A road mobile source pollution hotspot identification method based on satellite data dynamic analysis

Abstract

The application provides a road mobile source pollution hotspot identification method based on satellite data dynamic analysis, and relates to the technical field of environmental monitoring, which comprises the following steps: S1, constructing a multi-source collaborative observation dataset and performing pretreatment; S2, calculating a dynamic background field and extracting a road-related pollution differential signal; S3, based on historical data offline training, constructing a road mobile source pollution identification prior knowledge base and model; S4, based on real-time data and prior knowledge, generating a road network pollution contribution spatial distribution map; and S5, performing multi-dimensional verification on the hotspot result, and calculating a comprehensive confidence score based on the verification result to grade the hotspot. The application realizes dynamic and accurate identification of road mobile source pollution hotspots, reduces the dependence on high-cost data, and significantly improves the spatiotemporal resolution and reliability of the monitoring result.

Description

Technical Field

[0001] This invention relates to the field of environmental monitoring technology, specifically to a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data. Background Technology

[0002] Accurately identifying high-emission road sections (i.e., "hotspots") is a prerequisite for effective traffic pollution control. Existing technologies have inherent bottlenecks in achieving dynamic monitoring.

[0003] Existing methods for identifying roadside mobile source pollution hotspots have the following main shortcomings:

[0004] (1) Based on the "bottom-up" approach of real-time traffic flow, a model of "real-time traffic flow / vehicle type → real-time emissions" is established. However, in actual business, it is extremely difficult to continuously obtain large-scale, full-network, and high-time-efficiency vehicle characteristic data, resulting in problems such as incomplete coverage, high costs, or data lag.

[0005] (2) The "top-down" approach based on remote sensing observation: high temporal resolution satellites (such as geostationary satellites) can provide hourly pollution observations, but their spatial resolution is coarse (kilometer level); high spatial resolution satellites can identify road vehicles, but the revisit cycle is long (several days), which cannot meet the needs of dynamic monitoring. Existing remote sensing methods are mostly limited to regional pollution situation analysis or rely on simple spatial overlay, lacking effective means to quantitatively analyze and locate the contribution of road mobile sources from mixed satellite signals.

[0006] (3) The method based on road mobile monitoring also suffers from problems such as low monitoring frequency and high cost. In addition, existing identification methods often output a single result layer, lacking cross-validation and confidence assessment of the reliability of the results, making it difficult to directly apply to high-requirement decision-making scenarios.

[0007] Therefore, how to achieve high spatiotemporal resolution road mobile source hotspot monitoring and provide real-time and accurate data for management is an urgent problem to be solved. Summary of the Invention

[0008] In view of this, embodiments of this application provide a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, so as to achieve dynamic and accurate identification of road mobile source pollution hotspots and to evaluate the reliability of the identification results.

[0009] This application provides the following technical solution: a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, comprising: S1: Construct a multi-source collaborative observation dataset and perform preprocessing; the multi-source collaborative observation dataset includes geostationary satellite atmospheric pollutant concentration data, road traffic flow and vehicle characteristic data, ground mobile measurement data, and auxiliary data; S2: Construct a dynamic background field based on a geographic weighted regression model. Based on the difference between the original satellite observation field and the dynamic background field, and combined with road spatial constraints, extract the differential signal of road-related pollution to obtain the differential signal of the dynamic background field. S3: Based on the multi-source collaborative observation dataset and the dynamic background field differential signal, construct a prior knowledge base for identifying road mobile source pollution; S4: Based on real-time geostationary satellite atmospheric pollutant concentration data and real-time meteorological data, combined with the prior knowledge base, calculate the real-time road pollution contribution index and generate a spatial distribution map of road network pollution contribution for real-time hotspot dynamic identification. S5: Perform multidimensional verification on the hotspot results identified based on the spatial distribution map of pollution contribution from the road network, calculate a comprehensive confidence score based on the verification results, and classify the road mobile source pollution hotspots based on the comprehensive confidence score.

[0010] According to one embodiment of this application, in step S2, the dynamic background field is constructed in the following manner: Based on the digital road network, buffer zones of different radii are generated according to road grade. Satellite pixels outside the buffer zones are used as training sample points, and a geographic weighted regression model is employed to construct the dynamic background field. ; The road-related pollution differential signal Extract using the following formula:

[0011] Where t is the observation time, and (x, y) are the geographic coordinates of the center point of any satellite pixel within the study area. This is near-ground pollutant concentration data from geostationary satellites. For dynamic background fields, x,y Distance weight This indicates pixel-by-pixel multiplication.

[0012] According to one embodiment of this application, in step S3, the prior knowledge base includes a microscale vehicle-emission localization model, which is constructed in the following manner: Using historical road traffic flow and vehicle characteristic data, as well as spatiotemporally matched ground mobile measurement data, the net increase in road pollution is calculated; A quantitative relationship between the net increase in road pollution and vehicle density and the proportion of heavy vehicles is established by using a geographically weighted regression or linear mixed-effects model. This quantitative relationship is the microscale vehicle-emission localization model.

[0013] According to one embodiment of this application, in step S3, the prior knowledge base includes a function of macro-scale dynamic adjustment coefficients, which is constructed in the following manner: Using mobile field concentration data, geostationary satellite near-surface pollutant concentration data, and meteorological data, a statistical model is established, and a function of a macro-scale dynamic adjustment coefficient is defined to quantify the impact of meteorological conditions on pollution accumulation and satellite observations. The function of the macro-scale dynamic adjustment coefficient is defined as follows:

[0014] Where s represents the road grid and t represents the observation time. This is a dynamic adjustment coefficient on a macroscopic scale. This is data from mobile surveys of concentration. This is near-ground pollutant concentration data from geostationary satellites. For meteorological data, the function f is fitted using a multivariate nonlinear regression or machine learning model.

[0015] According to one embodiment of this application, in step S3, the prior knowledge base includes a road emission contribution weight lookup table, which is constructed in the following manner: Based on historical road traffic flow and vehicle characteristic data, and the aforementioned microscale vehicle-emission localization model, the prior emission intensity potential index for each road r is calculated. ; For each road grid s, calculate the sum of the prior emission intensity potential indices of all roads within the grid. ; Define the prior weight of road r's emission contribution in the road grid s to which road r belongs. And store it as a weight lookup table.

[0016] According to one embodiment of this application, in step S4, the road real-time pollution contribution index Calculated using the following formula:

[0017] in, For real-time observation, For real-time calculation with the current road Spatial matching road grid, A grid calculated based on real-time geostationary satellite atmospheric pollutant concentration data. The road-related pollution differential signal, This is a real-time macroscopic dynamic adjustment coefficient. Contribute prior weights to road emissions.

[0018] According to one embodiment of this application, in step S5, the multidimensional verification includes: Time-series filtering: Check whether the real-time pollution contribution index of the road continuously exceeds the set threshold within multiple consecutive set time periods, and obtain the quantitative value of the time-series filtering verification result based on the check results; Pollutant ratio fingerprint verification: Calculate the characteristic ratio of pollutants observed by satellite in hotspot areas during periods of increased pollution, and determine whether the ratio falls within the typical characteristic range of road mobile source emissions in order to obtain a pollutant ratio fingerprint verification score; Weekly cycle verification of time pattern: Compare the real-time pollution contribution index of hot road sections on set weekdays and rest days. Use statistical hypothesis testing to determine the significance difference between the average real-time pollution contribution index on weekdays and the average real-time pollution contribution index on rest days. If the average index on weekdays is higher than the average index on rest days and the significance difference is less than the preset threshold, then the weekly cycle verification is passed to obtain the weekly cycle verification score of time pattern. Standardized hotspot index verification: Based on the real-time road pollution contribution index and the historical mobile source pollution hotspot index of the current road section, the standardized hotspot index is calculated. If the standardized hotspot index is greater than 1, it is determined to be higher than the historical normal.

[0019] According to one embodiment of this application, in step S5, the comprehensive confidence score is calculated using the following weighted scoring function:

[0020] in, This is the quantized value of the time-series filtering verification result. The fingerprint verification score is the pollutant ratio. The score is used to verify the weekly cycle of the time pattern. To standardize the hot topic index, , , , These are the weighting coefficients.

[0021] According to one embodiment of this application, in step S5, the pollution hotspots of road mobile sources are classified into three levels: high-confidence hotspots, medium-confidence hotspots, and low-confidence / potential hotspots, based on the comprehensive confidence score and a set threshold range.

[0022] According to one embodiment of this application, in step S1, the road traffic flow and vehicle feature data are automatically extracted by processing sub-meter level satellite image sequences and using a pre-trained deep learning model for vehicle detection.

[0023] Compared with existing technologies, the beneficial effects achieved by at least one of the above-mentioned technical solutions adopted in the embodiments of this specification include at least the following: By constructing a multi-source collaborative observation system, establishing a dynamic background field difference mechanism and a dual-scale prior knowledge base, and introducing multi-dimensional verification and confidence leveling strategies, the embodiments of this invention solve the core contradiction of existing technologies in achieving both high spatiotemporal resolution and high analytical accuracy. The embodiments of this invention achieve dynamic and accurate identification of road mobile source pollution hotspots, significantly improving the spatiotemporal resolution and reliability of monitoring results, reducing reliance on high-cost data, and providing highly reliable and operationally feasible real-time decision support capabilities for road traffic pollution control. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a flowchart illustrating a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, according to the present invention. Detailed Implementation

[0026] The embodiments of this application will now be described in detail with reference to the accompanying drawings.

[0027] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. This application can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, in the absence of conflict, the following embodiments and features in the embodiments can be combined with each other. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0028] like Figure 1 As shown, this embodiment of the invention provides a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, including: S1: Construct a multi-source collaborative observation dataset and perform preprocessing; the multi-source collaborative observation dataset includes geostationary satellite atmospheric pollutant concentration data, road traffic flow and vehicle characteristic data, ground mobile measurement data, and auxiliary data; S2: Construct a dynamic background field based on a geographic weighted regression model. Based on the difference between the original satellite observation field and the dynamic background field, and combined with road spatial constraints, extract the differential signal of road-related pollution to obtain the differential signal of the dynamic background field. S3: Based on the multi-source collaborative observation dataset and the dynamic background field differential signal, construct a prior knowledge base for identifying road mobile source pollution; The prior knowledge base includes a microscale vehicle-emission localization model, a macroscale dynamic adjustment coefficient function, and a road emission contribution weight lookup table. S4: Based on real-time geostationary satellite atmospheric pollutant concentration data and real-time meteorological data, combined with the prior knowledge base, calculate the real-time road pollution contribution index and generate a spatial distribution map of road network pollution contribution for real-time hotspot dynamic identification. S5: Perform multidimensional verification on the hotspot results identified based on the spatial distribution map of pollution contribution from the road network, calculate a comprehensive confidence score based on the verification results, and classify the road mobile source pollution hotspots based on the comprehensive confidence score.

[0029] In its specific implementation, this embodiment mainly includes the following: In step S1: constructing and preprocessing a multi-source collaborative observation dataset.

[0030] Obtain four types of data for the target area within the same or similar time period.

[0031] (1) High spatial resolution geostationary satellite atmospheric pollutant concentration data: Obtain atmospheric pollutant column concentration data from geostationary satellites, and improve the spatial resolution of geostationary satellite data to 1 km based on machine learning or geographic weighted downscaling methods. Further combine atmospheric vertical profile data from ground-based remote sensing monitoring or atmospheric chemical model simulation to correct the pollutant column concentration data to near-ground concentration data, i.e., satellite near-ground pollutant concentration data.

[0032] (2) Road traffic flow and vehicle characteristic data: Obtain time-series data reflecting the traffic activity intensity of each road segment within a preset observation period in the target area. The data should at least include vehicle density or flow rate of each road segment at different times, as well as vehicle type composition information (such as the proportion of heavy vehicles). As a preferred implementation, the above information can be automatically extracted by processing high temporal resolution sub-meter level commercial satellite image sequences and using a pre-trained deep learning model for vehicle detection to generate a road vehicle density distribution map and a vehicle type composition information map.

[0033] (3) Ground mobile monitoring data: During the observation period, select typical time periods and conduct vehicle-mounted mobile monitoring along typical roads in the target area to obtain continuous data on road pollution concentration with high spatiotemporal resolution.

[0034] (4) Auxiliary data: including land use data of the study area, high-precision road vector data, a list of large stationary pollution sources such as thermal power plants, chemical plants, large industrial parks, population density raster data (as a proxy indicator of residential area density), etc.

[0035] In step S2: calculate the dynamic background field and extract the differential signal of road-related pollution.

[0036] S2.1 Dynamic Background Field Construction For any observation time t, the geographic coordinates of the center point of any satellite pixel within the study area are (x, y). The observed near-surface pollutant concentration at this point is denoted as... .

[0037] To isolate the influence of non-road sources (regional background and stationary sources), a geographically weighted regression model (GWR) is used to construct a dynamic background field. The modeling process is as follows: Based on a high-precision digital road network, buffer zones of different radii are generated according to road grade. Effective satellite pixels outside the buffer zones are used as training sample points. For observation time t, the model is established as follows:

[0038] in Let represent the geographic coordinates of the i-th sample pixel, and p be the number of covariates. It is the local regression coefficient of the k-th covariate. It is the first Indicator variables for land use types include, but are not limited to: the reciprocal of the distance from the cell center to a large stationary emission source, used to simulate the attenuation effect of power sources with distance, the percentage of industrial land within the cell, the percentage of urban built-up area within the cell, population density, and altitude. It is the intercept. It is a random error term.

[0039] For each training sample point at each time step, the GWR model assigns distance-related weights to its neighboring training sample points and fits the coefficients of that point using weighted least squares. After traversing all training sample points, the local coefficient surface of each covariate is obtained. The fitted GWR model is then applied to each pixel in the study area. For any pixel to be predicted, the corresponding covariate values ​​are extracted based on its geographical location, and the predicted background concentration is calculated using the local coefficients fitted from neighboring sample points at that location. Finally, the dynamic background field of the entire study area is obtained. This field reflects the spatial distribution of background concentrations at time t, determined by land use patterns, stationary source distribution, and regional transport under meteorological conditions.

[0040] S2.2 Extraction of road-related pollution signals.

[0041] The difference between the original observation field and the dynamic background field is used to obtain the preliminary differential signal field: , It primarily characterizes the contribution of near-ground dynamic emissions from road mobile sources.

[0042] To further focus on roads, a spatially constrained "road-related pollution differential signal" extraction method is introduced:

[0043] in This indicates pixel-by-pixel multiplication. (x, y) represents the distance weights, which are higher within the road buffer zone and decrease smoothly from 1 to 0 with increasing distance. The resulting road-associated pollution differential signal is... This is the dynamic background field differential signal, which can be interpreted as an enhancement of pollution in the vicinity of the road that exceeds the local dynamic background and may be related to road activity.

[0044] This invention constructs a dynamic background field using a geographically weighted regression model, effectively eliminating interference from regional background and stationary source emissions (non-road sources) on satellite observation signals. Furthermore, by constructing a dual-scale prior knowledge base comprising a "micro-scale vehicle-emission model" and a "macro-scale meteorological-observation coupling model," a quantitative correlation is established between vehicle characteristics, meteorological conditions, and satellite observation signals. This mechanism enables the invention to accurately analyze the actual contribution of road mobile sources to total observed pollution, solving the technical challenge of quantitatively identifying road source contributions from mixed signals in existing technologies.

[0045] In step S3: Prior knowledge base construction - offline construction and encapsulation of dual-scale association model.

[0046] S3.1 Construct a microscale correlation model to establish a localized quantitative relationship between road vehicle characteristics and pollution increment.

[0047] The training data includes: historical traffic flow and vehicle characteristic data (vehicle density) obtained from S1. Proportion of heavy vehicles Spatiotemporal matching of measured concentrations during mobile surveys and background pollutant concentration data Calculate the net increase in road pollution: - ,in The data can be obtained through the dynamic background field in step S2.1 or by interpolating pollutant concentration observations from ground-based air quality monitoring stations. A geographically weighted regression or linear mixed-effects model capable of handling spatial heterogeneity is used for training. The model is established in the following form:

[0048] in, For error terms, , , , These are the parameter values ​​to be solved. After training, the model is obtained, denoted as M. micro ( ).

[0049] S3.2 Construct a macro-scale model to establish the coupling relationship between ground emissions, satellite observation signals, and meteorological conditions at the grid scale.

[0050] Concentration measured by mobile survey (s,t) and the near-surface concentration of a geostationary satellite at the same time and location and meteorological data Matching factors such as wind speed, boundary layer height, and humidity is performed. A statistical model is established, defining a function of the macro-scale dynamic adjustment coefficient to quantify the impact of meteorological conditions on pollution accumulation and satellite observations. The function definition of the macro-scale dynamic adjustment coefficient is as follows:

[0051] Where s represents the road grid and t represents the observation time. This is a dynamic adjustment coefficient on a macroscopic scale. This is data from mobile surveys of concentration. This is near-ground pollutant concentration data from geostationary satellites. For meteorological data, the function f is fitted using a multivariate nonlinear regression or machine learning model.

[0052] S3.3 Calculate the prior weights of road emission contributions Based on historical road traffic flow and vehicle characteristic data, and the microscale vehicle-emission localization model M from step S3.1 micro Calculate the prior emission intensity potential index for each road r. : E base (r)=M micro (ρ base (r),ω base (r))×L(r) in, For road markings, ρ base ω represents the average density of historical vehicles. base This represents the historical average proportion of heavy-duty vehicles. This refers to the length of the road segment.

[0053] For each road grid s, calculate the sum of the prior emission intensity potential indices of all roads within it. :

[0054] Here, 's' specifically represents the spatial grid where road 'r' resides. Defining a road Prior weights of emission contributions in its grid s:

[0055] The weight represents the relative contribution of the road to the total pollution of its grid under historical normal conditions, and is stored as a weight lookup table.

[0056] S3.4 Prior Knowledge Base and Model Encapsulation. The microscale vehicle-emissions localization model M trained in S3.1 is then encapsulated. micro S3.2 calibrated macroscopic scale dynamic adjustment coefficient function S3.3 generated road emission contribution weight lookup table and dataset The knowledge base was integrated and packaged into a "preliminary knowledge base for identifying pollution from mobile sources on roads".

[0057] This invention constructs and encapsulates a priori knowledge base, requiring only conventional geostationary satellite data and meteorological data as input during the real-time identification phase to quickly calculate the real-time road pollution contribution index. Compared to existing technologies that rely on real-time traffic flow data across the entire road network or costly mobile monitoring, this invention significantly reduces dependence on high-cost, difficult-to-obtain data, improves the deployability and operational efficiency of the method in operational monitoring systems, and possesses good economic efficiency and scalability.

[0058] In step S4: Calculation of the real-time pollution contribution index of roads.

[0059] Real-time geostationary satellite atmospheric pollutant concentration data is acquired, near-ground pollutant concentration data is obtained through step S1, and road-related pollution differential signals are obtained through step S2. Then the road , Inputting real-time meteorological data into the S3.4 encapsulated prior knowledge base and model, road calculations are performed. Road real-time pollution contribution index , is represented as:

[0060] in, For real-time observation, For real-time calculation with the current road Spatial matching road grid, Road grid calculated based on real-time geostationary satellite atmospheric pollutant concentration data The road-related pollution differential signal, This is a real-time macroscopic dynamic adjustment coefficient. Contribute prior weights to road emissions.

[0061] Calculate the H value for all roads to form a spatial distribution map of the pollution contribution of the road network.

[0062] In step S5: road hotspot identification and multi-dimensional verification, and classification.

[0063] S5.1 Multi-level verification and filtering. (1) Time-series filtering, checking the exponent. Is it in a continuous If the threshold is continuously exceeded within a set time period (e.g., 3 hours), a Boolean value will be output. . If the hotspot passes the first-level verification, it is considered non-transient noise; otherwise, it is filtered. (2) Pollutant ratio fingerprint verification: For hotspots that pass the first-level verification, the satellite-observed pollutant characteristic ratio of their area during the period of significant pollution enhancement is calculated. (or other characteristic ratios, such as) ).like Falling within the typical characteristic range of road mobile source emissions in this region If it is within the acceptable range, it is considered to be in compliance. Verification score. It can be designed as a continuous value from 0 to 1, or simplified to a Boolean value. (3) Time pattern weekly cycle verification: Extract all recent (e.g., the past 4 weeks) weekdays and weekends / holidays of this road segment. Sequence. Calculate the significance value p for the difference between the two sets of data (e.g., Mann-Whitney U test). If the mean real-time pollution contribution index on weekdays is significantly greater than that on rest days (p value < 0.05), it is considered to conform to the weekly cycle pattern of road mobile sources. Output validation score. Similarly, it can be designed as a continuous value from 0 to 1 or a Boolean value. (4) Calculate the standardized hotspot index: .in, This is the hotspot index for this road segment during the same historical period (e.g., the same time period over the past 30 days). Median represents the median. This indicates a level higher than the historical norm.

[0064] S5.2 Comprehensive confidence quantification and classification.

[0065] Calculate the weighted scoring function for the overall confidence score:

[0066] in, This is the quantized value of the time-series filtering verification result. The fingerprint verification score is the pollutant ratio. The score is used to verify the weekly cycle of the time pattern. To standardize the hot topic index, , , , These are the weighting coefficients. Used to convert the value of the standardized hotspot index into a logarithmic value to smooth the impact of extreme values; to The weights are set according to business needs, and If any of the above items cannot be carried out due to practical difficulties and is therefore excluded from the calculation, its weight can be set to 0. Based on the distribution of scores, a threshold is defined to categorize hotspots into three distinct levels: High-confidence hotspots: scores ≥ Typically, it requires passing both time-duration verification and scoring highly in at least one of the ratio fingerprint or weekly cycle verification. Greater than 1. Medium confidence level hotspots: ≤ fraction < It may have passed basic verification, but a key piece of evidence (such as a chemical fingerprint) is weak or missing. Low confidence / potential hotspot: score < A single high score may lack supporting evidence, or there may be contradictions among the evidence, requiring continued monitoring.

[0067] After identifying potential hotspots, this invention introduces a multi-dimensional verification mechanism and a comprehensive confidence quantification and grading system. By cross-validating pollutant characteristics, temporal persistence, and traffic activity patterns, it effectively filters out instantaneous noise and false alarms, and outputs hotspot results with high, medium, and low confidence levels. This mechanism provides data products with clear reliability indicators for precise environmental control decisions, significantly enhancing the practical application value and decision support capabilities of the technical solution.

[0068] The present invention provides a method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, which enables dynamic and accurate identification of road mobile source pollution hotspots and assesses the reliability of the identification results.

[0069] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data, characterized in that, include: S1: Construct a multi-source collaborative observation dataset and perform preprocessing; the multi-source collaborative observation dataset includes geostationary satellite atmospheric pollutant concentration data, road traffic flow and vehicle characteristic data, ground mobile measurement data, and auxiliary data; S2: Construct a dynamic background field based on a geographic weighted regression model. Based on the difference between the original satellite observation field and the dynamic background field, and combined with road spatial constraints, extract the differential signal of road-related pollution to obtain the differential signal of the dynamic background field. S3: Based on the multi-source collaborative observation dataset and the dynamic background field differential signal, construct a prior knowledge base for identifying road mobile source pollution; S4: Based on real-time geostationary satellite atmospheric pollutant concentration data and real-time meteorological data, combined with the prior knowledge base, calculate the real-time road pollution contribution index and generate a spatial distribution map of road network pollution contribution for real-time hotspot dynamic identification. S5: Perform multidimensional verification on the hotspot results identified based on the spatial distribution map of pollution contribution from the road network, calculate a comprehensive confidence score based on the verification results, and classify the road mobile source pollution hotspots based on the comprehensive confidence score.

2. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 1, characterized in that, In step S2, the dynamic background field is constructed in the following way: Based on the digital road network, buffer zones of different radii are generated according to road grade. Satellite pixels outside the buffer zones are used as training sample points, and a geographic weighted regression model is employed to construct the dynamic background field. ; The road-related pollution differential signal Extract using the following formula: Where t is the observation time, and (x, y) are the geographic coordinates of the center point of any satellite pixel within the study area. This is near-ground pollutant concentration data from geostationary satellites. For dynamic background fields, x,y Distance weights This indicates pixel-by-pixel multiplication.

3. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 1, characterized in that, The prior knowledge base includes a microscale vehicle-emissions localization model, which is constructed in the following manner: Using historical road traffic flow and vehicle characteristic data, as well as spatiotemporally matched ground mobile measurement data, the net increase in road pollution is calculated; A quantitative relationship between the net increase in road pollution and vehicle density and the proportion of heavy vehicles is established by using a geographically weighted regression or linear mixed-effects model. This quantitative relationship is the microscale vehicle-emission localization model.

4. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 3, characterized in that, The prior knowledge base includes a function of macro-scale dynamic adjustment coefficients, which are constructed in the following manner: Using mobile field concentration data, geostationary satellite near-surface pollutant concentration data, and meteorological data, a statistical model is established, and a function of a macro-scale dynamic adjustment coefficient is defined to quantify the impact of meteorological conditions on pollution accumulation and satellite observations. The function of the macro-scale dynamic adjustment coefficient is defined as follows: Where s represents the road grid and t represents the observation time. This is a dynamic adjustment coefficient on a macroscopic scale. This is data from mobile surveys of concentration. This is near-ground pollutant concentration data from geostationary satellites. For meteorological data, the function f is fitted using a multivariate nonlinear regression or machine learning model.

5. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 4, characterized in that, The prior knowledge base includes a road emission contribution weight lookup table, which is constructed in the following manner: Based on historical road traffic flow and vehicle characteristic data, and the aforementioned microscale vehicle-emission localization model, the prior emission intensity potential index for each road r is calculated. ; For each road grid s, calculate the sum of the prior emission intensity potential indices of all roads within the grid. ; Define the prior weight of road r's emission contribution in the road grid s to which road r belongs. And store it as a weight lookup table.

6. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 5, characterized in that, In step S4, the real-time road pollution contribution index Calculated using the following formula: in, For real-time observation, For real-time calculation with the current road Spatial matching road grid, Road grid calculated based on real-time geostationary satellite atmospheric pollutant concentration data The road-related pollution differential signal, This is a real-time macroscopic dynamic adjustment coefficient. Contribute prior weights to road emissions.

7. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 1, characterized in that, In step S5, the multidimensional verification includes: Time-series filtering: Check whether the real-time pollution contribution index of the road continuously exceeds the set threshold within multiple consecutive set time periods, and obtain the quantitative value of the time-series filtering verification result based on the check results; Pollutant ratio fingerprint verification: Calculate the characteristic ratio of pollutants observed by satellite in hotspot areas during periods of increased pollution, and determine whether the ratio falls within the typical characteristic range of road mobile source emissions in order to obtain a pollutant ratio fingerprint verification score; Weekly cycle verification of time pattern: Compare the real-time pollution contribution index of hot road sections on set weekdays and rest days. Use statistical hypothesis testing to determine the significance difference between the average real-time pollution contribution index on weekdays and the average real-time pollution contribution index on rest days. If the average index on weekdays is higher than the average index on rest days and the significance difference is less than the preset threshold, then the weekly cycle verification is passed to obtain the weekly cycle verification score of time pattern. Standardized hotspot index verification: Based on the real-time road pollution contribution index and the historical mobile source pollution hotspot index of the current road section, the standardized hotspot index is calculated. If the standardized hotspot index is greater than 1, it is determined to be higher than the historical normal.

8. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 7, characterized in that, In step S5, the overall confidence score is calculated using the following weighted scoring function: in, This is the quantized value of the time-series filtering verification result. The fingerprint verification score is the pollutant ratio. The score is used to verify the weekly cycle of the time pattern. To standardize the hot topic index, , , , These are the weighting coefficients.

9. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 1, characterized in that, In step S5, the pollution hotspots of road mobile sources are classified, including: according to the comprehensive confidence score and the set threshold range, the pollution hotspots of road mobile sources are divided into three levels: high confidence hotspots, medium confidence hotspots, and low confidence / potential hotspots.

10. The method for identifying road mobile source pollution hotspots based on dynamic analysis of satellite data according to claim 1, characterized in that, In step S1, the road traffic flow and vehicle feature data are automatically extracted by processing sub-meter level satellite image sequences and using a pre-trained deep learning model for vehicle detection.

Images