Regional atmospheric pollution source distribution and influence evaluation system

Abstract

The invention discloses a regional atmospheric pollution source distribution and influence evaluation system, and relates to the technical field of atmospheric pollution monitoring and governance, and the system recognizes the spatial distribution of agricultural wastes such as straws, withered grass and livestock excrement through a remote sensing technology, can precisely recognize the spatial distribution characteristics of pollution sources, and improves the reliability of the pollution sources. And based on static analysis and dynamic monitoring means, the combustion emission characteristics and diffusion influence of the pollution source are comprehensively evaluated. And a scientific and reasonable pollution prevention and control strategy is formulated according to different pollution risk levels. Besides, the temperature acquisition unit and the thermal inversion layer identification unit identify a thermal inversion phenomenon in an area, calculate a thermal inversion layer effect factor, a vertical wind shear factor and a rainfall washing effect factor in combination with wind speed data and rainfall data, and calculate dynamic diffusion coefficient classification evaluation in combination with a dimensionless processing method. And a targeted regulation and control strategy is generated, so that the accuracy of pollution diffusion prediction is further improved.

Description

Technical Field

[0001] The present invention relates to the technical field of air pollution monitoring and control, and in particular to a regional air pollution source distribution and impact assessment system. Background Art

[0002] With the acceleration of industrialization, agricultural pollution has gradually attracted attention. In particular, the disposal of straw, dead grass, and livestock manure in large-scale agricultural production is increasingly showing its potential impact on the environment. In areas with dense agricultural production, farmers often resort to incineration to dispose of these agricultural wastes. However, the harmful gases and particulate matter released during incineration cause serious atmospheric pollution, affecting air quality and endangering the ecological environment and human health.

[0003] However, existing technologies have many shortcomings for identifying and assessing pollution sources in agricultural areas. First, traditional pollution source monitoring relies primarily on manual inspections and ground-based monitoring stations, which cannot obtain extensive and accurate spatial data in real time, resulting in incomplete identification of pollution sources. This is especially true for the burning of agricultural waste, such as straw, dead grass, and livestock manure, as their spatial distribution and combustion conditions are difficult to accurately monitor. Furthermore, many agricultural areas, particularly in rural or remote areas, lack effective technical means to monitor these pollution sources in real time, making it difficult to fully assess the specific impact of the burning process on air pollution.

[0004] At the same time, in some regions, changes in meteorological conditions, such as the formation of inversion layers, can inhibit the diffusion of atmospheric pollutants, causing them to stagnate at low altitudes and further exacerbate the cumulative effects of pollution. However, the emergence and impact of inversion layers are often difficult to monitor and assess in real time. Inversion layers can cause pollutants to remain in the lower atmosphere for extended periods, preventing them from effectively dispersing. This significantly impacts the transport and diffusion of atmospheric pollutants, leading to deterioration of air quality in certain areas. Therefore, identifying and analyzing the effects of inversion layers and integrating this with the distribution of atmospheric pollution sources has become a major challenge in the field of atmospheric pollution monitoring. Summary of the Invention

[0005] In view of the deficiencies of the prior art, the present invention provides a regional air pollution source distribution and impact assessment system to solve the problems mentioned in the background technology.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: A regional air pollution source distribution and impact assessment system, comprising:

[0007] The remote sensing identification module is used to divide the target area into several monitoring sub-areas and identify pollution sources in the monitoring sub-areas through remote sensing technology. It obtains the spatial distribution data of pollution source clusters within the monitoring sub-areas, including dead grass clusters, straw clusters, and livestock manure clusters.

[0008] The pollution source static analysis module is used to build a prediction pollution model using a convolutional neural network based on the pollution source aggregation point data obtained by the remote sensing identification module, analyze the distribution characteristics of each type of pollution source, calculate the pollution potential of each type of pollution source in the monitoring sub-area based on the distribution characteristics, and generate the corresponding pollution source distribution map. It simulates the pollution generated during the incineration process of each pollution source and predicts the combustion pollution coefficient Rs of the i-th monitoring sub-area. i , and Rs i Perform classification, output a first classification result, and generate a corresponding strategy based on the first classification result;

[0009] The dynamic monitoring module collects temperature data at different heights of the i-th monitoring sub-area after the corresponding strategy is executed, draws the temperature-height profile of the i-th monitoring sub-area, identifies the inversion layer, and collects wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area. r , establish a data set of thermospheric effects;

[0010] Dynamic analysis module, used to construct the inversion layer effect factor ITEF of the i-th monitoring sub-area based on the thermosphere effect data set i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i , and correlate to obtain the dynamic diffusion coefficient Ks of the i-th monitoring sub-region i , and evaluate to output the second classification result and generate the corresponding strategy.

[0011] Preferably, the remote sensing identification module includes a region segmentation unit, a remote sensing data acquisition unit, an image preprocessing unit and a pollution source identification unit;

[0012] The region segmentation unit is used to collect and identify the geographical features and landforms of the target area using the GIS geographic information system, establish a three-dimensional model, and divide the target area into several monitoring sub-areas, which are marked as {Zq1, Zq2, Zq3, ..., Zq n}; n represents the total number of monitoring sub-areas;

[0013] The remote sensing data acquisition unit is used to acquire multi-band images of the target area using a multispectral camera mounted on an unmanned aerial vehicle, and to establish a multi-band image data set, wherein the multi-band image data set includes multi-band images including visible light, near infrared, and short-wave infrared;

[0014] Image preprocessing unit, used to perform image registration on multi-band image data sets, align the data of each band, perform radiation correction, and remove noise using mean filtering, Gaussian filtering, and wavelet transform;

[0015] The pollution source identification unit is used to extract spectral index features from the multi-band image data set and generate distribution maps of different pollution types.

[0016] Preferably, the specific steps of generating distribution maps of different pollution types include:

[0017] Step Aa.1: Calculate the Normalized Difference Vegetation Index (NDVI), which is used to distinguish between green vegetation and dead grass. The expression is as follows:

[0018]

[0019] Wherein, NIR represents the value of the near-infrared band, which indicates the health of plants. Healthy vegetation has a high NIR reflectivity. Red represents the value of the infrared band, which indicates the absorption characteristics of plants. Healthy vegetation has a high absorption rate in the red light band.

[0020] Step Aa.2: Preliminary screening of dead grass areas: When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.5-0.9, it is identified as a healthy vegetation area and marked as a green area;

[0021] When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.11-0.3, it is identified as a dead grass area and marked as a red area;

[0022] When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.05-0.1, it is identified as a straw area and marked as the first gray area; when the NDVI is between 0.05-0.1, it means that the vegetation has completely lost its vitality but still retains a certain organic structure;

[0023] When the Normalized Difference Vegetation Index (NDVI) is within the range of -0.2-0.09, it is identified as a bare soil area and marked as a brown area;

[0024] When the Normalized Difference Vegetation Index (NDVI) is less than -0.1, it is identified as a water area and marked as an orange area;

[0025] The normalized difference vegetation index (NDVI) value of each pixel was calculated, and then the areas with 0.11≤NDVI≤0.3 were selected as dead grass areas, and those with 0.05≤NDVI≤0.1 were selected as straw areas, and healthy vegetation points with NDVI>0.5 were excluded.

[0026] Step Aa.3: Select the kth window in the i-th monitoring sub-area in the image data set, with a window size of 3×3 or 5×5; calculate the frequency of occurrence of the grayscale values ​​of each pair of pixels in the kth window; the grayscale value of each pair of pixels includes the first grayscale value I of the h-th pixel point h and the second grayscale value I of the f-th pixel f ;

[0027] Record the first grayscale value I of the hth pixel h and the second grayscale value I of the f-th pixel f The frequency I that appears in the kth window f , and normalize the frequency value to the probability value to obtain the gray-level symbiotic combination frequency P(I h ,i f );

[0028] Step Aa.4: According to the gray level symbiotic combination frequency P(I h ,I f ) Calculate the texture roughness Contrast of the kth window k and the entropy of the kth window Entropy k :

[0029] Cntrast k =∑ j,f P(I h ,I f )×(I h -I f ) 2 ;

[0030] Entropy k =-∑ j,f P(I h ,I f )×logP(I h ,I f );

[0031] Wherein, (I h -I f ) 2 Used to calculate the hth first grayscale value I of the jth pixel h The trivial difference with the f-th second gray value measures the grayscale difference;

[0032] Step Aa.5: Setting a threshold for the roughness of dead grass texture and a threshold for the entropy of dead grass to secondary screen the dead grass area;

[0033] When identifying the texture roughness Contrast of the kth window k ≥ the roughness threshold of dead grass texture, it is marked as the second gray area;

[0034] When identifying the texture roughness Contrast of the kth window k <Threshold of roughness of dead grass texture, no marking;

[0035] When identifying the entropy value of the k-th window Entropy k ≥ the threshold of the entropy value of dead grass, it is marked as the third gray area;

[0036] When identifying the entropy value of the k-th window Entropy k < the entropy value threshold of the grass, no marking;

[0037] When the intersection area of ​​the first yellow area, the second yellow area and the third yellow area is identified, the final dead grass area is obtained;

[0038] The final dead grass area = the first yellow area ∩ the second gray area ∩ the third gray area.

[0039] Preferably, the specific steps of generating distribution maps of different pollution types further include:

[0040] Step Aa.6: Identify the land surface temperature LST;

[0041] LST = BT × (1 + W × ln (ε));

[0042] Where BT is the brightness temperature, which is the ground radiation temperature measured by the sensor, in Kelvin. W is the atmospheric water vapor influence coefficient. ε is the surface emissivity, which reflects the absorption and emission capacity of the ground object to thermal radiation. kn(ε) is the natural logarithm of the surface reflectivity, which is used to correct the surface radiation error. The ground emissivity is the absorption and emission capacity of the ground object surface to infrared radiation, ranging from 0 ≤ ε ≤ 1. Different ground objects have different emissivities, including vegetation, soil, buildings, garbage, and livestock manure. Buildings: ε = 0.98; vegetation: ε = 0.96; soil: ε = 0.92-0.95; livestock manure: ε = 0.89-0.91.

[0043] Step Aa.7: Identify livestock manure areas:

[0044] When the surface temperature LST is 15-25℃, it is identified as soil and marked as brown area;

[0045] When the surface temperature LST is 30-40℃, it is identified as livestock manure and marked as black area;

[0046] When the surface temperature LST is greater than 40°C, it is identified as a building area and marked as a yellow area;

[0047] According to the markings of the final dead grass area, orange area, red area, brown area, green area, black area and yellow area in the three-dimensional model, the final dead grass area, red area and black area are extracted to generate the dead grass aggregation point, straw aggregation point and livestock manure aggregation point corresponding to the i-th monitoring sub-area.

[0048] Preferably, the pollution source static analysis module includes a first extraction unit and a model building unit;

[0049] The first extraction unit is used to identify the dead grass gathering points, straw gathering points and livestock feces gathering points in the i-th monitoring sub-area, and count the total number of pixels N of the j-th type of pollution source in the i-th monitoring sub-area i,j , and combined with the UAV image resolution analysis to obtain the pixel area S pixel The total area A of the jth type of pollution source in the i-th monitoring sub-area is calculated using the following formula: i,j :

[0050] A i,j =N i,j ×S pixel ;

[0051] The experimental density ρ of the jth type of pollution source j , calculate the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,j :

[0052] M i,j =A i,j ×ρ j ;

[0053] Among them, the experimental density ρ of the jth type of pollution source j Obtained based on historical reference data, including:

[0054] Straw density: 0.5-1.2kg / m 2 ;Desert density: 0.3-0.7kg / m 2 ;Livestock manure density: 0.8-2.0kg / m 2 ;

[0055] Extract the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,j Conduct an incineration experiment and collect the combustion emission factor and combustion time ratio of the jth type of pollution source in the i-th monitoring sub-area during the incineration experiment to establish a combustion experiment data set;

[0056] The model establishment unit is used to construct an initial convolutional neural network model by using a convolutional neural network, simulate, test, and train the initial convolutional neural network model with a combustion experiment data set, and use the trained initial convolutional neural network model as a prediction pollution model to simulate and obtain the combustion pollution coefficient Rs of the i-th monitoring sub-region i :

[0057] The combustion pollution coefficient Rs of the i-th monitoring sub-region i is calculated and obtained through the following formula:

[0058]

[0059] In the formula, j represents the index of pollution source type, including straw, dry grass h, and livestock manure; m represents the specific number of aggregation points of the j-th pollution source;

[0060] E i,j represents the combustion emission factor of the j-th type of pollution source in the i-th monitoring sub-region, M i,j represents the unit area weight of the j-th type of pollution source in the i-th monitoring sub-region, Pr i,j represents the combustion time ratio of the j-th type of pollution source in the i-th monitoring sub-region.

[0061] Preferably, the pollution source static analysis module further includes a first evaluation unit and a first strategy unit;

[0062] The first evaluation unit is used to establish a risk threshold X, and compare and evaluate the combustion pollution coefficient Rs of the i-th monitoring sub-region i with the risk threshold X to judge the risk level of the impact of the pollution source in the i-th monitoring sub-region on air pollution after combustion, and output a first classification result, including:

[0063] The risk threshold X includes a first threshold X1 and a second threshold X2, and the first threshold X1 is greater than the second threshold X2;

[0064] When Rs i < X2, it means that the distribution of pollution sources in this monitoring sub-region is dispersed, no obvious pollution aggregation area is formed, and the comprehensive impact on air pollution is limited, generating a first low-risk level;

[0065] When X2 ≤ Rs i ≤ X1, it means that the distribution of pollution sources in this monitoring sub-region is dispersed, but individual pollution aggregation areas have been formed, and the comprehensive impact of the emissions on air pollution is relatively significant, generating a second medium-risk level;

[0066] When Rs i>X1, indicating that the pollution sources in the monitoring sub-area are densely distributed, and the emissions from multiple pollution sources exceed expectations, requiring special attention, generating the third highest risk level;

[0067] The first strategy unit is used to summarize the monitoring sub-areas of the second medium risk level and the third high risk level and generate corresponding control strategies, including:

[0068] The first strategy for the second medium risk level includes: mechanized return of 50-60% of straw in the monitored sub-area to the fields or composting; conversion of 50-60% of dead grass in the monitored sub-area into organic fertilizer; and conversion of 50-60% of livestock manure in the monitored sub-area into organic fertilizer using aerobic fermentation technology;

[0069] The second strategy for the third highest risk level includes: mechanized return of 61-90% of the straw in the monitored sub-area to the fields or composting; converting 61-90% of the dead grass in the monitored sub-area into organic fertilizer, and converting 5-10% of the dead grass into biomass energy; using aerobic fermentation technology to convert 61-70% of the livestock manure in the monitored sub-area into organic fertilizer, and grading 20-30% of the livestock manure through anaerobic microorganisms to generate methane energy to supply the biogas power generation system.

[0070] Preferably, the dynamic monitoring module includes a temperature acquisition unit and an inversion layer identification unit;

[0071] The temperature acquisition unit is used to collect temperature data at different heights of the i-th monitoring sub-area and draw a temperature-height profile of the i-th monitoring sub-area; within the inversion layer, as the height increases, the temperature increases instead, presenting a positive temperature gradient, forming a temperature inversion area;

[0072] When drawing the temperature-height profile of the i-th monitoring sub-area, the temperature on the y-axis and the height on the x-axis are matched to obtain the relationship between temperature and height. The measured temperature values ​​of the vertical t-th height point and the r-th height point in the i-th monitoring sub-area are extracted, and the temperature difference γ between the t-th height point and the r-th height point is calculated by the following formula: t,r :

[0073]

[0074] Where, T t Indicates the temperature at the t-th height point, T rrepresents the temperature at the r-th height point; Δz represents the height difference between the t-th height point and the r-th height point; if the temperature gradient is negative, it means that the temperature decreases with increasing height; if the temperature gradient is positive, it means that the temperature increases with increasing height, then the inversion layer recognition unit identifies it as an inversion phenomenon, marks it as an inversion layer, and collects wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area r , establish a thermospheric effect data set.

[0075] Preferably, the dynamic analysis module includes an inversion layer effect factor calculation unit, a vertical internal shear factor calculation unit, and a precipitation washing effect factor calculation unit;

[0076] The temperature layer effect factor calculation unit is used to mark the temperature inversion layer of the ith monitoring sub-area and calculate the temperature inversion layer effect factor ITEF of the ith monitoring sub-area using the following formula: i :

[0077]

[0078] Where, T env Indicates the ambient temperature of the monitoring sub-area, T TILL represents the temperature in the inversion layer;

[0079] The vertical shear factor calculation unit is used to extract the wind speed data above and below the inversion layer in the thermospheric effect data set after marking the inversion layer of the i-th monitoring sub-area, and calculate the vertical shear factor VSCF of the i-th monitoring sub-area using the following formula: i :

[0080]

[0081] Where V wind (Z top ) is the wind speed above the inversion layer, V wind (Z bottom ) is the wind speed below the inversion layer; Z top and Z bottom are the upper and lower boundary heights of the inversion layer, respectively;

[0082] The precipitation washing effect factor calculation unit is used to extract the daily average precipitation JP of the i-th monitoring sub-area in the thermosphere effect data set. r The precipitation washing effect factor PWEF of the i-th monitoring sub-area is calculated by the following formula: i :

[0083] PWEF i =α×JP r ;

[0084] In the formula, α represents a constant factor used to represent the average effect of precipitation on pollutant removal, which is estimated through experimental data or empirical values; the precipitation washing effect refers to the ability of rainwater to wash pollutants in the air, thereby reducing the concentration of pollutants in the atmosphere; the greater the precipitation, the stronger the removal effect of pollutants in the air.

[0085] Preferably, the dynamic analysis module further includes an association unit, a second evaluation unit, and a second strategy unit;

[0086] The association unit is used to extract the inversion layer effect factor ITEF of the i-th monitoring sub-region i , the vertical shear factor VSCF i and the precipitation washing effect factor PWEF i . After dimensionless processing, the dynamic diffusion coefficient Ks of the i-th monitoring sub-region is calculated through the following formula i :

[0087]

[0088] In the formula, b1, b2, and b3 respectively represent the inversion layer effect factor ITEF i , the vertical shear factor VSCF i and the precipitation washing effect factor PWEF i weight coefficients, and the sum of the weights is 1; the precipitation washing effect factor is in an inverse relationship.

[0089] Preferably, the second evaluation unit is used to preset a diffusion threshold Y, and compare the dynamic diffusion coefficient Ks of the i-th monitoring sub-region i with the diffusion threshold Y to judge the diffusion impact level of the pollution source in the i-th monitoring sub-region on the adjacent region after combustion, and output a second classification result, including:

[0090] The diffusion threshold Y includes a first diffusion threshold Y1 and a second diffusion threshold Y2, and the first diffusion threshold Y1 is greater than the second diffusion threshold Y2;

[0091] When Ks i < Y2, it means that during the incineration process of the pollution source in this monitoring sub-region, the pollution source has stagnated in the inversion layer within the region and there is no diffusion risk, forming a first stagnant risk region level, and a third strategy is generated through the second strategy unit, including: when identifying that the pollution source is burning, 1-2 drones equipped with spray dust suppression equipment are arranged above the inversion layer in the upwind direction of the pollution source burning for 15-20 minutes of operation; the spray interval is set to once every 20 minutes, and the spray duration is 5 minutes;

[0092] When Y2 ≤ Ks i≤Y1, indicating that there is a risk of diffusion during the incineration process of the pollution source in the monitoring sub-area, but the risk is within expectations. The second low diffusion risk area level is generated, and the fourth strategy is generated through the second strategy unit, including: when the pollution source is identified for incineration, three drones equipped with spray dust suppression equipment are deployed at an altitude of 26-50 meters above the ground in the upwind direction of the pollution source for 21-30 minutes; the spray interval is set to once every 15 minutes, and the spray duration is 5 minutes;

[0093] When Ks i >Y1, indicating that the pollution source in the monitoring sub-area has a diffusion risk during the incineration process, and the diffusion risk will reach a farther range in a short period of time, generating the third highest diffusion risk area level; and generating the fifth strategy through the second strategy unit, including: when the pollution source is identified for incineration, 4-5 drones equipped with spray dust reduction equipment are deployed at an altitude of 40-50 meters above the ground in the upwind direction of the pollution source, and the operation lasts for 31 minutes to 40 minutes; the spray interval is set to once every 10 minutes, and the spray duration is 10 minutes.

[0094] The present invention provides a regional air pollution source distribution and impact assessment system. It has the following beneficial effects:

[0095] (1) Through the combination of remote sensing technology and deep learning models, the system can accurately identify the spatial distribution of agricultural waste such as straw, dead grass and livestock manure. This identification method can efficiently cover a wide range of target areas, and is particularly suitable for remote areas and agriculturally intensive areas, overcoming the limitations of traditional methods that rely on manual inspections and ground stations. The remote sensing identification module can provide real-time and accurate data on the location and type of pollution sources, providing solid data support for subsequent pollution assessments. The remote sensing data acquisition unit uses a multispectral camera equipped with an unmanned aerial vehicle for image acquisition, which can obtain image data in multiple bands including visible light, near infrared and short-wave infrared. These multi-band images have strong spatial and spectral information, and can provide the reflection characteristics of different pollution sources (such as straw, dead grass and livestock manure) in different bands, facilitating more accurate identification of pollution sources and their distribution.

[0096] (2) By marking each area in the three-dimensional model and combining it with the aforementioned steps (NDVI, texture analysis, LST identification, etc.), the spatial distribution map of various pollution sources can be accurately drawn. For example, pollution types such as healthy vegetation, dead grass, straw, livestock manure, and bare land can be visualized through different color markers (such as green, red, black, brown, etc.). The aggregation points of different pollution types (such as dead grass, straw, livestock manure, etc.) can be extracted and mapped in the monitoring sub-area, which can intuitively show the distribution trend of pollution sources and facilitate further analysis and decision-making.

[0097] (3) This regional atmospheric pollution source distribution and impact assessment system uses a predictive pollution model built by a convolutional neural network (CNN) to conduct in-depth learning and analysis of the distribution characteristics of different types of pollution sources. Through detailed calculations of pollution sources in each monitoring sub-area, the system can accurately assess the pollution potential of each type of pollution source, thereby generating an accurate pollution source distribution map. This method has significant advantages over traditional manual analysis and can quickly and efficiently process and analyze large-scale data. The pollution source static analysis module achieves accurate identification, quantification and prediction of pollution sources (dead grass, straw, livestock manure, etc.) by combining remote sensing image analysis with deep learning technology. The identification accuracy of pollution sources is improved, providing an important basis for subsequent pollution source management and preventive measures; the combustion pollution coefficient Rs of the pollution source is accurately simulated by training the incineration experiment data with a convolutional neural network. i , effectively support environmental protection and governance work.

[0098] (4) The dynamic monitoring module collects temperature data, wind speed data, and daily average precipitation data at different altitudes to draw a temperature-altitude profile, and combines it with the inversion layer effect for real-time monitoring. This enables the system to promptly identify the presence of an inversion layer and assess its impact on pollutant diffusion. Under the influence of the inversion layer, pollutants often stagnate at low altitudes. The system can adjust its monitoring strategy in real time, avoiding the neglect of the inversion layer effect in traditional methods and improving its response to dynamic changes in atmospheric pollution. BRIEF DESCRIPTION OF THE DRAWINGS

[0099] Figure 1 The figure is a flow chart of a regional air pollution source distribution and impact assessment system according to the present invention. DETAILED DESCRIPTION

[0100] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0101] Example 1

[0102] The present invention provides a regional air pollution source distribution and impact assessment system, please refer to Figure 1 ,include:

[0103] The remote sensing identification module is used to divide the target area into several monitoring sub-areas and identify pollution sources in the monitoring sub-areas through remote sensing technology. It obtains the spatial distribution data of pollution source clusters within the monitoring sub-areas, including dead grass clusters, straw clusters, and livestock manure clusters.

[0104] The pollution source static analysis module is used to build a prediction pollution model using a convolutional neural network based on the pollution source aggregation point data obtained by the remote sensing identification module, analyze the distribution characteristics of each type of pollution source, calculate the pollution potential of each type of pollution source in the monitoring sub-area based on the distribution characteristics, and generate the corresponding pollution source distribution map. It simulates the pollution generated during the incineration process of each pollution source and predicts the combustion pollution coefficient Rs of the i-th monitoring sub-area. i , and Rs i Perform classification, output a first classification result, and generate a corresponding strategy based on the first classification result;

[0105] The dynamic monitoring module collects temperature data at different heights of the i-th monitoring sub-area after the corresponding strategy is executed, draws the temperature-height profile of the i-th monitoring sub-area, identifies the inversion layer, and collects wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area. r , establish a data set of thermospheric effects;

[0106] Dynamic analysis module, used to construct the inversion layer effect factor ITEF of the i-th monitoring sub-area based on the thermosphere effect data set i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i , and correlate to obtain the dynamic diffusion coefficient Ks of the i-th monitoring sub-region i , and evaluate to output the second classification result and generate the corresponding strategy.

[0107] In this embodiment, using remote sensing technology, the system can accurately identify the spatial distribution of agricultural waste such as straw, dead grass, and livestock manure. This identification method can efficiently cover a wide range of target areas and is particularly suitable for remote areas and areas with dense agricultural production. It overcomes the limitations of traditional methods that rely on manual inspections and ground-based monitoring stations. The remote sensing identification module can provide real-time, accurate data on the location and type of pollution sources, providing solid data support for subsequent pollution assessments. The predictive pollution model constructed using a convolutional neural network (CNN) can conduct in-depth learning and analysis of the distribution characteristics of different types of pollution sources. By calculating the pollution sources in each monitoring sub-area in detail, the system can accurately assess the pollution potential of each type of pollution source, thereby generating an accurate pollution source distribution map. This method has significant advantages over traditional manual analysis and can quickly and efficiently process and analyze large-scale data. The dynamic monitoring module collects temperature data, wind speed data, and daily average precipitation data at different altitudes to create temperature-altitude profiles, and combines them with the inversion layer effect for real-time monitoring. This enables the system to promptly identify the presence of an inversion layer and assess its impact on pollutant diffusion. Under the influence of the inversion layer, pollutants often stagnate at low altitudes. The system can adjust the monitoring strategy in real time, avoiding the neglect of the inversion layer effect in traditional methods and improving the response capability to dynamic changes in atmospheric pollution.

[0108] Example 2

[0109] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically, the remote sensing identification module includes a region segmentation unit, a remote sensing data acquisition unit, an image preprocessing unit and a pollution source identification unit;

[0110] The region segmentation unit is used to collect and identify the geographical features and landforms of the target area using the GIS geographic information system, establish a three-dimensional model, and divide the target area into several monitoring sub-areas, which are marked as {Zq1, Zq2, Zq3, ..., Zq n}; n represents the total number of monitoring sub-areas;

[0111] The remote sensing data acquisition unit is used to acquire multi-band images of the target area using a multispectral camera mounted on an unmanned aerial vehicle, and to establish a multi-band image data set, wherein the multi-band image data set includes multi-band images including visible light, near infrared, and short-wave infrared;

[0112] Image preprocessing unit, used to perform image registration on multi-band image data sets, align the data of each band, perform radiation correction, and remove noise using mean filtering, Gaussian filtering, and wavelet transform;

[0113] The pollution source identification unit is used to extract spectral index features from the multi-band image data set and generate distribution maps of different pollution types.

[0114] In this embodiment, the regional segmentation unit uses GIS technology, based on the geographical features and topographic information of the target area, to accurately divide the target area into multiple sub-areas, and mark them in the three-dimensional model to form a detailed division of the distribution of atmospheric pollution sources. In this way, the system can achieve accurate regional division and spatial analysis, and provide basic data for subsequent pollution source identification and evaluation. The remote sensing data acquisition unit uses a multi-spectral camera equipped with an unmanned aerial vehicle for image acquisition, and can obtain image data of multiple bands including visible light, near infrared and short-wave infrared. These multi-band images have strong spatial and spectral information, and can provide the reflection characteristics of different pollution sources (such as straw, dead grass and livestock manure) in different bands, which is convenient for more accurate identification of pollution sources and their distribution.

[0115] Example 3

[0116] This embodiment is explained in Example 2, please refer to Figure 1 ,Specifically, the steps for generating distribution maps of different ,pollution types include:

[0117] Step Aa.1: Calculate the Normalized Difference Vegetation Index (NDVI), which is used to distinguish between green vegetation and dead grass. The expression is as follows:

[0118]

[0119] Wherein, NIR represents the value of the near-infrared band, which indicates the health of plants. Healthy vegetation has a high NIR reflectivity. Red represents the value of the infrared band, which indicates the absorption characteristics of plants. Healthy vegetation has a high absorption rate in the red light band.

[0120] Step Aa.2: Preliminary screening of dead grass areas: When the Normalized Difference Vegetation Index (NDVI) is between 0.5 and 0.9, it is identified as a healthy vegetation area and marked as green. The chlorophyll in green plants absorbs red light and strongly reflects near-infrared light (NIR), resulting in a high NDVI. An NDVI > 0.5 indicates high chlorophyll content, healthy plants, and vigorous growth.

[0121] When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.11-0.3, it is identified as a dead grass area and marked in red. Although dead grass has decayed, it still has some intact cells. Dead grass is vegetation that has decayed but still retains some cellular structure. Its chlorophyll content is significantly reduced, resulting in an NDVI of 0.11-0.3. The NDVI value decreases due to reduced near-infrared reflectance and weaker red light absorption. Dead grass is usually lower than healthy vegetation in NDVI, but still higher than bare soil or straw.

[0122] When the Normalized Difference Vegetation Index (NDVI) is between 0.05 and 0.1, it is identified as a straw area and marked as the first gray area. Straw is the remaining part of a crop after harvest and has almost no chlorophyll. Its NDVI is slightly higher than that of bare ground but lower than that of dead grass because its internal fiber structure can still reflect some infrared light. An NDVI between 0.05 and 0.1 indicates that the vegetation has completely lost its vitality but still retains a certain organic structure.

[0123] When the Normalized Difference Vegetation Index (NDVI) is between -0.2 and 0.09, it is identified as a bare soil area and marked as a brown area. Bare soil, sandy soil, and rocky areas, which do not contain vegetation, reflect visible and infrared light more evenly, resulting in an NDVI close to 0 or a negative value. An NDVI between -0.2 and 0.09 indicates that there is basically no vegetation cover.

[0124] When the Normalized Difference Vegetation Index (NDVI) is less than -0.1, it is identified as a water area and marked in orange. Water strongly absorbs infrared light (NIR), causing the NDVI calculated value to become negative. When NDVI is less than -0.1, it indicates that the area is a water body, such as a lake, river, or wetland.

[0125] The normalized difference vegetation index (NDVI) value of each pixel was calculated, and then the areas with 0.11≤NDVI≤0.3 were selected as dead grass areas, and those with 0.05≤NDVI≤0.1 were selected as straw areas, and healthy vegetation points with NDVI>0.5 were excluded.

[0126] Step Aa.3: Select the kth window in the i-th monitoring sub-area in the image data set, with a window size of 3×3 or 5×5; calculate the frequency of occurrence of the grayscale values ​​of each pair of pixels in the kth window; the grayscale value of each pair of pixels includes the first grayscale value I of the h-th pixel point h and the second grayscale value I of the f-th pixel f ;

[0127] Record the first grayscale value I of the hth pixel h and the second grayscale value I of the f-th pixel f The frequency I that appears in the kth window f , and normalize the frequency value to the probability value to obtain the gray-level symbiotic combination frequency P(I h ,I f ); it can effectively describe the texture features in the image. This helps to understand the spatial distribution and texture differences of different pollution types (such as dead grass, straw, etc.).

[0128] Step Aa.4: Calculate the texture roughness Contrast of the kth window based on the grayscale co-occurrence combination frequency P(i,j) k and the entropy of the kth window Entropy k :

[0129] Contrast k =∑ j,f P(I j ,I f )×(I j -I f ) 2 ;

[0130] Entropy k =-∑ j,f P(I j ,I f )×logP(I j ,I f );

[0131] Wherein, (I j -I f ) 2 Used to calculate the trivial difference between the first grayscale value of the h-th pixel and the second grayscale value of the f-th pixel, measuring the grayscale difference;

[0132] Texture roughness Contrast of the kth window k High value → rough texture (such as dead grass);

[0133] Texture roughness Contrast of the kth window k Low value → smooth texture (such as healthy vegetation);

[0134] Entropy of the kth window k High value → high texture complexity (such as dead grass);

[0135] Entropy of the kth window k Low value → uniform texture (such as healthy vegetation);

[0136] Step Aa.5: Setting a threshold for the roughness of dead grass texture and a threshold for the entropy of dead grass to secondary screen the dead grass area;

[0137] When identifying the texture roughness Contrast of the kth window k ≥ the roughness threshold of dead grass texture, it is marked as the second gray area;

[0138] When identifying the texture roughness Contrast of the kth window k <Threshold of roughness of dead grass texture, no marking;

[0139] When identifying the entropy value of the k-th window Entropy k ≥ the threshold of the entropy value of dead grass, it is marked as the third gray area;

[0140] When identifying the entropy value of the k-th window Entropy k < the entropy value threshold of the grass, no marking;

[0141] When the intersection area of ​​the first yellow area, the second yellow area and the third yellow area is identified, the final dead grass area is obtained;

[0142] The final dead grass area = the first yellow area ∩ the second gray area ∩ the third gray area.

[0143] In this embodiment, the Normalized Difference Vegetation Index (NDVI) is calculated to effectively distinguish between different types of ground cover, such as green vegetation, dead grass, straw, and bare soil. The NDVI value reflects the health of vegetation and provides an efficient, contactless monitoring method, facilitating remote sensing data processing and pollution source identification.

[0144] Based on the NDVI value range, different categories such as healthy vegetation, dead grass, straw, bare land and water bodies can be accurately distinguished, forming different regional labels (such as green, red, gray, etc.). This screening can improve the accuracy of subsequent pollution source identification and promote the accurate identification of dead grass and straw areas in agricultural waste monitoring;

[0145] By calculating texture features, the system can distinguish complex areas (such as dead grass and straw) from smooth areas (such as healthy vegetation) in the image, providing data support for subsequent texture analysis. The choice of window size (such as 3×3 or 5×5) allows for flexible adjustment of analysis accuracy, enhancing the adaptability and processing capabilities of remote sensing data.

[0146] By calculating the texture roughness and entropy value, we can gain a deeper understanding of the texture complexity and uniformity of different areas. For example, the texture of dead grass and straw is usually rough and complex, while healthy vegetation is relatively smooth and uniform. This method can identify the type of pollution source through image texture features, further improving the accuracy of pollution source identification, especially in complex environments. By setting the texture roughness and entropy value thresholds, areas that meet specific texture characteristics (such as dead grass areas) can be effectively screened out. This screening method can further refine the identification of pollution sources, eliminate some interference factors, and improve recognition accuracy.

[0147] Example 4

[0148] This embodiment is explained in Example 3, please refer to Figure 1 Specifically, the steps of generating distribution maps of different pollution types also include:

[0149] Step Aa.6: Identify the land surface temperature LST;

[0150] LST = BT × (1 + W × ln (ε));

[0151] Where BT represents brightness temperature, the ground object radiation temperature measured by the sensor, in degrees Kelvin. W represents the atmospheric water vapor influence coefficient. W is obtained by inverting the atmospheric water vapor content from remote sensing imagery. Generally, this requires combining ground-based meteorological station data or global climate models to estimate atmospheric water vapor content. The relationship between water vapor content and remote sensing imagery data can be used to infer the atmospheric water vapor influence coefficient.

[0152] ε represents the surface emissivity, reflecting the ability of the ground to absorb and emit thermal radiation; ln(ε) represents the natural logarithm of the surface reflectivity, which is used to correct surface radiation errors; the ground emissivity is the ability of the ground surface to absorb and emit infrared radiation, ranging from 0 ≤ ε ≤ 1; vegetation, soil, buildings, garbage and livestock manure have different emissivities, including: buildings: ε = 0.98; vegetation: ε = 0.96; soil: ε = 0.92-0.95; livestock manure ε = 0.89-0.91;

[0153] Step Aa.7: Identify livestock manure areas:

[0154] When the surface temperature LST is 15-25℃, it is identified as soil and marked as brown area;

[0155] When the surface temperature LST is 30-40℃, it is identified as livestock manure and marked as black area;

[0156] When the surface temperature LST is greater than 40°C, it is identified as a building area and marked as a yellow area;

[0157] According to the markings of the final dead grass area, orange area, red area, brown area, green area, black area and yellow area in the three-dimensional model, the final dead grass area, red area and black area are extracted to generate the dead grass aggregation point, straw aggregation point and livestock manure aggregation point corresponding to the i-th monitoring sub-area.

[0158] In this embodiment, land surface temperature (LST) identification provides a detailed understanding of the ground's thermal state. By comprehensively calculating brightness temperature (BT) and the atmospheric water vapor influence coefficient (W), it accurately reflects the thermal radiation characteristics of different ground features (such as soil, buildings, vegetation, and livestock manure). By measuring the surface's radiant temperature, this step effectively identifies different types of ground cover, particularly those associated with pollution, such as livestock manure and buildings.

[0159] By labeling each area in the 3D model and combining it with the aforementioned steps (NDVI, texture analysis, LST identification, etc.), the spatial distribution of various pollution sources can be accurately mapped. For example, pollution types such as healthy vegetation, dead grass, straw, livestock manure, and bare land can all be visualized using different color markers (such as green, red, black, and brown). By extracting and mapping the aggregation points of different pollution types (such as dead grass, straw, and livestock manure) in the monitoring sub-areas, the distribution trends of pollution sources can be intuitively displayed, facilitating further analysis and decision-making.

[0160] Example 5

[0161] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically, the pollution source static analysis module includes a first extraction unit and a model building unit;

[0162] The first extraction unit is used to identify the dead grass gathering points, straw gathering points and livestock feces gathering points in the i-th monitoring sub-area, and count the total number of pixels N of the j-th type of pollution source in the i-th monitoring sub-area i,j , and combined with the UAV image resolution analysis to obtain the pixel area S pixel The total area A of the jth type of pollution source in the i-th monitoring sub-area is calculated using the following formula: i,j :

[0163] A i,j =N i,j ×S pixel ;

[0164] The experimental density ρ of the jth type of pollution source j , calculate the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,j :

[0165] M i,j =A i,j ×ρ j ;

[0166] Among them, the experimental density ρ of the jth type of pollution source j Obtained based on historical reference data, including:

[0167] Straw density: 0.5-1.2kg / m 2 ;Desert density: 0.3-0.7kg / m 2 ;Livestock manure density: 0.8-2.0kg / m 2 ;

[0168] Extract the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,jConduct an incineration experiment and collect the combustion emission factor and combustion time ratio of the jth type of pollution source in the i-th monitoring sub-area during the incineration experiment to establish a combustion experiment data set;

[0169] The model building unit is used to use a convolutional neural network to construct an initial convolutional neural network model, and simulate, test and train the initial convolutional neural network model with a combustion experiment data set, and use the trained initial convolutional neural network model as a prediction pollution model to simulate and obtain the combustion pollution coefficient Rs of the i-th monitoring sub-area i :

[0170] The combustion pollution coefficient Rs of the i-th monitoring sub-area i Calculated using the following formula:

[0171]

[0172] Where j represents the pollution source type index, including straw, dead grass h, and livestock manure; m represents the number of specific gathering points of the jth pollution source;

[0173] E i,j M represents the combustion emission factor of the jth type of pollution source in the i-th monitoring sub-area, i,j represents the unit area weight of the jth type of pollution source in the i-th monitoring sub-area, Pr i,j It represents the burning time ratio of the j-th type of pollution source in the i-th monitoring sub-area.

[0174] The pollutant combustion emission factors are determined by experimental data. Typical values ​​for different pollution sources are shown in the following chart (unit: g / kg):

[0175] Pollution source category <![CDATA[CO2(g / kg)]]> NOx (g / kg) <![CDATA[SO2(g / kg)]]> PM2.5 (g / kg) straw burning 1512 4.2 0.8 12.6 Burning of dead grass 1385 3.6 0.7 10.2 Livestock manure burning 1205 2.9 1.4 14.8

[0176] The following is the combustion pollution coefficient Rs of the simulated Zq1 monitoring sub-area i Experimental diagram:

[0177]

[0178] Assume that monitoring sub-area Zq1 burns: straw: 500 kg, burning time 30 minutes, pollution source burning time ratio 0.5;

[0179] Dry grass: 300 kg burned, burning time 20 minutes, pollution source burning time ratio 0.33;

[0180] Livestock manure: 200 kg burned, burning time 40 minutes, pollution source burning time ratio 0.67;

[0181] This shows that during the 60-minute monitoring period, straw burned 50% of the time, dead grass burned 33% of the time, and livestock manure burned 67% of the time;

[0182] Calculate CO2 emissions:

[0183] FCO2,5=(1512×500×0.5)+(1385×300×0.33)+(1205×200×0.67)=676.585g=676.6kg;

[0184] Calculate PM2.5 emissions:

[0185] PM2.5,5=(12.6×500×0.5)+(10.2×300×0.33)+(14.8×200×0.67)=6.143g=6.14kg;

[0186] Finally, the total emissions of various pollutants can be calculated to obtain the combustion pollution coefficient Rs i .

[0187] In this embodiment, the pollution source static analysis module achieves accurate identification, quantification, and prediction of pollution sources (dead grass, straw, livestock manure, etc.) by combining remote sensing image analysis with deep learning technology. This improves the accuracy of pollution source identification and provides an important basis for subsequent pollution source management and prevention measures. By training the incineration experiment data through a convolutional neural network, the combustion pollution coefficient Rs of the pollution source is accurately simulated. i , effectively support environmental protection and governance work.

[0188] Example 6

[0189] This embodiment is explained in Example 5, please refer to Figure 1 ,Specifically, the pollution source static analysis module also includes a first evaluation unit and a first strategy unit;

[0190] The first evaluation unit is used to establish a risk threshold X and calculate the combustion pollution coefficient Rs of the i-th monitoring sub-area i Compare and evaluate with the risk threshold X to determine the risk level of the pollution source in the i-th monitoring sub-area that will affect air pollution after combustion, and output the first classification result, including:

[0191] The risk threshold X includes a first threshold X1 and a second threshold X2, and the first threshold X1 is greater than the second threshold X2;

[0192] When Rs i<X2 indicates that the distribution of pollution sources in this monitoring sub-region is dispersed, without forming an obvious pollution aggregation area, and has limited comprehensive impact on air pollution, generating the first low-risk level; the pollution sources in this monitoring sub-region are sparse, and the mutual influence between pollution sources is small, so the overall air pollution risk is low;

[0193] When X2 ≤ Rs i ≤ X1 indicates that the distribution of pollution sources in this monitoring sub-region is dispersed, but individual pollution aggregation areas have been formed, and the comprehensive impact of emissions on air pollution is relatively significant, generating the second medium-risk level; although the distribution of pollution sources in this monitoring sub-region is relatively dispersed, the emissions of local pollution sources are relatively high, resulting in a certain risk of air pollution;

[0194] When Rs i > X1 indicates that the distribution of pollution sources in this monitoring sub-region is intensive, and the emissions of multiple pollution sources exceed expectations, with a relatively high overall air pollution risk, which needs to be focused on, generating the third high-risk level; the distribution of pollution sources in this monitoring sub-region is very intensive, and the emissions of multiple pollution sources are relatively high, having a significant impact on air pollution and a high risk level;

[0195] The first strategy unit is used to summarize the monitoring sub-regions of the second medium-risk level and the third high-risk level, and generate corresponding regulation strategies, including:

[0196] Generate the first strategy for the second medium-risk level, including: returning 50 - 60% of the straw in this monitoring sub-region to the field by mechanization or composting; converting 50 - 60% of the dry grass in this monitoring sub-region into organic fertilizer, and converting 50 - 60% of the livestock manure in this monitoring sub-region into organic fertilizer by aerobic fermentation technology;

[0197] Generate the second strategy for the third high-risk level, including: returning 61 - 90% of the straw in this monitoring sub-region to the field by mechanization or composting; converting 61 - 90% of the dry grass in this monitoring sub-region into organic fertilizer, and converting 5 - 10% of the dry grass into biomass energy; converting 61 - 70% of the livestock manure in this monitoring sub-region into organic fertilizer by aerobic fermentation technology, and grading 20 - 30% of the livestock manure through anaerobic microorganisms to generate methane energy for supplying the biogas power generation system.

[0198] In this embodiment, through the collaborative work of the first evaluation unit and the first strategy unit, the pollution source static analysis module realizes the accurate evaluation and effective regulation of different pollution sources. By establishing reasonable risk thresholds and evaluation methods, it can conduct risk grading on the monitoring sub-regions and propose targeted regulation strategies for different risk levels. This method not only helps to control and reduce air pollution, but also improves the sustainability of agricultural production and promotes the protection of the ecological environment by converting waste into organic fertilizer or biomass energy.

[0199] Example 7

[0200] This embodiment is explained in Example 1, please refer to Figure 1 ,Specifically, the dynamic monitoring module includes a temperature acquisition unit and an ,inversion layer identification unit;

[0201] The temperature acquisition unit is used to collect temperature data at different heights of the i-th monitoring sub-area and draw a temperature-height profile of the i-th monitoring sub-area; within the inversion layer, as the height increases, the temperature increases instead, presenting a positive temperature gradient, forming a temperature inversion area;

[0202] When drawing the temperature-height profile of the i-th monitoring sub-area, the temperature on the y-axis and the height on the x-axis are matched to obtain the relationship between temperature and height. The measured temperature values ​​of the vertical t-th height point and the r-th height point in the i-th monitoring sub-area are extracted, and the temperature difference γ between the t-th height point and the r-th height point is calculated by the following formula: t,r :

[0203]

[0204] Where, T t Indicates the temperature at the t-th height point, T r represents the temperature at the r-th height point; Δz represents the height difference between the t-th height point and the r-th height point;

[0205] If the temperature gradient is negative, it means that the temperature decreases with increasing altitude, which is a normal phenomenon. If the temperature gradient is positive, it means that the temperature increases with increasing altitude. Then, the inversion layer identification unit identifies it as an inversion phenomenon and marks it as an inversion layer. The wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area are collected. r , establish a thermosphere effect data set, and obtain it through temperature sensor, wind speed sensor and rainfall sensor strategy; the daily average precipitation JP of the i-th monitoring sub-area r It can also be obtained by analyzing historical rainfall data.

[0206] In this embodiment, the dynamic monitoring module, through the collaborative work of the temperature acquisition unit and the inversion layer identification unit, accurately depicts the relationship between temperature and altitude, promptly identifies inversion phenomena, and, combined with data such as wind speed and precipitation, provides important information for air pollution management. Through this module's monitoring and analysis, we can effectively grasp environmental and meteorological changes, optimize pollution source monitoring and control strategies, and improve environmental management and pollution prevention and control effectiveness.

[0207] Example 8

[0208] This embodiment is explained in Example 7, please refer to Figure 1 ,Specifically, the dynamic analysis module includes an inversion layer effect factor calculation unit, a ,vertical internal shear factor calculation unit and a precipitation washing effect ,factor calculation unit;

[0209] The temperature layer effect factor calculation unit is used to mark the temperature inversion layer of the ith monitoring sub-area and calculate the temperature inversion layer effect factor ITEF of the ith monitoring sub-area using the following formula: i :

[0210]

[0211] Where, T env Indicates the ambient temperature of the monitoring sub-area, T TILL Indicates the temperature in the inversion layer. If the temperature of the inversion layer is high, pollutants will be retained in the area, and their diffusion capacity will be inhibited, resulting in higher pollution concentrations.

[0212] The vertical shear factor calculation unit is used to extract the wind speed data above and below the inversion layer in the thermospheric effect data set after marking the inversion layer of the i-th monitoring sub-area, and calculate the vertical shear factor VSCF of the i-th monitoring sub-area using the following formula: i :

[0213]

[0214] Where V wind (Z top ) is the wind speed above the inversion layer, V wind (Z bottom ) is the wind speed below the inversion layer; Z top and Z bottom are the upper and lower boundary heights of the inversion layer respectively; if the vertical wind speed changes greatly (shear is stronger), the vertical diffusion ability of pollutants is stronger;

[0215] The precipitation washing effect factor calculation unit is used to extract the daily average precipitation JP of the i-th monitoring sub-area in the thermosphere effect data set. r The precipitation washing effect factor PWEF of the i-th monitoring sub-area is calculated by the following formula: i :

[0216] PWEF i =α×JP r ;

[0217] Where α is a constant factor representing the average effect of precipitation on pollutant removal, estimated through experimental data or empirical values. The precipitation washing effect refers to the ability of rainwater to clean pollutants from the air, thereby reducing the concentration of pollutants in the atmosphere. The greater the precipitation, the stronger the effect of removing pollutants from the air.

[0218] In this embodiment, the inversion layer effect is quantified: Through the calculation of this factor, the effect of the inversion layer on pollutant retention can be quantified, thereby helping to predict the impact of the inversion phenomenon on regional atmospheric pollution. The introduction of the inversion layer effect factor makes the environmental prediction system more accurate, and can reflect the impact of temperature reversal on pollutant distribution in real time, providing an important basis for pollution management. The introduction of the vertical internal shear factor helps to assess the impact of wind speed changes on the vertical diffusion capacity of pollutants. If the wind speed changes significantly, pollutants may be more likely to diffuse to higher or lower altitudes, reducing the accumulation of local pollution concentrations.

[0219] By calculating the precipitation washing effect factor, we can quantitatively analyze the pollutant removal effect of precipitation and help assess its impact on air pollution. This factor enables environmental monitoring and air pollution prediction models to consider the role of precipitation, providing a dynamic adjustment basis for environmental management.

[0220] Example 8

[0221] This embodiment is explained in Example 7, please refer to Figure 1 ,Specifically, the dynamic analysis module further includes an ,associating unit, a second evaluation unit, and a second ,strategy unit;

[0222] The associated unit is used to extract the inversion layer effect factor ITEF of the i-th monitoring sub-area i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i After dimensionless processing, the dynamic diffusion coefficient Ks of the i-th monitoring sub-area is calculated by the following formula: i :

[0223]

[0224] Where b1, b2 and b3 represent the inversion layer effect factor ITEF of the i-th monitoring sub-area, respectively. i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i Weight coefficient, and the sum of the weights is 1; the precipitation washing effect factor is inversely proportional.

[0225] Example 2: This example is explained in Example 1. Figure 1, specifically, the second evaluation unit is used to preset a diffusion threshold Y and compare the dynamic diffusion coefficient Ks of the i-th monitoring sub-region i with the diffusion threshold Y to determine the diffusion impact level of the pollution source in the i-th monitoring sub-region on adjacent regions after combustion, and output a second classification result, including:

[0226] The diffusion threshold Y includes a first diffusion threshold Y1 and a second diffusion threshold Y2, and the first diffusion threshold Y1 is greater than the second diffusion threshold Y2;

[0227] When Ks i < Y2, it means that during the incineration of the pollution source in this monitoring sub-region, the pollution source has stagnated in the inversion layer within the region and there is no diffusion risk, forming a first stagnant risk region level, and a third strategy is generated through the second strategy unit, including: when identifying that the pollution source is being incinerated, 1-2 drones equipped with spray dust suppression equipment are arranged above the inversion layer in the upwind direction of the pollution source incineration for 15-20 minutes of operation; the spray interval is set to once every 20 minutes, and the spray duration is 5 minutes;

[0228] When Y2 ≤ Ks i ≤ Y1, it means that during the incineration of the pollution source in this monitoring sub-region, there is a diffusion risk, but the risk is within the expected range, generating a second low diffusion risk region level, and a fourth strategy is generated through the second strategy unit, including: when identifying that the pollution source is being incinerated, 3 drones equipped with spray dust suppression equipment are arranged at a height of 26-50 meters from the ground in the upwind direction of the pollution source incineration for 21-30 minutes of operation; the spray interval is set to once every 15 minutes, and the spray duration is 5 minutes;

[0229] When Ks i > Y1, it means that during the incineration of the pollution source in this monitoring sub-region, there is a diffusion risk, and the diffusion risk will reach a relatively far range in a short time, generating a third high diffusion risk region level; and a fifth strategy is generated through the second strategy unit, including: when identifying that the pollution source is being incinerated, 4-5 drones equipped with spray dust suppression equipment are arranged at a height of 40-50 meters from the ground in the upwind direction of the pollution source incineration for 31-40 minutes of operation; the spray interval is set to once every 10 minutes, and the spray duration is 10 minutes.

[0230] In this embodiment, accurate response strategies are generated according to different diffusion risk levels to ensure that effective pollution control measures can be taken in a timely manner when the pollution source is incinerated. The strategy of drones equipped with spray equipment can dynamically adjust the time, interval and duration of spray operations, effectively reduce pollution diffusion, and improve the flexibility and efficiency of air quality management.

[0231] This dynamic analysis module accurately identifies and addresses pollution source spread risks through multi-level assessment and strategy generation. Its core advantages include providing dynamic and accurate assessments of pollution spread by integrating the effects of inversion layers, wind shear, and precipitation washing. By comparing the diffusion coefficient with preset thresholds, the system accurately classifies the spread risk level and promptly adjusts response strategies. Based on different risk levels, drones and spray equipment can be flexibly configured to optimize pollution control effectiveness and ensure that air quality is not affected by the spread of pollution sources.

[0232] The threshold is set to facilitate comparison. The size of the threshold depends on the amount of sample data and the number of bases set by technicians in this field for each set of sample data; as long as it does not affect the proportional relationship between the parameter and the quantized value.

[0233] The above formulas are obtained by collecting a large amount of data and performing software simulation, and a formula close to the actual value is selected. The coefficients in the formula are set by those skilled in the art according to actual conditions. The above is only a preferred specific implementation method of the present invention, but the protection scope of the present invention is not limited to this. Any technician familiar with this technical field, within the technical scope disclosed by the present invention, can make equivalent replacements or changes based on the technical solution and inventive concept of the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A regional air pollution source distribution and impact assessment system, characterized in that: include: The remote sensing identification module is used to divide the target area into several monitoring sub-areas and identify pollution sources in the monitoring sub-areas through remote sensing technology. It obtains the spatial distribution data of pollution source clusters within the monitoring sub-areas, including dead grass clusters, straw clusters, and livestock manure clusters. The pollution source static analysis module is used to build a prediction pollution model using a convolutional neural network based on the pollution source aggregation point data obtained by the remote sensing identification module, analyze the distribution characteristics of each type of pollution source, calculate the pollution potential of each type of pollution source in the monitoring sub-area based on the distribution characteristics, and generate the corresponding pollution source distribution map. It simulates the pollution generated during the incineration process of each pollution source and predicts the combustion pollution coefficient Rs of the i-th monitoring sub-area. i , and Rs i Perform classification, output a first classification result, and generate a corresponding strategy based on the first classification result; The dynamic monitoring module collects temperature data at different heights of the i-th monitoring sub-area after the corresponding strategy is executed, draws the temperature-height profile of the i-th monitoring sub-area, identifies the inversion layer, and collects wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area. r , establish a data set of thermospheric effects; Dynamic analysis module, used to construct the inversion layer effect factor ITEF of the i-th monitoring sub-area based on the thermosphere effect data set i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i , and correlate to obtain the dynamic diffusion coefficient Ks of the i-th monitoring sub-region i , and evaluate to output the second classification result and generate the corresponding strategy.

2. A regional air pollution source distribution and impact assessment system according to claim 1, characterized in that: The remote sensing identification module includes a region segmentation unit, a remote sensing data acquisition unit, an image preprocessing unit and a pollution source identification unit; The region segmentation unit is used to collect and identify the geographical features and landforms of the target area using the GIS geographic information system, establish a three-dimensional model, and divide the target area into several monitoring sub-areas, which are marked as {Zq1, Zq2, Zq3, ..., Zq n }; n represents the total number of monitoring sub-areas; The remote sensing data acquisition unit is used to acquire multi-band images of the target area using a multispectral camera mounted on an unmanned aerial vehicle, and to establish a multi-band image data set, wherein the multi-band image data set includes multi-band images including visible light, near infrared, and short-wave infrared; Image preprocessing unit, used to perform image registration on multi-band image data sets, align the data of each band, perform radiation correction, and remove noise using mean filtering, Gaussian filtering, and wavelet transform; The pollution source identification unit is used to extract spectral index features from the multi-band image data set and generate distribution maps of different pollution types.

3. A regional air pollution source distribution and impact assessment system according to claim 2, characterized in that: The specific steps to generate distribution maps of different pollution types include: Step Aa.1: Calculate the Normalized Difference Vegetation Index (NDVI), which is used to distinguish between green vegetation and dead grass. The expression is as follows: Where NIR represents the near-infrared band value, which indicates plant health. Healthy vegetation has high NIR reflectivity. Red represents the infrared band value, which is the absorption characteristic of plants. Healthy vegetation has a high absorption rate in the red light band. Step Aa.2: Preliminary screening of dead grass areas: When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.5-0.9, it is identified as a healthy vegetation area and marked as a green area; When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.11-0.3, it is identified as a dead grass area and marked as a red area; When the Normalized Difference Vegetation Index (NDVI) is within the range of 0.05-0.1, it is identified as a straw area and marked as the first gray area; when the NDVI is between 0.05-0.1, it means that the vegetation has completely lost its vitality but still retains a certain organic structure; When the Normalized Difference Vegetation Index (NDVI) is within the range of -0.2-0.09, it is identified as a bare soil area and marked as a brown area; When the Normalized Difference Vegetation Index (NDVI) is less than -0.1, it is identified as a water area and marked as an orange area. The Normalized Difference Vegetation Index (NDVI) value of each pixel is calculated, and then 0.11≤NDVI≤0.3 is selected as the dead grass area, 0.05≤NDVI≤0.1 is selected as the straw area, and healthy vegetation points with NDVI>0.5 are excluded. Step Aa.3: Select the kth window in the i-th monitoring sub-area in the image data set, with a window size of 3×3 or 5×5; calculate the frequency of occurrence of the grayscale values ​​of each pair of pixels in the kth window; the grayscale value of each pair of pixels includes the first grayscale value I of the h-th pixel point h and the second grayscale value I of the f-th pixel f ; Record the first grayscale value I of the hth pixel h and the second grayscale value I of the f-th pixel f The frequency I that appears in the kth window f , and normalize the frequency value to the probability value to obtain the gray-level symbiotic combination frequency P(I h ,I f ); Step Aa.4: According to the gray level symbiotic combination frequency P(I h ,I f ) Calculate the texture roughness Contrast of the kth window k and the entropy of the kth window Entropy k : Contrast k =∑ j,f P(I h ,I f )×(I h -I f ) 2 ; Entropy k =-∑ j,f P(I h ,I f )×logP(I h ,I f ); Wherein, (I h -I f ) 2 Used to calculate the hth first grayscale value I of the jth pixel h The trivial difference with the f-th second gray value measures the grayscale difference; Step Aa.5: Setting a threshold for the roughness of dead grass texture and a threshold for the entropy of dead grass to secondary screen the dead grass area; When identifying the texture roughness Contrast of the kth window k ≥ the roughness threshold of dead grass texture, it is marked as the second gray area; When identifying the texture roughness Contrast of the kth window k <Threshold of roughness of dead grass texture, no marking; When identifying the entropy value of the k-th window Entropy k ≥ the threshold of the entropy value of dead grass, it is marked as the third gray area; When identifying the entropy value of the k-th window Entropy k < the entropy value threshold of the grass, no marking; When the intersection area of ​​the first yellow area, the second yellow area and the third yellow area is identified, the final dead grass area is obtained; The final dead grass area = the first yellow area ∩ the second gray area ∩ the third gray area.

4. A regional air pollution source distribution and impact assessment system according to claim 3, characterized in that: The specific steps to generate distribution maps of different pollution types also include: Step Aa.6: Identify the land surface temperature LST; LST = BT × (1 + W × ln (ε)); Where BT is the brightness temperature, which is the ground radiation temperature measured by the sensor, in Kelvin. W is the atmospheric water vapor influence coefficient. ε is the surface emissivity, which reflects the ground's ability to absorb and emit thermal radiation. ln(ε) is the natural logarithm of the surface reflectivity, which is used to correct the surface radiation error. The ground emissivity is the ability of the ground surface to absorb and emit infrared radiation, ranging from 0 ≤ ε ≤ 1. Different ground objects, such as vegetation, soil, buildings, garbage, and livestock manure, have different emissivities, including: buildings: ε = 0.98; vegetation: ε = 0.96; soil: ε = 0.92-0.95; livestock manure: ε = 0.89-0.

91. Step Aa.7: Identify livestock manure areas: When the surface temperature LST is 15-25℃, it is identified as soil and marked as brown area; When the surface temperature LST is 30-40℃, it is identified as livestock manure and marked as black area; When the surface temperature LST is greater than 40°C, it is identified as a building area and marked as a yellow area; According to the markings of the final dead grass area, orange area, red area, brown area, green area, black area and yellow area in the three-dimensional model, the final dead grass area, red area and black area are extracted to generate the dead grass aggregation point, straw aggregation point and livestock manure aggregation point corresponding to the i-th monitoring sub-area.

5. A regional air pollution source distribution and impact assessment system according to claim 1, characterized in that: The pollution source static analysis module includes a first extraction unit and a model building unit; The first extraction unit is used to identify the dead grass gathering points, straw gathering points and livestock feces gathering points in the i-th monitoring sub-area, and count the total number of pixels N of the j-th type of pollution source in the i-th monitoring sub-area i,j , and combined with the UAV image resolution analysis to obtain the pixel area S pixel The total area A of the jth type of pollution source in the i-th monitoring sub-area is calculated using the following formula: i,j : A i,j =N i,j ×S pixel ; The experimental density ρ of the jth type of pollution source j , calculate the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,j : M i,j =A i,j ×ρ j ; Among them, the experimental density ρ of the jth type of pollution source j Obtained based on historical reference data, including: Straw density: 0.5-1.2kg / m 2 ;Desert density: 0.3-0.7kg / m 2 ;Livestock manure density: 0.8-2.0kg / m 2 ; Extract the unit area weight M of the j-th type of pollution source in the i-th monitoring sub-area i,j Conduct an incineration experiment and collect the combustion emission factor and combustion time ratio of the jth type of pollution source in the i-th monitoring sub-area during the incineration experiment to establish a combustion experiment data set; The model building unit is used to use a convolutional neural network to construct an initial convolutional neural network model, and simulate, test and train the initial convolutional neural network model with a combustion experiment data set, and use the trained initial convolutional neural network model as a prediction pollution model to simulate and obtain the combustion pollution coefficient Rs of the i-th monitoring sub-area i : The combustion pollution coefficient Rs of the i-th monitoring sub-area i Calculated using the following formula: Where j represents the pollution source type index, including straw, dead grass h, and livestock manure; m represents the number of specific gathering points of the jth pollution source; E i,j M represents the combustion emission factor of the jth type of pollution source in the i-th monitoring sub-area, i,j represents the unit area weight of the jth type of pollution source in the i-th monitoring sub-area, Pr i,j It represents the burning time ratio of the j-th type of pollution source in the i-th monitoring sub-area.

6. A regional air pollution source distribution and impact assessment system according to claim 1, characterized in that: The pollution source static analysis module also includes a first evaluation unit and a first strategy unit; The first evaluation unit is used to establish a risk threshold X and calculate the combustion pollution coefficient Rs of the i-th monitoring sub-area i Compare and evaluate with the risk threshold X to determine the risk level of the pollution source in the i-th monitoring sub-area that will affect air pollution after combustion, and output the first classification result, including: The risk threshold X includes a first threshold X1 and a second threshold X2, and the first threshold X1 is greater than the second threshold X2; When Rs i < X2, it indicates that the pollution sources in this monitoring sub-region are distributed dispersedly and no obvious pollution aggregation area is formed, and the first low-risk level is generated; when X2 ≤ Rs i ≤ X1, it indicates that the pollution sources in this monitoring sub-region are distributed dispersedly, but individual pollution aggregation areas have been formed, and the second medium-risk level is generated; when Rs i > X1, it indicates that the pollution sources in this monitoring sub-region are distributed densely, and the emissions of multiple pollution sources exceed the expectations, which need to be focused on, and the third high-risk level is generated; The first strategy unit is used to summarize the monitoring sub-areas of the second medium risk level and the third high risk level and generate corresponding control strategies, including: The first strategy for the second medium risk level includes: mechanized return of 50-60% of straw in the monitored sub-area to the fields or composting; conversion of 50-60% of dead grass in the monitored sub-area into organic fertilizer; and conversion of 50-60% of livestock manure in the monitored sub-area into organic fertilizer using aerobic fermentation technology; The second strategy for the third highest risk level includes: mechanized return of 61-90% of the straw in the monitored sub-area to the fields or composting; converting 61-90% of the dead grass in the monitored sub-area into organic fertilizer, and converting 5-10% of the dead grass into biomass energy; using aerobic fermentation technology to convert 61-70% of the livestock manure in the monitored sub-area into organic fertilizer, and grading 20-30% of the livestock manure through anaerobic microorganisms to generate methane energy to supply the biogas power generation system.

7. A regional air pollution source distribution and impact assessment system according to claim 1, characterized in that: The dynamic monitoring module includes a temperature acquisition unit and an inversion layer identification unit; The temperature acquisition unit is used to collect temperature data at different heights of the i-th monitoring sub-area and draw a temperature-height profile of the i-th monitoring sub-area; within the inversion layer, as the height increases, the temperature increases instead, presenting a positive temperature gradient, forming a temperature inversion area; When drawing the temperature-height profile of the i-th monitoring sub-area, the temperature on the y-axis and the height on the x-axis are matched to obtain the relationship between temperature and height. The measured temperature values ​​of the vertical t-th height point and the r-th height point in the i-th monitoring sub-area are extracted, and the temperature difference γ between the t-th height point and the r-th height point is calculated by the following formula: t,r : Where, T t Indicates the temperature at the t-th height point, T r represents the temperature at the r-th height point; Δz represents the height difference between the t-th height point and the r-th height point; if the temperature gradient is negative, it means that the temperature decreases with increasing height; if the temperature gradient is positive, it means that the temperature increases with increasing height, then the inversion layer recognition unit identifies it as an inversion phenomenon, marks it as an inversion layer, and collects wind speed data above and below the inversion layer and the daily average precipitation JP of the i-th monitoring sub-area r , establish a thermospheric effect data set.

8. A regional air pollution source distribution and impact assessment system according to claim 7, characterized in that: The dynamic analysis module includes an inversion layer effect factor calculation unit, a vertical internal shear factor calculation unit and a precipitation washing effect factor calculation unit; The temperature layer effect factor calculation unit is used to mark the temperature inversion layer of the ith monitoring sub-area and calculate the temperature inversion layer effect factor ITEF of the ith monitoring sub-area using the following formula: i : Where, T env Indicates the ambient temperature of the monitoring sub-area, T TILL represents the temperature in the inversion layer; The vertical shear factor calculation unit is used to extract the wind speed data above and below the inversion layer in the thermospheric effect data set after marking the inversion layer of the i-th monitoring sub-area, and calculate the vertical shear factor VSCF of the i-th monitoring sub-area using the following formula: i : Where V wind (Z top ) is the wind speed above the inversion layer, V wind (Z bottom ) is the wind speed below the inversion layer; Z top and Z bottom are the upper and lower boundary heights of the inversion layer, respectively; The precipitation washing effect factor calculation unit is used to extract the daily average precipitation JP of the i-th monitoring sub-area in the thermosphere effect data set. r The precipitation washing effect factor PWEF of the i-th monitoring sub-area is calculated by the following formula: i : PWEF i =α×JP r ; Where α is a constant factor, which is used to represent the average effect of precipitation on pollutant removal.

9. A regional air pollution source distribution and impact assessment system according to claim 8, characterized in that: The dynamic analysis module further includes a correlation unit, a second evaluation unit and a second strategy unit; The associated unit is used to extract the inversion layer effect factor ITEF of the i-th monitoring sub-area i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i After dimensionless processing, the dynamic diffusion coefficient Ks of the i-th monitoring sub-area is calculated by the following formula: i : Where b1, b2 and b3 represent the inversion layer effect factor ITEF of the i-th monitoring sub-area, respectively. i , vertical internal shear factor VSCF i and precipitation washing effect factor PWEF i Weight coefficient, and the sum of the weights is 1; the precipitation washing effect factor is inversely proportional.

10. A regional air pollution source distribution and impact assessment system according to claim 9, characterized in that: The second evaluation unit is used to preset the diffusion threshold Y and calculate the dynamic diffusion coefficient Ks of the i-th monitoring sub-area. i Compare it with the diffusion threshold Y to determine the diffusion impact level of the pollution source in the i-th monitoring sub-area on the adjacent area after combustion, and output the second classification result, including: The diffusion threshold Y includes a first diffusion threshold Y1 and a second diffusion threshold Y2, and the first diffusion threshold Y1 is greater than the second diffusion threshold Y2; When Ks i <Y2, indicating that during the incineration process of the pollution source in the monitored sub-region, the pollution source has stagnated in the temperature inversion layer within the region and there is no diffusion risk, forming the first stagnation risk area level. And the third strategy is generated through the second strategy unit, including: when identifying the incineration of the pollution source, 1-2 drones equipped with spray dust suppression equipment are arranged above the temperature inversion layer in the upwind direction of the pollution source incineration for 15-20 minutes of operation; the spray interval is set to once every 20 minutes, and the spray duration is 5 minutes; When Y2≤Ks i ≤Y1, indicating that there is a risk of diffusion during the incineration process of the pollution source in the monitoring sub-area, but the risk is within expectations. The second low diffusion risk area level is generated, and the fourth strategy is generated through the second strategy unit, including: when the pollution source is identified for incineration, three drones equipped with spray dust suppression equipment are deployed at an altitude of 26-50 meters above the ground in the upwind direction of the pollution source for 21-30 minutes; the spray interval is set to once every 15 minutes, and the spray duration is 5 minutes; When Ks i >Y1, indicating that the pollution source in the monitoring sub-area has a diffusion risk during the incineration process, and the diffusion risk will reach a farther range in a short period of time, generating the third highest diffusion risk area level; and generating the fifth strategy through the second strategy unit, including: when the pollution source is identified for incineration, 4-5 drones equipped with spray dust reduction equipment are deployed at an altitude of 40-50 meters above the ground in the upwind direction of the pollution source, and the operation lasts for 31 minutes to 40 minutes; the spray interval is set to once every 10 minutes, and the spray duration is 10 minutes.

Images