Multi-component insulating gas leakage infrared detection system and method based on multiple detectors

Abstract

A multi-detector-based infrared detection system and method for multi-component insulating gas leakage belongs to the technical field of insulating gas leakage detection. It solves the problems of large size, high cost, low cost-effectiveness of existing devices and how to improve the infrared imaging detection effect. The present invention adopts a low-cost multivariate filter + infrared focal plane detector combination design for spectrometry and data processing to achieve simultaneous imaging detection of multi-component insulating gas, reduce equipment costs, and realize portable detection. The grayscale value obtained by the pixel point is matched with the current Gaussian mixture model, the weight parameter is updated, and the mean and variance of the model corresponding to the pixel point are updated. All models are re-sorted in order, and the model that meets the weight threshold is selected as the background. The current pixel point value is matched with the background model to determine whether it is the background or foreground. The gas diffusion display is enhanced by subtracting the complex dynamic background, improving the detection accuracy and automatically identifying the gas diffusion form.

Description

Technical Field

[0001] The present invention belongs to the technical field of insulator gas leakage detection, and is an infrared detection system and method for multi-component insulating gas leakage based on multiple detectors. Background Art

[0002] To achieve energy conservation and emission reduction, new environmentally friendly insulating gases, such as perfluoroisobutyronitrile (PFIBN) mixed gas, are gradually replacing sulfur hexafluoride (SF6) in power generation. The use of PFIBN in ring main units (RMUs) is steadily increasing, and its application in GIL and GIS busbars is also on the horizon. Pure PFIBN gas has a high boiling point and must be mixed with a buffer gas such as air or CO2 to form a PFIBN mixed insulating gas. This mixture exhibits excellent insulating properties. An 18% to 20% PFIBN-CO2 mixture has an electrical strength comparable to that of SF6 gas and similar arc-interrupting capability. PFIBN mixed insulating gas also exhibits excellent energy conservation and environmental protection properties. A 6% PFIBN mixture has a Global Warming Potential (GWP) of approximately 2% that of pure SF6 gas. However, PFIBN's acute toxicity is classified as Category 4 under the Globally Harmonized System of Classification and Labelling of Chemicals (GHS), and its leakage poses a health hazard to maintenance personnel. Currently, the technology and equipment for perfluoroisobutyronitrile gas leak detection are still in the research and development stage, and the market demand is urgent.

[0003] Among the many optical remote sensing imaging technologies for insulating gases, those based on narrow spectral bands (infrared imaging) or single spectral bands (laser imaging) acquire single-spectral information, which is significantly affected by environmental interference, resulting in low sensitivity and unstable operation. Differential absorption lidar technology, while highly sensitive, can suffer from laser instability and wavelength drift, leading to detection failure. In contrast, methods based on infrared multi- / hyperspectral methods (Fourier transform infrared spectroscopy) acquire multi-spectral information, enabling the complete collection of gas characteristic spectra, resulting in higher sensitivity and enhanced resistance to environmental interference. Fourier transform infrared spectroscopy remote sensing technology has been applied to insulating gas leak detection, but due to limitations in size and device cost, it is bulky, heavy, and expensive. Furthermore, Fourier transform infrared spectrometers acquire hyperspectral infrared data, capable of detecting thousands of gas components. This makes them less cost-effective when used to detect only a single gas. Summary of the Invention

[0004] The technical solution of the present invention is used to solve the problems of large size, high cost and low cost performance of existing devices and how to improve the infrared imaging detection effect.

[0005] The present invention solves the above technical problems through the following technical solutions:

[0006] A multi-component insulating gas leakage infrared detection system based on multiple detectors, comprising: a beam splitter (10), a visible light camera (11), a first flat plate spectroscope (12), a first filter (13), a first infrared focal plane detector (14), a second flat plate spectroscope (15), a second filter (16), a second infrared focal plane detector (17), and an image processing unit (18); the beam splitter (10), the first flat plate spectroscope (12), and the second flat plate spectroscope (15) are arranged in sequence along the horizontal direction; the beam splitter ( The angle between the reflecting surface of the first flat plate beam splitter (12) and the horizontal light path is 45°, and the beam splitter (10) is coated with a semi-transparent and semi-reflective film for transmitting infrared light and reflecting visible light at the same time; the optical path of the visible light camera (11) is perpendicular to the horizontal light path, and the visible light incident on the beam splitter (10) enters the visible light camera (11) after vertical reflection; the reflecting surfaces of the first flat plate beam splitter (12) and the second flat plate beam splitter (15) are both 45° with the horizontal light path, and the function of the flat plate beam splitter is to reflect half of the incident infrared light and transmit half; the The optical paths of the first infrared focal plane detector (14) and the second infrared focal plane detector (17) are both perpendicular to the horizontal optical path. The first filter (13) and the second filter (16) are respectively arranged on the lenses of the first infrared focal plane detector (14) and the second infrared focal plane detector (17). The first filter (13) and the second filter (16) are used to process the infrared spectrum information of different insulating gases to be tested. Preferably, the first filter (13) and the second filter (16) are used to process the infrared spectrum information of sulfur hexafluoride gas and Infrared spectrum information of perfluoroisobutyronitrile gas; infrared light passing through the beam splitter (10) is vertically reflected by the first flat-plate spectroscope (12) and enters the first infrared focal plane detector (14) for detection; infrared light passing through the first flat-plate spectroscope (12) is vertically reflected by the second flat-plate spectroscope (15) and enters the second infrared focal plane detector (17) for detection; the first infrared focal plane detector (14) and the second infrared focal plane detector (17) send the detected infrared spectrum signals to the image processing unit (18) for image processing.

[0007] Furthermore, the coaxial registration method of the field of view of the infrared focal plane detector and the visible light camera is as follows:

[0008] (1) Move the entire system so that the central axis of the field of view of the infrared focal plane detector is aligned with the heating target plate;

[0009] (2) Adjust the height of the visible light camera and the angle of the beam splitter so that the central axis of the visible light camera's field of view is aligned with the heating target plate.

[0010] An image processing method applied to the above-mentioned multi-component insulating gas leakage infrared detection system based on multiple detectors includes the following steps:

[0011] Step 1: Initialize model parameters;

[0012] Select continuous frame images of the infrared detector and initialize the mean matrix and variance matrix of each Gaussian mixture model;

[0013] Step 2: Update model parameters;

[0014] Step 2.1, obtain the video frame at time t from the real-time video of the infrared detector, and convert the gray value of the pixel point (x, y) into Matching judgment with the current Gaussian mixture model, if If it matches the jth model, then mark If there is no match with the jth model, mark ;

[0015] Step 2.2: After the matching is completed, the weight parameters of the j-th Gaussian mixture model are calculated based on each tag. Make updates;

[0016] Step 2.3: If there is a mark , then the mean and variance of the j-th Gaussian mixture model corresponding to the pixel are updated; if the pixel value If no match is found with any Gaussian mixture model, the pixel value is used Replace the mean of the corresponding position of the last Gaussian mixture model and redistribute the variance and weight;

[0017] Step 3: Model motion detection;

[0018] Step 3.1, normalize the weight of the Gaussian mixture model of each pixel;

[0019] Step 3.2, follow Rearrange all Gaussian mixture models in order of size, and select the first B Gaussian mixture models that are greater than or equal to the weight threshold as the background;

[0020] Step 3.3: Set the current pixel value Match it with the B background models obtained in step 3.2 to determine whether it is background information or motion foreground, and classify the pixels of background information and motion foreground, thereby enhancing the infrared image of insulating gas leakage.

[0021] Furthermore, the formulas for initializing the mean matrix and variance matrix of each Gaussian mixture model are as follows:

[0022]

[0023]

[0024] Where, represents the mean matrix of the j-th Gaussian mixture model, represents the variance matrix of the j-th Gaussian mixture model, N represents the number of consecutive frame images selected, I(i) represents the video sequence, where i=1, 2, ..., N; during parameter initialization, the video sequence I(i) is cumulatively sampled. Indicates the number of mean matrices and variance matrices.

[0025] Furthermore, the grayscale value obtained by the pixel point (x, y) The formula for matching with the current Gaussian mixture model is as follows:

[0026]

[0027] Where, is the mean matrix of the j-th Gaussian mixture model at time t-1, is the standard deviation matrix of the j-th Gaussian mixture model at time t-1, and c is the empirical parameter.

[0028] Furthermore, the weight parameters of the j-th Gaussian mixture model according to each tag are The update formula is as follows:

[0029]

[0030] Among them, α is the learning rate of weight, α∈[0,1], is the weight parameter of the j-th Gaussian mixture model at time t, is the weight parameter of the j-th Gaussian mixture model at time t-1, is the matching label of the j-th Gaussian mixture model.

[0031] Furthermore, the formula for updating the mean and variance of the j-th Gaussian mixture model corresponding to the pixel point is as follows:

[0032]

[0033]

[0034] Where β is the learning rate of the mean and variance, β∈[0,1], is the mean matrix of the j-th Gaussian mixture model at time t-1, is the grayscale value of the i-th frame image, is the variance matrix of the j-th Gaussian mixture model at time t-1.

[0035] Furthermore, the formula for selecting the first B Gaussian mixture models that are greater than or equal to the weight threshold as the background model is as follows:

[0036]

[0037] Where, τ is the weight threshold, is the minimum function.

[0038] Furthermore, the formula for performing matching judgment is as follows:

[0039]

[0040] in, is the mean matrix of the j-th Gaussian mixture model at time t.

[0041] A storage medium stores a computer program, which executes the steps of the above-mentioned image processing method when executed by a processor.

[0042] The advantages of the present invention are:

[0043] The present invention adopts a low-cost multivariate filter + infrared focal plane detector combination design for spectrometry and data processing to achieve simultaneous imaging detection of multi-component insulating gases. This not only reduces equipment cost and size, but also enables rapid data processing, realizing portable, multi-component insulating gas leakage imaging detection. Compared with the prior art, the device is smaller in size, lighter in weight, and lower in cost, improving the cost-effectiveness of gas leakage detection. The image processing method of the present invention selects continuous frame images from the infrared detector, initializes the mean matrix and variance matrix of each Gaussian mixture model, matches the grayscale value obtained by the pixel point with the current Gaussian mixture model, updates the weight parameters, and updates the mean and variance of the Gaussian mixture model corresponding to the pixel point. The weights are normalized, all Gaussian mixture models are reordered in order, and a Gaussian mixture model that meets the weight threshold is selected as the background. The current pixel value is matched with the background model to determine whether it is background information or a moving foreground. The present invention enhances the gas diffusion display by subtracting complex dynamic background, improving detection accuracy while automatically identifying the gas diffusion morphology, further improving the infrared imaging detection effect. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 This is a structural diagram of a multi-component insulating gas leakage infrared detection system based on multiple detectors of the present invention;

[0045] Figure 2 It is a heating target plate diagram of the infrared focal plane detector of the present invention and the field of view of the visible light camera coaxially aligned;

[0046] Figure 3 This is a diagram of an experimental scene of infrared image acquisition of insulating gas leakage according to the present invention;

[0047] Figure 4 This is a diagram showing the infrared imaging processing result of insulating gas leakage according to the present invention. DETAILED DESCRIPTION

[0048] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0049] The technical solution of the present invention is further described below with reference to the accompanying drawings and specific embodiments:

[0050] Example 1

[0051] 1. System structure

[0052] like Figure 1As shown, a multi-component insulating gas leakage infrared detection system based on multiple detectors is constructed, and the multi-component insulating gas leakage infrared detection system based on multiple detectors includes: a beam splitter (10), a visible light camera (11), a first flat plate spectroscope (12), a first filter (13), a first infrared focal plane detector (14), a second flat plate spectroscope (15), a second filter (16), a second infrared focal plane detector (17), and an image processing unit (18); the beam splitter (10), the first flat plate spectroscope (12), the second flat plate spectroscope ( 15) are arranged in sequence along the horizontal direction; the angle between the reflection surface of the beam splitter (10) and the horizontal light path is 45°, and the beam splitter (10) is coated with a semi-transparent and semi-reflective film for transmitting infrared light and reflecting visible light at the same time; the optical path of the visible light camera (11) is perpendicular to the horizontal light path, and the visible light incident on the beam splitter (10) enters the visible light camera (11) after vertical reflection; the angle between the reflection surface of the first flat beam splitter (12) and the second flat beam splitter (15) and the horizontal light path is 45°, and the function of the flat beam splitter is to reflect the incident infrared light. The light is half reflected and half transmitted; the optical paths of the first infrared focal plane detector (14) and the second infrared focal plane detector (17) are perpendicular to the horizontal optical path, the first filter (13) and the second filter (16) are respectively arranged on the lenses of the first infrared focal plane detector (14) and the second infrared focal plane detector (17), and the first filter (13) and the second filter (16) are used to process the infrared spectrum information of different insulating gases to be tested respectively. Preferably, the first filter (13) and the second filter (16) are used to process six Infrared spectrum information of sulfur fluoride gas and perfluoroisobutyronitrile gas; infrared light passing through the beam splitter (10) is vertically reflected by the first flat-plate spectroscope (12) and enters the first infrared focal plane detector (14) for detection; infrared light passing through the first flat-plate spectroscope (12) is vertically reflected by the second flat-plate spectroscope (15) and enters the second infrared focal plane detector (17) for detection; the first infrared focal plane detector (14) and the second infrared focal plane detector (17) send the detected infrared spectrum signals to the image processing unit (18) for image processing.

[0053] 2. Field of view coaxial registration

[0054] The goal of field of view coaxial alignment is to make the central axis of the visible light camera coincide with the central axis of the infrared focal plane detector. Therefore, the center of the field of view optical axis can be found and adjusted through the visible light and infrared light images of the same target. Since the infrared focal plane is sensitive to heat, it is necessary to make a thermally stable and regularly shaped target plate to ensure that the target is prominent and stable in the infrared image.

[0055] The coaxial registration method of the infrared focal plane detector and the visible light camera's field of view is as follows:

[0056] like Figure 2 As shown, this embodiment comprehensively considers the detection distance and the camera field of view angle factors, and uses heatable ceramic pieces to splice together to make a heating target plate with a size of 1.2*1.2m as the field of view coaxial alignment target.

[0057] The coaxial adjustment steps are as follows:

[0058] (1) Move the entire system so that the central axis of the field of view of the infrared focal plane detector is aligned with the heating target plate (the heating target plate is imaged in the center of the infrared image);

[0059] (2) Adjust the height of the visible light camera and the angle of the beam splitter so that the central axis of the visible light camera's field of view is aligned with the heating target plate.

[0060] The two-straight-line method is used to calibrate the optical axis coaxially. The heating target plates are placed 20 meters, 40 meters, and 100 meters in front of the system structure, and steps (1) and (2) are repeated. The 20 meters and 100 meters are used for calibration, and the 40 meter is used for verification.

[0061] 3. Image processing methods

[0062] The Gaussian mixture model (GMM) is a background modeling method based on pixel sample statistics. It assumes that the color information between pixels is independent. Therefore, the change in the value of each pixel in a video image can be viewed as a random process that continuously generates new pixel values. Statistically, this random process exhibits a Gaussian distribution. In real-world flame video images, due to lighting variations and slight object motion, the distribution of each pixel typically follows a multimodal Gaussian distribution. The motion detection algorithm process includes GMM parameter initialization, model parameter update, and moving object detection.

[0063] (1) Model parameter initialization

[0064] In motion detection, the video image information within a period of time is usually counted to initialize the parameters of each Gaussian model, as follows:

[0065] Select continuous frame images of the infrared detector and initialize the mean matrix and variance matrix of each Gaussian mixture model. The formula is as follows:

[0066] (1)

[0067] (2)

[0068] Where, represents the mean matrix of the j-th Gaussian mixture model, represents the variance matrix of the j-th Gaussian mixture model, N represents the number of consecutive frame images selected, I(i) represents the video sequence, where i=1, 2, ..., N; during parameter initialization, the video sequence I(i) is cumulatively sampled. Indicates the number of mean matrices and variance matrices.

[0069] (2) Model parameter update

[0070] After the GMM model parameter initialization state is completed, the next step of moving target detection begins. Get the video frame at time t from the real-time video of the infrared detector, and convert the grayscale value of the pixel point (x, y) Matching judgment with the current Gaussian mixture model, if If it matches the jth model, then mark ; If there is no match with the jth model, mark , the matching judgment formula is as follows:

[0071] (3)

[0072] Where, is the mean matrix of the j-th Gaussian mixture model at time t-1, is the standard deviation matrix of the j-th Gaussian mixture model at time t-1, c is an empirical parameter, and generally taking 2 to 3 will give a more ideal detection effect.

[0073] After the matching is completed, the weight parameters of the j-th Gaussian mixture model are adjusted according to each mark. Update according to formula (4).

[0074] (4)

[0075] Among them, α is the learning rate of weight, α∈[0,1], is the weight parameter of the j-th Gaussian mixture model at time t, is the weight parameter of the j-th Gaussian mixture model at time t-1, is the matching label of the j-th Gaussian mixture model.

[0076] If there is a tag , then the mean and variance of the j-th Gaussian mixture model corresponding to the pixel are updated according to formula (5) and formula (6); if the pixel value If no match is found with any Gaussian mixture model, the pixel value is used Replace the mean of the corresponding position of the last Gaussian mixture model and redistribute the variance and weight.

[0077] (5)

[0078] (6)

[0079] Where β is the learning rate of the mean and variance, β∈[0,1], is the mean matrix of the j-th Gaussian mixture model at time t-1, is the grayscale value of the i-th frame image, is the variance matrix of the j-th Gaussian mixture model at time t-1.

[0080] (3) Model motion detection

[0081] In the mixed Gaussian model, in order to obtain the background model at time t, it is necessary to adjust all j Gaussian models that have completed parameter updates at the current time t. The specific method is as follows:

[0082] First, the weight of the Gaussian mixture model of each pixel is normalized;

[0083] Secondly, according to All Gaussian models are reordered in order of size, and the first B models that satisfy the following formula (7) are selected as background.

[0084] (7)

[0085] Among them, τ is the weight threshold, is the minimum function;

[0086] Finally, the current pixel value The obtained B backgrounds are matched and judged according to formula (8). If a model j is found that satisfies formula (8), it is considered that the pixel point presents background information at time t; otherwise, the point is judged to be a moving foreground. The pixels of background information and moving foreground are classified to achieve the enhancement of the infrared image of insulating gas leakage.

[0087] (8)

[0088] in, is the mean matrix of the j-th Gaussian mixture model at time t.

[0089] 4. Experiments and Results

[0090] 4.1 Data Collection

[0091] An insulating gas (sulfur hexafluoride) leakage simulation experiment was carried out, and more than 700 frames of continuous diffusion infrared image data of insulating gas leakage under natural background were collected. The experimental scene is shown in Figure 3 .

[0092] The experiment used the Telops FAST V1k long-wave thermal imager, and the instrument parameters are shown in Table 1. A gas cylinder filled with sulfur hexafluoride was used to simulate a gas leak, and the detection distance was 50 meters.

[0093] Table 1 Instrument parameters

[0094]

[0095] The experimental parameters are shown in Table 2.

[0096] Table 2 Experimental parameters

[0097]

[0098] 4.2 Experimental Results and Discussion

[0099] Following the insulating gas leak infrared recognition process, we processed the simulated sulfur hexafluoride gas leak infrared images, achieving real-time dynamic background update and subtraction to obtain insulating gas diffusion identification results. The following analysis includes data processing results for key frames of gas diffusion (covering the initial release phase, diffusion phase, and final phase).

[0100] like Figure 4 As shown in the figure, the original infrared images of insulating gas leakage diffusion at each stage of acquisition ((1) is the beginning of leakage diffusion, (2) is the middle of diffusion, and (3) is about to end of diffusion) and their subsequent sequences (1-1, 2-1, 3-1) are processed by Gaussian mixture model. Through dynamic background recognition and updating, the foreground insulating gas diffusion results (1-2, 2-2, 3-2) are extracted. The identified gas leakage area is further subjected to morphological filtering to remove noise interference and obtain the filtered gas diffusion results (1-3, 2-3, 3-3). Finally, the gas diffusion results are fused with the original infrared image and displayed with pseudo-color annotation to obtain the insulating gas leakage diffusion identification results.

[0101] Judging from the results of insulating gas diffusion identification, the method of the present invention effectively subtracts static background and larger dynamic targets (such as walking people and fluttering leaves), successfully captures the dynamic infrared signals caused by gas diffusion, and realizes the effective identification and enhanced display of insulating gas diffusion that cannot be distinguished by the naked eye.

[0102] Example 2

[0103] An electronic device includes a memory and a processor, wherein the memory is used to store a program that supports the processor to execute the image processing method in embodiment 1, and the processor is configured to execute the program stored in the memory.

[0104] Example 3

[0105] A storage medium stores a computer program, which, when executed by a processor, executes the steps of the image processing method in embodiment 1.

[0106] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A multi-component insulating gas leakage infrared detection system based on multiple detectors, characterized in that: include: A beam splitter (10), a visible light camera (11), a first flat beam splitter (12), a first filter (13), a first infrared focal plane detector (14), a second flat beam splitter (15), a second filter (16), a second infrared focal plane detector (17), and an image processing unit (18); the beam splitter (10), the first flat beam splitter (12), and the second flat beam splitter (15) are sequentially arranged along a horizontal direction; the angle between the reflection surface of the beam splitter (10) and the horizontal light path is 45°, The beam splitter (10) is coated with a semi-transparent and semi-reflective film for transmitting infrared light and reflecting visible light at the same time; the optical path of the visible light camera (11) is perpendicular to the horizontal optical path, and the visible light incident on the beam splitter (10) enters the visible light camera (11) after vertical reflection; the angle between the reflecting surface of the first flat beam splitter (12) and the second flat beam splitter (15) and the horizontal optical path is 45 degrees, and the function of the flat beam splitter is to reflect half of the incident infrared light and transmit half; the first infrared focal plane detector (1 4) The optical paths of the second infrared focal plane detector (17) are perpendicular to the horizontal optical path. The first filter (13) and the second filter (16) are respectively arranged on the lenses of the first infrared focal plane detector (14) and the second infrared focal plane detector (17). The first filter (13) and the second filter (16) are used to process the infrared spectrum information of different insulating gases to be measured respectively. The first filter (13) and the second filter (16) are used to process the infrared spectrum information of sulfur hexafluoride gas and perfluoroisobutyronitrile gas respectively. The infrared light passing through the beam splitter (10) is vertically reflected by the first flat plate spectroscope (12) and enters the first infrared focal plane detector (14) for detection. The infrared light passing through the first flat plate spectroscope (12) is vertically reflected by the second flat plate spectroscope (15) and enters the second infrared focal plane detector (17) for detection. The first infrared focal plane detector (14) and the second infrared focal plane detector (17) send the detected infrared spectrum signals to the image processing unit (18) for image processing.

2. The multi-component insulating gas leakage infrared detection system based on multiple detectors according to claim 1 is characterized in that: The coaxial registration method of the infrared focal plane detector and the visible light camera's field of view is as follows: (1) Move the entire system so that the central axis of the field of view of the infrared focal plane detector is aligned with the heating target plate; (2) Adjust the height of the visible light camera and the angle of the beam splitter so that the central axis of the visible light camera's field of view is aligned with the heating target plate.

3. An image processing method for the multi-detector-based multi-component insulating gas leakage infrared detection system according to any one of claims 1 to 2, comprising the following steps: Step 1: Initialize model parameters; Select continuous frame images of the infrared detector and initialize the mean matrix and variance matrix of each Gaussian mixture model; Step 2: Update model parameters; Step 2.1, obtain the video frame at time t from the real-time video of the infrared detector, and convert the gray value of the pixel point (x, y) into Matching judgment with the current Gaussian mixture model, if If it matches the jth model, then mark If there is no match with the jth model, mark ; Step 2.2: After the matching is completed, the weight parameters of the j-th Gaussian mixture model are calculated based on each tag. Make updates; Step 2.3: If there is a mark , then the mean and variance of the j-th Gaussian mixture model corresponding to the pixel are updated; if the pixel value If no match is found with any Gaussian mixture model, the pixel value is used Replace the mean of the corresponding position of the last Gaussian mixture model and redistribute the variance and weight; Step 3: Model motion detection; Step 3.1, normalize the weight of the Gaussian mixture model of each pixel; Step 3.2, follow Rearrange all Gaussian mixture models in order of size, and select the first B Gaussian mixture models that are greater than or equal to the weight threshold as the background; among them, Represents the variance matrix of the j-th Gaussian mixture model at time t; Step 3.3: Set the current pixel value Match it with the B background models obtained in step 3.2 to determine whether it is background information or motion foreground, and classify the pixels of background information and motion foreground, thereby enhancing the infrared image of insulating gas leakage.

4. The image processing method according to claim 3, wherein: The formulas for initializing the mean matrix and variance matrix of each Gaussian mixture model are as follows: Where, represents the mean matrix of the j-th Gaussian mixture model, represents the variance matrix of the j-th Gaussian mixture model, N represents the number of consecutive frame images selected, I(i) represents the video sequence, where i=1, 2, ..., N; during parameter initialization, the video sequence I(i) is cumulatively sampled. Indicates the number of mean matrices and variance matrices.

5. The image processing method according to claim 4, characterized in that The gray value obtained by the pixel point (x, y) The formula for matching with the current Gaussian mixture model is as follows: Where, is the mean matrix of the j-th Gaussian mixture model at time t-1, is the standard deviation matrix of the j-th Gaussian mixture model at time t-1, and c is the empirical parameter.

6. The image processing method according to claim 5, characterized in that The weight parameters of the j-th Gaussian mixture model according to each mark The update formula is as follows: Among them, α is the learning rate of weight, α∈[0,1], is the weight parameter of the j-th Gaussian mixture model at time t, is the weight parameter of the j-th Gaussian mixture model at time t-1, is the matching label of the j-th Gaussian mixture model.

7. The image processing method according to claim 6, characterized in that: The formula for updating the mean and variance of the j-th Gaussian mixture model corresponding to the pixel is as follows: Where β is the learning rate of the mean and variance, β∈[0,1], is the mean matrix of the j-th Gaussian mixture model at time t-1, is the grayscale value of the i-th frame image, is the variance matrix of the j-th Gaussian mixture model at time t-1.

8. The image processing method according to claim 7, wherein: The formula for selecting the first B Gaussian mixture models that are greater than or equal to the weight threshold as the background model is as follows: Where, τ is the weight threshold, is the minimum function.

9. The image processing method according to claim 8, characterized in that: The formula for matching judgment is as follows: in, is the mean matrix of the j-th Gaussian mixture model at time t.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the image processing method according to any one of claims 3 to 9 are executed.

Images