Petrochemical VOCs concentration pixel estimation method and device based on radiation model

Abstract

The invention discloses a radiation model-based petrochemical VOCs concentration pixel estimation method and device, and belongs to the technical field of concentration estimation. The petrochemical VOCs concentration pixel estimation method based on the radiation model comprises the following steps: acquiring a visible light smoke image and an infrared light smoke image corresponding to a petrochemical volatile organic compound; respectively calculating a visible light atmospheric light value corresponding to the visible light smoke image and an infrared light atmospheric light value corresponding to the infrared light smoke image; based on the visible light atmospheric light value and the infrared light atmospheric light value, respectively calculating a visible light transmittance graph corresponding to the visible light smoke image and an infrared light non-gas transmittance graph corresponding to the non-gas area; based on the infrared atmospheric light value, the visible light transmittance graph and the infrared non-gas transmittance graph, obtaining a radiation intensity image corresponding to the gas region; and inputting the radiation intensity image into the target concentration prediction model to obtain the concentration information of the petrochemical volatile organic compound output by the target concentration prediction model.

Description

Technical Field

[0001] The present application belongs to the technical field of concentration estimation, and in particular, relates to a method and device for estimating the pixel concentration of petrochemical VOCs based on a radiation model. Background Art

[0002] Petrochemical volatile organic compounds may be harmful to the human body. When measuring personnel in a petrochemical volatile organic compound environment, directly measuring the concentration of petrochemical volatile organic compounds affects the health of the measuring personnel. In the related art, the concentration of petrochemical volatile organic compounds is obtained mainly by emitting multi-spectral infrared waves from an infrared light source and inverting the column concentration. However, the above method has high hardware equipment cost and low measurement accuracy. Summary of the invention

[0003] The present application aims to solve at least one of the technical problems existing in the related art. To this end, the present application proposes a method and device for estimating the concentration pixel of petrochemical VOCs based on a radiation model, which obtains pixel-level concentration information of petrochemical volatile organic compounds, obtains more accurate concentration information, and does not require the use of expensive spectral equipment, effectively reducing the cost of concentration measurement.

[0004] In a first aspect, the present application provides a method for estimating the concentration of petrochemical VOCs pixels based on a radiation model, the method comprising:

[0005] Obtaining visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds; the infrared light smoke images include: non-gas areas and gas areas;

[0006] Respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image;

[0007] Based on the visible light atmospheric light value and the infrared light atmospheric light value, respectively calculating a visible light transmittance map corresponding to the visible light smoke image and an infrared light non-gas transmittance map corresponding to the non-gas area;

[0008] Based on the infrared light atmospheric light value, the visible light transmittance map, and the infrared light non-gas transmittance map, a radiation intensity image corresponding to the gas region is obtained;

[0009] Inputting the radiation intensity image into a target concentration prediction model to obtain concentration information of the petrochemical volatile organic compounds output by the target concentration prediction model;

[0010] The target concentration prediction model is trained by taking the sample radiation intensity image as a sample and taking the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

[0011] According to the method for pixel estimation of petrochemical VOCs concentration based on radiation model of the present application, by obtaining visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds, the visible light smoke images and infrared light smoke images are processed to obtain visible light transmittance maps and infrared light non-gas transmittance maps. The visible light transmittance maps and infrared light non-gas transmittance maps are used to indicate the processing process of obtaining the radiation intensity image of the gas area in the infrared light smoke image, and a more accurate radiation intensity image corresponding to the gas area excluding atmospheric interference is obtained. The radiation intensity image is regressed and predicted through the target concentration prediction model to obtain pixel-level concentration information of petrochemical volatile organic compounds, and more accurate concentration information is obtained without using expensive spectral equipment, effectively reducing the cost of concentration measurement.

[0012] According to the method for estimating the concentration pixel of petrochemical VOCs based on the radiation model of the present application, the visible light atmospheric light value corresponding to the visible light smog image and the infrared light atmospheric light value corresponding to the infrared light smog image are calculated respectively, including:

[0013] Processing the visible light smoke image and the infrared light smoke image respectively through a dark channel prior model to obtain a visible light dark channel image corresponding to the visible light smoke image and an infrared light dark channel image corresponding to the infrared light smoke image;

[0014] The maximum intensity value is selected from the visible light dark channel image and the infrared light dark channel image respectively to obtain the visible light atmospheric light value and the infrared light atmospheric light value.

[0015] According to the method for estimating the concentration pixel of petrochemical VOCs based on the radiation model of the present application, the visible light transmittance map corresponding to the visible light smoke image and the infrared light non-gas transmittance map corresponding to the non-gas area are calculated based on the visible light atmospheric light value and the infrared light atmospheric light value, respectively, including:

[0016] Based on the visible light atmospheric light value and the infrared light atmospheric light value, the dark channel prior model is processed to obtain a transmittance map calculation model;

[0017] The visible light transmittance map and the infrared light non-gas transmittance map are obtained based on the transmittance map calculation model, the visible light dark channel image, and the non-gas area in the infrared light dark channel image.

[0018] According to the method for estimating the concentration pixel of petrochemical VOCs based on the radiation model of the present application, the visible light transmittance map and the infrared light non-gas transmittance map are obtained based on the transmittance map calculation model, the visible light dark channel image and the non-gas area in the infrared light dark channel image, including:

[0019] Inputting the non-gas regions in the visible light dark channel image and the infrared light dark channel image into the transmittance map calculation model respectively, to obtain an initial visible light transmittance map and an initial infrared non-gas transmittance map output by the transmittance map calculation model;

[0020] Image enhancement processing is performed on the initial visible light transmittance map and the initial infrared light non-gas transmittance map respectively to obtain the visible light transmittance map and the infrared light non-gas transmittance map.

[0021] According to the method for estimating the concentration pixel of petrochemical VOCs based on the radiation model of the present application, the radiation intensity image corresponding to the gas area is obtained based on the infrared atmospheric light value, the visible light transmittance map and the infrared non-gas transmittance map, including:

[0022] Based on the visible light transmittance map and the infrared light non-gas transmittance map, constructing a first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region;

[0023] Based on the first association relationship model, interpolation processing is performed on each pixel point in the gas area to obtain an infrared light gas transmittance map corresponding to the gas area;

[0024] Based on the infrared gas transmittance map, the gas area in the infrared smoke image and the infrared atmospheric light value, a dark channel prior model is processed to obtain a radiation intensity image corresponding to the gas area in the infrared smoke image.

[0025] According to the method for estimating the concentration pixel of petrochemical VOCs based on the radiation model of the present application, the sample radiation intensity image includes a sample synthetic image and a sample real image; the target concentration prediction model includes a first layer, a second layer and a third layer, the input end of the second layer and the input end of the third layer are respectively connected to the output end of the first layer, and the output end of the second layer and the output end of the third layer are connected in parallel; the target concentration prediction model is trained based on the following steps:

[0026] Acquire a plurality of the sample radiation intensity images and target features corresponding to each of the sample radiation intensity images; the target features include: general features, gas relative path concentration features, and actual maximum path concentration features;

[0027] training the first layer based on the sample radiation intensity image and the common features corresponding to the sample radiation intensity image;

[0028] Training the second layer based on each of the sample composite images and the gas relative path concentration features corresponding to each of the sample composite images;

[0029] The third layer is trained based on each of the sample real images and the actual maximum path concentration features corresponding to each of the sample real images.

[0030] According to the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model of the present application, the infrared smoke image is obtained based on the following steps:

[0031] Acquiring an initial infrared smoke image corresponding to the petrochemical volatile organic compound;

[0032] Segmenting the initial infrared light smoke image to obtain a gas region in the initial infrared light smoke image and a non-gas region in the initial infrared light smoke image;

[0033] The gas region in the initial infrared light smoke image is determined as the gas region included in the infrared light smoke image; and the non-gas region in the initial infrared light smoke image is determined as the non-gas region included in the infrared light smoke image.

[0034] In a second aspect, the present application provides a device for estimating the concentration of petrochemical VOCs pixels based on a radiation model, the device comprising:

[0035] The first processing module is used to obtain a visible light smoke image and an infrared light smoke image corresponding to petrochemical volatile organic compounds; the infrared light smoke image includes: a non-gas area and a gas area;

[0036] A second processing module, used for respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image;

[0037] A third processing module, configured to calculate, based on the visible light atmospheric light value and the infrared light atmospheric light value, a visible light transmittance map corresponding to the visible light smoke image and an infrared light non-gas transmittance map corresponding to the non-gas area;

[0038] A fourth processing module, configured to obtain a radiation intensity image corresponding to the gas region based on the infrared atmospheric light value, the visible light transmittance map, and the infrared non-gas transmittance map;

[0039] a fifth processing module, configured to input the radiation intensity image into a target concentration prediction model to obtain the concentration information of the petrochemical volatile organic compounds output by the target concentration prediction model;

[0040] The target concentration prediction model is trained by taking the sample radiation intensity image as a sample and taking the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

[0041] According to the petrochemical VOCs concentration pixel estimation device based on the radiation model of the present application, by obtaining visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds, the visible light smoke images and infrared light smoke images are processed to obtain visible light transmittance maps and infrared light non-gas transmittance maps. The visible light transmittance maps and infrared light non-gas transmittance maps are used to indicate the processing process of obtaining the radiation intensity image of the gas area in the infrared light smoke image, and a more accurate radiation intensity image corresponding to the gas area excluding atmospheric interference is obtained. The radiation intensity image is regressed and predicted through the target concentration prediction model to obtain pixel-level concentration information of petrochemical volatile organic compounds, thereby obtaining more accurate concentration information without the use of expensive spectral equipment, effectively reducing the concentration measurement cost.

[0042] In a third aspect, the present application provides an electronic device comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein when the processor executes the computer program, the method for pixel estimation of petrochemical VOCs concentration based on the radiation model as described in the first aspect above is implemented.

[0043] In a fourth aspect, the present application provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for pixel estimation of petrochemical VOCs concentration based on a radiation model as described in the first aspect above.

[0044] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the method for pixel estimation of petrochemical VOCs concentration based on a radiation model as described in the first aspect above.

[0045] The above one or more technical solutions in the embodiments of the present application have at least one of the following technical effects:

[0046] Visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds are obtained, and the visible light smoke images and infrared light smoke images are processed to obtain visible light transmittance maps and infrared light non-gas transmittance maps. The visible light transmittance maps and infrared light non-gas transmittance maps are used to indicate the processing process of obtaining the radiation intensity image of the gas area in the infrared light smoke image, and a more accurate radiation intensity image corresponding to the gas area excluding atmospheric interference is obtained. The radiation intensity image is regressed and predicted through a target concentration prediction model to obtain pixel-level concentration information of petrochemical volatile organic compounds, thereby obtaining more accurate concentration information without using expensive spectral equipment and effectively reducing the cost of concentration measurement.

[0047] Furthermore, the dark channel prior model is processed through the visible light atmospheric light values ​​and the infrared light atmospheric light values ​​to obtain a transmittance map calculation model that can restore the image details, so that the visible light dark channel image and the infrared light dark channel image are respectively input into the transmittance map calculation model to obtain clearer and more realistic visible light transmittance map and infrared non-gas transmittance map, which provide precise image support for the subsequent prediction of concentration information and improve the accuracy of the predicted concentration information.

[0048] Furthermore, through the visible light transmittance map and the infrared non-gas transmittance map, a first correlation model between the visible light transmittance and the infrared light transmittance is constructed, and the transmittance change trend of the gas area of ​​the visible light smoke image is effectively used to guide the calculation process of the infrared light gas transmittance map of the gas area of ​​the infrared light smoke image, so as to obtain the infrared light gas transmittance map that retains the attenuation effect of petrochemical volatile organic compounds, effectively reduce the atmospheric interference on the gas path, and thus improve the image quality of the obtained radiation intensity image.

[0049] Furthermore, by acquiring multiple sample radiation intensity images and the target features corresponding to each sample radiation intensity image, a rich training set of the target concentration prediction model is effectively obtained. The first layer, the second layer and the third layer of the target concentration prediction model are trained respectively through the acquired training set, which effectively makes the target concentration prediction model suitable for the concentration prediction of petrochemical volatile organic compounds, improves the accuracy of concentration prediction of the target concentration prediction model, and provides pixel-level concentration estimation.

[0050] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the description of the embodiments in conjunction with the following drawings, in which:

[0052] Figure 1 This is one of the flow diagrams of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0053] Figure 2 This is one of the principle schematic diagrams of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0054] Figure 3 This is the second principle schematic diagram of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0055] Figure 4This is the third principle schematic diagram of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0056] Figure 5 This is the fourth principle schematic diagram of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0057] Figure 6 This is the second flow chart of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model provided in the embodiment of the present application;

[0058] Figure 7 It is a structural schematic diagram of a petrochemical VOCs concentration pixel estimation device based on a radiation model provided in an embodiment of the present application;

[0059] Figure 8 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0060] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0061] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, and are not used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable under appropriate circumstances, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship.

[0062] In combination with the accompanying drawings, the following detailed description is made of a method for estimating the concentration pixel of petrochemical VOCs based on a radiation model, a device for estimating the concentration pixel of petrochemical VOCs based on a radiation model, an electronic device and a readable storage medium provided in an embodiment of the present application through specific embodiments and their application scenarios.

[0063] Among them, the petrochemical VOCs concentration pixel estimation method based on the radiation model can be applied to the terminal, and can be specifically executed by hardware or software in the terminal.

[0064] The terminal includes but is not limited to portable communication devices such as mobile phones or tablet computers. It should also be understood that in some embodiments, the terminal may not be a portable communication device, but a desktop computer.

[0065] In the following various embodiments, a terminal including a display and a touch-sensitive surface is described. However, it should be understood that the terminal may include one or more other physical user interface devices such as a physical keyboard, a mouse and a joystick.

[0066] The embodiment of the present application provides a method for estimating the concentration of petrochemical VOCs pixels based on a radiation model. The execution subject of the method for estimating the concentration of petrochemical VOCs pixels based on a radiation model can be an electronic device or a functional module or functional entity in an electronic device that can implement the method for estimating the concentration of petrochemical VOCs pixels based on a radiation model. The electronic devices mentioned in the embodiment of the present application include but are not limited to mobile phones, tablet computers, computers, cameras, and wearable devices. The method for estimating the concentration of petrochemical VOCs pixels based on a radiation model provided in the embodiment of the present application is described below using an electronic device as an example of the execution subject.

[0067] like Figure 1 As shown, the petrochemical VOCs concentration pixel estimation method based on the radiation model includes: step 110, step 120, step 130, step 140 and step 150.

[0068] Step 110, obtaining a visible light smoke image and an infrared light smoke image corresponding to petrochemical volatile organic compounds; the infrared light smoke image includes: a non-gas area and a gas area;

[0069] In this step, petrochemical volatile organic compounds (petrochemical VOCs) are gases containing volatile organic substances.

[0070] Petrochemical volatile organic compounds have a small molecular weight, are easily vaporized under normal conditions, and are relatively active. They are a type of gas that is relatively active and may be harmful to human health.

[0071] Petrochemical volatile organic compounds may include: alkane gases, olefin gases, aromatic gases, alcohol gases, etc.

[0072] The specific categories of petrochemical volatile organic compounds can be determined based on actual conditions and are not limited in this application.

[0073] Visible light smog images are of petrochemical volatile organic compounds captured under visible light.

[0074] Visible light smoke images have higher resolution and contrast, and can better present information such as the outline and details of petrochemical volatile organic compounds.

[0075] Infrared smog images are of petrochemical volatile organic compounds captured under infrared light.

[0076] Infrared smoke images can reflect the temperature distribution on the surface of petrochemical volatile organic compounds. The temperature changes produced during the volatilization of petrochemical volatile organic compounds can be captured and presented through infrared images.

[0077] Visible light smoke images and infrared light smoke images can be acquired through image sensors.

[0078] The visible light smoke image acquired by the image sensor can be Figure 2 shown.

[0079] The infrared smoke image acquired by the image sensor can be Figure 3 shown.

[0080] The image sensor may be an integrated infrared and visible light multispectral image sensor.

[0081] The image sensor may also be a visible light image sensor and an infrared light image sensor.

[0082] The specific type of image sensor can be determined based on actual conditions and is not limited in this application.

[0083] The image sensor can be pre-set in a scene where petrochemical volatile organic compounds exist, and detect whether the petrochemical volatile organic compounds are leaking through gas sensors and odor sensors. When the leakage of petrochemical volatile organic compounds is detected, the image sensor is controlled to automatically capture images.

[0084] like Figure 4 As shown, the infrared image sensor can receive the atmospheric radiation, gas radiation and background attenuation radiation emitted or reflected by the VOCs air mass, and collect the infrared smog image.

[0085] In some embodiments, the image sensor can capture multiple visible light smoke images and multiple infrared light smoke images, and can screen out visible light smoke images and infrared light smoke images with higher image quality such as higher clarity and a larger proportion of gas in the picture from the multiple visible light smoke images and multiple infrared light smoke images.

[0086] It should be noted that the spectral absorption peak of petrochemical volatile organic compounds is located in the mid-wave infrared band of 3-5μm. By analyzing the spectral data in this band, the concentration of petrochemical volatile organic compounds can be inverted and calculated.

[0087] In the actual implementation process, in the scenario of petrochemical volatile organic compound leakage, passive infrared imaging and visible light multispectral cameras can be used to collect video frames in real time to obtain automatically aligned visible light smoke images and infrared light smoke images corresponding to the petrochemical volatile organic compounds of the same scene at the same time, which can be used to estimate the concentration information of petrochemical volatile organic compounds in the future.

[0088] In the actual implementation process, an image sensor can be set in the leaked petrochemical volatile organic compound scene to collect visible light smoke images and infrared light smoke images.

[0089] When the image sensors are a visible light image sensor and an infrared light image sensor respectively, the shooting angles of the visible light image sensor and the infrared light image sensor can be adjusted so that the shooting angles of the visible light image sensor and the infrared light image sensor are the same or close.

[0090] In the actual implementation process, after obtaining the visible light smoke image and the infrared light smoke image, the visible light smoke image and the infrared light smoke image can be aligned so that the imaging optical axes of the visible light smoke image and the infrared light smoke image are parallel, that is, the same pixel position corresponds to the same position of the petrochemical volatile organic compounds.

[0091] Step 120, respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image;

[0092] In this step, the visible light atmospheric light value is calculated based on the visible light smoke image, and represents the value of the atmospheric light intensity or background radiation intensity at an infinite distance.

[0093] The infrared atmospheric light value is calculated based on the infrared smog image, and represents the value of the atmospheric light intensity or background radiation intensity at infinite distance.

[0094] During the actual execution process, the pixel value of each pixel in the visible light smoke image can be processed through the dark channel prior model or the model based on the sky area, and the visible light atmospheric light value corresponding to the visible light smoke image can be calculated; the pixel value of each pixel in the infrared light smoke image can be processed, and the infrared light atmospheric light value corresponding to the infrared light smoke image can be calculated.

[0095] Of course, in the actual implementation process, visible light smoke images and infrared light smoke images can also be processed by any other feasible methods, such as image segmentation algorithms and machine learning algorithms, to obtain visible light atmospheric light values ​​and infrared light atmospheric light values, which is not limited in this application.

[0096] Step 130, based on the visible light atmospheric light value and the infrared light atmospheric light value, respectively calculating a visible light transmittance map corresponding to the visible light smoke image and an infrared light non-gas transmittance map corresponding to the non-gas area;

[0097] In this embodiment, the visible light transmittance image is an image describing the transmittance value corresponding to each pixel in the visible light smoke image.

[0098] The infrared non-gas transmittance image is an image that describes the transmittance value corresponding to each pixel point in the non-gas area in the infrared smoke image.

[0099] The transmittance is the ratio of the radiant energy projected onto the petrochemical volatile organic compounds through the petrochemical volatile organic compounds to the total radiant energy projected onto the petrochemical volatile organic compounds.

[0100] The transmittance map can identify the areas with weaker light in the image, enhance the areas with weaker light, and improve the overall visual effect of the image.

[0101] In the actual implementation process, the dark channel prior model can be processed by the visible light atmospheric light value, and the dark channel prior model after the visible light atmospheric light value processing can be transformed to obtain a visible light transmittance map.

[0102] The dark channel priori model can be processed by infrared light atmospheric light values, and data transformation can be performed on the dark channel priori model after the infrared light atmospheric light values ​​are processed to obtain an infrared light non-gas transmittance map.

[0103] Data transformation may include: division transformation and logarithmic transformation, etc.

[0104] The actual type of data transformation can be determined based on actual conditions and is not limited in this application.

[0105] Step 140, obtaining a radiation intensity image corresponding to the gas region based on the infrared atmospheric light value, the visible light transmittance map, and the infrared non-gas transmittance map;

[0106] In this step, the radiation intensity image can show the radiation intensity distribution of petrochemical volatile organic compounds in different areas.

[0107] Theoretically, the radiation intensity of petrochemical volatile organic compounds and the path concentration information of petrochemical volatile organic compounds satisfy the Lambert-Beer law.

[0108] That is, the higher the path concentration of petrochemical volatile organic compounds, the lower the radiation intensity; the lower the path concentration of petrochemical volatile organic compounds, the higher the radiation intensity.

[0109] The correlation between the radiation absorption intensity of petrochemical volatile organic compounds and the concentration information of petrochemical volatile organic compounds can be determined based on actual conditions, such as determining the correlation between the radiation absorption intensity of petrochemical volatile organic compounds and the concentration information of petrochemical volatile organic compounds based on influencing factors such as temperature, pressure and gas type.

[0110] It should be noted that the grayscale value of the gas area in the infrared smoke image is closely related to the concentration information of the gas area. Calculating the transmittance map of the gas area through the traditional dark channel prior model will change the grayscale value of the gas area and make the texture disappear.

[0111] In order to improve the accuracy of the transmittance map of the calculated gas area, the accuracy of the infrared gas transmittance map corresponding to the gas area in the infrared smoke image can be improved by using the transmittance change trend of the visible light gas area in the visible light transmittance map.

[0112] In the actual implementation process, after obtaining the infrared light gas transmittance map, the infrared light gas transmittance map and the infrared light atmospheric light value are used to process the dark channel prior model to obtain the radiation intensity image corresponding to the gas area.

[0113] In the actual implementation process, after obtaining the infrared light gas transmittance map, the infrared light gas transmittance map can be processed based on the spectral characteristics of the infrared light to calculate the radiation intensity image corresponding to the gas area.

[0114] Step 150: inputting the radiation intensity image into a target concentration prediction model to obtain concentration information of petrochemical volatile organic compounds output by the target concentration prediction model;

[0115] In this step, the target concentration prediction model is a model that predicts the concentration information of petrochemical volatile organic compounds in the input image based on the input image.

[0116] The target concentration prediction model can be a convolutional neural network model, a machine learning model, an artificial intelligence model, etc.

[0117] The concentration information is the gas concentration of petrochemical volatile organic compounds in the predicted radiation intensity image.

[0118] The concentration information may be pixel-level concentration information.

[0119] Pixel-level concentration information is the light intensity or color intensity information carried by each pixel in the radiation intensity image, which is used to characterize the concentration information of petrochemical volatile organic compounds.

[0120] The concentration information may be an actual concentration value or a visualized image.

[0121] like Figure 5 As shown, when the concentration information is a visualized image, different concentrations may correspond to different colors.

[0122] The target concentration prediction model is trained using the sample radiation intensity image as a sample and the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

[0123] In the actual implementation process, after obtaining the radiation intensity image, the radiation intensity image can be input into the target concentration prediction model. The target concentration prediction model can predict the concentration information of petrochemical volatile organic compounds by analyzing and processing the radiation intensity image.

[0124] In the actual implementation process, the above-mentioned concentration prediction method can be used to quickly discover leaked petrochemical volatile organic compounds and the leakage of petrochemical volatile organic compounds, obtain pixel-level concentration information of leaked petrochemical volatile organic compounds, and enhance the efficiency and effectiveness of environmental safety management.

[0125] The inventors discovered during the research and development process that in the related technology, the concentration of petrochemical volatile organic compounds is obtained mainly by inverting column concentration using an infrared light source to emit multi-spectral infrared waves; the above method has high hardware equipment costs and low measurement accuracy.

[0126] In addition, the inventors also found that the concentration of petrochemical volatile organic compounds obtained by the method of emitting multi-spectral infrared waves through an infrared light source and inverting column concentration is a single-point gas concentration, which has less concentration information and lower accuracy.

[0127] The present application obtains visible light smoke images and infrared light smoke images of petrochemical volatile organic compounds under visible light and infrared light, effectively capturing the contours and temperature distribution information of petrochemical volatile organic compounds during the diffusion process, without requiring measurement personnel to be in the petrochemical volatile organic compound diffusion scene, thereby ensuring the safety of measurement personnel; by calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image, the absorption intensity of light by the petrochemical volatile organic compounds collected in the visible light smoke image and the infrared light smoke image is determined; and further based on the visible light atmospheric light value and the infrared light atmospheric light value, a visible light transmittance map and Infrared non-gas transmittance map is used to obtain an image that can indirectly reflect the concentration information of petrochemical volatile organic compounds, so that based on the infrared atmospheric light value, visible light transmittance map and infrared non-gas transmittance map, the gas area in the infrared smog image is processed to obtain a more accurate radiation intensity image corresponding to the gas area, so that the target concentration prediction model can predict the concentration information of petrochemical volatile organic compounds based on the radiation intensity image that accurately retains the characteristics of petrochemical volatile organic compounds, obtain pixel-level concentration information, simplify the operation process, and obtain more accurate, higher resolution and more informative concentration information without the need to use expensive spectral equipment, effectively reducing the cost of concentration measurement.

[0128] According to the method for estimating the pixel concentration of petrochemical VOCs based on the radiation model provided in the embodiment of the present application, by obtaining visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds, the visible light smoke images and infrared light smoke images are processed to obtain a visible light transmittance map and an infrared light non-gas transmittance map. The visible light transmittance map and the infrared light non-gas transmittance map are used to indicate the processing process of obtaining the radiation intensity image of the gas area in the infrared light smoke image, and a more accurate radiation intensity image corresponding to the gas area excluding atmospheric interference is obtained. The radiation intensity image is regressed and predicted by the target concentration prediction model to obtain pixel-level concentration information of petrochemical volatile organic compounds in the radiation intensity image, so as to obtain more accurate concentration information without using expensive spectral equipment, thereby effectively reducing the cost of concentration measurement.

[0129] In some embodiments, the infrared smoke image may be obtained based on the following steps:

[0130] Obtaining initial infrared smoke images corresponding to petrochemical volatile organic compounds;

[0131] Segmenting the initial infrared smoke image to obtain a gas region in the initial infrared smoke image and a non-gas region in the initial infrared smoke image;

[0132] The gas region in the initial infrared light smoke image is determined as the gas region included in the infrared light smoke image; and the non-gas region in the initial infrared light smoke image is determined as the non-gas region included in the infrared light smoke image.

[0133] In actual implementation, the initial infrared smoke image is an image of petrochemical volatile organic compounds captured by an image sensor under infrared illumination.

[0134] The gas region in the initial infrared smoke image is the foreground plume of petrochemical volatile organic compounds included in the initial infrared smoke image.

[0135] The non-gas region in the initial infrared smog image is a background region in the initial infrared smog image that does not include petrochemical volatile organic compounds.

[0136] During the actual execution process, after obtaining the initial infrared light smoke image, the initial infrared light smoke image can be segmented through the smoke segmentation model to obtain the foreground feather mask of the gas area in the initial infrared light smoke image and the non-gas area in the initial infrared light smoke image. The gas area in the initial infrared light smoke image is determined as the gas area included in the infrared light smoke image, and the non-gas area in the initial infrared light smoke image is determined as the non-gas area included in the infrared light smoke image.

[0137] The smoke segmentation model may include: a threshold-based segmentation model, a threshold-based segmentation model, and a fully convolutional neural network segmentation model, etc.

[0138] In the actual implementation process, a suitable smoke segmentation model can be selected based on the initial infrared smoke image features.

[0139] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, by performing image segmentation on the initial infrared light smog image, the gas area included in the infrared light smog image and the non-gas area included in the infrared light smog image are effectively obtained, which facilitates the subsequent different processing of the non-gas area and the gas area included in the infrared light smog image, improves the processing efficiency, and reduces invalid processing operations.

[0140] In some embodiments, step 120 may further include:

[0141] The visible light smoke image and the infrared light smoke image are processed respectively by the dark channel prior model to obtain a visible light dark channel image corresponding to the visible light smoke image and an infrared light dark channel image corresponding to the infrared light smoke image;

[0142] The maximum intensity value is selected from the visible light dark channel image and the infrared light dark channel image respectively to obtain the visible light atmospheric light value and the infrared light atmospheric light value.

[0143] In this embodiment, the dark channel prior model is based on an atmospheric scattering model, which describes the process of scattering and attenuation of light when propagating in the atmosphere.

[0144] The dark channel prior model can be used to dehaze visible light smoke images and infrared light smoke images to improve the image clarity and realism of visible light smoke images and infrared light smoke images.

[0145] The dark channel prior model can be determined based on the following formula:

[0146] I(x)=J(x)t(x)+A(1-t(x))

[0147] Among them, x is the position of the pixel in the smoke image; I(x) is the smoke image collected by the image sensor; J(x) is the radiation intensity at the target position, that is, the target image without atmospheric interference; t(x) is the transmittance, which is a manifestation of the light penetration ability; A is the atmospheric light value.

[0148] The visible light dark channel image is a dark channel image of the visible light smoke image obtained by calculating the minimum value of each pixel of the visible light smoke image in the three RGB channels.

[0149] The infrared dark channel image is a dark channel image of the infrared smoke image obtained by calculating the minimum value of the regional window of each pixel point of the infrared smoke image in the infrared channel.

[0150] The maximum intensity value is the value with the highest brightness or color intensity in the dark channel image.

[0151] According to the dark channel prior model, the visible light smoke image can be processed according to the following formula:

[0152]

[0153] The visible light dark channel image corresponding to the visible light smoke image is obtained.

[0154] Among them, I dark (x) is the visible light dark channel image; I(y) is the visible light smoke image; Ω(x) is a regional square window centered at the pixel position x; c is one of the three color channels r, g and b.

[0155] According to the dark channel prior model, the infrared smoke image can be processed according to the following formula:

[0156]

[0157] The infrared dark channel image corresponding to the infrared smoke image is obtained.

[0158] Among them, Idark (x) is an infrared dark channel image; I(y) is an infrared smoke image; Ω(x) is a square window centered at the pixel position x.

[0159] After obtaining the visible light dark channel image and the infrared light dark channel image, pixels with intensity heights of the front target number in the visible light dark channel image are selected, and the highest intensity value is selected from the pixels with intensity heights of the front target number as the visible light atmospheric light value.

[0160] Similarly, pixels with intensity heights of the number of front targets in the infrared dark channel image are selected, and the highest intensity value is selected from the pixels with intensity heights of the number of front targets as the infrared atmospheric light value.

[0161] The specific value of the target number can be determined based on actual conditions. For example, the target number can be 5 or 10, which is not limited in this application.

[0162] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, the visible light smoke image and the infrared light smoke image are processed through the dark channel prior model to effectively obtain the visible light dark channel image and the infrared light dark channel image, so as to select the maximum intensity value from the pixel intensities of the visible light dark channel image and the infrared light dark channel image, respectively, to obtain the visible light atmospheric light value and the infrared light atmospheric light value, effectively reflecting the brightest parts in the visible light smoke image and the infrared light smoke image, simplifying the calculation process of the atmospheric light value, and improving the robustness.

[0163] In some embodiments, step 130 may further include:

[0164] Based on the visible light atmospheric light value and the infrared light atmospheric light value, the dark channel prior model is processed to obtain the transmittance map calculation model;

[0165] Based on the transmittance map calculation model, the visible light dark channel image and the infrared light dark channel image, the visible light transmittance map and the infrared light non-gas transmittance map are obtained.

[0166] In this embodiment, the transmittance map calculation model is a model for calculating a transmittance map.

[0167] In the actual implementation process, the transmittance map calculation model can be obtained based on the following steps:

[0168] We can perform a minimum operation on both sides of the dark channel prior model equation to obtain

[0169]

[0170] Among them, y is the position of the pixel in the smoke image; I(y) is the smoke image collected by the image sensor; J(y) is the radiation intensity at the target position, that is, the target image without atmospheric interference; t(x) is the transmittance, which is a manifestation of the light penetration ability; A is the atmospheric light value.

[0171] Will

[0172]

[0173] Divide the atmospheric light value A on both sides of the equation and perform the minimum operation again to obtain:

[0174]

[0175] In the dark channel prior model, it is assumed that the pixel value of the dark channel image J(y) obtained by assuming that the image is not disturbed by atmospheric light is close to 0, that is: min y∈Ω(x) (min c∈{r，g，b} J(y))→0; min y∈Ω(x) (min c∈{r，g，b} J(y))→0; Substitute Can get

[0176] It should be noted that during the transformation process, the physical meaning of each letter remains unchanged, that is, y is the position of the pixel in the smoke image; I(y) is the smoke image collected by the image sensor; J(y) is the radiation intensity at the target position, that is, the target image without atmospheric interference; t(x) is the transmittance, which is a manifestation of the ability of light to penetrate; A is the atmospheric light value; c is one of the three color channels of r, g and b.

[0177] The atmospheric light value is a constant. Substituting it into the dark channel prior model, the transmittance map calculation model can be obtained, that is:

[0178] Where, t(x) is the transmittance diagram; I dark (x) is the dark channel image; A is the atmospheric light value.

[0179] In the actual implementation process, after obtaining the transmittance map calculation model, the visible light dark channel image and the visible light atmospheric light value can be substituted into the transmittance map calculation model to obtain the visible light transmittance map.

[0180] Similarly, after obtaining the infrared dark channel image, the infrared dark channel image can be segmented to obtain the infrared non-gas area dark channel image and the infrared gas area dark channel image. The infrared non-gas area dark channel image and the infrared atmospheric light value can be substituted into the transmittance map calculation model to obtain the infrared non-gas transmittance map.

[0181] In the actual implementation process, the infrared dark channel image can be segmented through the image segmentation model or the image segmentation tool.

[0182] The infrared light non-gas region dark channel image is an image that does not include a gas region in the infrared light dark channel image.

[0183] The infrared light gas region dark channel image is an image including the gas region in the infrared light dark channel image.

[0184] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, the dark channel prior model is processed through the visible light atmospheric light value and the infrared light atmospheric light value to obtain a transmittance map calculation model that can restore the image details, so that the visible light dark channel image and the infrared light dark channel image are respectively input into the transmittance map calculation model to obtain clearer and more realistic visible light transmittance map and infrared light non-gas transmittance map, thereby providing accurate image support for the subsequent prediction of concentration information and improving the accuracy of the predicted concentration information.

[0185] like Figure 6 As shown, in some embodiments, based on the transmittance map calculation model, the visible light dark channel image and the non-gas area in the infrared light dark channel image, obtaining the visible light transmittance map and the infrared light non-gas transmittance map may also include:

[0186] The non-gas regions in the visible light dark channel image and the infrared light dark channel image are respectively input into the transmittance map calculation model to obtain an initial visible light transmittance map and an initial infrared light non-gas transmittance map output by the transmittance map calculation model;

[0187] The initial visible light transmittance map and the initial infrared light non-gas transmittance map are respectively subjected to image enhancement processing to obtain a visible light transmittance map and an infrared light non-gas transmittance map.

[0188] In this embodiment, the initial visible light transmittance map is a transmittance map obtained by processing a visible light dark channel image with a transmittance map calculation model.

[0189] The initial infrared non-gas transmittance map is a transmittance map obtained by processing the transmittance map calculation model into a non-gas region in the infrared dark channel image.

[0190] Image enhancement is the operation of enhancing image features such as brightness, clarity, and color.

[0191] In the actual implementation process, different image enhancements may be performed on the initial visible light transmittance map and the initial infrared light non-gas transmittance map based on the actual image qualities of the initial visible light transmittance map and the initial infrared light non-gas transmittance map.

[0192] In the actual implementation process, image enhancement processing can be performed through multi-scale Retinex (MSR) algorithms, spatial domain enhancement algorithms, frequency domain enhancement algorithms, and generative adversarial network algorithms.

[0193] The following uses the multi-scale Retinex (MSR) algorithm to perform image enhancement processing on the initial visible light transmittance map and the initial infrared non-gas transmittance map respectively, and the process of obtaining the visible light transmittance map and the infrared non-gas transmittance map will be described.

[0194] Based on the Retinex theory, the luminance component (L(x)) of an image determines the overall brightness of an object, while the reflection component (R(x)) represents the color or texture of the object's surface.

[0195] The product of the two gives the final light I(x) received by the image sensor, that is, the brightness and color information in the image. Removing the illumination component in the image and only using the reflection component can enhance the true color and scene details of the image.

[0196] That is, the final light received by the image sensor can be determined based on the following formula:

[0197] I(x) = R(x) × L(x)

[0198] Among them, I(x) is the final light received by the image sensor; R(x) is the reflection component of the image; L(x) is the luminance component of the image.

[0199] Taking the logarithm of the above formula to separate the reflection component and the luminance component, we can get:

[0200] logI(x) = logR(x) + logL(x)

[0201] Among them, L(x) can be estimated by convolving the Gaussian function G(x) with the given image, (x).

[0202] The specific estimation method is as follows:

[0203] R SSR (x) = logI(x) - log[I(x) × G(x)]

[0204]

[0205] Among them, R SSR (x) is a single-scale image; σ is the scale factor; k is a value determined by normalization.

[0206] Usually, three scales are considered, including small scale 0 < σ < 500, medium scale 50 ≤ σ < 100, and large scale σ ≥ 100.

[0207] Multi-scale Retinex (MSR) can be obtained by weighted summation of multiple scales. The specific operation is as follows:

[0208]

[0209] Among them, R MSR (x) is the converted multi-scale image; n is the number of scales used; w n is the weight, the sum of all weights is equal to 1; R SSR (x) is a single-scale image.

[0210] Will Determine the MSR algorithm and get t MSR (x) = MSR(t(x)).

[0211] In the actual implementation process, the initial visible light transmittance map and the initial infrared light non-gas transmittance map can be input into t MSR (x)=MSR(t(x)); the initial visible light transmittance map and the initial infrared light non-gas transmittance map are subjected to detail enhancement processing to obtain a visible light transmittance map and an infrared light non-gas transmittance map.

[0212] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, by performing image enhancement processing on the initial visible light transmittance map and the initial infrared light non-gas transmittance map, the visual effects of the obtained visible light transmittance map and the infrared light non-gas transmittance map are effectively improved, and the readability and comprehensibility are enhanced.

[0213] In some embodiments, step 140 may further include:

[0214] Based on the visible light transmittance map and the infrared light non-gas transmittance map, construct a first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region;

[0215] Based on the first correlation model, interpolation processing is performed on each pixel point in the gas area to obtain an infrared light gas transmittance map corresponding to the gas area;

[0216] Based on the infrared gas transmittance map, the gas area in the infrared smoke image and the infrared atmospheric light value, the dark channel prior model is processed to obtain the radiation intensity image corresponding to the gas area in the infrared smoke image.

[0217] In this embodiment, the first correlation relationship model is a correlation relationship model of a visible light transmittance map obtained by transforming the infrared light transmittance map, or a correlation relationship model of a visible light transmittance map obtained by transforming the infrared light transmittance map.

[0218] The infrared gas transmittance image is an image that describes the transmittance value corresponding to each pixel point in the gas area in the infrared smoke image.

[0219] It should be noted that the grayscale value of the VOCs gas area in the infrared smoke image is closely related to its concentration. Using the traditional dark channel prior model to obtain the transmittance map of the VOCs gas area in the infrared smoke image will cause the grayscale of the VOCs gas to change and the texture to disappear.

[0220] According to the Lambert-Beer law, the total transmittance formula of the background radiation from the VOCs gas area reaching the image sensor can be obtained as follows:

[0221]

[0222] Among them, μ VOCs is the attenuation coefficient of VOCs gas; μ a is the attenuation coefficient caused by the atmosphere; L is the equivalent distance of the background radiation.

[0223] It can be seen that the transmittance components of the VOCs gas area in the infrared smoke image are relatively complex. If the infrared gas transmittance map obtained directly using the dark channel prior model is likely to remove the attenuation effect of VOCs, resulting in inaccurate subsequent concentration estimation.

[0224] The transmittance variation trend of the visible light gas area can be used to guide the interpolation algorithm for processing the transmittance image of the gas area in the infrared light smoke image, so as to calculate the infrared light gas transmittance map corresponding to the gas area in the infrared light smoke image.

[0225] The interpolation algorithm is a method for estimating the transmittance values ​​of unknown pixels through the transmittance values ​​of known pixels in the infrared smoke image.

[0226] In the process of calculating the transmittance image of the gas area in the infrared smoke image, the interpolation algorithm can estimate the transmittance of other wavelengths based on the transmittance data of known wavelengths.

[0227] Interpolation algorithms may include: Lagrange interpolation, polynomial interpolation, and piecewise linear interpolation.

[0228] In the actual implementation process, the first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region can be constructed through the following steps:

[0229] It should be noted that in the VOCs leakage scenario, the leakage point is usually a pipeline valve, etc. The distance from the VOCs gas to the background target is very short relative to the imaging path. It can be considered that the atmospheric path of the VOCs gas area is equal to the surrounding atmospheric path.

[0230] In the dark channel prior model, it is usually assumed that the ratio of atmospheric components in small-scale areas is uniform, that is:

[0231] C ir =kC vis

[0232] Among them, C ir is the density of atmospheric components that absorb infrared radiation, that is, the absorption coefficient corresponding to infrared light; C vis is the density of atmospheric components that absorb visible light radiation, that is, the visible light absorption coefficient corresponding to visible light; k is the first coefficient.

[0233] Further based on the calculation formula of the absorption coefficient:

[0234] The correlation between the absorption coefficient and gas density can be obtained when the temperature is fixed.

[0235] Among them, μ is the absorption coefficient; p is the gas pressure; k is the Boltzmann constant; T is the temperature, S(T) is the spectral line intensity of a single molecule per unit volume; f(T, p) is the spectral line profile function.

[0236] It should be noted that the gas pressure p can be equated with the gas density C. Based on the calculation formula of the absorption coefficient, μ is obtained. vis =mC vis

[0237] μ ir =nC ir

[0238] Among them, μ vis is the visible light absorption coefficient; C vis is the density of atmospheric components that absorb visible light radiation, i.e., the visible light absorption coefficient corresponding to visible light; m is the second coefficient; μ vis is the infrared light absorption coefficient; C ir is the density of atmospheric components that absorb infrared radiation, that is, the absorption coefficient corresponding to infrared light; n is the third coefficient.

[0239] Substituting the relationship between the absorption coefficient and the gas concentration into the transmittance formula of the Lambert-Beer law, we can obtain:

[0240]

[0241] Among them, t vis is the visible light transmittance; μ vis is the visible light absorption coefficient; C vis is the density of atmospheric components that absorb visible light radiation, i.e., the visible light absorption coefficient corresponding to visible light; L is the equivalent distance of background radiation; t ir is the infrared light transmittance; μ visis the infrared light absorption coefficient; C ir is the density of atmospheric components that absorb infrared radiation, that is, the absorption coefficient corresponding to infrared light.

[0242] Further build up vis and t ir The relationship between:

[0243]

[0244] C vis Substitute t ir : get

[0245]

[0246] in, is a constant; hence we get t ir and t vis The correlation relationship between them is obtained, that is, the first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region is obtained.

[0247] It should be noted that, in the non-gas region, one or more fitting λ values ​​in the visible light transmittance map and the infrared non-gas transmittance map obtained in step 130 may be used.

[0248] In the process of guiding the interpolation algorithm for processing the gas area transmittance image in the infrared smoke image through the transmittance change trend of the visible light gas area, the t values ​​of the four nearest neighboring non-VOCs pixels of each pixel in the VOCs area can be used by standard bilinear interpolation. ir The values ​​are weighted averaged according to their spatial positions.

[0249] For example, the four non-VOCs neighboring points around the interpolation point (x, y) are (x 1 ,y 1 )、(x 2 ,y 1 )、(x 1 ,y 2 ) and (x 2 ,y 2 ), the corresponding t ir t 11 ,t 21 ,t 12 and t 22 , then the bilinear interpolation formula can be determined based on the following formula:

[0250]

[0251] At the same time, the interpolation point t is treated by the first association model ir By estimating, we can get

[0252] Considering spatial proximity and t vis The changing trend of , using the weighted average method:

[0253]

[0254] Obtain infrared light gas transmittance diagram.

[0255] Among them, α is the weight coefficient, between 0 and 1; This is the infrared light gas transmittance diagram; It is the transmittance map obtained based on the bilinear interpolation formula.

[0256] It is an infrared gas transmittance map that combines the transmittance variation trend of the gas area in the visible light smoke image.

[0257] By substituting the infrared gas transmittance map, the infrared atmospheric light value and the gas area in the infrared smoke image into the dark channel prior model, a radiation intensity image excluding atmospheric interference can be obtained.

[0258] The radiation intensity image can be calculated based on the following formula:

[0259]

[0260] Where J(x) is the radiation intensity image; I(x) is the gas area in the infrared smoke image; is the infrared light gas transmittance diagram; A is the infrared light atmospheric light value.

[0261] like Figure 6 As shown, the visible light smoke image is first subjected to atmospheric interference feature extraction, and then the image after the atmospheric interference feature extraction is subjected to feature enhancement.

[0262] The infrared smoke sensor first performs smoke image segmentation, and then fuses the gas area and non-gas area included in the segmented infrared smoke image with the visible light smoke image after feature enhancement. Through the dark channel prior algorithm, the radiation intensity image is obtained for concentration information prediction.

[0263] By performing multiple detection operations on the radiation intensity image, the visualization results of the pixel-level concentration information of petrochemical volatile organic compounds in the infrared smoke image can be obtained.

[0264] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, a first correlation relationship model between the visible light transmittance and the infrared light transmittance is constructed through the visible light transmittance map and the infrared light non-gas transmittance map, and the transmittance change trend of the gas area of ​​the visible light smoke image is effectively used to guide the calculation process of the infrared light gas transmittance map of the gas area of ​​the infrared light smoke image, so as to obtain the infrared light gas transmittance map that retains the attenuation effect of petrochemical volatile organic compounds, effectively reduce the atmospheric interference on the gas path, and thus improve the image quality of the obtained radiation intensity image.

[0265] In some embodiments, the sample radiation intensity image may include a sample synthetic image and a sample real image; the target concentration prediction model may include a first layer, a second layer, and a third layer, the input end of the second layer and the input end of the third layer are respectively connected to the output end of the first layer, and the output end of the second layer and the output end of the third layer are connected in parallel; the target concentration prediction model may be trained based on the following steps:

[0266] Acquire multiple sample radiation intensity images and target features corresponding to each sample radiation intensity image;

[0267] Training the first layer based on the sample radiation intensity image and the common features corresponding to the sample radiation intensity image;

[0268] Training the second layer based on each sample synthetic image and the gas relative path concentration features corresponding to each sample synthetic image;

[0269] The third layer is trained based on the real images of each sample and the actual maximum path concentration features corresponding to the real images of each sample.

[0270] In this embodiment, the sample radiation intensity images are pre-collected radiation intensity images of different types of petrochemical volatile organic compounds.

[0271] In the actual implementation process, after obtaining sample visible light smoke images and sample infrared light smoke images of different types of sample petrochemical volatile organic compounds, the sample visible light smoke images and sample infrared light smoke images can be processed through the method of steps 110 to 140 to obtain a sample radiation intensity image.

[0272] The sample composite image is a simulated radiation intensity image generated by a computer algorithm.

[0273] The sample real image is a radiation intensity image obtained based on actual scene photography.

[0274] The target features are features in the sample radiation intensity image that are related to the predicted sample petrochemical volatile organic compound concentration information.

[0275] The target features include: general features, gas relative path concentration features, and actual maximum path concentration features.

[0276] Common features may include: color features and structural features, etc.

[0277] The relative path concentration characteristics of the gas are the relative concentration characteristics between the area where the sample petrochemical volatile organic compounds are located and the image sensor.

[0278] The actual maximum path concentration characteristic is the actual maximum path concentration value between the area where the sample petrochemical volatile organic compounds are located and the image sensor.

[0279] The first layer is used to extract common features of the sample radiation intensity image.

[0280] The second layer is used to learn the relationship between the background radiation and the gas relative path concentration characteristics in the sample radiation intensity image.

[0281] The third layer is used to predict the absolute concentration scaling factor based on the global background radiation information, including a global average pooling layer to extract global features. The global features are mapped to a scalar scaling factor through a fully connected layer. The relative concentration map is scaled by the scaling factor to obtain an absolute concentration with physical significance.

[0282] The target concentration prediction model can predict the relative concentration distribution through the second layer, predict the global scaling factor through the third layer, and then use the predicted global scaling factor to convert the relative concentration map into an absolute concentration map to obtain visualized concentration information.

[0283] In the actual implementation process, the sample radiation intensity image can be input into the first layer, the second layer and the third layer respectively to train the target concentration prediction model.

[0284] The following takes the target concentration prediction model as a multi-scale two-branch fully connected deep learning network as an example to illustrate the acquisition of the target concentration prediction model training set, the construction of the target concentration prediction model, and the training process of the target concentration prediction model.

[0285] First, acquisition of target concentration prediction model training set

[0286] In the actual implementation process, the sample radiation intensity image can be divided into two parts. One part is the background radiation data on the gas path simulated by the gas diffusion model. This part of the data has the relative distribution data of the gas relative path concentration characteristics, which is used to learn the relationship between the background radiation intensity and the gas relative path concentration characteristics; the other part is the methane gas leakage data of different concentrations collected under real experimental conditions. By controlling the methane concentration, the actual maximum path concentration characteristics of the densest point of the gas plume are obtained, which is used to explore the relationship between the overall background radiation and the maximum concentration on the gas path.

[0287] Blender software can generate various forms of smoke data based on the gas diffusion model (i.e., dark channel prior model). Therefore, Blender software is used to generate a gas dispersion model in 17 directions in the horizontal plane, including front, back, left, and right, to simulate VOCs gas, and the background radiation data on the VOCs gas path is generated by projecting it onto different backgrounds.

[0288] During the Blender rendering process, three-dimensional density data of the gas is generated. The density data of each point is regarded as the column density of the gas at that point. The gas path concentration is obtained by integrating along the imaging path of the image sensor according to the following gas path concentration formula:

[0289] C g =∫ 0 l CL g dl

[0290] Where: CL g is the gas column density; C g is the gas path integrated concentration; l is the gas path length; d is the gas density.

[0291] Density data in Blender has no actual physical meaning, so the C of each pixel is g The values ​​were normalized to [0,1] as relative concentration values.

[0292] In the process of collecting real images of samples, 10% concentration was used as the starting value, and the methane gas concentration was increased by 10% to generate ten groups of experimental gases with different concentrations. A circular nozzle was used to release the gas under different backgrounds, and an infrared camera was used to collect background radiation data through the gas.

[0293] If the gas concentration near the nozzle is the experimental gas concentration, combined with the diameter of the circular nozzle, the actual gas path concentration at the gas most concentrated point can be calculated using the same gas concentration path calculation formula.

[0294] The gas data rendered by Blender and the actually collected gas data were randomly divided into a training validation set and a test set at a ratio of 9:1, and then the training validation set was randomly divided into a training set and a validation set at a ratio of 9:1 to obtain the gas image-relative concentration distribution data set and the gas image-maximum concentrated point physical concentration data set.

[0295] Second, the construction of the target concentration prediction model

[0296] In order to learn the relative concentration distribution characteristics of gases and the physical concentration data characteristics of the most concentrated point, a multi-scale dual-branch fully connected deep learning network is designed.

[0297] The main framework of the multi-scale dual-branch fully connected deep learning network is based on the U-Net network, which consists of a contraction path and two expansion paths.

[0298] The basic U-Net network has the following structure:

[0299] 1) Contraction path (Encoder): Similar to a typical convolutional neural network, it is used to extract contextual information of the image; it contains a series of convolutional layers (usually 3x3 convolutions), each followed by a ReLU activation function; every few layers, a maximum pooling layer (2x2) is used for downsampling, gradually reducing the spatial dimension of the feature map while increasing the depth of the feature.

[0300] 2) Expansion path (Decoder): Use upsampling (usually transposed convolution) to increase the size of the feature map; concatenate (skip connections) the feature maps of the corresponding layers of the contraction path, which helps to combine high-resolution features and high-level semantic information.

[0301] The target concentration model is an improved multi-scale two-branch fully connected deep learning network structure. The target concentration model changes the encoder and decoder structure of Unet into a two-branch decoding structure of a shared encoder (i.e., the first layer), a relative concentration decoder (i.e., the second layer), and a scaling factor decoder (i.e., the third layer).

[0302] The shared encoder is used to extract common features of smoke images, using the standard U-Net encoder structure.

[0303] The relative concentration decoder focuses on learning the correlation between background radiation and gas relative path concentration features. It contains a standard U-Net decoder structure, and the last layer uses 1x1 convolution and Sigmoid activation function to output a relative concentration map with the same resolution as the input image.

[0304] The Sigmoid activation function can be determined based on the following formula:

[0305]

[0306] The output range of the Sigmoid activation function is between 0 and 1, showing an S-shaped curve shape, which makes the Sigmoid function suitable for tasks that require normalizing the output to a probability value.

[0307] The scaling factor decoder predicts the absolute concentration scaling factor based on the global background radiation information. It includes a global average pooling layer to extract global features. The global features are mapped to a scalar scaling factor through a fully connected layer. The relative concentration map is scaled by the scaling factor to obtain the absolute concentration with physical meaning.

[0308] In addition, a convolutional block attention module (CBAM) is added to the jump connection part of U-Net to help the model automatically learn to focus on the feature areas of different channels and spaces of important feature maps and improve the prediction accuracy.

[0309] Third, the training process of the target concentration prediction model

[0310] In the actual training process, sample synthetic images and sample real images are used to train the improved U-Net network in stages.

[0311] In the pre-training stage, the relative concentration decoder is trained using sample synthetic images. The synthesized sample synthetic images are input and the relative distribution map of gas relative path concentration features is output. The loss function uses the mean square error (MSE) to measure the difference between the predicted gas relative path concentration features and the actual maximum path concentration features.

[0312] The mean square error can be determined based on the following formula:

[0313]

[0314] Where MSE is the mean square error; yi is the actual value of the i-th sample; is the predicted value of the i-th sample; n is the number of sample synthetic images.

[0315] During the training of the relative concentration decoder, the scaling factor decoder remains frozen.

[0316] In the model fine-tuning stage, the collected sample real images are used, the input is the actual collected sample real images, and the output is the actual maximum path concentration feature of the most concentrated point as the scaling factor.

[0317] Freeze the encoder and relative concentration decoder, and adjust the scaling factor decoder to calibrate the accuracy of the absolute concentration predictions.

[0318] The predicted scaling factor, i.e., the concentration prediction feature of the most concentrated point, and the actual maximum path concentration feature of the real gas data are used to calculate the mean square error loss and train the scaling factor decoder.

[0319] During model inference, the relative concentration distribution is predicted by the relative concentration decoder, the global scaling factor is predicted by the scaling factor decoder, and then the relative concentration map is converted into an absolute concentration map using the predicted scaling factor.

[0320] The model reasoning process can be determined based on the following formula:

[0321] C absolute (x, y) = C relative(x,y)×Scaling

[0322] Among them, C absolute is the absolute concentration diagram; C relative is a relative concentration graph; Scaling is a scaling factor.

[0323] According to the petrochemical VOCs concentration pixel estimation method based on the radiation model provided in the embodiment of the present application, by acquiring multiple sample radiation intensity images and the target features corresponding to each sample radiation intensity image, a rich training set of the target concentration prediction model is effectively obtained. The first layer, the second layer and the third layer of the target concentration prediction model are trained respectively through the acquired training set, which effectively makes the target concentration prediction model suitable for the concentration prediction of petrochemical volatile organic compounds, improves the accuracy of concentration prediction of the target concentration prediction model, and provides pixel-level concentration estimation.

[0324] The method for estimating the concentration pixel of petrochemical VOCs based on the radiation model provided in the embodiment of the present application can be executed by a device for estimating the concentration pixel of petrochemical VOCs based on the radiation model. In the embodiment of the present application, the device for estimating the concentration pixel of petrochemical VOCs based on the radiation model is taken as an example to illustrate the device for estimating the concentration pixel of petrochemical VOCs based on the radiation model provided in the embodiment of the present application.

[0325] The embodiment of the present application also provides a petrochemical VOCs concentration pixel estimation device based on a radiation model.

[0326] like Figure 7 As shown, the petrochemical VOCs concentration pixel estimation device based on the radiation model includes: a first processing module 710, a second processing module 720, a third processing module 730, a fourth processing module 740 and a fifth processing module 750.

[0327] The first processing module 710 is used to obtain a visible light smoke image and an infrared light smoke image corresponding to petrochemical volatile organic compounds; the infrared light smoke image includes: a non-gas area and a gas area;

[0328] The second processing module 720 is used to respectively calculate the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image;

[0329] The third processing module 730 is used to calculate the visible light transmittance map corresponding to the visible light smoke image and the infrared light non-gas transmittance map corresponding to the non-gas area based on the visible light atmospheric light value and the infrared light atmospheric light value;

[0330] A fourth processing module 740 is used to obtain a radiation intensity image corresponding to the gas area based on the infrared atmospheric light value, the visible light transmittance map and the infrared non-gas transmittance map;

[0331] The fifth processing module 750 is used to input the radiation intensity image into the target concentration prediction model to obtain the concentration information of petrochemical volatile organic compounds output by the target concentration prediction model;

[0332] The target concentration prediction model is trained using the sample radiation intensity image as a sample and the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

[0333] According to the petrochemical VOCs concentration pixel estimation device based on the radiation model provided in the embodiment of the present application, by

[0334] In some embodiments, the second processing module 720 may also be used to:

[0335] The visible light smoke image and the infrared light smoke image are processed respectively by the dark channel prior model to obtain a visible light dark channel image corresponding to the visible light smoke image and an infrared light dark channel image corresponding to the infrared light smoke image;

[0336] The maximum intensity value is selected from the visible light dark channel image and the infrared light dark channel image respectively to obtain the visible light atmospheric light value and the infrared light atmospheric light value.

[0337] In some embodiments, the third processing module 730 may also be used to:

[0338] Based on the visible light atmospheric light value and the infrared light atmospheric light value, the dark channel prior model is processed to obtain the transmittance map calculation model;

[0339] Based on the transmittance map calculation model, the visible light dark channel image and the non-gas area in the infrared light dark channel image, a visible light transmittance map and an infrared light non-gas transmittance map are obtained.

[0340] In some embodiments, the third processing module 730 may also be used to:

[0341] The non-gas regions in the visible light dark channel image and the infrared light dark channel image are respectively input into the transmittance map calculation model to obtain an initial visible light transmittance map and an initial infrared light non-gas transmittance map output by the transmittance map calculation model;

[0342] The initial visible light transmittance map and the initial infrared light non-gas transmittance map are respectively subjected to image enhancement processing to obtain a visible light transmittance map and an infrared light non-gas transmittance map.

[0343] In some embodiments, the fourth processing module 740 may also be used to:

[0344] Based on the visible light transmittance map and the infrared non-gas transmittance map, construct a first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region;

[0345] Based on the first correlation relationship model, perform interpolation processing on each pixel point in the gas region to obtain the infrared gas transmittance map corresponding to the gas region;

[0346] Based on the infrared gas transmittance map, the gas region in the infrared smoke image, and the infrared atmospheric light value, process the dark channel prior model to obtain the radiation intensity image corresponding to the gas region in the infrared smoke image.

[0347] In some embodiments, the sample radiation intensity image includes a sample synthesis image and a sample real image; the target concentration prediction model includes a first layer, a second layer, and a third layer. The input end of the second layer and the input end of the third layer are respectively connected to the output end of the first layer, and the output end of the second layer and the output end of the third layer are in parallel; the device may further include a sixth processing module for:

[0348] Obtain a variety of sample radiation intensity images and the target features corresponding to each sample radiation intensity image; the target features include: general features, gas relative path concentration features, and actual maximum path concentration features;

[0349] Based on the sample radiation intensity image and the general features corresponding to the sample radiation intensity image, train the first layer;

[0350] Based on each sample synthesis image and the gas relative path concentration features corresponding to each sample synthesis image, train the second layer;

[0351] Based on each sample real image and the actual maximum path concentration features corresponding to each sample real image, train the third layer.

[0352] In some embodiments, the device may further include a seventh processing module for:

[0353] Obtain the initial infrared smoke image corresponding to the petrochemical volatile organic compound;

[0354] Segment the initial infrared smoke image to obtain the gas region in the initial infrared smoke image and the non-gas region in the initial infrared smoke image;

[0355] Determine the gas region in the initial infrared smoke image as the gas region included in the infrared smoke image; determine the non-gas region in the initial infrared smoke image as the non-gas region included in the infrared smoke image.

[0356] The petrochemical VOCs concentration pixel estimation device based on the radiation model in the embodiment of the present application can be an electronic device, or a component in an electronic device, such as an integrated circuit or a chip. The electronic device can be a terminal, or it can be other devices other than a terminal. Exemplarily, the electronic device can be a mobile phone, a tablet computer, a laptop computer, a PDA, a car electronic device, a mobile Internet device (Mobile Internet Device, MID), an augmented reality (augmentedreality, AR) / virtual reality (virtual reality, VR) device, a robot, a wearable device, an ultra-mobile personal computer (ultra-mobile personal computer, UMPC), a netbook or a personal digital assistant (personaldigital assistant, PDA), etc., and can also be a server, a network attached storage (Network AttachedStorage, NAS), a personal computer (personal computer, PC), a television (television, TV), a teller machine or a self-service machine, etc., which is not specifically limited in the embodiment of the present application.

[0357] The device for estimating the concentration of petrochemical VOCs pixels based on the radiation model in the embodiment of the present application may be a device having an operating system. The operating system may be an Android operating system, an IOS operating system, or other possible operating systems, which are not specifically limited in the embodiment of the present application.

[0358] The petrochemical VOCs concentration pixel estimation device based on the radiation model provided in the embodiment of the present application can achieve Figures 1 to 6 To avoid repetition, the various processes implemented by the method embodiment are not described here.

[0359] In some embodiments, Figure 8 As shown, the embodiment of the present application also provides an electronic device 800, including a processor 801, a memory 802, and a computer program stored in the memory 802 and executable on the processor 801. When the program is executed by the processor 801, each process of the above-mentioned embodiment of the method for estimating the concentration of petrochemical VOCs pixels based on the radiation model is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0360] It should be noted that the electronic devices in the embodiments of the present application include the mobile electronic devices and non-mobile electronic devices mentioned above.

[0361] The embodiment of the present application also provides a non-transitory computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the various processes of the above-mentioned petrochemical VOCs concentration pixel estimation method embodiment based on the radiation model are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0362] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0363] An embodiment of the present application also provides a computer program product, including a computer program, which, when executed by a processor, implements the above-mentioned petrochemical VOCs concentration pixel estimation method based on the radiation model.

[0364] The processor is the processor in the electronic device described in the above embodiment. The readable storage medium includes a computer readable storage medium, such as a computer read-only memory ROM, a random access memory RAM, a magnetic disk or an optical disk.

[0365] An embodiment of the present application further provides a chip, which includes a processor and a communication interface, wherein the communication interface is coupled to the processor, and the processor is used to run programs or instructions to implement the various processes of the above-mentioned embodiment of the petrochemical VOCs concentration pixel estimation method based on the radiation model, and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0366] It should be understood that the chip mentioned in the embodiments of the present application can also be called a system-level chip, a system chip, a chip system or a system-on-chip chip, etc.

[0367] It should be noted that, in this article, the terms "comprise", "include" or any other variant thereof are intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise one..." do not exclude the presence of other identical elements in the process, method, article or device including the element. In addition, it should be pointed out that the scope of the methods and devices in the embodiments of the present application is not limited to performing functions in the order shown or discussed, and may also include performing functions in a substantially simultaneous manner or in reverse order according to the functions involved, for example, the described method may be performed in an order different from that described, and various steps may also be added, omitted, or combined. In addition, the features described with reference to certain examples may be combined in other examples.

[0368] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the relevant technology, can be embodied in the form of a computer software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0369] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

[0370] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials, or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0371] Although the embodiments of the present application have been shown and described, those skilled in the art will appreciate that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present application, and that the scope of the present application is defined by the claims and their equivalents.

Claims

1. A method for estimating the concentration of petrochemical VOCs pixels based on a radiation model, characterized in that: include: Obtain visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds; The infrared smoke image includes: a non-gas area and a gas area; Respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image; Based on the visible light atmospheric light value and the infrared light atmospheric light value, respectively calculating a visible light transmittance map corresponding to the visible light smoke image and an infrared light non-gas transmittance map corresponding to the non-gas area; Based on the infrared light atmospheric light value, the visible light transmittance map, and the infrared light non-gas transmittance map, a radiation intensity image corresponding to the gas region is obtained; Inputting the radiation intensity image into a target concentration prediction model to obtain concentration information of the petrochemical volatile organic compounds output by the target concentration prediction model; The target concentration prediction model is trained by taking the sample radiation intensity image as a sample and taking the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

2. The method for estimating the concentration of petrochemical VOCs pixels based on the radiation model according to claim 1 is characterized in that: The respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image comprises: Processing the visible light smoke image and the infrared light smoke image respectively through a dark channel prior model to obtain a visible light dark channel image corresponding to the visible light smoke image and an infrared light dark channel image corresponding to the infrared light smoke image; The maximum intensity value is selected from the visible light dark channel image and the infrared light dark channel image respectively to obtain the visible light atmospheric light value and the infrared light atmospheric light value.

3. The method for estimating the concentration of petrochemical VOCs pixels based on the radiation model according to claim 2 is characterized in that: The method of calculating the visible light transmittance map corresponding to the visible light smoke image and the infrared light non-gas transmittance map corresponding to the non-gas area based on the visible light atmospheric light value and the infrared light atmospheric light value includes: Based on the visible light atmospheric light value and the infrared light atmospheric light value, the dark channel prior model is processed to obtain a transmittance map calculation model; The visible light transmittance map and the infrared light non-gas transmittance map are obtained based on the transmittance map calculation model, the visible light dark channel image, and the non-gas area in the infrared light dark channel image.

4. The method for estimating the concentration of petrochemical VOCs pixels based on the radiation model according to claim 3 is characterized in that: The method of obtaining the visible light transmittance map and the infrared light non-gas transmittance map based on the transmittance map calculation model, the visible light dark channel image, and the non-gas area in the infrared light dark channel image comprises: Inputting the non-gas regions in the visible light dark channel image and the infrared light dark channel image into the transmittance map calculation model respectively, to obtain an initial visible light transmittance map and an initial infrared non-gas transmittance map output by the transmittance map calculation model; Image enhancement processing is performed on the initial visible light transmittance map and the initial infrared light non-gas transmittance map respectively to obtain the visible light transmittance map and the infrared light non-gas transmittance map.

5. The method for estimating the concentration of petrochemical VOCs in pixels based on a radiation model according to any one of claims 1 to 4, characterized in that: The obtaining of the radiation intensity image corresponding to the gas area based on the infrared light atmospheric light value, the visible light transmittance map and the infrared light non-gas transmittance map comprises: Based on the visible light transmittance map and the infrared light non-gas transmittance map, constructing a first correlation relationship model between the visible light transmittance and the infrared light transmittance in the gas region; Based on the first association relationship model, interpolation processing is performed on each pixel point in the gas area to obtain an infrared light gas transmittance map corresponding to the gas area; Based on the infrared gas transmittance map, the gas area in the infrared smoke image and the infrared atmospheric light value, a dark channel prior model is processed to obtain a radiation intensity image corresponding to the gas area in the infrared smoke image.

6. The method for estimating the concentration of petrochemical VOCs in pixels based on a radiation model according to any one of claims 1 to 4, characterized in that: The sample radiation intensity image includes a sample synthetic image and a sample real image; the target concentration prediction model includes a first layer, a second layer and a third layer, the input end of the second layer and the input end of the third layer are respectively connected to the output end of the first layer, and the output end of the second layer and the output end of the third layer are connected in parallel; the target concentration prediction model is trained based on the following steps: Acquire a plurality of the sample radiation intensity images and target features corresponding to each of the sample radiation intensity images; the target features include: general features, gas relative path concentration features, and actual maximum path concentration features; training the first layer based on the sample radiation intensity image and the common features corresponding to the sample radiation intensity image; Training the second layer based on each of the sample composite images and the gas relative path concentration features corresponding to each of the sample composite images; The third layer is trained based on each of the sample real images and the actual maximum path concentration features corresponding to each of the sample real images.

7. The method for estimating the concentration of petrochemical VOCs in pixels based on a radiation model according to any one of claims 1 to 4, characterized in that: The infrared smoke image is obtained based on the following steps: Acquiring an initial infrared smoke image corresponding to the petrochemical volatile organic compound; Segmenting the initial infrared light smoke image to obtain a gas region in the initial infrared light smoke image and a non-gas region in the initial infrared light smoke image; The gas region in the initial infrared light smoke image is determined as the gas region included in the infrared light smoke image; and the non-gas region in the initial infrared light smoke image is determined as the non-gas region included in the infrared light smoke image.

8. A device for estimating the concentration of petrochemical VOCs pixels based on a radiation model, characterized in that: include: The first processing module is used to obtain visible light smoke images and infrared light smoke images corresponding to petrochemical volatile organic compounds; The infrared smoke image includes: a non-gas area and a gas area; A second processing module, used for respectively calculating the visible light atmospheric light value corresponding to the visible light smoke image and the infrared light atmospheric light value corresponding to the infrared light smoke image; A third processing module, configured to calculate, based on the visible light atmospheric light value and the infrared light atmospheric light value, a visible light transmittance map corresponding to the visible light smoke image and an infrared light non-gas transmittance map corresponding to the non-gas area; A fourth processing module, configured to obtain a radiation intensity image corresponding to the gas region based on the infrared atmospheric light value, the visible light transmittance map, and the infrared non-gas transmittance map; a fifth processing module, configured to input the radiation intensity image into a target concentration prediction model to obtain the concentration information of the petrochemical volatile organic compounds output by the target concentration prediction model; The target concentration prediction model is trained by taking the sample radiation intensity image as a sample and taking the sample concentration information corresponding to the sample petrochemical volatile organic compounds included in the sample radiation intensity image as a sample label.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the method for estimating the petrochemical VOCs concentration pixel based on the radiation model as described in any one of claims 1 to 7 is implemented.

10. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for estimating the concentration of petrochemical VOCs pixels based on a radiation model as described in any one of claims 1 to 7 is implemented.

11. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for estimating the petrochemical VOCs concentration pixel based on a radiation model as described in any one of claims 1 to 7 is implemented.

Images