Precise inversion method of highway vegetation carbon storage based on multi-modal remote sensing and tail gas stress correction

Abstract

The application provides a multi-modal remote sensing highway vegetation carbon storage accurate inversion method based on tail gas stress correction, and belongs to the field of remote sensing images; comprising the following steps: obtaining satellite remote sensing data, unmanned aerial vehicle data, ground measured data and traffic flow data; multi-element data fusion vegetation biomass inversion, including single tree scale, sample area scale and road area scale biomass inversion; converting road area biomass into road area carbon storage; calculating traffic pollutant emissions; simulating the daily and hourly pollutant concentration of the vegetation sample site according to the traffic pollutant emissions; constructing a carbon storage dynamic correction model, correcting the road area carbon storage based on photosynthesis inhibition effect and respiration promotion effect; the application fuses satellite remote sensing, unmanned aerial vehicle remote sensing and ground measured data, constructs a multi-modal data closed loop, solves the limitations of single data source in terms of spatial and temporal resolution and coverage, realizes multi-scale biomass inversion from single tree to road area, and improves the carbon storage estimation accuracy.

Description

TECHNICAL FIELD

[0001] The present application relates to the cross field of remote sensing technology and environmental monitoring technology, in particular to a multi-modal remote sensing highway vegetation carbon storage accurate inversion method based on tail gas stress correction. BACKGROUND

[0002] Highway vegetation carbon storage monitoring is an important technical means to achieve the carbon peak and carbon neutralization target in the transportation field. By constructing a multi-scale and multi-dimensional monitoring system, the carbon sink efficiency of vegetation can be quantitatively evaluated, the ecological engineering design can be optimized, the green layout can be optimized, and the stability of the ecological system can be improved. It is of great significance to promote the efficient use of road carbon sink resources, optimize road ecological system carbon management, and support green development of the transportation industry.

[0003] At present, the monitoring of highway vegetation carbon storage has multi-dimensional technical bottlenecks, which restrict the improvement of monitoring accuracy and efficiency, which is specifically manifested as follows: (1) The time and space coverage efficiency is limited: the traditional ground sample survey method is limited by manual operation characteristics, and has low efficiency, high cost and long cycle, which is difficult to meet the continuous linear monitoring demand of highway across complex terrain. (2) The multi-source data fusion efficiency is insufficient: the existing monitoring method relies on a single remote sensing data source. Although satellite remote sensing has the advantage of macro coverage, it is difficult to capture the dynamic changes of vegetation in the growing season due to the limitation of time resolution. Although unmanned aerial vehicle monitoring can obtain centimeter-level images, the coverage range of a single flight is limited, and it is difficult to realize continuous monitoring of the road network. (3) The adaptability to complex environment is defective: the traditional model is mostly based on the parameters of plain areas, and has insufficient adaptability in complex topography conditions such as plateau and mountainous areas. The existing model is mostly based on general vegetation parameters, and does not fully consider the nonlinear influence of environmental stress factors such as automobile exhaust pollution on vegetation carbon sink capacity. Pollutants such as NOx and PM2.5 in exhaust gas inhibit chlorophyll synthesis and reduce photosynthetic efficiency, resulting in a decrease in vegetation carbon storage. SUMMARY

[0004] In order to solve the above technical problems, the present application proposes a multi-modal remote sensing highway vegetation carbon storage accurate inversion method based on tail gas stress correction, which adopts a multi-modal data collaboration technology that fuses satellite remote sensing data, unmanned aerial vehicle high-resolution samples and ground three-dimensional sample survey, and couples a tail gas stress response model, so as to realize accurate estimation of road vegetation carbon storage.

[0005] The technical scheme adopted by the present application is as follows: a multi-modal remote sensing highway vegetation carbon storage accurate inversion method based on tail gas stress correction, comprising the following steps:

[0006] S1: Multi-modal data acquisition and preprocessing: the multi-modal data includes satellite remote sensing data, unmanned aerial vehicle data, ground measured data and traffic flow data;

[0007] S2: Vegetation biomass inversion based on multi-source data fusion, including single tree scale, sample plot scale and road scale biomass inversion, wherein the single tree scale is used to invert the biomass of a single tree to obtain single tree biomass, the sample plot scale is used to invert the biomass of the UAV detected sample plot according to the single tree biomass to obtain sample plot biomass, and the road scale is used to invert the biomass on the highway according to the sample plot biomass to obtain road scale biomass;

[0008] S3: Dynamic correction of carbon storage under pollution stress, including:

[0009] S31: converting road scale biomass into road scale carbon storage;

[0010] S32: constructing a tail gas pollution stress response model, including:

[0011] S321: calculating traffic pollutant emissions;

[0012] S322: simulating the daily and hourly pollutant concentration of the vegetation sample plot according to the traffic pollutant emissions;

[0013] S33: constructing a carbon storage dynamic correction model based on photosynthesis inhibition effect and respiration promotion effect to correct road scale carbon storage;

[0014] S4: Road vegetation carbon storage inversion and dynamic monitoring.

[0015] Further, the satellite remote sensing data in step S1 is used to delineate the target research area after preprocessing, the unmanned aerial vehicle data is used to obtain a multispectral data set and a standard point cloud data set after preprocessing, and the ground measured data includes single tree diameter at breast height, tree height, tree species type, coordinates, and vegetation photosynthetic rate and respiration rate; the traffic flow data includes real-time data recorded by highway entrances and exits, ETC systems, ground coils or video monitoring.

[0016] Further, the biomass inversion process at the single tree scale is as follows: the standard point cloud data collected by the unmanned aerial vehicle is normalized to obtain a canopy height model, single tree segmentation is performed based on a watershed algorithm, the tree height of the tree is extracted, the ground measured tree diameter at breast height is combined with the detected tree height during single tree segmentation, and an allometric growth equation is used to estimate the aboveground biomass of the single tree to obtain single tree biomass.

[0017] Further, the biomass inversion process at the sample plot scale is as follows: feature extraction is performed on the preprocessed unmanned aerial vehicle multispectral image to obtain vegetation index and texture features at the sample plot scale, the vegetation index and texture features extracted from the unmanned aerial vehicle multispectral image are used as independent variables, and single tree biomass is used as a dependent variable to perform variable screening, the screened feature combination is fitted with single tree biomass to construct a tree species biomass inversion model for the unmanned aerial vehicle monitored sample plot, and finally the entire sample plot biomass is inverted according to the model.

[0018] Further, the biomass inversion process in the road domain is as follows: based on the satellite multispectral image, the vegetation index is calculated, and the texture features are extracted, and the slope, aspect and elevation terrain factors are generated combined with DEM; the biomass of the unmanned aerial vehicle monitoring sample area is matched with the corresponding satellite pixel feature parameters to form a training sample; the machine learning algorithm is adopted to take the vegetation index, texture features and terrain factors extracted from the satellite multispectral image as independent variables, and the biomass of the sample area as the dependent variable to construct a “satellite pixel-biomass” model, realize the biomass inversion of the road vegetation, and obtain the biomass of the road domain.

[0019] Further, the traffic pollutant emission amount is calculated by using the COPERT model in step S321, and the emission factor in the COPERT model is dynamically modified, and the expression of the dynamically modified emission factor is as follows:

[0020] ;

[0021] Among them, is the emission factor of the i-th vehicle type; is the reference emission factor of the vehicle type i; is the vehicle speed correction function; is the emission standard correction coefficient; is the temperature correction coefficient, which is corrected by using the Arrhenius formula:

[0022] ;

[0023] Among them, is the activation energy; R is the gas constant; is the standard temperature; is the real-time atmospheric temperature;

[0024] Therefore, the total emission amount calculated according to the above emission factor dynamic correction formula is:

[0025] ;

[0026] Among them, is the total emission amount of the pollutant of the road section j; is the total driving mileage of the vehicle type i on the road section j.

[0027] Further, in step S322, the CALPUFF Lagrangian atmospheric diffusion model is adopted, the emission inventory output by the COPERT model is taken as the pollution source intensity input, the daily and hourly pollutant concentration at the vegetation sample site is simulated, and the pollutant concentration includes a plurality of single pollutant concentrations.

[0028] Further, the Logistic inhibition model is adopted to construct the carbon storage dynamic correction model in step S33, and the pollutant concentration simulated by the CALPUFF Lagrangian atmospheric diffusion model is used to correct the Logistic inhibition model, wherein the Logistic inhibition model includes photosynthetic inhibition effect and respiratory promotion effect;

[0029] The photosynthetic inhibition effect correction step is: the Logistic inhibition model is adopted to quantify the inhibition effect of single pollutant on photosynthetic rate, the toxicity equivalent method is used to determine the weight and weight coupling of multi-pollutant synergistic effect, the photosynthetic rate reduction ratio is calculated, and the carbon storage correction under single pollutant and multi-pollutant synergistic effect is realized.

[0030] The respiratory promotion effect correction step is: the Logistic inhibition model is adopted to quantify the nonlinear relationship between respiratory rate and pollutant concentration, the toxicity equivalent method is used to determine the weight and weight coupling of multi-pollutant synergistic effect, and the carbon storage correction under single pollutant and multi-pollutant synergistic effect is realized.

[0031] A computer device comprising a memory, a processor and a computer program stored on the memory, the processor executing the computer program to implement the steps of the method.

[0032] A computer readable storage medium having stored thereon a computer program / instructions, which, when executed by a processor, implement the steps of the method.

[0033] The present application has the beneficial effects relative to the prior art:

[0034] (1) The present application integrates satellite remote sensing, unmanned aerial vehicle remote sensing and ground measurement data, constructs a multi-modal data closed loop, solves the limitations of single data source in spatial resolution and coverage, realizes multi-scale biomass inversion from single tree to road domain, and improves the accuracy of carbon storage estimation.

[0035] (2) The present application breaks through the bottleneck of complex environment monitoring, covers typical landforms by unmanned aerial vehicle sampling, dynamically adjusts the flight height of unmanned aerial vehicle by terrain slope recognition algorithm, and ensures the effective data collection rate of highway mountain section. In the road domain carbon storage model, the terrain factor is input as an independent characteristic variable to improve the terrain adaptability of the model, and meet the continuous linear monitoring needs of highway across complex terrain.

[0036] (3) The method corrects the pollution stress and solves the non-linear influence. The tail gas stress response model is introduced for the first time to quantify the inhibition effect of pollutants such as NOx and PM2.5 on the photosynthetic rate of vegetation and the promotion effect of high-concentration CO2 on respiration consumption. The traditional model overestimates the problem of not considering environmental stress by coupling the synergistic effect of multiple pollutants through the Logistic dose-response model and the toxicity equivalent method, which is especially suitable for road environments with significant traffic pollution.

[0037] (4) The application provides a highway road vegetation carbon storage dynamic monitoring and engineering application intelligent decision support system. The system realizes carbon storage interannual variation analysis and model dynamic updating through multi-source data fusion and three-dimensional visualization technology, combined with time series satellite data and annual unmanned aerial vehicle inspection. Meanwhile, 10m resolution carbon storage raster map, single kilometer statistical report and dynamic early warning report are output, providing efficient decision support for highway carbon sink management and double carbon target achievement. BRIEF DESCRIPTION OF DRAWINGS

[0038] The application will be further described below in conjunction with the drawings:

[0039] Fig. 1 The application provides a highway road vegetation carbon storage monitoring method combining satellite remote sensing data, unmanned aerial vehicle high-resolution samples and machine learning algorithm;

[0040] Fig. 2 The application provides a highway road vegetation carbon storage inversion map combining satellite remote sensing data, high-resolution samples and machine learning;

[0041] Fig. 3 The application provides a structure block diagram of a system for monitoring vegetation carbon storage. DETAILED DESCRIPTION

[0042] As shown in Figs. 1 to 3 The application provides a multi-modal remote sensing highway vegetation carbon storage precise inversion method based on tail gas stress correction, including the following steps:

[0043] S1: Multi-modal data acquisition and preprocessing, wherein the multi-modal data includes satellite remote sensing data, unmanned aerial vehicle remote sensing data, ground measured data and traffic flow data;

[0044] Satellite remote sensing data: Sentinel-2A / B dual satellite networking is adopted to obtain multi-spectral data containing 13 bands, with a spatial resolution of 10m and a cloud cover of less than 5%. The terrain auxiliary data chain is constructed by synchronously combining DEM digital elevation model. The satellite image data is subjected to radiation calibration, atmospheric correction and orthorectification, and the left and right 200m buffer zones of the highway vector boundary are cropped to generate ROI.

[0045] UAV data: Use UAV platform equipped with multispectral camera and high-precision POS module to obtain the original multispectral data set of the sample area. After converting the DN value (Digital Number) of the original band gray image into the reflectivity of the ground object band using the calibration board image, perform stitching to obtain the multispectral data set. Use UAV platform equipped with laser radar and high-precision POS module, dynamically adjust the flight height through terrain slope recognition algorithm, combine with point cloud density adaptive control technology to improve the effective data acquisition rate, obtain the original point cloud data set of the sample area, and obtain the standard point cloud data set after point cloud solving, denoising filtering, strip stitching, coordinate conversion, etc.

[0046] Ground measurement data: Set up UAV monitoring sample area (sample area) along the main line of the expressway at every 10 km interval, use spatial stratified sampling method to cover typical landform units such as mountains, plains and hills, and ensure that the sample library contains key heterogeneity factors such as vegetation types and terrain complexity. Use standard plot method to select representative 20m x 20m fixed quadrats in the UAV monitoring sample area, record the diameter at breast height (1.3m height), tree height and tree species of the trees in the sample plot range, and record the coordinates of the single tree with RTK. Deploy weather stations and high-precision multi-gas analyzers in the UAV monitoring sample area to monitor environmental factors such as wind speed, wind direction, temperature, humidity and CO2, NOx concentration in real time, and determine physiological parameters such as photosynthetic rate and respiration rate of vegetation with portable photosynthetic instrument.

[0047] Traffic flow data: Obtain real-time data recorded by ETC system, ground coil or video monitoring at expressway entrance and exit from expressway operation company. Collect vehicle type, traffic volume of each vehicle type and average speed of road section in detail, collect corresponding fuel types of each vehicle type, convert data into CSV or database format, fields include time, road section ID, vehicle type, traffic volume, speed, etc., as the core input of tail gas emission model.

[0048] S2: Vegetation biomass inversion of multi-source data fusion, including single tree scale, sample area scale and road domain scale biomass inversion, wherein:

[0049] Single tree scale: Normalize the standard point cloud data collected by UAV to obtain canopy height model (CHM), perform single tree segmentation based on watershed algorithm, extract tree height, use ground measured tree diameter, combine with detected tree height during single tree segmentation, use allometric growth equation to estimate the aboveground biomass of single tree, and obtain single tree biomass.

[0050] Sample plot scale: After pretreatment of the unmanned aerial vehicle multispectral image, the feature extraction is performed to obtain the vegetation index and texture features at the sample plot scale. The vegetation index and texture features extracted from the unmanned aerial vehicle multispectral image are used as independent variables, and the single-tree biomass is used as the dependent variable to perform variable screening (in which the vegetation index and texture features are used as independent variables, and the correlation between the single-tree biomass and the independent variables is calculated to perform screening). The screened features (optimal features) are combined with the single-tree biomass to perform fitting to construct a tree species biomass inversion model for the unmanned aerial vehicle monitoring sample plot. Finally, the biomass of the entire sample plot is inversed according to the model.

[0051] Road domain scale: The vegetation index is calculated based on the satellite multispectral image, and eight texture features are extracted. The slope, aspect, and elevation terrain factors are generated in combination with DEM.

[0052] The biomass of the unmanned aerial vehicle monitoring sample plot is matched with the feature parameters of the corresponding satellite pixels to form a training sample. A machine learning algorithm is used to construct a “satellite pixel-biomass” model with the vegetation index, texture features, terrain factors, and other feature variables extracted from the satellite multispectral image as independent variables, and the sample plot biomass as the dependent variable. The road domain vegetation biomass is inversed at a resolution of 10 m to obtain the road domain biomass.

[0053] S3: Dynamic correction of carbon storage under pollution stress, including:

[0054] S31: Quantitative calculation of the background of vegetation carbon storage:

[0055] The road domain scale vegetation biomass inversion result is converted into vegetation carbon storage to obtain the basic carbon storage of the entire study area (i.e., the 200-meter buffer cropped ROI on the satellite remote sensing image):

[0056] ;

[0057] Wherein, is the basic carbon storage of the entire study area; is the biomass; is the carbon conversion factor. The carbon conversion factor (CF) adopts the general coefficient 0.5 recommended by the IPCC “National Greenhouse Gas Inventory Guidelines”. This coefficient comprehensively considers the average content of organic carbon in plant dry matter (45%-50%) and is suitable for carbon storage estimation of forest, shrub, and herbaceous vegetation types, ensuring the standardization and comparability of cross-scale calculation.

[0058] S32: Construction of tail gas pollution stress response model:

[0059] The vehicle type ratio, fuel type, average vehicle speed and meteorological data are input into the COPERT model, the daily calculation time scale is selected, the exhaust emission type (CO2, NOx, PM2.5, etc.) is selected, and the exhaust emission amount of each section is output in time periods. The CALPUFF Lagrangian atmospheric diffusion model is used to simulate the daily and hourly pollutant concentration at the vegetation plot, and the specific steps are as follows.

[0060] S321: Calculate the traffic pollutant emission amount based on the COPERT model:

[0061] 1) The input parameters of the COPERT model include vehicle type ratio, fuel type, average vehicle speed and meteorological data, the daily calculation time scale is selected, the exhaust emission type (CO2, NOx, PM2.5, etc.) is selected, and the exhaust emission amount of each section is output in time periods.

[0062] The input vehicle type ratio includes: being divided into small passenger cars, large passenger cars, light trucks, heavy trucks, motorcycles, etc. according to vehicle type, the vehicle type ratio is real-time counted by the vehicle type identification module of the highway ETC system or video monitoring, and is dynamically updated daily.

[0063] The fuel type is marked as gasoline, diesel, natural gas, electric, etc. The CO2 emission of electric vehicles needs to be converted into indirect emission of upstream power production.

[0064] The average vehicle speed is based on real-time vehicle speed data monitored by inductive coils or microwave radars, and the average vehicle speed is counted by lane according to road section, and the low-speed emission scene during congestion is corrected combined with traffic flow.

[0065] The meteorological data includes real-time wind speed, wind direction, temperature, atmospheric pressure and relative humidity, which are obtained through meteorological stations deployed along the highway, and the time resolution is consistent with the traffic flow data.

[0066] 2) COPERT emission calculation process: first, the emission factor in the COPERT model is dynamically corrected, and the dynamic correction formula of the emission factor is as follows:

[0067] ;

[0068] Wherein, is the reference emission factor of vehicle type i, which adopts the vehicle type reference emission data in the European COPERT database; is the vehicle speed correction function, which uses a segmented function to simulate the nonlinear change of the emission factor with vehicle speed; is the emission standard correction coefficient, which is obtained according to the vehicle registration year corresponding to the emission standard (such as national five, national six a, national six b); is the temperature correction coefficient, mainly for nitrogen oxides NOx emission correction, using Arrhenius formula for correction:

[0069] ;

[0070] wherein, is the activation energy (take 80 kJ / mol); R is the gas constant; is the standard temperature (298 K); is the real-time atmospheric temperature (K).

[0071] Then the total emissions calculated according to the above emission factor dynamic correction formula is:

[0072] ;

[0073] wherein, is the total emissions of pollutants (exhaust gas) of road section j; is the emission factor of the i-th vehicle type; is the total mileage of vehicle type i on road section j.

[0074] 3) Output results: output the exhaust emissions of road sections by time period and type.

[0075] S322: Using CALPUFF Lagrangian atmospheric diffusion model, the emission inventory output by COPERT model is used as the pollution source intensity input, and the daily and hourly pollutant concentration at the vegetation plot is simulated. The specific implementation steps are as follows:

[0076] 1) First, synchronize the data time: unify the time of traffic flow, emissions, and meteorological data to the same resolution, and use linear interpolation to handle missing values.

[0077] 2) The core formula of CALPUFF Lagrangian atmospheric diffusion model is as follows:

[0078] ;

[0079] wherein, is the pollutant concentration (kg / m 3 ) at position at time t; is the emission rate of the p-th plume segment; is the time step; , is the diffusion parameter (related to atmospheric stability, terrain); is the wind speed; , is the position of the plume segment center in the horizontal and vertical directions;

[0080] ;

[0081] wherein, is the total emission of pollutant of road segment j in the time period T used in COPERT model calculation; is the time period used in COPERT model calculation; is the number of plume segments for road segment j (road segment j is divided into N plume segments with 2000 meters interval).

[0082] 3) Concentration data extraction:

[0083] Based on the concentration grid in.DAT format output by CALPUFF, the concentration value of the center point of each sample area is extracted in GIS. The concentration value contains the concentration of each single pollutant.

[0084] S33: Constructing the dynamic correction model of carbon storage, including:

[0085] S331: Constructing and parameterizing the Logistic inhibition model:

[0086] 1) Data normalization: Normalizing the measured values of vegetation photosynthetic rate and respiration rate in the sample area to relative values;

[0087] 2) The Logistic inhibition model includes:

[0088] Photosynthetic inhibition effect:

[0089] ;

[0090] Respiratory promotion effect:

[0091] ;

[0092] wherein, is the function describing the inhibition effect of limiting factors on the system; is the function describing the growth effect of limiting factors on the system; C is the exhaust pollutant concentration output by CALPUFF model; K is the half-inhibition concentration; k is the shape parameter.

[0093] 3) Parameterization: Using nonlinear least squares method to fit the experimental data, taking the exhaust pollutant concentration data output by CALPOST model as the independent variable and the simultaneously measured photosynthetic rate and respiration rate as the dependent variable to parameterize and determine the half-inhibition concentration K and the shape parameter k of the Logistic inhibition model.

[0094] S332: Photosynthesis inhibition effect correction: Based on the CALPUFF model simulation of tail gas pollutant concentration, the single pollutant inhibition effect on photosynthesis rate is quantified by using the Logistic inhibition model, and the weight is determined by the toxicity equivalent method Weighted coupling of multi-pollutant synergistic effect, calculation of photosynthesis rate decline ratio (P) ), realizing the correction of carbon storage under single pollutant and multi-pollutant synergistic effect, as follows:

[0095] 1) Single pollutant correction, the formula is as follows:

[0096] ;

[0097] Among them, is the actual photosynthesis rate under the action of the nth pollutant; is the basic photosynthesis rate.

[0098] 2) Multi-pollutant synergistic correction, including:

[0099] Weight calculation: The concentration of different pollutants is converted into equivalent toxicity equivalent to NOx, PM2.5, CO2 and other pollutants by using the toxicity equivalent method, and the weight is allocated according to its toxicity contribution to vegetation physiological indicators. The calculation formula is:

[0100] ;

[0101] Among them, represents the weight of the nth pollutant; is the concentration value of the nth pollutant leading to a specific effect, here taking the photosynthesis rate drop of 50%; a is the total number of pollutant species participating in synergistic effect; m is the serial number of the pollutant.

[0102] The formula for multi-pollutant synergistic correction is as follows:

[0103] ;

[0104] Among them, is the photosynthesis rate loss value under the stress of n kinds of pollutants (calculated from the loss value of CALPOST concentration); is the actual photosynthesis rate under the action of the nth pollutant.

[0105] 3) Photosynthesis inhibition carbon storage correction formula is as follows:

[0106] ;

[0107] Among them, is the corrected vegetation carbon storage; is the basic carbon storage; The photosynthetic rate loss value (loss value calculated by CALPOST concentration) under n kinds of pollutant stress; The basic photosynthetic rate.

[0108] S333: Respiratory promotion effect correction: In view of the promotion effect of tail gas pollutants such as high concentration CO2 on plant respiration consumption, a Logistic inhibition model is used to quantify the nonlinear relationship between respiration rate and pollutant concentration, and the weight is determined by toxicity equivalence method Weighted coupling of multi-pollutant synergistic effect, realizing the correction of carbon storage under single pollutant and multi-pollutant synergistic effect, as follows:

[0109] 1) Single pollutant correction, the formula is as follows:

[0110] ;

[0111] ;

[0112] Among them, is the increase of plant respiration rate in the polluted environment relative to the background respiration rate under the action of single pollutant; is the plant respiration rate in the polluted environment (calculated based on the experimental relationship between CALPOST concentration and respiration rate), is the background respiration rate.

[0113] 2) Multi-pollutant synergistic correction, including:

[0114] Weight calculation: The toxicity equivalence method is used to convert the concentrations of different pollutants into equivalent toxicity equivalents for NOx, PM2.5, CO2 and other pollutants, and the weights are allocated according to their toxicity contribution to vegetation physiological indicators. The calculation formula is:

[0115] ;

[0116] Among them, represents the weight of the nth pollutant; is the concentration value of the nth pollutant leading to a specific effect, and the respiration rate increases by 50%; a is the total number of pollutant species participating in synergistic effect; m is the serial number of the pollutant.

[0117] Multi-pollutant synergistic correction formula:

[0118] ;

[0119] Among them, is the total increase of plant respiration rate relative to background respiration rate under the action of multi-pollutant synergism; represents the weight of the nth pollutant; The increase of the plant respiration rate relative to the background respiration rate under the polluted environment when the nth pollutant acts alone.

[0120] 3) The respiration-promoting carbon storage correction formula is as follows:

[0121] ;

[0122] wherein, is the corrected vegetation carbon storage; is the basic carbon storage; is the total increase of the plant respiration rate relative to the background respiration rate under the synergistic action of multiple pollutants.

[0123] Based on the above correction, the carbon storage correction model containing the photosynthesis inhibition effect correction and the respiration promotion effect correction is obtained.

[0124] S4: Roadside vegetation carbon storage inversion and dynamic monitoring:

[0125] According to the carbon storage correction model, the road carbon storage is corrected, and is applied to the satellite image to generate a highway roadside carbon storage raster map, realize 10m spatial resolution carbon storage mapping, and count the single-kilometer carbon storage and total carbon storage; by using time series satellite data, the interannual variation of carbon storage is analyzed, and the model can be combined with an unmanned aerial vehicle to perform annual inspection on a key road section.

[0126] The embodiment of the application further provides a highway roadside vegetation carbon storage intelligent decision support system, the system comprises: a multi-modal data acquisition module: used for acquiring data associated with target vegetation in a target region; a multi-modal data deep fusion module: constructing a satellite-unmanned aerial vehicle-plot data closed loop; a stress effect double correction module: quantifying the synergistic influence of photosynthesis inhibition and respiration promotion, solving the overestimation problem of traditional methods in high-pollution areas; an engineering intelligent decision module: outputting a 10m resolution carbon storage raster map, a single-kilometer statistical report and a dynamic early warning report, providing efficient decision support for highway carbon sink management and double carbon target achievement.

[0127] The following takes a certain highway section in a certain province as an example to further illustrate the application.

[0128] Table 1 is a certain highway roadside common tree species biomass model in a certain province.

[0129] Table 1 is a certain highway roadside common tree species biomass model in a certain province.

[0130] .

[0131] In the table, D is the diameter at breast height; H is the tree height; W S is the stem biomass; W BBiomass of branches; W L Biomass of leaves; W P Biomass of roots; W T Total biomass (stem + branches + leaves); W R Underground biomass.

[0132] For the sample plot scale, firstly, the characteristic calculation is carried out, the unmanned aerial vehicle multi-spectral original band characteristics include five original single-band images of blue, green, red, red edge and near infrared after radiation correction, the vegetation index (Table 2) is obtained by calculation or statistical transformation of multiple spectral bands, and the texture factor (Table 3) is usually extracted based on the gray level co-occurrence matrix.

[0133] Table 2 Unmanned aerial vehicle vegetation index

[0134]

[0135] Table 3 Unmanned aerial vehicle texture characteristic factor

[0136]

[0137] Then, variable screening is carried out: Pearson correlation analysis is carried out on the variables, the Pearson correlation coefficient describes the degree of linear relationship between X and Y, the Pearson correlation coefficient, the value range is [-1, 1], the greater the absolute value indicates the stronger the correlation, less than 0 indicates negative correlation, greater than 0 indicates positive correlation. Set the correlation coefficient threshold to |r|>0.6, retain the variables with higher correlation with carbon storage than the threshold, the formula is as follows:

[0138]

[0139] Wherein, is the pth sample value of the characteristic variable; is the pth sample value of the sample plot carbon storage; is the average value of the characteristic variable; is the average value of the sample plot carbon storage; N is the sample number.

[0140] Then, model construction and biomass inversion are carried out: taking the screened characteristics as the independent variable and the single tree biomass as the dependent variable, a linear regression model is constructed, and the determination coefficient (R²) and the root mean square error (RMSE) are used to test the model accuracy. R 2 The closer to 1, the higher the model accuracy, the lower the RMSE value, and the more accurate the regression model. The trained carbon storage model is applied to the unmanned aerial vehicle sample plot.

[0141]

[0142] ​​​

[0143] wherein, Bm is the measured biomass of a single tree; Bm is the measured biomass of a single tree; Bm is the measured biomass of a single tree; P is the total number of samples.

[0144] For the road domain scale, the vegetation index was calculated based on satellite multispectral images (Table 4), the texture characteristic factor was extracted (Table 5), and the slope, aspect and elevation terrain factors were generated in combination with DEM. Pearson correlation analysis was performed on the variables, and the characteristic factors with a correlation coefficient threshold of |r|>0.6 were retained.

[0145] Table 4 Satellite vegetation index

[0146]

[0147] Table 5 Satellite texture characteristic factor

[0148]

[0149] The biomass of the UAV sample area was matched with the characteristic parameters of the corresponding satellite pixels to form a training sample; the random forest (RF) algorithm, the gradient boosting tree (XGBoost) algorithm and the K- nearest neighbor (KNN) algorithm were used to construct a "satellite pixel-biomass" model with the vegetation index factor, the texture factor and the terrain factor extracted from the satellite multispectral images as the independent variables and the biomass of the sample area as the dependent variable, so as to realize the inversion of the 10m resolution road domain vegetation biomass.

[0150] Finally, the dynamic correction and inversion of carbon storage under pollution stress were performed to obtain a 10m resolution carbon storage grid map, a single kilometer statistical report and a dynamic early warning report, which provided quantifiable basis for carbon sink trading and greening planning.

[0151] The method provided in the application overcomes the defects of the traditional ground sample survey method, such as low efficiency and high cost; the single satellite remote sensing is limited by spatial resolution and is difficult to accurately capture the vegetation heterogeneity of the road domain; the coverage range of a single UAV monitoring is limited, and it is difficult to realize continuous monitoring at the road network level; the traditional model has poor adaptability in complex terrain, and does not consider the nonlinear inhibition of exhaust pollution stress on the carbon sink capacity of vegetation.

[0152] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the application, and not to limit them; although the application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the application.

Claims

1. A multimodal remote sensing method for accurate inversion of vegetation carbon storage on highways based on exhaust gas stress correction, characterized in that: Includes the following steps: S1: Multimodal data acquisition and preprocessing: Multimodal data includes satellite remote sensing data, UAV data, ground measurement data, and traffic flow data; S2: Vegetation biomass inversion based on multivariate data fusion, including biomass inversion at the individual tree scale, sample area scale, and road area scale. The individual tree scale is used to invert the biomass of a single tree to obtain the individual tree biomass. The sample area scale is used to invert the biomass of the sample area detected by the UAV based on the individual tree biomass to obtain the sample area biomass. The road area scale is used to invert the biomass on the highway based on the sample area biomass to obtain the road area biomass. S3: Dynamic correction of carbon storage under pollution stress, including: S31: Convert roadside biomass into roadside carbon storage; S32: Construct a vehicle exhaust pollution stress response model, including: S321: Calculate traffic pollutant emissions; S322: Simulate daily and hourly pollutant concentrations at vegetation sample plots based on traffic pollutant emissions; S33: Construct a dynamic correction model for carbon storage, and correct roadside carbon storage based on photosynthetic inhibition and respiration promotion effects; S4: Inversion and dynamic monitoring of carbon storage in roadside vegetation.

2. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 1, is characterized in that: The satellite remote sensing data preprocessed in step S1 is used to delineate the target study area. The UAV data preprocessed to obtain a multispectral dataset and a standard point cloud dataset. The ground measured data includes tree diameter at breast height, tree height, tree species, coordinates, and vegetation photosynthetic rate and respiration rate. The traffic flow data includes real-time data recorded by highway entrances and exits, ETC system, inductive loops, or video surveillance.

3. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 2, is characterized in that: The biomass inversion process at the tree-scale is as follows: Standard point cloud data collected by UAV is normalized to obtain a canopy height model. Trees are segmented based on the watershed algorithm to extract tree height. Using the measured tree diameter at breast height on the ground and the tree height detected during tree segmentation, the aboveground biomass of a single tree is estimated using the allometric growth equation to obtain the tree biomass.

4. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 2, is characterized in that: The biomass inversion process at the sample area scale is as follows: Feature extraction is performed on the preprocessed UAV multispectral images to obtain vegetation index and texture features at the sample area scale. The vegetation index and texture features extracted from the UAV multispectral images are used as independent variables, and the biomass of individual trees is used as the dependent variable for variable selection. The selected feature combinations are fitted with the biomass of individual trees to construct a tree species biomass inversion model for the UAV-monitored sample area. Finally, the biomass of the entire sample area is inverted based on the model.

5. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 2, is characterized in that: The biomass inversion process at the road area scale is as follows: vegetation indices are calculated based on satellite multispectral images, and texture features are extracted. Slope, aspect, and elevation topographic factors are generated by combining DEM. The biomass of the sample area monitored by UAV is matched with the feature parameters of the corresponding satellite pixels to form training samples. A machine learning algorithm is used to construct a "satellite pixel-biomass" model with vegetation indices, texture features, and topographic factors extracted from satellite multispectral images as independent variables and sample area biomass as the dependent variable to realize the inversion of road area vegetation biomass and obtain the road area biomass.

6. A method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in any one of claims 3-5, characterized in that: In step S321, the COPERT model is used to calculate traffic pollutant emissions, and the emission factors in the COPERT model are dynamically modified. The expression for the dynamically modified emission factors is as follows: ； in, Let be the emission factor for vehicle type i; The baseline emission factor for vehicle model i; For vehicle speed correction function; This refers to the emission standard correction factor. The temperature correction factor is applied using the Arrhenius formula: ； in, is the activation energy; R is the gas constant; Standard temperature; Real-time atmospheric temperature; Therefore, the total emissions calculated using the above dynamic correction formula for emission factors are: ； in, The total amount of pollutants emitted from road segment j; Let represent the total mileage traveled by vehicle type i on road segment j.

7. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 6, is characterized in that: In step S322, the CALPUFF Lagrange atmospheric diffusion model is used, and the emission inventory output by the COPERT model is used as the source strength input to simulate the daily and hourly pollutant concentrations at the vegetation sample plots. The pollutant concentrations include the concentrations of various single pollutants.

8. The method for accurate inversion of highway vegetation carbon storage based on exhaust gas stress correction using multimodal remote sensing, as described in claim 7, is characterized in that: In step S33, a dynamic correction model for carbon storage is constructed using the Logistic inhibition model, and the Logistic inhibition model is corrected by the pollutant concentration simulated by the CALPUFF Lagrange atmospheric diffusion model. The Logistic inhibition model includes photosynthetic inhibition effect and respiration promotion effect. The steps for correcting the photosynthetic inhibition effect are as follows: the Logistic inhibition model is used to quantify the inhibitory effect of a single pollutant on the photosynthetic rate, the weights are determined by the toxicity equivalent method and the synergistic effect of multiple pollutants is weighted and coupled, the proportion of photosynthetic rate decrease is calculated, and the carbon storage is corrected under the single pollutant and the synergistic effect of multiple pollutants. The steps for correcting the respiratory enhancement effect are as follows: the Logistic inhibition model is used to quantify the nonlinear relationship between respiration rate and pollutant concentration, the weights are determined by the toxicity equivalent method and the synergistic effect of multiple pollutants is weighted and coupled to achieve carbon storage correction under single pollutant and multi-pollutant synergistic effects.

9. A computer device comprising a memory, a processor, and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1-8.

10. A computer-readable storage medium having a computer program / instructions stored thereon, characterized in that: When the computer program / instructions are executed by the processor, they implement the steps of the method according to any one of claims 1-8.

Images