A new pollutant dynamic monitoring method and system based on multi-scale remote sensing and mechanism constraint probability inversion

Abstract

This invention discloses a novel pollutant dynamic monitoring method and system based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, belonging to the field of remote sensing information processing technology. The method first collects and preprocesses multi-scale remote sensing data, then builds a pollutant migration and diffusion model based on water hydrodynamics and environmental parameters to obtain a spatial prior estimate of the pollutants. It then integrates remote sensing observations, diffusion mechanisms, and spatial constraints to construct a pollutant probabilistic state field, establishing an environmental disturbance response model based on this field and remote sensing characteristics. Finally, it constructs and solves a joint state estimation model incorporating multiple constraints, dynamically updating the probabilistic state field based on observation residual feedback. This invention does not rely on the pollutant's own spectrum, transforming traditional concentration inversion into spatial state estimation, significantly improving the spatial continuity, temporal stability, and robustness of monitoring weakly spectral new pollutants in complex aquatic environments. It is applicable to the dynamic monitoring of new pollutants such as microplastics, persistent organic pollutants, and endocrine disruptors.

Description

Technical Field

[0001] This invention relates to the field of remote sensing information processing technology, and in particular to a novel method and system for dynamic monitoring of pollutants based on the coupling of multi-scale remote sensing data with environmental behavior mechanisms. Background Technology

[0002] With industrial development and the widespread use of new chemicals, the types and levels of new pollutants in the aquatic environment are continuously increasing, posing potential risks to ecosystem safety and human health. These pollutants are typically characterized by low concentrations, high concealment, complex migration and diffusion processes, and high uncertainty in environmental behavior. Some pollutants also exhibit long-distance migration, bioaccumulation, and recalcitrant degradation characteristics, thus becoming an important research direction in the field of current water environment monitoring.

[0003] Existing methods for monitoring new pollutants mainly rely on manual sampling, laboratory analysis, and offline detection technologies. Although they can achieve high detection accuracy, they generally suffer from problems such as limited monitoring range, insufficient spatiotemporal continuity, long monitoring cycles, and difficulty in achieving large-scale dynamic monitoring. These methods are insufficient to meet the needs for real-time monitoring and dynamic tracking of new pollutants in complex aquatic environments.

[0004] With the development of remote sensing technology, water environment monitoring methods based on satellite and UAV remote sensing are gradually being applied to the field of water quality parameter inversion. Existing remote sensing monitoring technologies mainly establish empirical relationships between remote sensing reflectance and pollutant concentrations to achieve inversion analysis of parameters with obvious spectral response characteristics, such as chlorophyll, colored soluble organic matter, and suspended solids. In recent years, research has also been conducted on remote sensing monitoring of new pollutants. For example, CN113295835A discloses a remote sensing-based method for identifying the spatial distribution of microplastic pollution in water bodies, which identifies areas with high microplastic abundance by collecting water samples and screening remote sensing features; CN115266632B discloses a UAV hyperspectral remote sensing method for investigating water pollution sources, which establishes a quantitative water quality model and analyzes diffusion characteristics through hyperspectral imagery; CN112464746A and CN117592005B, among others, use machine learning algorithms to establish a mapping relationship between remote sensing data and pollutant concentrations, improving inversion efficiency.

[0005] However, most novel pollutants exist in water bodies in dissolved, colloidal, or micro-scale particulate forms, lacking stable, significant, and directly identifiable remote sensing spectral characteristics. This makes the aforementioned remote sensing inversion methods based on direct spectral responses difficult to apply. CN113295835A still relies on the spectral differences between microplastics and water bodies, limiting its applicability to novel pollutants with low concentrations and weak spectral responses. CN115266632B uses a single UAV platform, making it difficult to balance large-scale monitoring and fine identification, and it does not incorporate pollutant migration and diffusion mechanisms. Pure data-driven methods such as CN112464746A and CN117592005B lack physical mechanism constraints, easily leading to problems such as local abnormal fluctuations, spatial discontinuities, and insufficient temporal stability in the inversion results.

[0006] Furthermore, existing remote sensing monitoring methods typically only establish empirical mapping relationships between remote sensing observation data and pollutant concentrations, failing to effectively combine the migration and diffusion mechanisms of pollutants under hydrodynamic conditions, environmental behavior characteristics, and spatial continuity constraints. In particular, for pollutants with weak spectral response, they usually cannot form stable and identifiable direct response characteristics in the remote sensing observation space. Relying solely on the direct mapping relationship between pollutant concentrations and remote sensing information makes it difficult to achieve stable monitoring and dynamic tracking of new pollutants in complex aquatic environments. Summary of the Invention

[0007] The purpose of this invention is to overcome the above-mentioned shortcomings in the prior art and provide a new method and system for dynamic monitoring of pollutants based on multi-scale remote sensing and mechanism-constrained probabilistic inversion. This method and system can estimate the spatial state based on the environmental physical disturbance characteristics caused by pollutants without relying on the spectral characteristics of the pollutants themselves. This improves the spatial continuity, temporal stability and multi-source remote sensing collaborative monitoring capabilities of new pollutants in complex aquatic environments. It can also achieve continuous estimation and stable monitoring of the spatial existence state, migration trend and dynamic evolution process of pollutants under conditions where pollutants cannot be directly observed by remote sensing.

[0008] To achieve the above-mentioned technical features, the objective of this invention is as follows: a novel dynamic monitoring method for pollutants based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, comprising the following steps: S1: Acquire multi-scale remote sensing data of the target water area, preprocess the multi-scale remote sensing data and unify it to the same spatial reference grid to construct a multi-scale remote sensing observation dataset; S2: Based on the hydrodynamic conditions and environmental parameters of the target water area, a pollutant migration and diffusion model is constructed, and the prior estimate of the spatial state of pollutants is obtained through numerical calculation. S3: Based on the multi-scale remote sensing observation dataset, extract one or more remote sensing observation feature information from water surface texture gradient, water surface roughness, local thermal anomaly, water surface wave intensity and apparent reflectance change, and integrate pollutant observation information, environmental behavior mechanism information and spatial continuity constraints to construct a probabilistic state field to characterize the possibility of pollutant existence and spatial distribution. S4: Based on the remote sensing physical characteristics in the multi-scale remote sensing observation dataset, establish an environmental physical disturbance response relationship model between the probability state field and the remote sensing physical characteristics, and realize the mapping of pollutant state to remotely observable physical quantities. S5: Based on the multi-scale remote sensing observation dataset, probabilistic state field, and environmental behavior mechanism information, a joint state estimation model is constructed that integrates remote sensing observation constraints, migration and diffusion mechanism constraints, multi-scale structural consistency constraints, and spatial continuity constraints. The spatial state of pollutants is optimized and solved to obtain the optimal probabilistic state field. S6: Based on the multi-temporal remote sensing observation dataset and observation residual feedback mechanism in the multi-scale remote sensing observation dataset, the optimal probability state field is dynamically updated to realize continuous dynamic state monitoring of the pollutant migration and diffusion process.

[0009] Preferably, in step S1, the multi-scale remote sensing data includes satellite remote sensing data and UAV remote sensing data; the satellite remote sensing data is used to acquire macroscopic environmental information of the water body, including temperature field distribution, water flow field structure, and regional scale apparent reflectance information; the UAV remote sensing data is used to acquire local high-resolution water surface microstructure feature information, including water surface texture gradient, water surface wave intensity, local roughness changes, and microscale thermal anomaly distribution information; the preprocessing includes radiometric correction, geometric correction, and spatial registration processing.

[0010] Preferably, in step S3, the probability state field is constructed using the following formula: ; in: Indicates the spatial location of pollutants and time The probability state value; This represents the observation probability state field extracted based on multi-scale remote sensing observation data; This represents the prior probability state field obtained based on the migration-diffusion model; Represents a probability mapping function; Indicates the weighting coefficient of the observation information; The prior weight coefficients of the mechanism are represented.

[0011] Preferably, in step S4, the remote sensing physical characteristics include one or more of the following: water surface texture gradient, local thermal anomaly distribution, water surface roughness, water surface wave intensity, and apparent reflectivity; the environmental physical disturbance response model is established based on the indirect disturbances to the physical state, hydrodynamic structure, and thermodynamic characteristics of the water surface caused by pollutants during the migration and diffusion of pollutants in the water body.

[0012] Preferably, in step S5, the joint state estimation model is optimized using the following objective function: ; in: Describe the joint inversion objective function; A nonlinear mapping function representing the probability state field and remote sensing reflectivity; A nonlinear mapping function representing the probabilistic state field and water surface roughness; This represents the actual observed value of remotely sensed reflectance; This represents the actual observed value of water surface roughness; This represents the probability state field to be optimized. This represents the prior probability state field obtained based on the migration-diffusion model; Indicates the weighting coefficients of the mechanism constraints; Indicates the weighting coefficients of spatial continuity constraints; Represents spatial continuity constraints; This represents a multi-scale structural consistency constraint term, used to constrain the structural consistency between the probability state fields corresponding to remote sensing observations at different scales. The multi-scale structural consistency weight coefficient is represented; the optimal probability state field is obtained by minimizing the objective function. The multi-scale structural consistency constraint term is expressed as follows: ; in: This represents a regional-scale probabilistic state field constructed based on satellite remote sensing data; This represents a local-scale probabilistic state field constructed based on UAV remote sensing data; This represents a scale transformation operator used to map the UAV-scale probabilistic state field to the satellite-scale space.

[0013] Preferably, in step S6, the dynamic recursive update is implemented using the following state recursive model: ; in: express The probability field of pollutants at any given moment; express The probability field of pollutants at any given moment; Indicates the time step; Represents the dynamic evolution function of pollutants; , They represent , directional water flow velocity component; Indicates the diffusion coefficient; Indicates the ambient temperature parameter; Indicates actual remote sensing observation characteristics; This indicates that the model predicts remote sensing features; Represents the observation feedback gain coefficient. This is the observation residual feedback correction term, used to dynamically correct the probability state field based on the residual between the remote sensing observation features and the predicted remote sensing features.

[0014] Preferably, the spatial risk distribution map of pollutants and the pollutant migration trajectory results are generated based on the dynamically recursively updated probability state field.

[0015] Preferably, the new pollutant includes any one of microplastics, persistent organic pollutants, endocrine disruptors, and antibiotics.

[0016] Preferably, the probability state field is used to characterize the probability of the presence of pollutants in the target area, their spatial diffusion state, and their dynamic migration trend, and the characterization result is a probability state estimate of the pollutants.

[0017] Preferably, the environmental physical disturbance response model is established based on the changes in water surface texture structure, water surface roughness, local thermal anomalies, water surface wave intensity, and apparent reflectivity caused by the migration and diffusion of pollutants.

[0018] Preferably, the multi-scale remote sensing data further includes one or more of synthetic aperture radar data, thermal infrared remote sensing data, and hyperspectral remote sensing data; the synthetic aperture radar data is used to acquire information on water surface roughness and wave structure; the thermal infrared remote sensing data is used to acquire the distribution of local thermal anomalies; and the hyperspectral remote sensing data is used to assist in identifying the apparent reflectance characteristics of water bodies.

[0019] Preferably, the optimization solution method for the objective function includes any one of gradient descent, genetic algorithm, particle swarm optimization algorithm, variational method, Bayesian state estimation method, and Kalman recursive update method.

[0020] Preferably, another aspect of the present invention provides a novel pollutant dynamic monitoring system based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, wherein the system, based on the novel pollutant dynamic monitoring method using the multi-scale remote sensing and mechanism-constrained probabilistic inversion, includes: The data acquisition and preprocessing module is used to acquire multi-scale remote sensing data of the target water area, preprocess the multi-scale remote sensing data and unify it to the same spatial reference grid, and construct a multi-scale remote sensing observation dataset. The prior estimation module is used to construct a pollutant migration and diffusion model based on the hydrodynamic conditions and environmental parameters of the target water area, and obtain the prior estimation results of the spatial state of pollutants through numerical calculation. The probability field construction module is used to construct a probability state field that characterizes the possibility of the existence and spatial distribution of pollutants based on the observation probability state field extracted from multi-scale remote sensing observation data and the prior probability state field obtained from the migration and diffusion model. The response model construction module is used to establish a model of the environmental physical disturbance response relationship between the probabilistic state field and the remote sensing physical characteristics, so as to realize the mapping of pollutant state to remotely observable physical quantities. The joint inversion module is used to construct a joint state estimation model that integrates remote sensing observation constraints, migration and diffusion mechanism constraints, multi-scale structural consistency constraints, and spatial continuity constraints, and optimizes the spatial state of pollutants to obtain the optimal probabilistic state field. The dynamic update module is used to dynamically update the optimal probability state field based on multi-temporal remote sensing observation data and observation residual feedback mechanism, so as to realize continuous dynamic state monitoring of pollutant migration and diffusion process.

[0021] Preferably, the data acquisition and preprocessing module includes a satellite data acquisition unit, a UAV data acquisition unit, and a preprocessing unit; the satellite data acquisition unit is used to acquire satellite remote sensing data, the UAV data acquisition unit is used to acquire UAV remote sensing data, and the preprocessing unit is used to perform radiometric correction, geometric correction, and spatial registration processing on the remote sensing data.

[0022] The present invention has the following beneficial effects: 1. This invention transforms the traditional pollutant concentration inversion problem into a pollutant spatial state estimation problem. By constructing a mechanism that integrates remote sensing observation information, prior information on environmental behavior mechanisms, and a probabilistic state field expression mechanism, it achieves stable state expression and dynamic characterization of new pollutants with low concentration and weak spectral response, effectively solving the core problem that traditional methods cannot monitor pollutants without obvious spectral characteristics.

[0023] 2. This invention abandons the direct identification method based on the spectral characteristics of pollutants themselves. Instead, it establishes an indirect perturbation coupling relationship between pollutants and environmental physical parameters such as water surface roughness, surface texture structure, local thermal anomalies, water surface wave intensity, and apparent reflectivity. This enables state estimation of pollutants that cannot be directly observed by remote sensing, significantly expanding the applicability of remote sensing monitoring methods in the field of new pollutants.

[0024] 3. This invention introduces the pollutant migration and diffusion mechanism constraint based on the convection-diffusion equation, so that the pollutant probabilistic state field satisfies mass conservation, hydrodynamic continuity and spatial diffusion law during the optimization process. This fundamentally avoids the non-physical oscillation problem that is prone to occur in traditional empirical models and pure data-driven models, and ensures that the state estimation results conform to the actual environmental evolution law.

[0025] 4. By constructing a multi-scale remote sensing data fusion mechanism based on a joint optimization framework, the regional-scale macroscopic information from satellite remote sensing and the local-scale fine information from UAV remote sensing are coordinated and constrained within a unified state space. This achieves an organic combination of macroscopic coverage and local fine monitoring, significantly improving the consistency and stability of cross-scale remote sensing information, as well as the ability to identify and spatially resolve the spatial state of pollutants in complex aquatic environments.

[0026] 5. By introducing a dynamic update mechanism based on state recursion and feedback correction of observation residuals, the spatial state of pollutants is extended from single-phase static estimation to modeling of a continuous time evolution process. This enables real-time dynamic tracking of pollutant migration trends and diffusion paths, and accurate identification of high-risk accumulation areas, thus realizing continuous dynamic monitoring and risk warning of pollutant migration and diffusion processes.

[0027] 6. This invention comprehensively considers the hydrodynamic conditions, environmental parameter changes, and observation noise in complex aquatic environments. Through multi-constraint joint optimization and dynamic correction mechanisms, it effectively improves the robustness of monitoring results under complex working conditions and can meet the needs of large-scale, long-term, and high-precision dynamic monitoring of new pollutants in practical engineering. Attached Figure Description

[0028] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0029] Figure 1 This is a flowchart illustrating the novel pollutant dynamic monitoring method of the present invention.

[0030] Figure 2 This is a schematic diagram illustrating the multi-scale remote sensing data and environmental status information fusion processing of the present invention.

[0031] Figure 3 This is a schematic diagram illustrating the construction of the spatial state parameters and probability state field of pollutants in this invention.

[0032] Figure 4 This is a schematic diagram of the joint state estimation model of multi-source remote sensing and environmental behavior mechanism of the present invention.

[0033] Figure 5 This is a schematic diagram illustrating the remote sensing response relationship of pollutants based on environmental physical disturbances, as presented in this invention.

[0034] Figure 6This is a schematic diagram of the time-series dynamic update of the pollutant probability state field of the present invention.

[0035] Figure 7 This is a schematic diagram of the pollutant dynamic state correction based on observation feedback according to the present invention. Detailed Implementation

[0036] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0037] Example 1: S1: Multi-scale Remote Sensing Data Acquisition and Processing Acquire multi-source remote sensing data of the target water area, including satellite remote sensing data and UAV remote sensing data; The remote sensing data is subjected to radiometric correction, geometric correction and spatial registration, and unified to the same spatial reference grid to construct a multi-scale remote sensing observation dataset; Among them, satellite remote sensing data is used to obtain macro-environmental information of water bodies, including temperature field distribution, water flow field structure and regional scale apparent reflectance information; UAV remote sensing data is used to obtain local high-resolution water surface microstructure characteristics, including water surface texture gradient, water surface wave intensity, local roughness changes and microscale thermal anomaly distribution information.

[0038] S2: Construction of Pollutant Migration and Diffusion Model Based on the hydrodynamic conditions and environmental parameters of the target water area, a pollutant migration and diffusion model is established to describe the transport, diffusion and attenuation process of pollutants in the water body. The prior estimate of the spatial state of pollutants is obtained through numerical calculation to characterize the potential existence area and diffusion direction of pollutants in the target area.

[0039] S3: Construction of the Probabilistic State Field Based on pollutant migration and diffusion models, hydrodynamic environmental parameters, local remote sensing physical disturbance characteristics, and prior state information of pollutants, an uncertain probability state field is constructed to describe the spatial existence state of pollutants. The probabilistic state field does not directly perform remote sensing inversion or simple normalization of pollutant concentration. Instead, it integrates pollutant observation information, prior information on environmental behavior mechanisms, spatial continuity constraints, and remote sensing observation physical disturbance information to express the probability of pollutant presence, migration trend, and spatial risk status in the target area in a state-based manner.

[0040] The probability state field is used to characterize the probability of existence, spatial diffusion state and dynamic migration trend of pollutants at different spatial locations, and serves as a unified state variable in the pollutant dynamic state estimation process, thereby realizing the coupled expression between the physical space of remote sensing observation and the state space of pollutant environmental behavior.

[0041] The remote sensing observation physical disturbance information includes one or more of the following: water surface texture gradient, water surface roughness change, local thermal anomaly distribution, water surface wave intensity, and apparent reflectance change.

[0042] The probability state field is used to characterize the dynamic state estimation result of the possibility of the existence of pollutants, rather than the actual concentration value of pollutants.

[0043] Based on multi-scale remote sensing observation data, remote sensing observation features such as water surface texture gradient, water surface roughness, local thermal anomaly distribution, water surface wave intensity, and apparent reflectance changes are extracted. A remote sensing observation probability state field is then constructed using normalized mapping, statistical probability mapping, or machine learning probability mapping methods. .

[0044] The probability state field is constructed using the following formula: ; in: Indicates the spatial location of pollutants and time The probability state value; This represents the observation probability state field extracted based on multi-scale remote sensing observation data, used to characterize the probability of pollutant presence reflected by remote sensing observation features; This represents the prior probability state field obtained based on the migration and diffusion model, used to characterize the theoretical distribution state of pollutants under the constraints of environmental behavior mechanisms; This represents a probability mapping function used to map state variables to a probability space; Indicates the weighting coefficient of the observation information; The prior weight coefficients of the mechanism are represented.

[0045] By observing the probability state field Mechanism prior probability state field By performing probability space fusion, a probabilistic state field of pollutants that takes into account both remote sensing observation information and environmental behavior mechanism constraints can be obtained. .

[0046] S4: Construction of Environmental Physical Disturbance Response Model Establish a model of the relationship between the probabilistic state field of pollutants and the physical characteristics observed by remote sensing, including the influence of pollutants on water surface roughness, surface texture structure, local thermal anomaly distribution, apparent reflectance and water surface wave intensity, so as to realize the mapping of pollutant state to remotely observable physical quantities. The environmental physical disturbance response model is not based on the direct identification of the pollutant's own spectral characteristics, but rather on the indirect disturbances to the physical state, hydrodynamic structure, and thermodynamic characteristics of the water surface caused by the pollutant's migration and diffusion in the water body, establishing a coupling relationship between the pollutant's state and remote sensing observation characteristics. The remote sensing physical features include one or more of the following: water surface texture gradient, local thermal anomaly, water surface roughness, water surface wave intensity, and apparent reflectivity.

[0047] The environmental physical disturbance response model is trained, calibrated, or optimized based on the multi-scale remote sensing observation dataset constructed in step S1 to establish a mapping relationship between pollutant state and remote sensing observation physical characteristics.

[0048] S5: Construction and Optimization Solution of Joint State Estimation Model A joint state estimation model integrating constraints from remote sensing observation, migration and diffusion mechanisms, spatial continuity, and multi-scale structural consistency is constructed to optimize the solution of the spatial state of pollutants. The following objective function is constructed for optimization: ; in: Describe the joint inversion objective function; A nonlinear mapping function representing the probability state field and remote sensing reflectivity; A nonlinear mapping function representing the probabilistic state field and water surface roughness; This represents the actual observed value of remotely sensed reflectance; This represents the actual observed value of water surface roughness; This represents the prior probability field obtained based on the migration-diffusion model; Indicates the weighting coefficients of the mechanism constraints; Indicates the weighting coefficients of spatial continuity constraints; Represents spatial continuity constraints; This represents a multi-scale structural consistency constraint term; This represents the weighting coefficient for multi-scale structural consistency.

[0049] Preferably, the spatial continuity constraint term can be expressed as: ; in: Represents a spatial neighborhood set; , This represents the probability state value corresponding to adjacent grid nodes.

[0050] Preferably, the multi-scale structural consistency constraint term can be expressed as: ; in: This represents a regional-scale probabilistic state field constructed based on satellite remote sensing data; This represents a local-scale probabilistic state field constructed based on UAV remote sensing data; This represents a scale transformation operator used to map the UAV-scale probabilistic state field to the satellite-scale space.

[0051] By minimizing the objective function, the consistency constraint between the physical characteristics of remote sensing observation and the dynamic state space of pollutants is achieved, thereby obtaining the optimal probabilistic state field distribution results that satisfy the environmental behavior mechanism, hydrodynamic continuity and remote sensing observation consistency, and generating the spatial risk distribution results of pollutants.

[0052] S6: Dynamic recursive update of the probability state field Based on multi-temporal remote sensing observation data and observation residual feedback mechanism, the optimal probability state field is dynamically recursively updated to realize dynamic state estimation of pollutant migration and diffusion process and analysis of pollutant migration trajectory evolution.

[0053] The following state recursion model is used to achieve dynamic recursive updates: ; in: express The probability field of pollutants at any given moment; express The probability field of pollutants at any given moment; Indicates the time step; Represents the dynamic evolution function of pollutants; , These represent the water flow velocity components; Indicates the diffusion coefficient; Indicates the ambient temperature parameter; Indicates actual remote sensing observation characteristics; This indicates that the model predicts remote sensing features; This represents the observation feedback gain coefficient.

[0054] Among them, by introducing the observation residual feedback term This enables dynamic correction of the probability state field, thereby improving the stability, temporal consistency, and reliability of pollutant migration trend prediction during dynamic pollutant monitoring.

[0055] The spatial risk distribution map and migration trajectory results of pollutants are generated based on the dynamically recursively updated probability state field.

[0056] Example 2: Based on the above embodiments, the present invention may also have the following optional embodiments: (1) The new pollutants may include, but are not limited to: persistent organic pollutants, such as perfluorinated and polyfluoroalkyl compounds (PFAS); endocrine disruptors; antibiotics; and microplastics. Different types of new pollutants correspond to different environmental behavior parameters (such as diffusion coefficient and attenuation coefficient) and remote sensing physical disturbance response parameters. The model parameter weights and the selection of remote sensing observation features can be adjusted according to the specific environmental behavior characteristics of the target pollutant.

[0057] (2) The multi-scale remote sensing data may also include: synthetic aperture radar data, used to acquire information on water surface roughness and wave structure; thermal infrared remote sensing data, used to acquire information on local thermal anomaly distribution; hyperspectral remote sensing data, used to assist in identifying the apparent reflectance characteristics of water bodies; and UAV low-altitude image data, used to acquire information on microscale water surface texture structure changes. The above-mentioned types of remote sensing data can be used individually or in combination according to monitoring needs to further improve the comprehensiveness and accuracy of multi-scale observation.

[0058] (3) The construction of the pollutant probabilistic state field can be achieved by: normalization mapping method, statistical probability mapping method, threshold function mapping method, state estimation method based on spatial neighborhood constraints, or recursive estimation method based on dynamic state space. All of the above methods can achieve the purpose of mapping the spatial state parameters of pollutants to the probability space. The appropriate construction method can be selected according to the noise level, data volume, and computing resources of the monitoring scenario.

[0059] (4) The objective function can be optimized using the following methods: gradient descent, genetic algorithm, particle swarm optimization, variational method, Bayesian state estimation method, or Kalman recursive update method. All of the above methods can minimize the joint inversion objective function. Gradient descent is suitable for fast solutions to large-scale data, while Bayesian state estimation method and Kalman recursive update method are more suitable for dynamic monitoring scenarios with significant uncertainties.

[0060] (5) The satellite remote sensing data is used for regional scale environmental state estimation, with a spatial scale of km; the UAV remote sensing data is used for local scale water surface microstructure state analysis, with a spatial scale of m; the two can be uniformly coupled through spatial interpolation, multi-resolution analysis or state-space fusion methods to achieve an organic combination of macro-regional coverage and local fine monitoring.

[0061] Example 3: Dynamic monitoring based on microplastic spatial state estimation In this embodiment, the new pollutant is specifically microplastics. Since microplastics typically exist in water bodies in particulate form, their spatial state can be characterized by the number of microplastic particles per unit volume, particle abundance, spatial aggregation degree, and migration trend. This embodiment does not directly perform remote sensing inversion of the actual concentration of microplastics, but rather, based on hydrodynamic environmental behavior, remote sensing physical disturbance characteristics, and a probabilistic state estimation model, it provides a state-based representation of the spatial existence and dynamic migration process of microplastics in the target water area.

[0062] S1: Multi-scale Remote Sensing Data Acquisition and Processing The following data were acquired: satellite remote sensing data, used to extract water temperature field and flow field information; UAV remote sensing data, used to extract water surface texture and reflectivity information; and field sampling data, used to obtain microplastic abundance parameters, particle size distribution, and local state calibration information.

[0063] The aforementioned remote sensing data were preprocessed, including atmospheric correction and radiometric correction; geometric correction and spatial registration; and all data were unified to the same spatial reference grid to construct a multi-scale remote sensing observation dataset.

[0064] S2: Construction and Calculation of Pollutant Migration and Diffusion Model Based on the hydrodynamic conditions and environmental parameters of the target water area, a pollutant migration and diffusion model is established to describe the transport, diffusion, and attenuation processes of pollutants in the water body. The pollutant migration process is described using a convection-diffusion model, the expression of which is as follows: ; in: Indicates spatial location and time The pollutant concentration is used to characterize the pollutant content in a unit volume of water. In the microplastic example, it can also be expressed as the abundance of microplastic particles per unit volume. , They represent direction and The directional velocity component of water flow is used to describe the horizontal flow characteristics of water bodies; It represents the diffusion coefficient of pollutants in water bodies and is used to characterize the diffusion ability of pollutants under the influence of turbulence and molecular diffusion. It represents the pollutant decay coefficient, used to characterize the degradation or transformation rate of microplastics under environmental influences; This represents the pollutant source term, used to characterize the intensity and spatial distribution of pollutant input, covering point source emissions, area source input, and internal biochemical processes. It can be quantified through fixed point source intensity, for sewage outlets, industrial wastewater discharge points, or river inlets, determined by the product of measured discharge flow and monitored concentration, using the following formula: ; in: The total source term intensity of all fixed point sources in the target water area represents the total mass of pollutants discharged into the water body by all fixed point sources per unit time. For spatial Dirac functions, used to define specific coordinates Apply input constraints at the location; Point source number, used to uniquely identify different fixed point sources; No. i The emission flow rate of a fixed point source, representing the emission flow rate per unit time from the first fixed point source. i The volume of wastewater discharged into the water body from a point source. No. i The pollutant emission concentration at a fixed point source outlet represents the mass of the target pollutant contained in a unit volume of wastewater.

[0065] Rainfall-driven non-point source loads: For pollutants carried by surface runoff (such as microplastics from urban roads), the pollutant flux entering water bodies with runoff is estimated using the rainfall runoff coefficient and land use type, combined with meteorological remote sensing data and the proportion of impervious surfaces, and calculated based on an empirical cumulative-scour model.

[0066] Boundary input and background values: For the upstream boundary of the study area, real-time hydrological station monitoring data or remote sensing inversion results of the upstream river section are used as inflow boundary conditions and transformed into equivalent source terms.

[0067] Source Item Spatial distribution Its intensity is determined by the geographical coordinates of the sewage outlet and varies with time. Driven by industrial emission cycles or rainfall events, and constrained by hydrodynamic mechanisms, probabilistic inversion models are guided to generate fluctuations that conform to physical laws on a spatiotemporal scale.

[0068] The migration-diffusion model equations were numerically solved using the finite difference method to obtain the spatial distribution of pollutant concentrations. Based on the spatial state parameters obtained from the migration-diffusion model, hydrodynamic constraints, and physical disturbance characteristics observed by remote sensing, a probabilistic state field for pollutants was constructed.

[0069] S3: Construction of the Probabilistic State Field Based on the spatial state parameters obtained from the migration-diffusion model, hydrodynamic constraints, and physical disturbance characteristics from remote sensing observations, an uncertain probabilistic state field is constructed to describe the spatial existence state of microplastics. The microplastic spatial state parameters are mapped to the probabilistic state field, and its expression is as follows: ; in: Represents the probability state field of pollutants; This represents the observation probability state field extracted based on multi-scale remote sensing observation data; This represents the prior probability state field obtained based on the migration-diffusion model; This represents a probability mapping function used to map state variables to the [0,1] probability space. The Sigmoid function or a piecewise linear normalization function is preferred. These are the weighting coefficients for the observation data, used to adjust the contribution of remote sensing observation state information to the probabilistic state field; The prior weight coefficients are used to adjust the contribution of prior information in the probabilistic state field of the migration-diffusion model.

[0070] By employing the above methods, the probabilistic state field of pollutants can simultaneously satisfy the constraints of remote sensing observation consistency, hydrodynamic migration continuity, and spatial smoothness, thereby improving the stability and noise resistance of spatial inversion of low-concentration new pollutants.

[0071] S4: Construction of Environmental Physical Disturbance Response Model Establish a model to model the relationship between the probabilistic state field of microplastics and the environmental physical disturbance response of remote sensing observations, for example: ; ; in: This represents the results of remote sensing reflectance response; This indicates the response results to water surface roughness. This represents a nonlinear mapping function between the probabilistic state field of pollutants and remote sensing reflectance. This represents a nonlinear mapping function between the probabilistic state field of pollutants and the surface roughness of water. Represents the probability state field of pollutants; The remote sensing observation features do not originate from the spectral characteristics of the microplastics themselves, but rather from the indirect perturbation response of the microplastics during migration and diffusion to water surface texture, surface roughness, local thermal anomalies, and apparent reflectivity. The nonlinear mapping function can be constructed using linear functions, polynomial functions, exponential functions, or machine learning fitting functions.

[0072] S5: Construction and Optimization Solution of Joint State Estimation Model A joint state estimation model is constructed that integrates constraints from remote sensing observation, migration and diffusion mechanisms, spatial continuity, and multi-scale structural consistency. The model is then optimized using the following objective function: ; in: Describe the joint inversion objective function; This represents the actual observed value of remotely sensed reflectance; This represents the actual observed value of water surface roughness; A nonlinear response mapping function representing the probability state field and remote sensing reflectivity; A nonlinear response mapping function representing the probabilistic state field and water surface roughness; This represents the prior probability field obtained based on the migration-diffusion model; This represents a spatial continuity constraint term, used to constrain the continuity of its spatial topology and suppress remote sensing observation noise; This represents a multi-scale structural consistency constraint term, used to constrain the structural consistency between the satellite-scale probabilistic state field and the UAV-scale probabilistic state field. This represents the weighting coefficient for multi-scale structural consistency. The mechanistic constraint weighting coefficient is used to adjust the contribution ratio of the prior probability field obtained based on the migration and diffusion model to the objective function; This represents the spatial regularization weight coefficient, used to adjust the optimization weights of the spatial constraint function to balance the goodness of fit of remote sensing data with spatial smoothness.

[0073] The first term constrains the consistency between pollutant state and remote sensing reflectance observation; the second term constrains the consistency between pollutant state and water surface roughness observation; the third term constrains the pollutant state to satisfy the migration and diffusion mechanism; the fourth term improves the continuity of spatial distribution and suppresses local anomalous noise; and the fifth term constrains the pollutant probability state fields corresponding to remote sensing observations at different spatial scales to maintain structural consistency, so that regional scale distribution characteristics and local scale distribution characteristics maintain a coordinated match in a unified probability state space, thereby improving the stability, spatial continuity, and multi-source remote sensing collaborative monitoring capabilities of cross-scale inversion results.

[0074] Preferably, the spatial continuity constraint term can be expressed as: ; in: Represents a spatial neighborhood set; , This represents the probability state value corresponding to adjacent grid nodes.

[0075] Preferably, the multi-scale structural consistency constraint term can be expressed as: ; in: This represents a regional-scale probabilistic state field constructed based on satellite remote sensing data; This represents a local-scale probabilistic state field constructed based on UAV remote sensing data; This represents a scale transformation operator used to map the UAV-scale probabilistic state field to the satellite-scale space.

[0076] By minimizing the objective function, the optimal pollutant probabilistic state field distribution results that satisfy the constraints of remote sensing observation consistency, hydrodynamic mechanism, and spatial continuity are obtained.

[0077] S6: Dynamic recursive update of the probability state field Based on multi-temporal remote sensing observation data and an observation residual feedback mechanism, the optimal probability state field is dynamically and recursively updated, and its expression is as follows: ; in: express The probability field of pollutants at any given moment; express The probability field of pollutants at any given moment; Indicates the time step; Represents the dynamic evolution function of pollutants; , These represent the water flow velocity components; Indicates the pollutant diffusion coefficient; Indicates the ambient temperature parameter; It represents the actual comprehensive remote sensing observation characteristics, including one or more of the following: reflectivity, temperature field, water surface roughness, and texture structure parameters; This represents the remote sensing feature results obtained based on the pollutant probability state field prediction; This represents the observation feedback gain coefficient, used to dynamically correct the pollutant probability state field based on the remote sensing observation residuals; in, This is the observation residual feedback correction term, used to dynamically correct the pollutant probability state field based on remote sensing observation errors.

[0078] By introducing an observation feedback correction mechanism, the dynamic evolution results of pollutants can simultaneously satisfy the constraints of historical state continuity, remote sensing observation consistency, and environmental behavior mechanism, thereby improving the temporal stability and prediction reliability of the pollutant dynamic monitoring process.

[0079] Based on the above dynamic update results of the probability state field, a spatial risk distribution map of microplastics, a migration trend map, and the identification results of high-risk pollutant aggregation areas can be further generated.

[0080] Example 4: PFAS Pollutant Monitoring The difference compared to Example 3 is as follows: PFAS (per- and polyfluoroalkyl compounds) mainly exist in dissolved form in water bodies, and their spatial migration is more easily affected by hydrodynamic conditions, temperature field changes, and interfacial transport processes. Therefore, this embodiment does not rely on direct remote sensing identification based on the spectral characteristics of PFAS themselves, but rather on the indirect physical disturbance characteristics of PFAS during migration and diffusion, which affect local temperature distribution, water surface microstructure, and hydrodynamic field changes, to perform probabilistic state estimation of the spatial existence state of PFAS.

[0081] In constructing the environmental physical disturbance response model, the weight of apparent reflectivity response is weakened, while the constraints of temperature field distribution, water surface structure parameters, and local dynamic texture features are strengthened. Specifically, temperature field parameters are preferably obtained from thermal infrared remote sensing data, and water surface structure parameters are preferably extracted from UAV low-altitude remote sensing imagery and synthetic aperture radar data.

[0082] In the pollutant migration and diffusion model, parameters such as the diffusion coefficient and attenuation coefficient are adjusted according to the environmental behavior characteristics of PFAS. The remaining steps are exactly the same as in Example 3.

[0083] Example 5: This embodiment provides a novel pollutant dynamic monitoring system based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, including: The data acquisition and preprocessing module is used to acquire multi-scale remote sensing data of the target water area, preprocess the multi-scale remote sensing data and unify it to the same spatial reference grid, and construct a multi-scale remote sensing observation dataset. The prior estimation module is used to construct a pollutant migration and diffusion model based on the hydrodynamic conditions and environmental parameters of the target water area, and obtain the prior estimation results of the spatial state of pollutants through numerical calculation. The probability field construction module is used to integrate pollutant observation information, environmental behavior mechanism information and spatial continuity constraints to construct a probability state field that characterizes the possibility of pollutant existence and spatial distribution. The response model construction module is used to establish an environmental physical disturbance response relationship model between the probability state field and the remote sensing physical characteristics, so as to realize the mapping of pollutant state to remotely observable physical quantities. The joint inversion module is used to construct a joint state estimation model that integrates remote sensing observation constraints, migration and diffusion mechanism constraints, multi-scale structural consistency constraints, and spatial continuity constraints, and optimizes the spatial state of pollutants to obtain the optimal probabilistic state field. The dynamic update module is used to dynamically update the optimal probability state field based on multi-temporal remote sensing observation data and observation residual feedback mechanism, so as to realize continuous dynamic state monitoring of pollutant migration and diffusion process.

[0084] Furthermore, the data acquisition and preprocessing module includes a satellite data acquisition unit, a UAV data acquisition unit, and a preprocessing unit; the satellite data acquisition unit is used to acquire satellite remote sensing data, the UAV data acquisition unit is used to acquire UAV remote sensing data, and the preprocessing unit is used to perform radiometric correction, geometric correction, and spatial registration processing on the remote sensing data.

[0085] Furthermore, the probability field construction module is specifically used to construct a probability state field using the following formula: ; in: Represents the probability state field of pollutants; This represents the observation probability state field extracted based on multi-scale remote sensing observation data; This represents the prior probability state field calculated based on the pollutant migration and diffusion mechanism model; Represents a probability mapping function; These are the weighting coefficients for the observed data; These are the prior weighting coefficients for the mechanism.

[0086] Furthermore, the joint inversion module is specifically used to optimize the solution using the following objective function: ; in: Describe the joint inversion objective function; This represents the actual observed value of remotely sensed reflectance; This represents the actual observed value of water surface roughness; A nonlinear response mapping function representing the probability state field and remote sensing reflectivity; A nonlinear response mapping function representing the probabilistic state field and water surface roughness; This represents the prior probability field obtained based on the migration-diffusion model; This represents a spatial continuity constraint term, used to constrain the continuity of its spatial topology and suppress remote sensing observation noise; This represents a multi-scale structural consistency constraint term, used to constrain the structural consistency between the satellite-scale probabilistic state field and the UAV-scale probabilistic state field. This represents the weighting coefficient for multi-scale structural consistency. The mechanistic constraint weighting coefficient is used to adjust the contribution ratio of the prior probability field obtained based on the migration and diffusion model to the objective function; This represents the spatial regularization weight coefficient, used to adjust the optimization weights of the spatial constraint function to balance the goodness of fit of remote sensing data with spatial smoothness.

[0087] Preferably, the multi-scale structural consistency constraint term can be expressed as: ; in: This represents a regional-scale probabilistic state field constructed based on satellite remote sensing data; This represents a local-scale probabilistic state field constructed based on UAV remote sensing data; This represents a scale transformation operator used to map the UAV-scale probabilistic state field to the satellite-scale space.

[0088] Furthermore, the dynamic update module is specifically used to implement dynamic recursive updates through the following state recursive model: ; in: express The probability field of pollutants at any given moment; express The probability field of pollutants at any given moment; Indicates the time step; Represents the dynamic evolution function of pollutants; , These represent the water flow velocity components; Indicates the pollutant diffusion coefficient; Indicates the ambient temperature parameter; It represents the actual comprehensive remote sensing observation characteristics, including one or more of the following: reflectivity, temperature field, water surface roughness, and texture structure parameters; This represents the remote sensing feature results obtained based on the pollutant probability state field prediction; This represents the observation feedback gain coefficient, used to dynamically correct the pollutant probability state field based on the remote sensing observation residuals.

[0089] in, This is the observation residual feedback correction term, used to dynamically correct the pollutant probability state field based on remote sensing observation errors.

[0090] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A novel dynamic monitoring method for pollutants based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, characterized in that, Includes the following steps: S1: Acquire multi-scale remote sensing data of the target water area, preprocess the multi-scale remote sensing data and unify it to the same spatial reference grid to construct a multi-scale remote sensing observation dataset; S2: Based on the hydrodynamic conditions and environmental parameters of the target water area, a pollutant migration and diffusion model is constructed, and the prior estimate of the spatial state of pollutants is obtained through numerical calculation. S3: Based on the multi-scale remote sensing observation dataset, extract one or more remote sensing observation feature information from water surface texture gradient, water surface roughness, local thermal anomaly, water surface wave intensity and apparent reflectance change, and integrate pollutant observation information, environmental behavior mechanism information and spatial continuity constraints to construct a probabilistic state field to characterize the possibility of pollutant existence and spatial distribution. S4: Based on the remote sensing physical characteristics in the multi-scale remote sensing observation dataset, establish an environmental physical disturbance response relationship model between the probability state field and the remote sensing physical characteristics, and realize the mapping of pollutant state to remotely observable physical quantities. S5: Based on the multi-scale remote sensing observation dataset, probabilistic state field, and environmental behavior mechanism information, a joint state estimation model is constructed that integrates remote sensing observation constraints, migration and diffusion mechanism constraints, multi-scale structural consistency constraints, and spatial continuity constraints. The spatial state of pollutants is optimized and solved to obtain the optimal probabilistic state field. S6: Based on the multi-temporal remote sensing observation dataset and observation residual feedback mechanism in the multi-scale remote sensing observation dataset, the optimal probability state field is dynamically updated to realize continuous dynamic state monitoring of the pollutant migration and diffusion process.

2. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, In step S1, the multi-scale remote sensing data includes satellite remote sensing data and UAV remote sensing data; the satellite remote sensing data is used to acquire macro-environmental information of the water body, including temperature field distribution, water flow field structure, and regional scale apparent reflectance information; the UAV remote sensing data is used to acquire local high-resolution water surface microstructure feature information, including water surface texture gradient, water surface wave intensity, local roughness changes, and microscale thermal anomaly distribution information; the preprocessing includes radiometric correction, geometric correction, and spatial registration processing.

3. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, In step S3, the probability state field is constructed using the following formula: ； in: Indicates the spatial location of pollutants and time The probability state value; This represents the observation probability state field extracted based on multi-scale remote sensing observation data; This represents the prior probability state field obtained based on the migration-diffusion model; Represents a probability mapping function; Indicates the weighting coefficient of the observation information; The prior weight coefficients of the mechanism are represented.

4. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, In step S4, the remote sensing physical characteristics include one or more of the following: water surface texture gradient, local thermal anomaly distribution, water surface roughness, water surface wave intensity, and apparent reflectivity; the environmental physical disturbance response model is established based on the indirect disturbances to the physical state, hydrodynamic structure, and thermodynamic characteristics of the water surface caused by pollutants during the migration and diffusion of pollutants in the water body.

5. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, In step S5, the joint state estimation model is optimized using the following objective function: ； in: Describe the joint inversion objective function; A nonlinear mapping function representing the probability state field and remote sensing reflectivity; A nonlinear mapping function representing the probabilistic state field and water surface roughness; This represents the actual observed value of remotely sensed reflectance; This represents the actual observed value of water surface roughness; This represents the probability state field to be optimized. This represents the prior probability state field obtained based on the migration-diffusion model; Indicates the weighting coefficients of the mechanism constraints; Indicates the weighting coefficients of spatial continuity constraints; Represents spatial continuity constraints; This represents a multi-scale structural consistency constraint term, used to constrain the structural consistency between the probability state fields corresponding to remote sensing observations at different scales. The multi-scale structural consistency weight coefficient is represented; the optimal probability state field is obtained by minimizing the objective function. The multi-scale structural consistency constraint term is expressed as follows: ； in: This represents a regional-scale probabilistic state field constructed based on satellite remote sensing data; This represents a local-scale probabilistic state field constructed based on UAV remote sensing data; This represents a scale transformation operator used to map the UAV-scale probabilistic state field to the satellite-scale space.

6. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, In step S6, the dynamic recursive update is implemented through the following state recursive model: ； in: express The probability field of pollutants at any given moment; express The probability field of pollutants at any given moment; Indicates the time step; Represents the dynamic evolution function of pollutants; , They represent , directional water flow velocity component; Indicates the diffusion coefficient; Indicates the ambient temperature parameter; Indicates actual remote sensing observation characteristics; This indicates that the model predicts remote sensing features; Represents the observation feedback gain coefficient. This is the observation residual feedback correction term, used to dynamically correct the probability state field based on the residual between the remote sensing observation features and the predicted remote sensing features.

7. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 6, characterized in that, The spatial risk distribution map and migration trajectory results of pollutants are generated based on the dynamically recursively updated probability state field.

8. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, The new pollutants include any one of microplastics, persistent organic pollutants, endocrine disruptors, and antibiotics.

9. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, The probability state field is used to characterize the probability of the presence of pollutants in the target area, their spatial diffusion state, and their dynamic migration trend. The characterization result is the probability state estimate of the pollutants.

10. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 1, characterized in that, The environmental physical disturbance response model is established based on the changes in water surface texture, water surface roughness, local thermal anomalies, water surface wave intensity, and apparent reflectivity caused by the migration and diffusion of pollutants.

11. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 2, characterized in that, The multi-scale remote sensing data also includes one or more of synthetic aperture radar data, thermal infrared remote sensing data, and hyperspectral remote sensing data; the synthetic aperture radar data is used to acquire information on water surface roughness and wave structure; the thermal infrared remote sensing data is used to acquire the distribution of local thermal anomalies; and the hyperspectral remote sensing data is used to assist in identifying the apparent reflectance characteristics of water bodies.

12. The novel pollutant dynamic monitoring method based on multi-scale remote sensing and mechanism-constrained probabilistic inversion according to claim 5, characterized in that, The optimization solution method for the objective function includes any one of the following: gradient descent, genetic algorithm, particle swarm optimization algorithm, variational method, Bayesian state estimation method, and Kalman recursive update method.

13. A novel pollutant dynamic monitoring system based on multi-scale remote sensing and mechanism-constrained probabilistic inversion, characterized in that, The system is implemented based on the method of any one of claims 1 to 12, comprising: The data acquisition and preprocessing module is used to acquire multi-scale remote sensing data of the target water area, preprocess the multi-scale remote sensing data and unify it to the same spatial reference grid, and construct a multi-scale remote sensing observation dataset. The prior estimation module is used to construct a pollutant migration and diffusion model based on the hydrodynamic conditions and environmental parameters of the target water area, and obtain the prior estimation results of the spatial state of pollutants through numerical calculation. The probability field construction module is used to construct a probability state field that characterizes the possibility of the existence and spatial distribution of pollutants based on the observation probability state field extracted from multi-scale remote sensing observation data and the prior probability state field obtained from the migration and diffusion model. The response model construction module is used to establish a model of the environmental physical disturbance response relationship between the probabilistic state field and the remote sensing physical characteristics, so as to realize the mapping of pollutant state to remotely observable physical quantities. The joint inversion module is used to construct a joint state estimation model that integrates remote sensing observation constraints, migration and diffusion mechanism constraints, multi-scale structural consistency constraints, and spatial continuity constraints, and optimizes the spatial state of pollutants to obtain the optimal probabilistic state field. The dynamic update module is used to dynamically update the optimal probability state field based on multi-temporal remote sensing observation data and observation residual feedback mechanism, so as to realize continuous dynamic state monitoring of pollutant migration and diffusion process.

14. The new pollutant dynamic monitoring system according to claim 13, characterized in that, The data acquisition and preprocessing module includes a satellite data acquisition unit, a UAV data acquisition unit, and a preprocessing unit; the satellite data acquisition unit is used to acquire satellite remote sensing data, the UAV data acquisition unit is used to acquire UAV remote sensing data, and the preprocessing unit is used to perform radiometric correction, geometric correction, and spatial registration processing on the remote sensing data.

Images