Gas plant station equipment safety risk monitoring method based on image processing

Abstract

The invention discloses a gas plant station equipment safety risk monitoring method based on image processing, and belongs to the technical field of gas plant station safety monitoring. The invention aims to solve the problem that an existing plant station equipment safety monitoring method is affected by environmental factors and is poor in data reliability. Comprising the steps of collecting and preprocessing a target surface source image and a reference surface source image to obtain a single-channel grey-scale map; calculating an average gray value and a gray standard deviation, calculating an illumination-dust coupling influence factor in combination with dust noise features, and further calculating a gray mean difference value of the target surface source single-channel gray-scale map; calculating a texture feature difference vector; calculating a morphological feature difference vector; calculating a comprehensive effective feature difference value; and comparing the comprehensive effective characteristic difference value with the comprehensive effective characteristic difference value and the dynamic safety threshold value of the strain key component determined according to the experiment, and determining the operation state of the strain key component. According to the invention, the accuracy and response efficiency of safety monitoring of the gas plant station equipment can be obviously improved.

Description

Technical Field

[0001] This invention relates to a method for monitoring safety risks of gas plant equipment based on image processing, belonging to the field of gas plant safety monitoring technology. Background Technology

[0002] As the central hub for storage, transmission, and pressure regulation, the safety of gas plants is directly linked to production and public safety. Current manual inspections and sensor-based methods are inefficient and highly subjective, failing to identify early defects such as cracks, aging seals, and damaged anti-corrosion layers online. Furthermore, when using sensors to monitor data from gas plants, environmental factors such as changes in lighting, dust accumulation, and fluctuations in temperature and humidity can easily distort the data, and there is a lack of effective error correction mechanisms.

[0003] The plant equipment is exposed to highly corrosive, extreme temperature and humidity, high-pressure pulsation, and dust erosion environments for extended periods, resulting in an average lifespan reduction of 40%, with critical components such as high-pressure pipeline welds experiencing a reduction of over 60%. Current risk control measures lack targeted monitoring of vulnerable components such as pipeline welds, flange joints, and shut-off valves, leading to insufficient monitoring focus and difficulty in accurately providing risk warnings.

[0004] Therefore, there is an urgent need for a safety monitoring technology for gas plant equipment that can accurately identify hidden risks in vulnerable components, eliminate environmental errors, and achieve automatic early warning, in order to address the shortcomings of existing technologies. Summary of the Invention

[0005] To address the problem that existing methods for monitoring the safety of gas plant equipment are susceptible to environmental factors and have poor data reliability, this invention provides a method for monitoring the safety risks of gas plant equipment based on image processing.

[0006] The present invention provides a method for monitoring safety risks of gas plant equipment based on image processing, comprising:

[0007] Based on historical failure data, key components for strain are identified; each key component for strain is used as a target surface source, and a corresponding reference surface source is selected; target surface source images and reference surface source images are acquired.

[0008] The target and reference surface source images are preprocessed to obtain single-channel grayscale images of the target and reference surfaces. The average grayscale value and standard deviation of each image are calculated. The illumination-dust coupling influence factor is calculated based on dust noise characteristics, leading to the calculation of the mean difference in grayscale values ​​of the target surface source's single-channel grayscale image. Simultaneously, the contrast, correlation, energy, and entropy values ​​of the target and reference surface source's single-channel grayscale images are calculated, and the texture feature difference vector of the target surface source's single-channel grayscale image is obtained. Finally, the gradient magnitude of the target surface source's single-channel grayscale image is calculated to determine the target contour, and the target contour area, length, and width are then calculated to obtain the morphological feature difference vector.

[0009] The comprehensive effective feature difference of the target surface source single-channel grayscale image is calculated based on the grayscale mean difference, texture feature difference vector, and morphological feature difference vector. The comprehensive effective feature difference of the target surface source single-channel grayscale image is compared with the comprehensive effective feature difference of the wear and tear key components determined by experiments and the dynamic safety threshold to determine the operating status of the wear and tear key components.

[0010] The method for monitoring safety risks of gas plant equipment based on image processing according to the present invention includes the following method for calculating the average gray value and gray standard deviation of the target area source single-channel grayscale image and the reference area source single-channel grayscale image:

[0011] ,

[0012] ,

[0013] In the formula The average grayscale value of the reference surface source single-channel grayscale image. The number of pixel rows in the reference surface source single-channel grayscale image. The number of pixel columns in the reference surface source single-channel grayscale image. The reference surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation is the reference surface source single-channel grayscale image.

[0014] ,

[0015] ,

[0016] In the formula The average grayscale value of the single-channel grayscale image of the target surface source. The number of pixel rows in the single-channel grayscale image of the target surface source. The number of pixel columns in the single-channel grayscale image of the target surface source. For the target surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation is the grayscale standard deviation of the target surface source single-channel grayscale image.

[0017] The method for safety risk monitoring of gas plant equipment based on image processing according to the present invention includes the following method for calculating the light-dust coupling influence factor:

[0018] ,

[0019] ,

[0020] In the formula Based on the baseline dust noise characteristics, for The pixel grayscale value after median filtering; The target dust noise characteristics, for The pixel grayscale value after median filtering;

[0021] Based on the grayscale ratio and noise energy ratio of the target surface source single-channel grayscale image and the reference surface source single-channel grayscale image, the coupling contribution is calculated using the entropy weight method. Then calculate the light-dust coupling effect factor. :

[0022] .

[0023] According to the image processing-based safety risk monitoring method for gas plant equipment of the present invention, the gray-level mean difference of the single-channel grayscale image of the target area source is expressed as: :

[0024] ,

[0025] In the formula For time, This is the net parameter for the grayscale mean.

[0026] .

[0027] According to the image processing-based gas plant equipment safety risk monitoring method of the present invention, the method for calculating the texture feature difference vector of the target surface source single-channel grayscale image is as follows:

[0028] ,

[0029] In the formula This represents the texture feature difference vector of the single-channel grayscale image of the target surface source. For target energy texture feature values Compared with the baseline energy texture feature value The difference, Target contrast texture feature value Contrast texture feature values ​​compared to the baseline The difference, For target-related texture feature values Texture feature values ​​correlated with the benchmark The difference, For target entropy texture feature value Compared with the baseline entropy value and texture feature value The difference.

[0030] According to the image processing-based gas plant equipment safety risk monitoring method of the present invention, the morphological feature difference vector is represented as... :

[0031] ,

[0032] In the formula The target contour area, The target contour length, The width of the target outline;

[0033] The target profile is determined based on the gradient magnitude.

[0034] According to the image processing-based safety risk monitoring method for gas plant equipment of the present invention, the comprehensive effective feature difference of the single-channel grayscale image of the target area source is expressed as: :

[0035] ,

[0036] In the formula This is the grayscale mean weighting coefficient. These are the texture feature weight coefficients. The morphological feature weighting coefficient. After smoothing , After smoothing , After smoothing .

[0037] According to the image processing-based gas plant equipment safety risk monitoring method of the present invention, the dynamic safety threshold is calculated as follows:

[0038] ,

[0039] In the formula For dynamic security threshold function, For dynamic security thresholds, For support vector machines, These are the training samples for the support vector machine in the experiment. For training sample labels, For symbolic functions, This represents the number of samples identified as support vectors in the training dataset. For Lagrange multipliers, for The tag, For radial basis kernel functions, For the first Samples of support vectors, For the current input sample, This is a bias term.

[0040] The method for determining the operating status of key components prone to wear and tear, based on image processing, according to the present invention, includes:

[0041] like The key components were determined to be operating normally after the wear and tear.

[0042] like The assessment determined that there was a safety risk in the critical components affected by the strain. The comprehensive effective characteristic difference of key components damaged by wear and tear;

[0043] like The failure of key components due to wear and tear was determined.

[0044] The image processing-based safety risk monitoring method for gas plant equipment according to the present invention includes preprocessing the target area source image and the reference area source image, comprising:

[0045] Median filtering, histogram equalization, cropping, and grayscale processing are performed on the target surface image and the reference surface image to obtain a single-channel grayscale image of the target surface image and a single-channel grayscale image of the reference surface image.

[0046] The key components affected by wear and tear include pipe welds, flange joints, tank welds, and anti-corrosion coatings.

[0047] The beneficial effects of this invention are as follows: The method of this invention is applicable to wear and tear monitoring and safety early warning of vulnerable metal parts. It achieves targeted screening and risk rating of vulnerable parts in gas plants through image recognition, and eliminates environmental errors such as lighting and dust by combining dual-camera synchronous acquisition and a comparative calibration mechanism. Accurate early warning is achieved based on image feature thresholds. A full-process monitoring system of image preprocessing + feature extraction + threshold determination is constructed, forming a closed-loop monitoring and control mechanism integrating latent defect identification, dynamic early warning, and shutdown management. This significantly improves the accuracy and response efficiency of safety monitoring of gas plant equipment. Attached Figure Description

[0048] Figure 1This is a flowchart of the image processing-based gas plant equipment safety risk monitoring method described in this invention;

[0049] Figure 2 This is a schematic diagram of the deployment of monitoring points for vulnerable components and control components. The vulnerable components are the key components prone to wear and tear. Detailed Implementation

[0050] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0051] Specific Implementation Method 1: Combination Figure 1 and Figure 2 As shown, this invention provides a method for monitoring safety risks of gas plant equipment based on image processing, including:

[0052] Based on historical failure data, key components for strain are identified; each key component for strain is used as a target surface source, and a corresponding reference surface source is selected; target surface source images and reference surface source images are acquired.

[0053] The target and reference surface source images are preprocessed to obtain single-channel grayscale images of the target and reference surfaces. The average grayscale value and standard deviation of each image are calculated. The illumination-dust coupling influence factor is calculated based on dust noise characteristics, leading to the calculation of the mean difference in grayscale values ​​of the target surface source's single-channel grayscale image. Simultaneously, the contrast, correlation, energy, and entropy values ​​of the target and reference surface source's single-channel grayscale images are calculated, and the texture feature difference vector of the target surface source's single-channel grayscale image is obtained. Finally, the gradient magnitude of the target surface source's single-channel grayscale image is calculated to determine the target contour, and the target contour area, length, and width are then calculated to obtain the morphological feature difference vector.

[0054] The comprehensive effective feature difference of the target surface source single-channel grayscale image is calculated based on the grayscale mean difference, texture feature difference vector, and morphological feature difference vector. The comprehensive effective feature difference of the target surface source single-channel grayscale image is compared with the comprehensive effective feature difference of the wear and tear key components determined by experiments and the dynamic safety threshold to determine the operating status of the wear and tear key components.

[0055] High-definition industrial cameras (target surface sources) are deployed in key monitoring areas of vulnerable metal components, with the lenses directly facing the wear-sensitive locations of the components. Near the vulnerable components, areas of the same material, with consistent surface conditions and no risk of wear are selected as control objects (reference surface sources), and industrial cameras of the same model are deployed there. Both cameras synchronously and continuously acquire image data.

[0056] Surface source compensation for illumination-dust adaptation. Experiments were conducted at different times and with different combinations of illumination intensity and dust concentration, recording the image recognition grayscale values ​​of the target surface source and the reference surface source without damage. By using the ratio of the mean grayscale values ​​of the monitored area and the control area, the illumination-dust coupling influence factor under the current environment was retrieved in real time; based on this factor, the grayscale values ​​of the monitored images were dynamically corrected so that the output results only reflect the true wear and tear evolution characteristics of the metal surface.

[0057] Furthermore, methods for calculating the average gray value and gray standard deviation of the target area source single-channel grayscale image and the reference area source single-channel grayscale image include:

[0058] ,

[0059] ,

[0060] In the formula The average grayscale value of the reference surface source single-channel grayscale image. The number of pixel rows in the reference surface source single-channel grayscale image. The number of pixel columns in the reference surface source single-channel grayscale image. The reference surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation is the reference surface source single-channel grayscale image. The pixel size of the reference surface source image;

[0061] ,

[0062] ,

[0063] In the formula The average grayscale value of the single-channel grayscale image of the target surface source. The number of pixel rows in the single-channel grayscale image of the target surface source. The number of pixel columns in the single-channel grayscale image of the target surface source. For the target surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation of the target surface source single-channel grayscale image. The pixel size of the target surface image.

[0064] The baseline dust noise feature is calculated to reflect the degree of interference of dust in the environment on the image.

[0065] Methods for calculating the light-dust coupling effect factor include:

[0066] ,

[0067] ,

[0068] In the formula Based on the baseline dust noise characteristics, for The pixel grayscale value after median filtering; Target dust noise characteristics, for The pixel grayscale value after median filtering;

[0069] Based on the grayscale ratio and noise energy ratio of the target surface source single-channel grayscale image and the reference surface source single-channel grayscale image, the coupling contribution is calculated using the entropy weight method. Then calculate the light-dust coupling effect factor. :

[0070] .

[0071] The grayscale ratio reflects the relative intensity of light interference.

[0072] This represents the noise energy ratio, reflecting the relative intensity of dust interference.

[0073] Calculate the grayscale ratio, noise energy ratio, and coupling contribution between the target surface source and the reference surface source. Then, the influence factor of light-dust coupling was determined. .

[0074] Coupling contribution Determined by the entropy weight method, the data is then standardized, shifted, and calculated sequentially. (Indicator Contribution) (Entropy) and W (weight). The specific formula is as follows:

[0075] Data standardization:

[0076] Grayscale ratio data standardization: ;

[0077] Noise energy ratio data standardization: ;

[0078] In the above formula, , Let be the initial values ​​for the grayscale ratio and noise energy ratio of the q-th sample. , This represents the standardized values ​​of the initial grayscale ratio and noise energy ratio for the q-th sample. and These represent the maximum values ​​of the grayscale ratio and the noise energy ratio, respectively. , These represent the minimum values ​​of the grayscale ratio and the noise energy ratio, respectively.

[0079] Contribution:

[0080] Grayscale ratio (illumination) contribution: ;

[0081] Noise energy contribution (dust): ;

[0082] In the formula Let q be the standardized grayscale ratio of the q-th sample. Let q be the standardized noise energy proportion of the q-th sample. , , respectively, represent the contribution of the grayscale ratio (light) and noise energy ratio (dust) of the q-th sample, where n is the number of sample data.

[0083] Entropy value:

[0084] Grayscale ratio (illumination) entropy value: ;

[0085] Noise energy ratio (dust) entropy value: ;

[0086] , These are the entropy values ​​of the grayscale ratio (light) and noise energy ratio (dust) of the q-th sample, respectively;

[0087] Weight:

[0088] Grayscale ratio (illumination) entropy value: ;

[0089] Noise energy ratio (dust) entropy value: ,

[0090] In the formula , These are the weights of the grayscale ratio (light) and noise energy ratio (dust) of the q-th sample, respectively.

[0091] Coupling contribution: ;

[0092] Coupling contribution The value was calculated using the entropy weight method based on historical data, and the range is 0.6-0.8.

[0093] The difference in texture features between the target surface source and the reference surface source is calculated to offset the interference of lighting and dust, so as to obtain the net feature parameters that only reflect metal wear.

[0094] The difference in the mean grayscale value of the single-channel grayscale image of the target surface source is expressed as: :

[0095] ,

[0096] In the formula For time, This is the net parameter for the grayscale mean.

[0097] .

[0098] Furthermore, the method for calculating the texture feature difference vector of the single-channel grayscale image of the target surface source is as follows:

[0099] ,

[0100] In the formula The texture feature difference vector of the single-channel grayscale image of the target surface source is used for multi-dimensional threshold determination; For target energy texture feature values Compared with the baseline energy texture feature value The difference, For target contrast texture feature values Contrast texture feature values ​​compared to the baseline The difference, For target-related texture feature values Texture feature values ​​correlated with the benchmark The difference, For target entropy texture feature value Compared with the baseline entropy value and texture feature value The difference.

[0101] The gray-level co-occurrence matrix algorithm is used to calculate contrast, correlation, energy, and entropy.

[0102] by Represents the elements of the gray-level co-occurrence matrix; grayscale value of image pixels ; This represents the pixel spacing, with a value of 1. It is the direction angle; The total number of pixel pairs that meet the conditions; For image dimensions. Calculation:

[0103] Contrast Ratio: ;

[0104] Correlation: ;

[0105] energy: ;

[0106] Entropy value: ;

[0107] in, , is the grayscale mean;

[0108] , where is the standard deviation of grayscale.

[0109] The contour area A, length L, and width W of cracks, dents, and corrosion points on the surface of the component are extracted using an edge detection algorithm.

[0110] ,

[0111] ,

[0112] ,

[0113] in, This represents the gradient response value in the horizontal direction. This represents the gradient response value in the vertical direction. This represents the gradient magnitude.

[0114] Horizontal Sobel operator;

[0115] Vertical Sobel operator;

[0116] Outline area: ;

[0117] Outline length: ;

[0118] Outline width: ;

[0119] in: Let Q be the coordinates of the k-th vertex of the contour, and let Q be the total number of vertices of the contour.

[0120] The morphological feature difference vector is represented as :

[0121] ,

[0122] In the formula The target contour area reflects the increase in the area of ​​cracks or corrosion; The target profile length reflects the increment in crack propagation length. The target profile width reflects the increment in the width of the crack or corrosion.

[0123] The target profile is determined based on the gradient magnitude.

[0124] The same moving average operation is performed on the texture feature difference and morphological feature difference to obtain the smoothed comprehensive effective feature difference; the comprehensive effective feature difference of the single-channel grayscale image of the target surface source is expressed as... :

[0125] ,

[0126] In the formula This is the grayscale mean weighting coefficient. These are the texture feature weight coefficients. The morphological feature weighting coefficient. After smoothing , After smoothing , After smoothing . , , Calculated using the entropy weight method.

[0127] The effective feature differences are smoothed, and a moving average algorithm is used to reduce random errors.

[0128] Gray-level mean difference and moving average correction formula:

[0129] ,

[0130] In the formula To adjust the sliding window size; for Average grayscale value at any given time.

[0131] ;

[0132] ;

[0133] Let be the texture feature difference vector at time I; Let be the morphological feature difference vector at time I.

[0134] By collecting historical fault data from gas plants and conducting artificial damage simulations on vulnerable components, the smoothed comprehensive effective feature difference corresponding to component failure was collected. This serves as the initial safety threshold. The upper limit of the maximum feature difference for vulnerable components that are not failed and can operate safely is collected, and the corresponding smoothed comprehensive effective feature difference is calculated. This serves as the baseline judgment threshold. The failure threshold is compared with the baseline judgment threshold to obtain the threshold safety factor. This reflects the redundancy of the safety threshold and is a fixed constant that does not change with operating conditions. Based on the calculation results and in conjunction with design specifications and the safety redundancy requirements of gas plants, the threshold safety factor ranges from 1.2 to 1.5.

[0135] Threshold safety factor: , The redundancy coefficient is a specific standard for reserving a safety margin in the quantification specification, with different vulnerable components corresponding to a single redundancy coefficient.

[0136] Calculate the baseline safety threshold As a fundamental indicator for risk assessment, the threshold is calibrated using a support vector machine optimization model to ultimately obtain a dynamic safety threshold. .

[0137] Baseline safety threshold: ;

[0138] The dynamic security threshold is calculated as follows:

[0139] ,

[0140] In the formula For dynamic security threshold function, For dynamic security thresholds, For support vector machines, These are the training samples for the support vector machine in the experiment. For training sample labels, failure = 1, normal = 0; For symbolic functions, This represents the number of samples identified as support vectors in the training dataset. For Lagrange multipliers, for The tag, For radial basis kernel functions, For the first Samples of support vectors, For the current input sample, This is a bias term.

[0141] Furthermore, methods for determining the operating status of critical components prone to wear and tear include:

[0142] The effective feature difference calculated in real time is compared with the dynamic security threshold:

[0143] like The key components damaged by the strain were determined to be operating normally, and monitoring will continue.

[0144] like If a safety risk is detected in a critical component due to overuse injury, it can trigger an audible and visual alarm and data storage. The comprehensive effective characteristic difference of key components damaged by wear and tear;

[0145] like The system determines that the critical components have failed due to wear and tear, triggering the highest level of warning and emergency shutdown command.

[0146] The early warning mechanism includes: sending audible and visual alarm signals to the gas plant monitoring center, displaying the location of the risky component, the risk type, and the characteristic difference; sending shutdown commands to the equipment control system to cut off the power supply or gas transmission channel of the equipment where the risky component is located; and automatically storing the current image data and characteristic parameters as a basis for fault analysis.

[0147] As an example, preprocessing of the target surface image and the reference surface image includes:

[0148] Median filtering, histogram equalization, cropping, and grayscale processing are performed on the target surface image and the reference surface image to obtain a single-channel grayscale image of the target surface image and a single-channel grayscale image of the reference surface image.

[0149] During preprocessing, a median filtering algorithm is used to eliminate dust noise in the image, with the filter window size set to 3×3. To eliminate brightness differences caused by changes in illumination, a histogram equalization algorithm is used to adjust the image brightness distribution. To remove background interference, the image is cropped, retaining only the effective image portions of the monitoring and control areas. The preprocessed image is then subjected to grayscale processing, converting the color image into a single-channel grayscale image with a grayscale range of 0-255.

[0150] Based on historical fault data, maintenance records, and equipment drawings of gas plant equipment, key components prone to wear and tear can be identified; these key components include pipe welds, flange joints, storage tank welds, and anti-corrosion coatings.

[0151] The monitoring system for implementing the method of the present invention may include:

[0152] The vulnerable component analysis module uses machine learning algorithms to screen vulnerable components and determine monitoring priorities.

[0153] The image acquisition module deploys industrial cameras at monitoring points for vulnerable components and control objects, simultaneously acquiring image data of both.

[0154] The image preprocessing module performs noise reduction, illumination correction, cropping, grayscale processing, and feature parameter extraction on the acquired images.

[0155] The environmental correction and feature extraction module cancels out the interference of light and dust, and outputs only the net feature parameters that reflect metal wear. At the same time, it extracts the morphological features of metal parts such as cracks and corrosion based on the edge detection algorithm.

[0156] The threshold determination module compares the valid feature data with the safety threshold and outputs the determination result.

[0157] The early warning control module controls the system based on the judgment result. When the threshold range is exceeded, it triggers an audible and visual alarm and sends a shutdown command to the equipment control system.

[0158] This implementation first filters vulnerable components and determines monitoring priorities by analyzing historical fault data. Second, it simultaneously acquires images of vulnerable components and undamaged control objects of the same material. After preprocessing, it calculates baseline grayscale values ​​and illumination compensation coefficients based on the principle of area source compensation. Feature differences are adaptively used to offset illumination and dust interference, resulting in net feature parameters that only reflect metal wear. Third, it extracts texture and morphological features using grayscale co-occurrence matrix and edge detection algorithms. After smoothing using a moving average algorithm, it calculates a comprehensive effective feature difference, which is then combined with a dynamic safety threshold to complete risk assessment. The entire process includes vulnerable component analysis, image acquisition, image preprocessing, environmental correction and feature extraction, threshold determination, and early warning control. It enables accurate identification and rapid early warning of hidden wear on vulnerable components, solving the problems of poor targeting, significant environmental error impact, and delayed early warning in existing monitoring technologies. This significantly improves the reliability and response efficiency of safety monitoring for gas plant equipment.

[0159] While the invention has been described herein with reference to specific embodiments, it should be understood that these embodiments are merely examples of the principles and applications of the invention. Therefore, it should be understood that many modifications can be made to the exemplary embodiments, and other arrangements can be designed without departing from the spirit and scope of the invention as defined by the appended claims. It should be understood that different dependent claims and features described herein can be combined in ways different from those described in the original claims. It is also understood that features described in conjunction with individual embodiments can be used in other described embodiments.

Claims

1. A method for monitoring safety risks of gas plant equipment based on image processing, characterized in that... include, Based on historical failure data, key components prone to wear and tear were identified. Each critical component damaged by overuse is used as the target surface source, and a corresponding reference surface source is selected; images of the target surface source and the reference surface source are acquired; Preprocessing is performed on the target surface image and the reference surface image to obtain a single-channel grayscale image of the target surface image and a single-channel grayscale image of the reference surface image; The average gray value and gray standard deviation of the target surface source single-channel grayscale image and the reference surface source single-channel grayscale image are calculated respectively. The illumination-dust coupling influence factor is calculated in combination with dust noise characteristics, and then the gray mean difference of the target surface source single-channel grayscale image is calculated. At the same time, the contrast, correlation, energy and entropy values ​​of the target surface source single-channel grayscale image and the reference surface source single-channel grayscale image are calculated, and the texture feature difference vector of the target surface source single-channel grayscale image is calculated. The gradient magnitude of the single-channel grayscale image of the target surface source is calculated to determine the target contour. Then, the target contour area, target contour length and target contour width are calculated to obtain the morphological feature difference vector. The comprehensive effective feature difference of the target surface source single-channel grayscale image is calculated based on the grayscale mean difference, texture feature difference vector, and morphological feature difference vector. The comprehensive effective feature difference of the target surface source single-channel grayscale image is compared with the comprehensive effective feature difference of the wear and tear key components determined by experiments and the dynamic safety threshold to determine the operating status of the wear and tear key components.

2. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 1, characterized in that, Methods for calculating the average gray value and gray standard deviation of the target area source single-channel grayscale image and the reference area source single-channel grayscale image include: ， ， In the formula The average grayscale value of the reference surface source single-channel grayscale image. The number of pixel rows in the reference surface source single-channel grayscale image. The number of pixel columns in the reference surface source single-channel grayscale image. The reference surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation is the reference surface source single-channel grayscale image. ， ， In the formula The average grayscale value of the single-channel grayscale image of the target surface source. The number of pixel rows in the single-channel grayscale image of the target surface source. The number of pixel columns in the single-channel grayscale image of the target surface source. For the target surface source single-channel grayscale image line, number The pixel grayscale value of the column; The grayscale standard deviation is the grayscale standard deviation of the target surface source single-channel grayscale image.

3. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 2, characterized in that, Methods for calculating the light-dust coupling effect factor include: ， ， In the formula Based on the baseline dust noise characteristics, for The pixel grayscale value after median filtering; The target dust noise characteristics, for The pixel grayscale value after median filtering; Based on the grayscale ratio and noise energy ratio of the target surface source single-channel grayscale image and the reference surface source single-channel grayscale image, the coupling contribution is calculated using the entropy weight method. Then calculate the light-dust coupling effect factor. : 。 4. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 3, characterized in that, The difference in the mean grayscale value of the single-channel grayscale image of the target surface source is expressed as: : ， In the formula For time, This is the net parameter for the grayscale mean. 。 5. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 4, characterized in that, The method for calculating the texture feature difference vector of the single-channel grayscale image of the target surface source is as follows: ， In the formula This represents the texture feature difference vector of the single-channel grayscale image of the target surface source. For target energy texture feature values Compared with the baseline energy texture feature value The difference, Target contrast texture feature value Contrast texture feature values ​​compared to the baseline The difference, For target-related texture feature values Texture feature values ​​correlated with the benchmark The difference, For target entropy texture feature value Compared with the baseline entropy value and texture feature value The difference.

6. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 5, characterized in that, The morphological feature difference vector is represented as : ， In the formula The target contour area, The target contour length, The width of the target outline; The target profile is determined based on the gradient magnitude.

7. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 6, characterized in that, The comprehensive effective feature difference of the single-channel grayscale image of the target surface source is expressed as: : ， In the formula This is the grayscale mean weighting coefficient. These are the texture feature weight coefficients. The morphological feature weighting coefficient. After smoothing , After smoothing , After smoothing .

8. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 7, characterized in that, The dynamic security threshold is calculated as follows: ， In the formula For dynamic security threshold function, For dynamic security thresholds, For support vector machines, These are the training samples for the support vector machine in the experiment. For training sample labels, For symbolic functions, This represents the number of samples identified as support vectors in the training dataset. For Lagrange multipliers, for The tag, For radial basis kernel functions, For the first Samples of support vectors, For the current input sample, This is a bias term.

9. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 8, characterized in that, Methods for determining the operating status of critical components prone to wear and tear include: like The key components were determined to be operating normally after the wear and tear. like The assessment determined that there was a safety risk in the critical components affected by the strain. The comprehensive effective characteristic difference of key components damaged by wear and tear; like The failure of key components due to wear and tear was determined.

10. The method for monitoring safety risks of gas plant equipment based on image processing according to claim 1, characterized in that, Preprocessing of the target surface image and the reference surface image includes: Median filtering, histogram equalization, cropping, and grayscale processing are performed on the target surface image and the reference surface image to obtain a single-channel grayscale image of the target surface image and a single-channel grayscale image of the reference surface image. The key components affected by wear and tear include pipe welds, flange joints, tank welds, and anti-corrosion coatings.

Images