is to provide a steam pipeline leakage monitoring system based on image recognition, which uses advanced image processing technology and deep learning algorithms to not only be able to detect leakage events in steam pipelines in real time, but also be able to judge the amount of leakage according to the image characteristics of the leakage. This system is specially designed to accurately monitor steam leakage under various environmental conditions, and effectively reduce the safety and economic risks brought by leakage.
[0007] To achieve the above object, the technical solution adopted by the present invention is as follows:
[0008] A steam pipeline leakage monitoring system based on image recognition includes an image acquisition unit for the steam leakage area, an environmental light adjustment unit, an image recognition and leakage judgment unit, and a data display and alarm feedback unit; the image acquisition unit for the steam leakage area captures images of the area around the steam pipeline in real time, and the image acquisition unit for the steam leakage area is configured with an environmental light adjustment unit, which is installed in key pipeline areas; the image recognition and leakage judgment unit optimizes the collected image data through preprocessing operations, performs image analysis using a deep learning model optimized for the characteristics of steam leakage images, and judges the presence and severity of steam leakage; the data display and alarm feedback unit timely feeds back the image recognition results to the operator, and when a leakage is detected, issues an alarm signal of the corresponding level according to the severity of the leakage.
[0009] The image acquisition unit for the steam leakage area is configured with a high-resolution camera, which is installed in key pipeline areas to ensure that the on-site situation of steam leakage can be clearly recorded under different environmental and lighting conditions.
[0010] The preprocessing operations of the image recognition and leakage judgment unit include image denoising, brightness adjustment, and contrast adjustment.
[0011] Images with high quality and accurate perspectives are selected for data annotation, and the annotation content includes the shape, density, and distribution information of the steam.
[0012] The data display and alarm feedback unit displays real-time monitoring images, leakage detection results, and historical data analysis through a graphical user interface. At the same time, when a leakage is detected, it can quickly notify the maintenance team to take countermeasures by means of sound, light signals, or sending electronic messages.
[0013] The contrast adjustment is achieved through the following formula: In the formula: I′ represents the image after contrast adjustment, I is the original image, represents the average brightness of the original image, and α is the contrast adjustment coefficient, which is used to control the increase or decrease of the contrast.
[0014] The preprocessing operations of the image recognition and leakage judgment unit include image sharpening, and the Laplacian operator is used for image sharpening: In the formula: I′ is the sharpened image, I represents the original image, λ is the sharpening intensity coefficient, represents the image obtained by applying the Laplacian operator.
[0015] The deep learning model is based on the Transformer architecture and further combines the optical flow method and multi-scale feature fusion technology.
[0016] The basic building block of the Transformer model is the self-attention mechanism, and its calculation formula is as follows:
[0017]
[0018] In the formula: Q, K, and V represent query, key, and value respectively. They learn the characteristics of steam leakage through different representations of the input image. d k represents the dimension of the key vector, which is used to adjust the scaling problem introduced by the dot product calculation;
[0019] The optical flow method provides important information about the speed and direction of object movement for the model by analyzing the movement of pixel points in consecutive image frames. In steam leakage detection, by calculating the movement of steam over time, the model can capture the dynamic change characteristics of steam. The optical flow vector calculation formula is as follows:
[0020]
[0021] In the formula: F(x, y, t) represents the optical flow vector at position F(x, y) and time t, V x , V y represent the velocity components in the horizontal and vertical directions respectively;
[0022] The multi-scale feature fusion technology ensures that the model can process and identify steam leaks of different sizes and shapes. By extracting image features at different resolutions and then fusing these features together, the model can comprehensively consider various forms and sizes of steam leaks. The formula for the fused features is expressed as:
[0023]
[0024] In the formula: F fusion represents the finally fused feature vector, F i is the feature vector extracted at the i-th scale, w i is the weight of the corresponding scale feature, obtained through learning, and n represents the total number of scales.
[0025] The deep learning model determines the existence and severity of leakage by analyzing the dynamic changes and morphological features of steam. The assessment of the leakage amount is based on the analysis of steam density and diffusion speed. The model can distinguish between minor, moderate, and severe leakage events and be updated in real time to adapt to environmental changes. The steam density in the model is an important static parameter for measuring the leakage amount, and its assessment is based on the comprehensive calculation of the pixel density in the steam area in the image. The calculation process uses the following formula:
[0026]
[0027] In the formula: D represents the steam density, N is the total number of pixel points in the steam area, and p i is the density value of the i-th pixel point. By accurately calculating the density of the steam area in the image, the model can obtain important clues about the size of the leakage amount; the diffusion speed of the steam, as another key dynamic parameter, is evaluated based on the velocity vector calculated by the optical flow method. The calculation of the diffusion speed uses the following formula:
[0028]
[0029] In the formula: S represents the steam diffusion speed, VV x , VV y respectively represent the changes in speed in the horizontal and vertical directions. Through in-depth analysis and real-time monitoring of the two key parameters of steam density and diffusion speed, the model can not only accurately evaluate the severity of the leakage under different environmental conditions, but also adjust the evaluation results in real time as the environment changes.
[0030] The beneficial effects achieved by the present invention are as follows:
[0031] The present invention significantly enhances the accuracy and efficiency of industrial steam pipeline leakage monitoring. By integrating cutting-edge image recognition technology and deep learning algorithms, the system can accurately identify steam leaks under changing lighting and environmental conditions, greatly reducing the risk of safety accidents caused by undetected leaks in a timely manner and reducing maintenance costs. The preprocessing module built into the image recognition and leakage judgment unit further improves the image quality, ensuring the accuracy and reliability of leakage detection. The display terminal in this invention not only provides an intuitive display of leakage monitoring data, but also has a data management function, enabling operators to easily understand the real-time status of the steam pipeline, supporting rapid decision-making based on data, and thus promoting the process of intelligent manufacturing and industrial automation. Brief Description of the Drawings
[0032] Figure 1 is a schematic diagram of a steam pipeline leakage monitoring system based on image recognition;
[0033] Figure 2 is a flowchart of a steam pipeline leakage monitoring system based on image recognition. Detailed Embodiments
[0034] The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
[0035] A steam pipeline leakage monitoring system combining image recognition technology and deep learning includes: a steam leakage area image acquisition unit, an environmental light adjustment unit, an image recognition and leakage judgment unit, and a data display and alarm feedback unit;
[0036] The steam leakage area image acquisition unit is used to capture the images of the area around the steam pipeline in real time, providing the original monitoring data for the system. This acquisition unit is equipped with a high-resolution camera and environmental light adjustment equipment, and is installed in key pipeline areas to ensure that the scene of steam leakage can be clearly recorded under different environmental and lighting conditions.
[0037] The image recognition and leakage judgment unit optimizes the collected image data through a series of preprocessing operations, which include image denoising, brightness adjustment, contrast adjustment, etc., to improve the recognizability of steam leakage features in the images. Next, this unit uses a deep learning model optimized for the features of steam leakage images to perform image analysis to judge the existence and severity of steam leakage. This model can effectively distinguish different scales and types of leakage events, and can accurately identify from minor steam leakage to large-scale steam leakage, ensuring high accuracy and reliability of leakage detection.
[0038] The data display and alarm feedback unit is responsible for timely feeding back the image recognition results to the operator, and when the system detects a leakage, it issues alarm signals of corresponding levels according to the severity of the leakage. This unit displays real-time monitoring images, leakage detection results and historical data analysis through a graphical user interface. At the same time, when a leakage is detected, it can quickly notify the maintenance team to take countermeasures by means of sound, light signals or sending electronic messages.
[0039] Through the collaborative work of the above components, the steam pipeline leakage monitoring system of the present invention can achieve real-time and accurate monitoring of steam pipeline leakage in the industrial field, significantly improve the safety management efficiency and response speed, and provide strong technical support for ensuring the safe and stable operation of industrial production.
[0040] A steam pipeline leakage monitoring method based on image recognition technology and deep learning includes the following steps:
[0041] The image acquisition unit is responsible for capturing real-time images of the steam pipeline area, mainly collecting camera screenshots around the pipeline to form a data set. This unit is equipped with high-definition imaging equipment to ensure that the steam leakage phenomenon can be clearly captured under various lighting conditions.
[0042] The image preprocessing module of the image recognition and leakage judgment unit is used to enhance the image quality and data screening, and screen out images with high quality and accurate perspectives from the collected pictures for data annotation. The annotation content includes key information such as the shape, density and distribution of steam. These are all important bases for the subsequent analysis of the deep learning model. To achieve this goal, the image processing module adopts the following technical means and related formulas:
[0043] 1. Contrast adjustment
[0044] The contrast adjustment is achieved through the following formula to enhance the distinction between the steam and the background in the image:
[0045]
[0046] In the formula:
[0047] I′ represents the image after contrast adjustment.
[0048] I is the original image.
[0049] represents the average brightness of the original image.
[0050] α is the contrast adjustment coefficient, which is used to control the increase or decrease of the contrast.
[0051] 2. Image sharpening
[0052] The Laplacian operator is used for image sharpening to enhance the clarity of the steam edges in the image:
[0053]
[0054] In the formula:
[0055] I′ is the sharpened image.
[0056] I represents the original image.
[0057] λ is the sharpening intensity coefficient.
[0058] represents the image obtained by applying the Laplacian operator, which is used to emphasize the high-frequency part of the image, i.e., the edges.
[0059] By applying the above technologies and formulas, the image preprocessing module not only improves the visual quality of the image, but also provides accurate basic data for the further identification and analysis of steam leaks. These preprocessing steps ensure that the subsequent deep learning model can effectively identify and evaluate the situation of steam leaks, thereby improving the performance and accuracy of the entire monitoring system.
[0060] The image recognition and leakage judgment unit adopts a novel deep learning model. This model is based on the Transformer architecture and further combines the optical flow method with multi-scale feature fusion technology, significantly improving the adaptability and accuracy of the algorithm for steam leakage recognition. The choice of the Transformer architecture stems from its excellent ability in processing sequence data, especially in capturing long-range dependencies. The basic building block of the Transformer model is the self-attention mechanism, which can effectively identify the complex relationships between various parts of the image, thereby extracting features crucial for steam leakage detection. The calculation formula of the self-attention mechanism is as follows:
[0061]
[0062] In the formula:
[0063] Q, K, and V represent Query, Key, and Value respectively, and they learn the features of steam leakage through different representations of the input image.
[0064] d k represents the dimension of the key vector, which is used to adjust the scaling problem introduced by the dot product calculation.
[0065] The application of the optical flow method further enhances the model's ability to capture dynamic features. The optical flow method provides important information about the speed and direction of object movement for the model by analyzing the movement of pixel points in consecutive image frames. In steam leakage detection, by calculating the movement of steam over time, the model can capture the dynamic change features of steam, which is crucial for identifying leakage events. The calculation formula of the optical flow vector is as follows:
[0066]
[0067] In the formula:
[0068] F(x, y, t) represents the optical flow vector at position F(x, y) and time t.
[0069] V x ,V y represent the velocity components in the horizontal and vertical directions respectively.
[0070] The introduction of the multi-scale feature fusion technology ensures that the model can process and identify steam leaks of different sizes and shapes. This technology extracts image features at different resolutions and then fuses these features together, enabling the model to comprehensively consider various forms and sizes of steam leaks. The formula for the fused features can be expressed as:
[0071]
[0072] Wherein:
[0073] F fusion represents the finally fused feature vector.
[0074] F i is the feature vector extracted at the i-th scale.
[0075] w i is the weight corresponding to the feature of the scale, obtained through learning.
[0076] n represents the total number of scales.
[0077] The model determines the existence and severity of leakage by analyzing the dynamic changes and morphological features of steam. The assessment of the leakage amount is based on the analysis of steam density and diffusion speed. The model can distinguish between minor, moderate, and severe leakage events and is updated in real time to adapt to environmental changes to ensure the accuracy of the monitoring results. The steam density in the model is an important static parameter for measuring the leakage amount, and its assessment is obtained based on the comprehensive calculation of the pixel density in the steam area of the image. The calculation process uses the following formula:
[0078]
[0079] Wherein:
[0080] D represents the steam density.
[0081] N is the total number of pixel points in the steam area.
[0082] p i is the density value of the i-th pixel point. By accurately calculating the density in the steam area of the image, the model can obtain important clues about the size of the leakage amount.
[0083] Furthermore, the diffusion speed of steam, as another key dynamic parameter, is evaluated based on the velocity vector calculated by the optical flow method. The calculation of the diffusion speed uses the following formula:
[0084]
[0085] Wherein:
[0086] S represents the steam diffusion speed.
[0087] VV x ,VV y respectively represent the change amounts of velocity in the horizontal and vertical directions. By accurately capturing and calculating the steam movement speed, the model further enhances the understanding and recognition ability of the dynamic changes of leakage.
[0088] Through in-depth analysis and real-time monitoring of these two key parameters, namely steam density and diffusion rate, the model can not only accurately evaluate the severity of the leakage under different environmental conditions, but also adjust the evaluation results in real time as the environment changes.
[0089] The display terminal integrates data display, warning issuance, and storage, achieving the secure storage of leakage data, threshold-triggered warnings, and intuitive display, ensuring efficient leakage monitoring and management. Through this method, the system can accurately locate leakage incidents and provide reliable information support for timely response measures.
[0090] The present invention provides a steam pipeline leakage monitoring system based on image recognition. Through advanced image processing technology and deep learning algorithms, the system realizes real-time and accurate monitoring and warning of leakage incidents in steam pipelines.
[0091] As Figure 1 shown, the system mainly includes four parts: a steam pipeline, an image acquisition unit for the steam pipeline area, an image recognition and leakage judgment unit, and a data display and alarm feedback unit.
[0092] The image acquisition unit for the steam leakage area is responsible for capturing images of the area around the steam pipeline in real time, providing the system with raw monitoring data. The acquisition unit is equipped with a high-resolution camera and environmental light adjustment equipment, and is installed in key pipeline areas to ensure the continuity and clarity of image acquisition. The image processing module adopts the following technical means and related formulas:
[0093] 1. Contrast adjustment
[0094] Contrast adjustment is achieved through the following formula to enhance the distinction between steam and the background in the image:
[0095] I′ = α(I - I) + I
[0096] Where:
[0097] I′ represents the image after contrast adjustment.
[0098] I is the original image.
[0099] I represents the average brightness of the original image.
[0100] α is the contrast adjustment coefficient, used to control the increase or decrease of contrast.
[0101] 2. Image sharpening
[0102] Image sharpening uses the Laplace operator to enhance the clarity of the steam edges in the image:
[0103]
[0104] In the formula:
[0105] I′ is the sharpened image.
[0106] I represents the original image.
[0107] λ is the sharpening intensity coefficient.
[0108] represents the image obtained by applying the Laplace operator, which is used to emphasize the high-frequency part of the image, i.e., the edges.
[0109] The image recognition and leakage judgment unit optimizes the collected image data through a series of preprocessing operations, including image denoising, brightness adjustment, and contrast adjustment, etc., to improve the recognizability of steam leakage features in the image. Next, the unit uses a deep learning model specifically for steam leakage image features to perform image analysis to judge the existence and severity of steam leakage. This model can effectively distinguish different scales and types of leakage events, ensuring high accuracy and reliability of leakage detection.
[0110] The image recognition and leakage judgment unit adopts an innovative deep learning model, which integrates advanced technologies based on the Transformer architecture and further integrates the optical flow method and multi-scale feature fusion technology. This unique combination significantly improves the adaptability and accuracy of the system in steam leakage recognition. The Transformer architecture is selected to process image sequence data and is particularly good at capturing long-range dependencies in the image sequence. The core of its self-attention mechanism lies in being able to accurately identify and analyze each detail of the image, revealing the complex interactions between key features in steam leakage detection. The self-attention mechanism is calculated by the following formula:
[0111]
[0112] In the formula:
[0113] Q, K, and V represent Query, Key, and Value respectively, and they learn the features of steam leakage through different representations of the input image.
[0114] d k represents the dimension of the key vector, which is used to adjust the scaling problem introduced by the dot product calculation.
[0115] The integration of the optical flow method enables the model to keenly capture the dynamic changes in steam movement. By analyzing the movement trajectories of pixel points in the image sequence, it provides important clues about the speed and direction of object movement. In the application of steam leak detection, the model can accurately capture the behavior pattern of steam when a leak occurs by calculating the dynamic changes of steam over time. The calculation of the optical flow vector can be achieved through the following formula:
[0116]
[0117] In the formula:
[0118] F(x, y, t) represents the optical flow vector at position F(x, y) and time t.
[0119] V x ,V y represent the velocity components in the horizontal and vertical directions respectively.
[0120] The multi-scale feature fusion technology can adapt to and identify steam leak phenomena of different sizes and shapes. This technology extracts image features at different resolution levels and integrates these features to obtain a more comprehensive understanding of the leak phenomenon. The calculation of the multi-scale fusion features is implemented through the following formula:
[0121]
[0122] In the formula:
[0123] F fusion represents the finally fused feature vector.
[0124] F i is the feature vector extracted at the i-th scale.
[0125] w i is the weight of the corresponding scale feature, obtained through learning.
[0126] n represents the total number of scales.
[0127] In the present invention, the deep learning model is carefully designed to capture and analyze the unique dynamic changes and morphological features of steam leaks, and then accurately judge the existence and severity of leaks. The model uniquely combines steam density and diffusion speed as key indicators to evaluate the levels of mild, moderate to severe for leak events, and has a real-time update function to adapt to environmental changes, thus ensuring a high degree of accuracy of the monitoring results.
[0128] Inside the model, steam density is regarded as the core static parameter for evaluating the leakage amount, and its evaluation is based on a comprehensive calculation method that obtains the result by analyzing the pixel density of the steam area in the image. Specifically, this calculation process is achieved through the following formula:
[0129]
[0130] In the formula:
[0131] D represents the steam density.
[0132] N is the total number of pixels in the steam area.
[0133] p i is the density value of the i-th pixel. By accurately calculating the density of the steam area in the image, the model can obtain important clues about the size of the leakage volume.
[0134] Furthermore, the diffusion speed of the steam is considered a key dynamic parameter for evaluating the dynamic changes of the leakage. Its calculation depends on the velocity vectors obtained by the optical flow method. The optical flow method provides important data on the diffusion speed and direction of the steam for the model by capturing the movement of pixels between consecutive image frames. The specific calculation method of the diffusion speed uses the following formula:
[0135]
[0136] In the formula:
[0137] S represents the steam diffusion speed.
[0138] VV x ,VV y represent the changes in velocity in the horizontal and vertical directions respectively. By accurately capturing and calculating the movement speed of the steam, the model further enhances its understanding and recognition ability of the dynamic changes of the leakage.
[0139] By deeply analyzing the two core parameters of steam density and diffusion speed and implementing real-time monitoring, this model can flexibly adapt to leakage detection in various environments and instantly adjust the evaluation results to match the environmental changes, thus ensuring the accuracy of the monitoring process.
[0140] The data display and alarm feedback unit is responsible for instantly feeding back the image analysis results to the operation team and issuing alarms of corresponding levels according to the severity of the leakage. This unit displays real-time monitoring images, leakage detection results, and historical data analysis through a graphical user interface, and can quickly notify the maintenance team to take countermeasures by means of sound, light signals, or sending electronic messages when a steam leakage in the steam pipeline is detected.
[0141] The steam pipeline leakage monitoring system of the present invention, by integrating advanced image recognition technology and deep learning algorithms, not only greatly improves the accuracy and efficiency of leakage detection, but also realizes the instant storage, analysis and intuitive display of leakage monitoring data through its data display and alarm feedback unit. When the system detects a leakage event, it can immediately warn the operator through various notification methods, thus greatly reducing the response time and improving the processing efficiency.
Claims
1. A steam pipeline leakage monitoring system based on image recognition, characterized in that: It includes a steam leakage area image acquisition unit, an ambient light adjustment unit, an image recognition and leakage judgment unit, and a data display and alarm feedback unit; the steam leakage area image acquisition unit captures images of the area around the steam pipe in real time, and the steam leakage area image acquisition unit is equipped with an ambient light adjustment unit, which is installed in key pipe areas; the image recognition and leakage judgment unit optimizes the collected image data through preprocessing operations, and uses a deep learning model optimized for steam leakage image features to perform image analysis to determine the existence and severity of steam leakage; the data display and alarm feedback unit promptly feeds back the image recognition results to the operator, and when a leakage is detected, sends an alarm signal of a corresponding level according to the severity of the leakage.
2. The steam pipeline leakage monitoring system based on image recognition according to claim 1 is characterized in that: The steam leakage area image acquisition unit is equipped with a high-resolution camera, which is installed in the key pipeline area to ensure that the on-site situation of steam leakage can be clearly recorded under different environmental and lighting conditions.
3. The steam pipeline leakage monitoring system based on image recognition according to claim 1 is characterized in that: The preprocessing operations of the image recognition and leakage judgment unit include image denoising, brightness adjustment and contrast adjustment.
4. The steam pipeline leakage monitoring system based on image recognition according to claim 3 is characterized in that: Images with high quality and accurate viewing angles are selected for data annotation, and the annotation content includes the shape, density and distribution information of the steam.
5. The steam pipeline leakage monitoring system based on image recognition according to claim 1 is characterized in that: The data display and alarm feedback unit displays real-time monitoring images, leak detection results and historical data analysis through a graphical user interface. At the same time, when a leak is detected, it can quickly notify the maintenance team to take countermeasures through sound, light signals or sending electronic messages.
6. The steam pipeline leakage monitoring system based on image recognition according to claim 3 is characterized in that: Contrast adjustment is achieved using the following formula: Where: I′ represents the image after contrast adjustment, I is the original image, It represents the average brightness of the original image, and α is the contrast adjustment coefficient, which is used to control the increase or decrease of the contrast.
7. The steam pipeline leakage monitoring system based on image recognition according to claim 1 is characterized in that: The preprocessing operation of the image recognition and leakage judgment unit includes image sharpening, which uses the Laplace operator: Where: I′ is the sharpened image, I represents the original image, λ is the sharpening strength coefficient, Represents the image obtained by applying the Laplacian operator.
8. The steam pipeline leakage monitoring system based on image recognition according to claim 1 is characterized in that: The deep learning model is based on the Transformer architecture, which further combines the optical flow method with multi-scale feature fusion technology.
9. The steam pipeline leakage monitoring system based on image recognition according to claim 8 is characterized in that: The basic building block of the Transformer model is the self-attention mechanism. The calculation formula of the self-attention mechanism is as follows: Where: Q, K, V represent query, key and value respectively, which learn the characteristics of steam leakage by different representations of the input image. k Represents the dimension of the health vector, which is used to adjust the scaling problem introduced by the dot product calculation; The optical flow method provides the model with important information about the speed and direction of the object's movement by analyzing the movement of pixels in consecutive image frames. In steam leak detection, by calculating the movement of steam over time, the model can capture the dynamic characteristics of steam. The optical flow vector calculation formula is as follows: Where: F(x,y,t) represents the optical flow vector at position F(x,y) and time t, V x ,V y Represent the velocity components in the horizontal and vertical directions respectively; Multi-scale feature fusion technology ensures that the model can process and identify steam leaks of different sizes and shapes. By extracting image features at different resolutions and then fusing these features together, the model can comprehensively consider the various shapes and sizes of steam leaks. The formula for fusion features is expressed as: Where: F fusion Represents the final fused feature vector, F i is the feature vector extracted at the i-th scale, w i is the weight of the corresponding scale feature, obtained through learning, and n represents the total number of scales.
10. The steam pipeline leakage monitoring system based on image recognition according to claim 9, characterized in that: The deep learning model determines the presence and severity of a leak by analyzing the dynamic changes and morphological characteristics of steam. The evaluation of the leak is based on the analysis of steam density and diffusion rate. The model can distinguish between minor, moderate and severe leaks and is updated in real time to adapt to environmental changes. The steam density in the model is an important static parameter for measuring the leak. Its evaluation is based on the comprehensive calculation of the pixel density of the steam area in the image. The calculation process uses the following formula: Where: D represents the steam density, N is the total number of pixels in the steam area, p i is the density value of the i-th pixel. By accurately calculating the density of the steam area in the image, the model can obtain important clues about the size of the leak. The diffusion rate of steam is another key dynamic parameter. Its evaluation depends on the velocity vector calculated by the optical flow method. The diffusion rate is calculated using the following formula: Where: S represents the vapor diffusion rate, VV x ,VV y Representing the velocity changes in the horizontal and vertical directions respectively, through in-depth analysis and real-time monitoring of two key parameters, steam density and diffusion velocity, the model can not only accurately assess the severity of the leak under different environmental conditions, but also adjust the assessment results in real time as the environment changes.