Gas station safety risk assessment method and system

Abstract

The present invention relates to the technical field of gas station safety risk assessment, and in particular to a gas station safety risk assessment method and system. The present invention deploys a multi-band infrared imaging array to acquire a full-area infrared image of the gas station in real time, combines dark current compensation, gain correction and atmospheric transmittance calibration, and significantly improves the accuracy of oil and gas concentration detection. Based on the dynamic coordinate alignment technology of the homography matrix, it utilizes the optimization of the reprojection error of the landmark points and the optical flow tracking of the feature points to solve the coordinate drift problem caused by thermal deformation, ensuring that the spatial positioning error is continuously less than 2 pixels. Combined with the time series concentration field construction and the hybrid prediction model, the present invention realizes the accurate deduction of the oil and gas concentration field at any moment in the future, realizes real-time monitoring of the gas station, and eliminates the explosion accidents caused by detection lag from the source. The present invention solves the technical problem that the current method of monitoring the oil and gas concentration of gas stations is difficult to detect potential safety risks in a timely manner.

Description

Technical Field

[0001] The present invention relates to the technical field of gas station safety risk assessment, and in particular to a gas station safety risk assessment method and system. Background Art

[0002] During the operation of gas stations, according to the provisions of the "Gas Station Air Pollutant Emission Standards", it is necessary to ensure that the oil and gas concentrations in various areas of the station meet safety requirements to prevent combustion and explosion accidents. At present, the detection of oil and gas concentrations at gas stations mainly relies on staff using portable LEL (lower explosion limit) detectors for periodic manual inspections and fixed-point monitoring through sensors.

[0003] However, manual inspections cannot achieve uninterrupted, full-coverage real-time monitoring of the entire gas station. There is a time interval between two inspections. Sudden oil and gas leaks that occur during this period (such as oil gun seal failure, pipeline rupture, or abnormal discharge from the breathing valve) cannot be captured in time, forming a blind spot for safety hazards. In addition, manual inspections can only reflect the local concentration conditions at the moment of the inspection. Although fixed-point monitoring methods can provide continuous monitoring, it is difficult to achieve full coverage of gas stations and predict oil and gas diffusion trends and future concentration field evolution. In addition, the gas station environment is complex and is affected by wind direction, wind speed, temperature, and operational activities. The oil and gas concentration changes rapidly, making it difficult to dynamically predict which areas will reach or exceed the dangerous threshold in the future (such as the next one or five minutes). This makes preventive measures delayed and difficult to issue early warnings and intervene in the embryonic stage of accidents. By the time the concentration exceeds the standard, the danger may have accumulated to a critical point, or the leakage source may have been releasing for a period of time, greatly increasing the risk of combustion and explosion accidents. Summary of the Invention

[0004] In response to the deficiencies of the existing technology, the present invention provides a gas station safety risk assessment method and system, which solves the technical problem that the current method of monitoring oil and gas concentration at gas stations is difficult to detect potential safety risks in a timely manner.

[0005] To solve the above technical problems, the present invention provides the following technical solutions: a gas station safety risk assessment method, the specific steps of the safety risk assessment method are:

[0006] S1. Deploy an infrared imaging array containing several bands in the gas station to acquire infrared images in real time, and preprocess and align the infrared images to obtain corrected infrared images;

[0007] S2. Calculate the oil and gas concentration at each location of the gas station in real time based on the corrected infrared image, and establish an oil and gas concentration field including the oil and gas concentration at each location of the gas station in real time;

[0008] S3. predicting the oil and gas concentration field at any time in the future based on the oil and gas concentration field established in real time;

[0009] S4. Selecting a potential risk point in the global coordinates where the oil and gas concentration exceeds the oil and gas concentration threshold based on the oil and gas concentration field at any future moment;

[0010] S5. Assess the safety risk of the gas station based on the area occupied by potential hazards.

[0011] Preferably, in step S1, the specific steps of preprocessing and coordinate aligning the infrared image to obtain the corrected infrared image are as follows:

[0012] S11. Correct the infrared image at each moment to obtain a corrected infrared image. The expression of the corrected infrared image is:

[0013]

[0014] In the above formula, express The radiation intensity at coordinate (x, y) of the corrected infrared image of the band at time t, It represents the radiation intensity of the acquired infrared image at the coordinate (x, y) at time t. represents the dark current matrix at the pixel level of the infrared detector that collects the infrared image, represents the gain correction coefficient at the pixel level, represents the atmospheric transmittance compensation factor at time t;

[0015] S12. Establishing a global coordinate system within the gas station and presetting a number of landmark points within the gas station;

[0016] S13. Establish a projection relationship between the corrected infrared image and the global coordinate system to complete coordinate alignment. The projection relationship satisfies:

[0017]

[0018] In the above formula, To correct the coordinates of any point in the infrared image in the global coordinate system, represents the image coordinates of any point in the corrected infrared image, represents the homography matrix;

[0019] S14. Setting the objective function for calculating the homography matrix with the goal of minimizing the reprojection error of all landmark points to obtain the homography matrix. The expression of the objective function is:

[0020]

[0021] in,

[0022]

[0023] In the above formula, M represents the number of landmark points, is the measured coordinate of the i-th marker point in the global coordinate system, represents the image coordinates of the i-th marker point in the corrected infrared image;

[0024] S15, calculating the reprojection error of each marker point in the infrared image according to the obtained homography matrix;

[0025] S16, determining whether the reprojection errors of each marker point are less than the reprojection error threshold;

[0026] If yes, then output the homography matrix and go to step S17;

[0027] If not, then remove the marker points whose reprojection error is greater than or equal to the reprojection error threshold, and then return to step S14;

[0028] S17. Extract feature points from the standard infrared image, and dynamically update the homography matrix according to the global coordinates of each feature point.

[0029] Preferably, in step S17, the steps for calculating the homography matrix are as follows:

[0030] S171, marking feature points in the corrected infrared image whose Hessian determinant values ​​are greater than a determinant threshold as standard feature points to obtain a feature point set;

[0031] S172. Calculate the displacement of each standard feature point in the feature point set using the optical flow method. The calculation formula is:

[0032]

[0033] In the above formula, Indicates the displacement of the standard feature point in the x direction, Indicates the displacement of the standard feature point in the y direction, represents the spatial gradient of the corrected infrared image in the x direction, Indicates the spatial gradient of the corrected infrared image in the y direction, Represents the gradient of the corrected infrared image in the time dimension, Indicates a non-zero minimum value;

[0034] S173, calculating the average displacement of each standard feature point;

[0035] S174, determining whether the average displacement is less than the average position threshold;

[0036] If yes, then output the homography matrix and go to step S2;

[0037] If not, proceed to step S175;

[0038] S175. Calculate the incremental homography matrix based on the displacement of each standard feature point. The calculation formula is:

[0039]

[0040] In the above formula, Represents the incremental homography matrix, that is, the function When taking the minimum value The value of and Respectively represent the x-coordinate and y-coordinate of feature point j in the current frame, and Respectively represent the x-coordinate and y-coordinate of feature point j in the previous frame;

[0041] S176. Calculate an updated homography matrix based on the incremental homography matrix. The expression of the updated homography matrix is:

[0042]

[0043]

[0044] Where S(T) represents the temperature compensation matrix, s represents the thermal expansion scaling factor, Indicates the thermal expansion coefficient of the object material, T indicates the current ambient temperature, Indicates the reference temperature during calibration.

[0045] Preferably, in step S2, the following steps are specifically included:

[0046] S21. For each feature point in the corrected infrared image, calculate the differential absorption ratio:

[0047]

[0048] In the above formula, represents the differential absorption ratio at the coordinate (x, y) in the corrected infrared image at time t, represents the radiation intensity of the reference band, Indicates the radiation intensity of the oil and gas absorption band;

[0049] S22. Invert the oil and gas concentration field of the gas station based on the differential absorption ratio of each feature point to obtain the oil and gas concentration at any position in the global coordinates. The calculation formula for the oil and gas concentration at any position in the global coordinates is:

[0050]

[0051] In the above formula, Represents the coordinates in the global coordinate system The oil and gas concentration at time t, represents the calibration absorption coefficient, L represents the optical path of binocular vision measurement, Represents the long-wave infrared background compensation term.

[0052] Preferably, in step S3, the following steps are specifically included:

[0053] S31. For each position in the global coordinates, extract the concentration time series data of the forward time window including the current moment, and the expression is as follows:

[0054]

[0055] In the above formula, it represents Represents the coordinates in the global coordinate system The concentration time series data of the forward time window at the current moment is included. Indicates the coordinates in the global coordinates at time t The oil and gas concentration at Indicates the current moment, represents the time span of the forward time window;

[0056] S32. Calculate the concentration gradient vector of each position at each moment in the concentration time series data of each position in the global coordinates in sequence. The expression of the concentration gradient vector is:

[0057]

[0058] In the above formula, Indicates the coordinates of the global coordinate system at time t The concentration gradient vector at 、 and Indicates that at time t, the global coordinate The concentration in the x direction, y direction and relative to the previous moment;

[0059] S33, obtaining wind direction and speed data at various locations within the gas station;

[0060] S34, identifying characteristic points where the oil and gas concentration values ​​are abnormal, and correcting the oil and gas concentration values ​​of the abnormal characteristic points by interpolation;

[0061] S35. Predict the oil and gas concentration and oil and gas concentration gradient at each location in the gas station at any time based on the wind direction and speed data. The prediction is publicly displayed as:

[0062]

[0063] In the above formula, Indicates The gas station concentration field at the moment, F represents the hybrid prediction operator integrating LSTM and CFD, represents the turbulent diffusion coefficient, C represents the concentration time series data, Representation and global coordinates Corresponding wind speed vector data.

[0064] Preferably, in step S34, the following steps are specifically included:

[0065] S341, sequentially establishing a window with any pixel point in the corrected infrared image as the center, and calculating the average oil and gas concentration of all pixels in the window excluding the center pixel point to obtain the average oil and gas concentration in the window;

[0066] S342. Calculate the absolute value of the difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window to obtain the oil and gas concentration difference:

[0067]

[0068] In the above formula, Indicates the oil and gas concentration difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration of the corresponding window. and Respectively represent the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window;

[0069] S343, setting the oil and gas concentration difference threshold corresponding to the central pixel point according to the wind direction and speed data and the concentration gradient vector;

[0070] S344, determining whether the oil-gas concentration difference corresponding to the central pixel point is less than the oil-gas concentration difference threshold;

[0071] If yes, mark the central pixel as a normal feature point;

[0072] If not, the central pixel is marked as an abnormal feature point;

[0073] S345. Correct the oil and gas concentration value of the abnormal feature point by interpolation.

[0074] Preferably, step S343 specifically includes the following steps:

[0075] S3431, obtaining wind speed and direction data corresponding to the central pixel point, and establishing a wind direction line projected onto the plane where the infrared image is located with the central pixel point as the starting point;

[0076] S3432. Obtain the pixel point where the wind direction line intersects the window boundary corresponding to the center pixel point, and mark it as the end pixel point. The end pixel point is located within the window boundary corresponding to the center pixel point.

[0077] S3433. Calculate the actual distance between the center pixel and the corresponding end pixel using the following formula:

[0078]

[0079] In the above formula, Indicates the actual distance between the center pixel and the corresponding end pixel. Indicates the pixel distance between the center pixel and the corresponding end pixel. Indicates the ratio coefficient between actual distance and pixel distance;

[0080] S3434. Calculate the diffusion velocity of oil and gas molecules based on wind direction and speed data. The calculation formula is:

[0081]

[0082] In the above formula, Indicates the diffusion speed of oil and gas molecules, represents the Boltzmann constant, represents the ambient temperature, m represents the average mass of oil and gas molecules, represents the gradient diffusion coefficient, Indicates the oil and gas concentration value of the central pixel point, Indicates the minimum value of oil and gas concentration in the pixel points other than the central pixel point in the window corresponding to the central pixel point. represents the turbulence coupling factor, Represents the wind speed vector at the center pixel point, Indicates the angle between the actual wind direction and the wind direction line. Represents the concentration gradient vector of the central pixel;

[0083] S3435. Calculate the diffusion time of oil and gas molecules from the center pixel to the corresponding end pixel. The calculation formula is:

[0084]

[0085] In the above formula, Indicates the diffusion time of oil and gas molecules from the central pixel to the corresponding end pixel. Indicates the actual distance between the center pixel and the corresponding end pixel;

[0086] S3436. Calculate the oil-gas concentration difference threshold based on Fick's second law.

[0087] Preferably, in step S3436, the calculation formula of the oil-gas concentration difference threshold is:

[0088]

[0089] In the above formula, Indicates the oil-gas concentration difference threshold, Indicates the acquisition time interval of infrared images, represents the turbulent diffusion coefficient, Represents the safety margin factor.

[0090] Preferably, in step S4, the following steps are specifically included:

[0091] S41, setting a first oil and gas concentration threshold, and marking pixels in the corrected infrared image whose oil and gas concentration values ​​are greater than the first oil and gas concentration threshold as suspicious pixels based on the oil and gas concentration field of the gas station;

[0092] S42: Obtain the global coordinates corresponding to the suspicious pixel points, and identify the global coordinates of the fuel pumps in the gas station that are refueling or have refueled but have not exceeded the evaporation time threshold, and mark the fuel pumps as working fuel pumps;

[0093] S43, calculating the Euclidean distance between the suspicious pixel point and the working refueling gun in the global coordinates;

[0094] S44, selecting suspicious pixel points whose Euclidean distance from the working refueling gun in the global coordinates is greater than the volatilization radius and marking them as potential risk points, and marking the remaining suspicious pixel points as pending pixel points;

[0095] S45 , setting a second oil and gas concentration threshold that is greater than the first oil and gas concentration threshold, and marking undetermined pixel points whose oil and gas concentration values ​​are greater than the second oil and gas concentration threshold as potential risk points.

[0096] The present invention also provides a gas station safety risk assessment system, comprising a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the gas station safety risk assessment method is implemented.

[0097] By means of the above technical solution, the present invention provides a gas station safety risk assessment method and system, which has at least the following beneficial effects:

[0098] 1. This invention deploys a multi-band infrared imaging array to acquire infrared images of the entire gas station in real time. Combined with dark current compensation, gain correction, and atmospheric transmittance calibration, it significantly improves the accuracy of oil and gas concentration detection. Based on the dynamic coordinate alignment technology of the homography matrix, it utilizes landmark point reprojection error optimization and feature point optical flow tracking to solve the coordinate drift problem caused by thermal deformation, ensuring that the spatial positioning error remains below 2 pixels. Combined with the construction of a time-series concentration field and a hybrid prediction model, it achieves accurate deduction of the oil and gas concentration field at any moment in the future, enabling real-time monitoring of gas stations and eliminating explosion accidents caused by delayed detection at the source.

[0099] 2. The differential absorption ratio-based inversion algorithm of the present invention effectively eliminates interference from ambient thermal radiation and atmospheric scattering by separating the radiation intensity of the oil and gas absorption band from that of the reference band. It introduces binocular visual optical path measurement and long-wave infrared background compensation terms to correct the nonlinear error of the Beer-Lambert law, reducing the absolute error of concentration inversion to less than 5%. Combined with laboratory calibration of the calibrated absorption coefficient, the system can adapt to different oil components and maintain concentration detection stability under complex lighting, temperature and humidity conditions, providing a reliable data basis for risk assessment.

[0100] 3. The present invention uses the LSTM-CFD hybrid prediction operator to synchronously integrate concentration time series data (C), turbulent diffusion coefficient and wind speed vector. The LSTM network learns the historical concentration change pattern, and the CFD model analyzes the fluid mechanics diffusion equation. The dual mechanisms work together to overcome the prediction failure problem of a single model under sudden wind speed changes. Through forward time window concentration gradient vector calculation and abnormal point interpolation correction, the diffusion path of oil and gas in strong wind disturbances is accurately simulated to achieve early warning.

[0101] 4. The present invention dynamically generates an oil and gas concentration difference threshold by calculating the diffusion velocity of oil and gas molecules and the diffusion time from the center pixel to the end pixel. This threshold integrates parameters such as the Boltzmann constant, ambient temperature, and turbulence coupling factor, and is derived based on Fick's second law. The threshold is adaptively adjusted with wind speed and concentration gradient. Compared with a fixed threshold, the false alarm rate is reduced by more than 60%, effectively distinguishing between real leaks and sensor noise.

[0102] 5. The present invention adopts a dual-threshold potential risk point screening strategy. The first oil and gas concentration threshold (e.g., 500 ppm) is used to initially screen suspicious points, and interference with normal operations is eliminated by combining the working status of the fuel gun and the volatilization radius (e.g., 3 meters). The second oil and gas concentration threshold (e.g., 1500 ppm) is used to identify high-risk leakage points. Through global coordinate mapping and area statistics of potential risk points, the risk is quantified into a specific value (e.g., a level 1 alarm is triggered when the potential risk point area accounts for >5%). This mechanism avoids the misjudgment of operational activities by traditional solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0103] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0104] Figure 1 This is a flow chart of the gas station safety risk assessment method of the present invention;

[0105] Figure 2 This is a schematic diagram of the present invention for correcting the wind direction line in the infrared image. DETAILED DESCRIPTION

[0106] To make the above-mentioned objectives, features, and advantages of the present invention more clearly understood, the present invention is further described below in detail with reference to the accompanying drawings and specific embodiments. This will enable a full understanding of how this application uses technical means to solve technical problems and achieve technical effects, and to implement the invention accordingly.

[0107] In order to solve the technical problem that the current method of monitoring oil and gas concentration at gas stations is difficult to achieve uninterrupted monitoring, the present invention provides a gas station safety risk assessment method. After arranging equipment inside and outside the gas station to obtain infrared images, the infrared images are detected and calculated to obtain the oil and gas concentration at various locations of the gas station in real time. Then, based on the wind direction and the oil and gas concentration change data at various locations that have been obtained, the oil and gas concentration at various locations in the gas station within a certain period of time in the future is predicted, and accordingly, real-time assessment and prediction of safety risk assessment in the gas station are achieved.

[0108] S1. First, an infrared imaging array containing several bands is deployed in key areas inside the gas station, such as the operation area and outside the wall, to obtain infrared images of the corresponding area in real time. The infrared images are obtained in sequence according to time to obtain an infrared image sequence, that is, the infrared image sequence of the same position is composed of infrared images obtained by the infrared imaging array at several consecutive moments. For example, it can be arranged on the top of the gas station island to obtain the infrared image sequence of the oil gun operation area to monitor the high-risk area of ​​the oil gun operation area, or arranged near the breathing valve of the oil tank to achieve monitoring, or arranged around the wall to monitor the unorganized emission area. It can monitor only some key areas according to needs, or it can achieve full monitoring inside and outside the gas station. For scene coverage monitoring, the infrared imaging array generally needs to include infrared detectors in the medium-wave infrared (MWIR, wavelength range 3-5μm), long-wave infrared (LWIR, wavelength range 8-12μm), and visible light bands. The medium-wave infrared is used for hydrocarbon gas absorption. For example, the wavelength of oil and gas is generally 3.3μm. The long-wave infrared is used for background temperature compensation. It is generally necessary to select the visible light band to assist in target recognition and to identify objects at various locations in the gas station. The infrared image sequence is then preprocessed to eliminate background interference and to align the global coordinates of the gas station with the coordinates of each location in the infrared image sequence. The oil and gas concentration field is then constructed through subsequent steps:

[0109] S11. In the absence of light, infrared detectors generate electron-hole pairs due to thermal stimulation, thus forming a base current that is unrelated to the signal. In addition, the differences in the detector pixel responsivity caused by the manufacturing process and the interference of environmental noise will affect the numerical value of the radiation intensity in the infrared image. Therefore, it is necessary to correct the infrared image at each moment in the infrared image sequence of each band to obtain a corrected infrared image to ensure the accuracy of the radiation intensity data. The expression for the corrected infrared image is:

[0110]

[0111] In the above formula, express The radiation intensity at coordinate (x, y) of the corrected infrared image of the band at time t, It represents the radiation intensity of the acquired infrared image at the coordinate (x, y) at time t. represents the dark current matrix at the pixel level of the infrared detector that collects the infrared image, represents the gain correction coefficient at the pixel level, represents the atmospheric transmittance compensation factor at time t;

[0112] S12. After that, different corrected infrared images at the same moment need to be aligned in space. When the nodes equipped with infrared sensors capture the same object, the coordinates corresponding to the object are inconsistent. Therefore, in order to ensure the consistency of the coordinates of the same object in each corrected infrared image, it is necessary to select a location point in the gas station as the origin to establish a global coordinate system, and arrange high-contrast markers on the ground, and measure the position coordinates of the marker points in the global coordinate system. For example, a ceramic target with an infrared reflectivity greater than 90% is set as the marker point. In order to meet the solution of the overdetermined equation, the number of marker points needs to be at least 12, which are mainly arranged in key areas for oil and gas concentration monitoring such as the gas island, oil tank area and wall boundary.

[0113] S13. Each infrared controller corresponding to the marker point position simultaneously captures an image containing the control point. For any node k, after obtaining the corrected infrared image, it is necessary to establish a projection relationship of the control point in the global coordinate system to achieve coordinate alignment between the corrected infrared image and the actual coordinates in the gas station. The projection relationship satisfies:

[0114]

[0115] In the above formula, To correct the coordinates of any point in the infrared image in the global coordinate system, Indicates the image coordinates of any point in the corrected infrared image. and are all known, so the homography matrix can be obtained , then the image coordinates of any position in any corrected infrared image can be substituted according to the projection relationship to obtain the coordinates of the point in the global coordinate system, that is, to complete the spatial registration of any corrected infrared image at the same time. However, this method mainly relies on manual calibration. When the spatial alignment of coordinates is achieved through the homography matrix, there is inevitably a deviation between the position of the object calculated by the corrected infrared image and the actual physical position, that is, the reprojection error. In addition, its accuracy may gradually decrease over time during long-term use. Therefore, the following steps are further set to continuously calibrate and automatically update the homography matrix during use. The specific process is as follows:

[0116] S14. First, calculate the homography matrix When , the objective function is set to minimize the reprojection error of all landmark points:

[0117]

[0118] in,

[0119]

[0120] In the above formula, M represents the number of landmark points, is the measured coordinate of the i-th marker point in the global coordinate system, Represents the image coordinates of the i-th marker point in the corrected infrared image, so that a preliminary homography matrix can be obtained .

[0121] S15. After that, it is necessary to calculate the value of each marker point measured in the global coordinate system and its image coordinate in the corrected infrared image according to the formula A value is calculated, that is, the reprojection error of each landmark point, to reflect the obtained homography matrix Whether qualified.

[0122] S16. In general, the reprojection error threshold is set to 2.0 pixels. If there are marker points with a reprojection error greater than the threshold, the marker points with a reprojection error greater than the threshold are eliminated to obtain standard marker points. Then, return to step S14 and substitute the data of the standard marker points into the objective function and the projection relationship to calculate the homography matrix. If all marker points are less than the reprojection error threshold, it means that the homography matrix Relatively good, that is, qualified.

[0123] S17, thus obtaining a relatively better homography matrix At this time, it is necessary to judge the accuracy of the homography matrix, by extracting the feature points in the corrected infrared image and verifying the stability of its global coordinates in the time dimension, and based on this, the homography matrix Perform dynamic updates.

[0124] S171. First, for the feature points in the corrected infrared image, extract the feature points whose Hessian determinant value is greater than the determinant threshold (generally set to 500) in the corrected infrared image and mark them as standard feature points. The larger the value, the more drastic the curvature change of the point, which generally corresponds to corner points or spot features, etc., and is suitable as a tracking target to obtain a feature point set.

[0125] S172. Then, the displacement of each standard feature point in the feature point set is calculated using the optical flow method. The calculation formula is:

[0126]

[0127] In the above formula, Indicates the displacement of the standard feature point in the x direction, Indicates the displacement of the standard feature point in the y direction, represents the spatial gradient of the corrected infrared image in the x direction, Indicates the spatial gradient of the corrected infrared image in the y direction, Represents the gradient of the corrected infrared image in the time dimension, Indicates a non-zero minimum value;

[0128] S173. The displacement of each standard feature point reflects the real-time reprojection error of the standard feature point, so the average displacement is calculated: ,in, That is the average displacement, the number of standard feature points is M, Indicates the displacement of the standard feature point in the x direction, Indicates the displacement of the standard feature point in the y direction;

[0129] S174 , if the average displacement is less than the average displacement threshold (generally 2.0 pixels), there is no need to update the homography matrix; otherwise, the homography matrix needs to be updated through step S175 .

[0130] S175. When updating the homography matrix, first calculate the incremental homography matrix according to the displacement of each standard feature point. The calculation formula is:

[0131]

[0132] In the above formula, Represents the incremental homography matrix, that is, the function When taking the minimum value The value of and Respectively represent the x-coordinate and y-coordinate of feature point j in the current frame, and Respectively represent the x-coordinate and y-coordinate of feature point j in the previous frame;

[0133] S176. The updated homography matrix is:

[0134]

[0135]

[0136] Where S(T) represents the temperature compensation matrix, s represents the thermal expansion scaling factor, Indicates the thermal expansion coefficient of the object material, T indicates the current ambient temperature, Indicates the reference temperature during calibration.

[0137] S2. After obtaining a relatively accurate homography matrix, the relatively accurate global coordinates of each feature point in the corrected infrared image can be obtained according to the projection relationship, and the spatiotemporal alignment of each feature point in the corrected infrared image is achieved to convert the corrected infrared image into a real-time concentration distribution field in the global coordinate system, thereby completing the construction of the concentration field in the global coordinate system. The specific method is as follows:

[0138] S21. First, for each feature point in the corrected infrared image, calculate the differential absorption ratio:

[0139]

[0140] In the above formula, represents the differential absorption ratio at the coordinate (x, y) in the corrected infrared image at time t, represents the radiation intensity of the reference band, It represents the radiation intensity of the oil and gas absorption band. The infrared absorption characteristics of oil and gas are strongest in specific bands (such as 3.3μm). By calculating the ratio of the radiation intensity of the absorption band to the reference band, common-mode interference such as background thermal radiation and atmospheric scattering can be eliminated, the oil and gas absorption effect can be highlighted, and the multi-band radiation data can be converted into a characteristic quantity related only to the oil and gas concentration.

[0141] S22. Then, the oil and gas concentration field of the gas station is inverted according to the differential absorption ratio of each feature point to obtain the oil and gas concentration of any feature point in the global coordinates. The calculation formula is:

[0142]

[0143] In the above formula, Represents the coordinates in the global coordinate system The oil and gas concentration at time t, It represents the calibration absorption coefficient, which is obtained through laboratory calibration. L represents the optical path measured by binocular vision, that is, the physical path length of infrared light passing through the oil and gas layer. represents the long-wave infrared background compensation term, which compensates for the ambient thermal radiation base. The theoretical relationship between the differential absorption ratio D and the concentration C follows the Beer-Lambert Law, but two corrections are required: first, the actual optical path is nonlinear, requiring binocular vision to reconstruct the spatial path; second, the ambient thermal radiation base B affects the absolute radiation value, requiring compensation in the long-wave infrared band. These two corrections improve the accuracy of oil and gas concentration calculations.

[0144] S3. After constructing the preliminary oil and gas concentration field in global coordinates, it is necessary to extract the characteristic data of the oil and gas concentration and concentration gradient at any moment in the oil and gas concentration field. Based on the real-time established oil and gas concentration field, the oil and gas concentration field at any moment in the future should be predicted. Due to the influence of wind, the effect of wind on the diffusion of oil and gas concentration needs to be considered in the prediction process. The specific steps are as follows:

[0145] S31, after the position in the global coordinates is matched with the feature point in the corrected infrared image, for each position in the global coordinates , extract the concentration time series data of the forward time window including the current moment, and its expression is as follows:

[0146]

[0147] In the above formula, it represents Represents the coordinates in the global coordinate system The concentration time series data of the forward time window at the current moment is included. Indicates the coordinates in the global coordinates at time t The oil and gas concentration at Indicates the current moment, represents the time span of the forward time window;

[0148] S32. Calculate the concentration gradient vector of each position at each moment in the concentration time series data of each position in the global coordinates in sequence. The expression of the concentration gradient vector is:

[0149]

[0150] In the above formula, Indicates the coordinates of the global coordinate system at time t The concentration gradient vector at 、 and Indicates that at time t, the global coordinate The concentration in the x direction, y direction and relative to the previous moment;

[0151] S33. Obtain wind direction and speed data at various locations within the gas station through anemometers. Depending on the measurement accuracy requirements, a corresponding number of anemometers can be selected. The more anemometers there are, the more accurate the wind direction and speed data at various locations within the gas station. At least one anemometer can be set as the wind direction and speed data for various locations within the gas station.

[0152] S34, identifying characteristic points of abnormal oil and gas concentration values, and correcting the oil and gas concentration values ​​of the abnormal characteristic points by interpolation. Based on the spatial continuity of the real concentration field, abnormal values ​​may be caused by sensor noise or occlusion, resulting in sudden changes. Therefore, the specific steps include the following:

[0153] S341, sequentially establish a window with any pixel point in the corrected infrared image as the center, and calculate the average oil and gas concentration of all pixels in the window except the central pixel point to obtain the average oil and gas concentration of the window, such as Figure 2 As shown in FIG, a schematic diagram of correcting pixels in an infrared image is shown, and the black bold part is a window established based on the central pixel.

[0154] S342. Calculate the absolute value of the difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window to obtain the oil and gas concentration difference:

[0155]

[0156] In the above formula, Indicates the oil and gas concentration difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration of the corresponding window. and Respectively represent the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window;

[0157] S343. Set the oil-gas concentration difference threshold corresponding to the central pixel point based on the wind direction and speed data and the concentration gradient vector. Within the window, the direction where the oil-gas gradient is the largest due to the diffusion of oil-gas concentration is the direction along the direction. Therefore, the maximum value of the oil-gas concentration difference threshold is calculated based on this. The specific steps are as follows:

[0158] S3431. Obtain wind speed and direction data corresponding to the central pixel point, and establish a wind direction line projected onto the plane of the infrared image with the central pixel point as the starting point. The wind direction line is the direction in which the oil and gas concentration diffuses fastest in the infrared image, thus generating a large oil and gas concentration gradient, which serves as the oil and gas concentration difference threshold.

[0159] S3432. Obtain the pixel point where the wind direction line intersects the window boundary corresponding to the center pixel point, and mark it as the end pixel point. The end pixel point is located within the window boundary corresponding to the center pixel point.

[0160] S3433. Calculate the actual distance between the center pixel and the corresponding end pixel using the following formula:

[0161]

[0162] In the above formula, Indicates the actual distance between the center pixel and the corresponding end pixel. Indicates the pixel distance between the center pixel and the corresponding end pixel. Indicates the ratio coefficient between actual distance and pixel distance;

[0163] S3434. Calculate the diffusion velocity of oil and gas molecules based on wind direction and speed data. The calculation formula is:

[0164]

[0165] In the above formula, Indicates the diffusion speed of oil and gas molecules, represents the Boltzmann constant, represents the ambient temperature, m represents the average mass of oil and gas molecules, represents the gradient diffusion coefficient, Indicates the oil and gas concentration value of the central pixel point, Indicates the minimum value of oil and gas concentration in the pixel points other than the central pixel point in the window corresponding to the central pixel point. represents the turbulence coupling factor, Represents the wind speed vector at the center pixel point, Indicates the angle between the actual wind direction and the wind direction line. Represents the concentration gradient vector of the central pixel;

[0166] S3435. Calculate the diffusion time of oil and gas molecules from the center pixel to the corresponding end pixel. The calculation formula is:

[0167]

[0168] In the above formula, Indicates the diffusion time of oil and gas molecules from the central pixel to the corresponding end pixel. Indicates the actual distance between the center pixel and the corresponding end pixel;

[0169] S3436. Calculate the oil-gas concentration difference threshold based on Fick's second law. The calculation formula is:

[0170]

[0171] In the above formula, Indicates the oil-gas concentration difference threshold, Indicates the acquisition time interval of infrared images, represents the turbulent diffusion coefficient, represents the safety margin coefficient. The traditional fixed threshold cannot adapt to the diffusion capacity under different wind speeds and concentrations. Through this method, the oil and gas concentration difference threshold during diffusion under different wind speeds and oil and gas concentrations can be dynamically calculated to reduce the misjudgment rate in step S344.

[0172] S344, determining whether the oil-gas concentration difference corresponding to the central pixel point is less than the oil-gas concentration difference threshold;

[0173] If yes, mark the central pixel as a normal feature point;

[0174] If not, the central pixel is marked as an abnormal feature point;

[0175] S345. Correct the oil and gas concentration value of the abnormal feature point by using interpolation methods such as weighted average of neighboring points.

[0176] S35. Predict the oil and gas concentration at each location in the gas station at any time based on wind direction and speed data, that is, build a diffusion-transport prediction model to predict the future period. Evolution of the concentration field inside:

[0177]

[0178] In the above formula, Indicates The concentration field of the gas station at the moment, F represents the hybrid prediction operator that integrates LSTM (long short-term memory network) and CFD (computational fluid dynamics model), represents the turbulent diffusion coefficient, C represents the concentration time series data, Representation and global coordinates Corresponding wind speed vector data.

[0179] S4. Based on the oil and gas concentration field at any future moment, select the potential risk points in the global coordinates where the oil and gas concentration exceeds the oil and gas concentration threshold, thereby screening the risk locations. However, during the refueling process at the gas station, the oil and gas concentration near the corresponding location will temporarily increase. These potential risk points need to be distinguished when identifying them. Therefore, the identification of potential risk points specifically includes the following steps:

[0180] S41 . Set a first oil and gas concentration threshold, and mark pixels in the corrected infrared image whose oil and gas concentration values ​​are greater than the first oil and gas concentration threshold as suspicious pixels based on the oil and gas concentration field of the gas station.

[0181] S42. Obtain the global coordinates corresponding to the suspicious pixel points, and identify the global coordinates of the gas pumps in the gas station that are currently refueling or have refueled but have not exceeded the evaporation time threshold (e.g., within 3 minutes), and mark the gas pumps as working gas pumps. Short-term evaporation near working gas pumps is normal and requires special treatment.

[0182] S43, calculating the Euclidean distance between the suspicious pixel point and the working refueling gun in the global coordinates;

[0183] S44. Select suspicious pixel points whose Euclidean distance from the working refueling gun in the global coordinates is greater than the volatilization radius and mark them as potential risk points. Mark the remaining suspicious pixel points as pending pixel points. The reason for the high oil and gas concentration near the pending pixel points may be that refueling is in progress or has just been completed, so further judgment is required.

[0184] S45. Set a second oil and gas concentration threshold that is greater than the first oil and gas concentration threshold, and mark the pending pixel points whose oil and gas concentration values ​​are greater than the second oil and gas concentration threshold as potential risk points to determine whether they are short-term high concentrations caused by normal operations. Mark the pending pixel points whose non-oil and gas concentration values ​​are greater than the second oil and gas concentration threshold as short-term high concentrations caused by abnormal operations, that is, potential risk points, such as oil and gas leakage caused by oil gun seal failure.

[0185] S5. Assess the safety risk of a gas station based on the area occupied by the potential danger points. For example, the danger level can be divided according to the area occupied by the potential danger points. The larger the area occupied by the potential danger points, the higher the danger level. Another example is to use the sigmoid function to determine whether it is dangerous.

[0186] The present invention also provides a gas station safety risk assessment system, comprising a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, a gas station safety risk assessment method is implemented.

[0187] Those skilled in the art will appreciate that all or part of the steps in the above-mentioned embodiment methods can be accomplished by instructing the relevant hardware through a program. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0188] Each embodiment in this specification is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to in detail. For the above embodiments, since they are basically similar to the method embodiments, the description is relatively simple. For relevant parts, please refer to the partial description of the method embodiments.

[0189] The above embodiments provide a detailed introduction to the present invention. Specific examples are used herein to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core ideas. At the same time, for those skilled in the art, according to the ideas of the present invention, there may be changes in the specific implementation methods and application scopes. In summary, the contents of this specification should not be understood as limiting the present invention.

Claims

1. A gas station safety risk assessment method, characterized in that: The specific steps of the security risk assessment method are: S1. Deploy an infrared imaging array containing several bands in the gas station to acquire infrared images in real time, and preprocess and align the infrared images to obtain corrected infrared images; S2. Calculate the oil and gas concentration of each characteristic point of the gas station in real time based on the corrected infrared image, and establish an oil and gas concentration field including the oil and gas concentration of each location of the gas station in real time; S3. predicting the oil and gas concentration field at any time in the future based on the oil and gas concentration field established in real time; In step S3, the following steps are specifically included: S31. For each position in the global coordinates, extract the concentration time series data of the forward time window including the current moment, and the expression is as follows: ； In the above formula, it represents Represents the coordinates in the global coordinate system The concentration time series data of the forward time window at the current moment is included. Indicates the coordinates in the global coordinates at time t The oil and gas concentration at Indicates the current moment, represents the time span of the forward time window; S32. Calculate the concentration gradient vector of each position at each moment in the concentration time series data of each position in the global coordinates in sequence. The expression of the concentration gradient vector is: ； In the above formula, Indicates the coordinates of the global coordinate system at time t The concentration gradient vector at 、 and Indicates that at time t, the global coordinate The concentration in the x direction, y direction and relative to the previous moment; S33, obtaining wind direction and speed data at various locations within the gas station; S34, identifying characteristic points where the oil and gas concentration values ​​are abnormal, and correcting the oil and gas concentration values ​​of the abnormal characteristic points by interpolation; S35. Predict the oil and gas concentration and oil and gas concentration gradient at each location in the gas station at any time based on the wind direction and speed data. The prediction formula is: ； In the above formula, Indicates The gas station concentration field at the moment, F represents the hybrid prediction operator integrating LSTM and CFD, represents the turbulent diffusion coefficient, C represents the concentration time series data, Representation and global coordinates Corresponding wind speed vector data; S4. Selecting a potential risk point in the global coordinates where the oil and gas concentration exceeds the oil and gas concentration threshold based on the oil and gas concentration field at any future moment; S5. Assess the safety risk of the gas station based on the area occupied by potential hazards.

2. The gas station safety risk assessment method according to claim 1, characterized in that: In step S1, the specific steps of preprocessing and coordinate alignment of the infrared image to obtain the corrected infrared image are as follows: S11. Correct the infrared image at each moment to obtain a corrected infrared image. The expression of the corrected infrared image is: ； In the above formula, express The radiation intensity at coordinate (x, y) of the corrected infrared image of the band at time t, It represents the radiation intensity of the acquired infrared image at the coordinate (x, y) at time t. represents the dark current matrix at the pixel level of the infrared detector that collects the infrared image, represents the gain correction coefficient at the pixel level, represents the atmospheric transmittance compensation factor at time t; S12. Establishing a global coordinate system within the gas station and presetting a number of landmark points within the gas station; S13. Establish a projection relationship between the corrected infrared image and the global coordinate system to complete coordinate alignment. The projection relationship satisfies: ； In the above formula, To correct the coordinates of any point in the infrared image in the global coordinate system, represents the image coordinates of any point in the corrected infrared image, represents the homography matrix; S14. Setting the objective function for calculating the homography matrix with the goal of minimizing the reprojection error of all landmark points to obtain the homography matrix. The expression of the objective function is: ； in, ； In the above formula, M represents the number of landmark points, is the measured coordinate of the i-th marker point in the global coordinate system, represents the image coordinates of the i-th marker point in the corrected infrared image; S15, calculating the reprojection error of each marker point in the infrared image according to the obtained homography matrix; S16, determining whether the reprojection errors of each marker point are less than the reprojection error threshold; If yes, then output the homography matrix and go to step S17; If not, then remove the marker points whose reprojection error is greater than or equal to the reprojection error threshold, and then return to step S14; S17. Extract feature points from the standard infrared image, and dynamically update the homography matrix according to the global coordinates of each feature point.

3. The gas station safety risk assessment method according to claim 2, characterized in that: In step S17, the steps for calculating the homography matrix are as follows: S171, marking feature points in the corrected infrared image whose Hessian determinant values ​​are greater than a determinant threshold as standard feature points to obtain a feature point set; S172. Calculate the displacement of each standard feature point in the feature point set using the optical flow method. The calculation formula is: ； In the above formula, Indicates the displacement of the standard feature point in the x direction, Indicates the displacement of the standard feature point in the y direction, represents the spatial gradient of the corrected infrared image in the x direction, Indicates the spatial gradient of the corrected infrared image in the y direction, Represents the gradient of the corrected infrared image in the time dimension, Indicates a non-zero minimum value; S173, calculating the average displacement of each standard feature point; S174, determining whether the average displacement is less than an average displacement threshold; If yes, then output the homography matrix and go to step S2; If not, proceed to step S175; S175. Calculate the incremental homography matrix based on the displacement of each standard feature point. The calculation formula is: ； In the above formula, Represents the incremental homography matrix, that is, the function When taking the minimum value The value of and Respectively represent the x-coordinate and y-coordinate of feature point j in the current frame, and Respectively represent the x-coordinate and y-coordinate of feature point j in the previous frame; S176. Calculate an updated homography matrix based on the incremental homography matrix. The expression of the updated homography matrix is: ； ； Where S(T) represents the temperature compensation matrix, s represents the thermal expansion scaling factor, Indicates the thermal expansion coefficient of the object material, T indicates the current ambient temperature, Indicates the reference temperature during calibration.

4. The gas station safety risk assessment method according to claim 1, characterized in that: In step S2, the following steps are specifically included: S21. For each feature point in the corrected infrared image, calculate the differential absorption ratio: ； In the above formula, represents the differential absorption ratio at the coordinate (x, y) in the corrected infrared image at time t, represents the radiation intensity of the reference band at the coordinate (x, y) in the corrected infrared image at time t, Indicates the radiation intensity of the oil and gas absorption band; S22. Invert the oil and gas concentration field of the gas station based on the differential absorption ratio of each feature point to obtain the oil and gas concentration at any position in the global coordinates. The calculation formula for the oil and gas concentration at any position in the global coordinates is: ； In the above formula, Represents the coordinates in the global coordinate system The oil and gas concentration at time t, represents the calibration absorption coefficient, Represents the coordinates in the global coordinate system The optical path measured by binocular vision at time t, Represents the coordinates in the global coordinate system The long-wave infrared background compensation term at time t.

5. The gas station safety risk assessment method according to claim 1, characterized in that: In step S34, the following steps are specifically included: S341, sequentially establishing a window with any pixel point in the corrected infrared image as the center, and calculating the average oil and gas concentration of all pixels in the window excluding the center pixel point to obtain the average oil and gas concentration in the window; S342. Calculate the absolute value of the difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window to obtain the oil and gas concentration difference: ； In the above formula, Indicates the oil and gas concentration difference between the oil and gas concentration value of the central pixel and the average oil and gas concentration of the corresponding window. and Respectively represent the oil and gas concentration value of the central pixel and the average oil and gas concentration value of the corresponding window; S343, setting the oil and gas concentration difference threshold corresponding to the central pixel point according to the wind direction and speed data and the concentration gradient vector; S344, determining whether the oil-gas concentration difference corresponding to the central pixel point is less than the oil-gas concentration difference threshold; If yes, mark the central pixel as a normal feature point; If not, the central pixel is marked as an abnormal feature point; S345. Correct the oil and gas concentration value of the abnormal feature point by interpolation.

6. The gas station safety risk assessment method according to claim 5, characterized in that: In step S343, the following steps are specifically included: S3431, obtaining wind speed and direction data corresponding to the central pixel point, and establishing a wind direction line projected onto the plane where the infrared image is located with the central pixel point as the starting point; S3432. Obtain the pixel point where the wind direction line intersects the window boundary corresponding to the center pixel point, and mark it as the end pixel point. The end pixel point is located within the window boundary corresponding to the center pixel point. S3433. Calculate the actual distance between the center pixel and the corresponding end pixel using the following formula: ； In the above formula, Indicates the actual distance between the center pixel and the corresponding end pixel. Indicates the pixel distance between the center pixel and the corresponding end pixel. Indicates the ratio coefficient between actual distance and pixel distance; S3434. Calculate the diffusion velocity of oil and gas molecules based on wind direction and speed data. The calculation formula is: ； In the above formula, Indicates the diffusion speed of oil and gas molecules, represents the Boltzmann constant, represents the ambient temperature, m represents the average mass of oil and gas molecules, represents the gradient diffusion coefficient, Indicates the oil and gas concentration value of the central pixel point, Indicates the minimum value of oil and gas concentration in the pixel points other than the central pixel point in the window corresponding to the central pixel point. represents the turbulence coupling factor, Represents the wind speed vector at the center pixel point, Indicates the angle between the actual wind direction and the wind direction line. Represents the concentration gradient vector of the central pixel; S3435. Calculate the diffusion time of oil and gas molecules from the center pixel to the corresponding end pixel. The calculation formula is: ； In the above formula, Indicates the diffusion time of oil and gas molecules from the central pixel to the corresponding end pixel. Indicates the actual distance between the center pixel and the corresponding end pixel; S3436. Calculate the oil-gas concentration difference threshold based on Fick's second law.

7. The gas station safety risk assessment method according to claim 6, characterized in that: In step S3436, the calculation formula for the oil-gas concentration difference threshold is: ； In the above formula, Indicates the oil-gas concentration difference threshold, Indicates the acquisition time interval of infrared images, represents the turbulent diffusion coefficient, Represents the safety margin factor.

8. The gas station safety risk assessment method according to claim 1, characterized in that: In step S4, the following steps are specifically included: S41, setting a first oil and gas concentration threshold, and marking pixels in the corrected infrared image whose oil and gas concentration values ​​are greater than the first oil and gas concentration threshold as suspicious pixels based on the oil and gas concentration field of the gas station; S42: Obtain the global coordinates corresponding to the suspicious pixel points, and identify the global coordinates of the fuel pumps at the gas station that are refueling or have refueled but have not exceeded the evaporation time threshold, and mark the fuel pumps as working fuel pumps; S43, calculating the Euclidean distance between the suspicious pixel point and the working refueling gun in the global coordinates; S44, selecting suspicious pixel points whose Euclidean distance from the working refueling gun in the global coordinates is greater than the volatilization radius and marking them as potential risk points, and marking the remaining suspicious pixel points as pending pixel points; S45 , setting a second oil and gas concentration threshold that is greater than the first oil and gas concentration threshold, and marking undetermined pixel points whose oil and gas concentration values ​​are greater than the second oil and gas concentration threshold as potential risk points.

9. A system for implementing the gas station safety risk assessment method according to any one of claims 1 to 8, characterized in that: The method comprises a processor and a memory, wherein the memory is used to store a computer program, and when the computer program is executed by the processor, the gas station safety risk assessment method according to any one of claims 1 to 8 is implemented.

Images