Early warning method and system based on ecological data processing

Abstract

The invention discloses an early warning method and system based on ecological data processing, and particularly relates to the field of ecological data early warning. According to the method, the value of the missing data point can be predicted more accurately by improving the data interpolation method, and the adaptability prediction capability of ecological meteorology is improved by reflecting the dynamic nature and complexity of an ecological system by adopting a more complex ecological model.

Description

Technical Field

[0001] The present invention relates to the field of ecological data early warning, and in particular to an early warning method and system based on ecological data processing. Background Art

[0002] Air pollution monitoring and management are key areas of environmental protection. These efforts rely mainly on remote sensing image data obtained by earth observation satellites, as well as atmospheric pollutant concentrations, meteorological data, and environmental quality indicators collected by ground monitoring stations. Satellite remote sensing technology is widely used because it can cover a wide area and provide continuous time series data. The integration of these data and advanced data processing methods provide atmospheric scientists with tools to deeply understand the dynamics and changing patterns of air pollution.

[0003] Advances in remote sensing technology, especially high-resolution imaging technology, have greatly improved the spatial resolution of data, allowing atmospheric scientists to more accurately monitor small-scale changes in atmospheric pollution. The application of multispectral and hyperspectral sensors has enhanced the ability to identify the types and concentrations of atmospheric pollutants and the assessment of environmental quality.

[0004] In terms of data processing, existing technologies are already able to process large-scale data sets and improve the efficiency of data processing through automated data processing processes. This process includes steps such as data preprocessing, classification, feature extraction and analysis. Data preprocessing involves operations such as denoising, correction and atmospheric correction to ensure data quality. Classification technology is used to convert remote sensing image data into useful information, such as the distribution of atmospheric pollutants and the identification of emission sources. Feature extraction technology further analyzes the classification results and extracts parameters useful for atmospheric pollution research, such as air quality index, pollutant emissions and emission rates.

[0005] Although existing technologies have made significant progress in atmospheric pollution data processing, some challenges still exist. First, the quality of remote sensing images may be affected by factors such as cloud occlusion, sensor noise, and atmospheric conditions, resulting in discontinuity and incompleteness of the data. Second, the distribution of ground monitoring sites may be uneven, especially in remote or difficult-to-reach areas, which limits the spatial representativeness of the data. In addition, existing data processing methods, such as traditional spatial interpolation and temporal interpolation techniques, may not be able to fully handle the complexity and heterogeneity of the data, especially when faced with large-scale and high-dimensional data sets, and face the problem of low computational efficiency.

[0006] To overcome these challenges, the field of air pollution monitoring and management needs to further develop and improve remote sensing technology, ground monitoring networks and data processing algorithms to improve the accuracy, representativeness and processing efficiency of data, so as to better respond to air pollution problems and protect and improve environmental quality. Summary of the invention

[0007] The purpose of the present invention is to provide an early warning method and system based on ecological data processing, and to improve the adaptive prediction ability of ecological meteorology by constructing an ecological niche suitability model and an atmospheric dynamic model through machine learning technology.

[0008] The present invention is achieved through the following technical solutions:

[0009] An early warning method based on ecological data processing comprises the following steps:

[0010] S1. Acquire multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm;

[0011] S2. Define spatial data points and their spatial positions, and fill in missing values ​​in the data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the values ​​of unknown points for blank areas in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the values ​​of missing points for missing data points in a time series through known time series data according to cubic spline interpolation;

[0012] S3. Extract features from the data after time data interpolation, calculate the air quality index through atmospheric parameters as environmental variable data; establish an ecological niche suitability model through environmental variable data, calculate the ecological niche width and predict the ecological niche suitability through the ecological niche suitability model;

[0013] S4. Establish an atmospheric dynamic model based on atmospheric parameter data and predict atmospheric change trends through the atmospheric dynamic model;

[0014] S5. Based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model as features, the features are enhanced through generative adversarial networks, and the atmospheric change patterns are identified through deep neural networks.

[0015] Furthermore, the step S1 specifically includes the following sub-steps:

[0016] S101. Adaptively crop the remote sensing image data by using GIS;

[0017] S102. Reducing spatial deviation by adjusting multi-phase images;

[0018] S103. Correcting the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data by an atmospheric correction algorithm;

[0019] S104. Detect and process clouds and shadows using a pixel reflectance value threshold method.

[0020] Furthermore, the step S2 specifically includes the following sub-steps:

[0021] S201. Define spatial data point Z = {z1, z2…z n} and its spatial position S = {s1,s2…s n}, wherein Z represents a spatial data point, n represents the total number of known data points, and S represents the spatial position of the spatial data point;

[0022] S202. Identify missing areas in spatial data and perform spatial interpolation using Kriging: Wherein, Z(s0) represents the predicted value at the unknown position, and Z(s i ) represents the predicted value at the known position, the λ i represents the weight calculated according to the spatial correlation, and n represents the total number of known data points;

[0023] S203. Identify missing points in the time series and perform time interpolation by cubic spline interpolation: Wherein, S(t) represents the interpolation function at time point t, t represents the time point, and a i represents the coefficient of the polynomial, i ranges from 0, 1, 2, and 3, corresponding to the constant term, the first term, the second term, and the third term, respectively. j Represents the spline basis function B 3,j (t), the j represents the weight of the basis function, the B 3,j (t) represents the cubic B-spline basis function, which is used to create a smooth curve between data points.

[0024] Furthermore, feature extraction is performed on the data after time data interpolation processing, and the atmosphere quality index is calculated through atmospheric parameters as environmental variable data, which specifically includes the following sub-steps:

[0025] S3011. Calculate the air quality index through atmospheric parameters. The process is as follows:

[0026] AQI=max(I p );

[0027]

[0028] Among them, the I p represents a sub-index of pollutant p, where p represents a pollutant that affects air quality, and C prepresents the actual concentration of pollutant p, the B high and B low Respectively represent the breakpoint value of pollutant concentration, the I high and I low Respectively represent the corresponding B high and B low Air quality index;

[0029] S3012. Wavelet transform to extract features of atmospheric parameter time series: Among them, the W f (a, b) represents the wavelet transform coefficients, the ψ * represents the complex conjugate of the wavelet function, a represents a scaling parameter, b represents a translation parameter, f(t) represents a signal, and t represents a time point.

[0030] Furthermore, in step S3, an ecological niche suitability model is established through environmental variable data, and ecological niche suitability prediction is performed through the ecological niche suitability model, which specifically includes the following sub-steps:

[0031] S3021. Estimating niche breadth from standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche suitability model;

[0032] S3022. Estimate the parameters of the logistic regression model based on maximum likelihood, and establish a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x1, x2, ..., x p}, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β0, β1…β p represents the model parameters, x1, x2…x p represents a subset of the niche breadth X.

[0033] Furthermore, the step S4 specifically includes the following sub-steps:

[0034] S401. Through the Lotka-Volterra equation, a mathematical model describing atmospheric dynamics is established:

[0035]

[0036]

[0037]

[0038] Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state;

[0039] S402. Predict the future trend of atmospheric parameters using the Euler algorithm: Wherein, X(t+h) represents the predicted atmospheric parameters at the future time point t+h, X(t) represents the actual atmospheric parameters at the time point t, Represents the atmospheric parameter vector [T,H,P,…] at time t n The rate of change of , and h represents the time step.

[0040] Furthermore, the step S5 specifically includes:

[0041] S501. Obtain the niche width prediction results of the niche suitability model and the atmospheric change trend prediction results of the atmospheric dynamics model, and construct a training set;

[0042] S502. Perform data enhancement on the features in the training set by minimizing the loss between the generator and the discriminator through the generative adversarial network: Wherein, G represents the generator network, D represents the discriminator network, and P data Represents the probability distribution of real data, the P z represents the probability distribution of random noise, V(D,G) represents the value function, x represents the real sample, Represents the real data distribution P data (x), the expectation of the sample x under Represents the potential space distribution z~P z (z) is the expectation of sample x under G z represents the generator output, and z represents the distribution from the latent space P z (z) samples drawn from;

[0043] S503. Build and train a deep learning model through the training set processed by data enhancement, and identify atmospheric change patterns.

[0044] Furthermore, an early warning system based on ecological data processing is proposed. The system is implemented based on any of the above-mentioned early warning methods based on ecological data processing, and includes:

[0045] The data acquisition module is used to obtain multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm;

[0046] A data preprocessing module defines spatial data points and their spatial positions, and fills in missing values ​​in a data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the values ​​of unknown points for blank areas in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the values ​​of missing points for missing data points in a time series through known time series data according to cubic spline interpolation;

[0047] The feature extraction module is used to extract features from the data after time data interpolation processing, and calculate the atmospheric quality index as environmental variable data through atmospheric parameters;

[0048] Model building module, used to build ecological niche suitability model and atmospheric dynamics prediction model respectively;

[0049] The atmospheric change pattern prediction module is used to enhance the features through generative adversarial networks based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model, and to identify the atmospheric change pattern through deep neural networks;

[0050] The model building module specifically includes:

[0051] The niche fitness prediction unit is used to establish the niche fitness model through environmental variable data, calculate the niche width and predict the niche fitness through the niche fitness model;

[0052] The atmospheric dynamic model prediction unit is used to establish an atmospheric dynamic model through the niche width prediction results and environmental variable data, and to predict the atmospheric change trend through the atmospheric dynamic model.

[0053] Furthermore, the ecological niche suitability prediction unit specifically includes:

[0054] The niche breadth prediction subunit is used to estimate the niche breadth using the standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche suitability model;

[0055] The model building subunit is used to estimate the parameters of the logistic regression model based on maximum likelihood and to build a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x1, x2, ..., x p}, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β0, β1…β p represents the model parameters, x1, x2…x p represents a subset of the niche breadth X.

[0056] Furthermore, the atmospheric dynamic model prediction unit specifically includes:

[0057] The atmospheric dynamics model building subunit is used to build a mathematical model describing the atmospheric dynamics through the Lotka-Volterra equation:

[0058]

[0059]

[0060]

[0061] Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state;

[0062] The change trend prediction subunit is used to predict the change trend of future atmospheric parameters through the Euler algorithm: Wherein, X(t+h) represents the predicted atmospheric parameters at the future time point t+h, X(t) represents the actual atmospheric parameters at the time point t, Represents the atmospheric parameter vector [T,H,P,…] at time t n The rate of change of , and h represents the time step.

[0063] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0064] The present invention can more accurately predict the values ​​of missing data points by improving the data interpolation method, and improve the adaptive prediction ability of ecological meteorology by adopting a more complex ecological model to reflect the dynamics and complexity of the ecosystem. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] The drawings described herein are used to provide a further understanding of the embodiments of the present invention, constitute a part of this application, and do not constitute a limitation of the embodiments of the present invention. In the drawings:

[0066] Figure 1 The present invention is a method flow chart of an early warning method based on ecological data processing. DETAILED DESCRIPTION

[0067] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings, but the protection scope of the present invention is not limited to the following.

[0068] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention is further described in detail in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.

[0069] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely represents selected embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative work are within the scope of protection of the present invention. It should be noted that relational terms such as the terms "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations.

[0070] Moreover, the terms "comprises", "includes" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or apparatus that includes a list of elements includes not only those elements, but also includes other elements not explicitly listed, or also includes elements inherent to such process, method, article or apparatus. In the absence of more restrictions, an element defined by the phrase "comprises a ..." does not exclude the presence of additional identical elements in the process, method, article or apparatus that includes the element.

[0071] The features and performance of the present invention are further described in detail below in conjunction with the embodiments.

[0072] An early warning method based on ecological data processing, such as Figure 1 , including the following steps:

[0073] S1. Acquire multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm;

[0074] S2. Define spatial data points and their spatial positions, and fill in missing values ​​in the data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the values ​​of unknown points for blank areas in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the values ​​of missing points for missing data points in a time series through known time series data according to cubic spline interpolation;

[0075] S3. Extract features from the data after time data interpolation, calculate the air quality index through atmospheric parameters as environmental variable data; establish an ecological niche suitability model through environmental variable data, calculate the ecological niche width and predict the ecological niche suitability through the ecological niche suitability model;

[0076] S4. Establish an atmospheric dynamic model based on atmospheric parameter data and predict atmospheric change trends through the atmospheric dynamic model;

[0077] S5. Based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model as features, the features are enhanced through generative adversarial networks, and the atmospheric change patterns are identified through deep neural networks.

[0078] Furthermore, the step S1 specifically includes the following sub-steps:

[0079] S101. Adaptively crop the remote sensing image data by using GIS;

[0080] S102. Reducing spatial deviation by adjusting multi-temporal images;

[0081] S103. Correcting the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data by an atmospheric correction algorithm;

[0082] S104. Detect and process clouds and shadows using a pixel reflectance value threshold method.

[0083] Furthermore, the step S2 specifically includes the following sub-steps:

[0084] S201. Define spatial data point Z = {z1, z2…z n} and its spatial position S = {s1,s2…s n}, wherein Z represents a spatial data point, n represents the total number of known data points, and S represents the spatial position of the spatial data point;

[0085] S202. Identify missing areas in spatial data and perform spatial interpolation using Kriging: Wherein, Z(s0) represents the predicted value at the unknown position, and Z(s i ) represents the predicted value at the known position, the λ i represents the weight calculated according to the spatial correlation, and n represents the total number of known data points;

[0086] S203. Identify missing points in the time series and perform time interpolation by cubic spline interpolation: Wherein, S(t) represents the interpolation function at time point t, t represents the time point, and a i represents the coefficient of the polynomial, i ranges from 0, 1, 2, and 3, corresponding to the constant term, the first term, the second term, and the third term, respectively. j Represents the spline basis function B 3,j (t), the j represents the weight of the basis function, the B 3,j (t) represents the cubic B-spline basis function, which is used to create a smooth curve between data points.

[0087] Specifically, for the above embodiment, for spatial interpolation, the missing spatial data points are estimated by the Kriging method, and the values ​​of unknown points are predicted by known data points based on the spatial correlation between the data points. Furthermore, time interpolation processes the missing values ​​in the time series, and fills the data through time series prediction models such as linear interpolation and polynomial interpolation.

[0088] Furthermore, in step S3, feature extraction is performed on the data after the time data interpolation processing, and the air quality index is calculated by the atmospheric parameters as the environmental variable data, which specifically includes the following sub-steps:

[0089] S3011. Calculate the air quality index through atmospheric parameters. The process is as follows:

[0090] AQI=max(I p );

[0091]

[0092] Among them, the I p represents a sub-index of pollutant p, where p represents a pollutant that affects air quality, and C p represents the actual concentration of pollutant p, the B high and B low Respectively represent the breakpoint value of pollutant concentration, the I high and I low Respectively represent the corresponding B high and B lowThe air quality index, wherein the breakpoint values ​​are used to determine the concentration limits of different AQI levels, each pollutant has a series of breakpoint values, which divide the pollutant concentration range into different intervals, each interval corresponds to a specific AQI value according to; and B high and B low and its corresponding I high and I low Determined based on historical data, scientific research, health impact assessments, and environmental protection standards;

[0093] S3012. Wavelet transform to extract features of atmospheric parameter time series: Among them, the W f (a, b) represents the wavelet transform coefficients, the ψ * represents the complex conjugate of the wavelet function, a represents a scaling parameter, b represents a translation parameter, f(t) represents a signal, specifically represents a sequence data of atmospheric parameters changing with time, and t represents a time point;

[0094] S3013. Establish a feature set based on the features extracted by wavelet transform, other environmental parameters and the calculated air quality index, and use the feature set as the environmental variable data env, wherein the other environmental parameters include at least temperature, humidity and atmospheric pressure.

[0095] Specifically, the atmospheric health status and pollutant concentrations are evaluated by calculating the normalized pollutant index, and further, the multi-scale features in the time series data are extracted through wavelet transform.

[0096] Furthermore, in step S3, an ecological niche suitability model is established through environmental variable data, and ecological niche suitability prediction is performed through the ecological niche suitability model, which specifically includes the following sub-steps:

[0097] S3021. Estimating niche breadth from standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche suitability model;

[0098] S3022. Estimate the parameters of the logistic regression model based on maximum likelihood, and establish a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x1, x2, ..., x p}, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β0, β1…β p represents the model parameters, x1, x2…x p represents a subset of the niche breadth X.

[0099] Specifically, for the above-mentioned embodiments, by constructing a mathematical relationship between environmental variables and atmospheric states, the variability or adaptability of the atmosphere in the environmental variable space is quantified. The model uses environmental variables extracted from ground monitoring stations and meteorological data, such as temperature, humidity, atmospheric pressure, and pollutant concentrations, to estimate the niche width of the atmosphere by calculating the standard deviation of the variables. Niche width, as a key indicator, reflects the fluctuation range and adaptability of the atmospheric environment in the face of changes in environmental variables. Through the niche suitability model, the suitability of the atmosphere under different environmental conditions is predicted, which not only includes the current atmospheric state, but also covers potential trends and ecological risks. The model uses statistical methods such as logistic regression to link environmental variables with atmospheric suitability and predict the response of the atmosphere to environmental changes.

[0100] Furthermore, as a preferred implementation scheme of this embodiment, a specific implementation process is proposed for the above maximum likelihood estimation, wherein the maximum likelihood estimation is used to estimate the parameters of the logistic regression model: Said Represents the estimated value of the parameter after maximum likelihood estimation, the y i Indicates the atmospheric observation value. In addition, the specific process for the prediction of ecological niche suitability is expressed as: DegradationWarning = f(P i (x)), wherein f represents the early warning function, exemplarily, according to the niche suitability P i (x) Determine the warning level.

[0101] Furthermore, the step S4 specifically includes the following sub-steps:

[0102] S401. Through the Lotka-Volterra equation, a mathematical model describing atmospheric dynamics is established:

[0103]

[0104]

[0105]

[0106] Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state;

[0107] S402. Predict the future trend of atmospheric parameters using the Euler algorithm: Wherein, X(t+h) represents the predicted atmospheric parameters at the future time point t+h, X(t) represents the actual atmospheric parameters at the time point t, Represents the atmospheric parameter vector [T,H,P,…] at time t n The rate of change of , and h represents the time step.

[0108] Specifically, the atmospheric dynamics model is used to simulate and predict changes in the atmosphere in an ecosystem, for example, by simulating the interaction between environmental variables and atmospheric changes through the Lotka-Volterra equation.

[0109] Furthermore, the parameters of the influence function of temperature, humidity and atmospheric pressure on the atmospheric state in the above embodiment are determined by the nonlinear least squares method: Said represents the estimated value of the parameter after nonlinear least squares method, Represents the atmospheric pollutant concentration predicted by the model.

[0110] Furthermore, the step S5 specifically includes:

[0111] S501. Obtain the niche width prediction results of the niche suitability model and the atmospheric change trend prediction results of the atmospheric dynamics model, and construct a training set;

[0112] S502. Perform data enhancement on the features in the training set by minimizing the loss between the generator and the discriminator through the generative adversarial network: Wherein, G represents the generator network, D represents the discriminator network, and P data Represents the probability distribution of real data, the P z represents the probability distribution of random noise, V(D,G) represents the value function, x represents the real sample, Represents the real data distribution P data (x), the expectation of the sample x under Represents the potential space distribution z~P z (z) is the expectation of sample x under G z represents the generator output, and z represents the distribution from the latent space P z (z) samples drawn from;

[0113] S503. Build and train a deep learning model through the training set processed by data enhancement, and identify atmospheric change patterns.

[0114] Specifically, the deep learning model includes but is not limited to a convolutional neural network (CNN), a recurrent neural network (RNN), a long short-term memory network (LSTM) or a Transformer model. The niche width prediction results provide information on the adaptability of atmospheric parameters, while the atmospheric change trend prediction results provide dynamic information on the changes of atmospheric parameters over time. The niche width prediction results and the atmospheric change trend prediction results serve as input features of the deep learning model to help the model learn the complex patterns of atmospheric states. The deep learning model identifies and predicts atmospheric change patterns, such as the development of pollution events and extreme weather events, through the above features.

[0115] Furthermore, as a preferred implementation scheme of the above embodiment, a Markov random field optimization prediction method is proposed for identifying ecological degradation patterns after prediction by a deep learning model, which specifically includes the following steps:

[0116] S601. Fusion of atmospheric parameter features predicted by the deep learning model with atmospheric environmental variable data obtained by wavelet transform and AQI calculation;

[0117] S602. Based on the integrated feature data, define an energy function of the Markov random field, wherein the energy function includes node potential energy and interaction terms between adjacent nodes;

[0118] S603. According to the energy function, a Markov random field model based on the probability of the atmospheric state is established. The Markov random field model is trained by the Gibbs sampling algorithm to obtain the probability distribution of the spatial data;

[0119] S604. Predict atmospheric conditions using a Markov random field model, including but not limited to the changing trend of the air quality index.

[0120] Furthermore, the interaction term between the node potential energy and the adjacent nodes is specifically expressed as:

[0121]

[0122] Wherein, E(x) represents the interaction term between the point potential energy and the adjacent nodes, and U(x) i ) represents the potential energy of node i, and J ij represents the interaction between nodes i and j, and the δ(·) represents the indicator function.

[0123] Furthermore, the Gibbs sampling algorithm for training the Markov random field model is specifically expressed as:

[0124]

[0125] Among them, the x i(t+1) represents the state of node i at time t+1, and P(x i |x \i ) means that given all other nodes x \i The probability distribution of node i in the state, the x \i represents the state of all nodes except node i, T represents the temperature parameter, and E(x i |x \i ) means that given all other nodes x \i The energy function of node i under the state.

[0126] Specifically, the above embodiment is used to simulate the spatial distribution and correlation of air pollution, and to generate spatial distribution samples of air pollution through Gibbs sampling, and the spatial distribution samples are used to predict and evaluate the risk of air pollution.

[0127] Furthermore, an early warning system based on ecological data processing is proposed. The system is implemented based on any of the above-mentioned early warning methods based on ecological data processing, and includes:

[0128] The data acquisition module is used to obtain multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm;

[0129] A data preprocessing module defines spatial data points and their spatial positions, and fills in missing values ​​in a data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the values ​​of unknown points for blank areas in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the values ​​of missing points for missing data points in a time series through known time series data according to cubic spline interpolation;

[0130] The feature extraction module is used to extract features from the data after time data interpolation processing, and calculate the atmospheric quality index as environmental variable data through atmospheric parameters;

[0131] Model building module, used to build ecological niche suitability model and atmospheric dynamics prediction model respectively;

[0132] The atmospheric change pattern prediction module is used to enhance the features through generative adversarial networks based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model, and to identify the atmospheric change pattern through deep neural networks;

[0133] The model building module specifically includes:

[0134] The niche fitness prediction unit is used to establish the niche fitness model through environmental variable data, calculate the niche width and predict the niche fitness through the niche fitness model;

[0135] The atmospheric dynamic model prediction unit is used to establish an atmospheric dynamic model through the niche width prediction results and environmental variable data, and to predict the atmospheric change trend through the atmospheric dynamic model.

[0136] Furthermore, the ecological niche suitability prediction unit specifically includes:

[0137] The niche breadth prediction subunit is used to estimate the niche breadth using the standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche suitability model;

[0138] The model building subunit is used to estimate the parameters of the logistic regression model based on maximum likelihood and to build a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x1, x2, ..., x p}, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β0, β1…β p represents the model parameters, x1, x2…x p represents a subset of the niche breadth X.

[0139] Furthermore, the atmospheric dynamic model prediction unit specifically includes:

[0140] The atmospheric dynamics model building subunit is used to build a mathematical model describing the atmospheric dynamics through the Lotka-Volterra equation:

[0141]

[0142]

[0143]

[0144] Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state;

[0145] The change trend prediction subunit is used to predict the change trend of future atmospheric parameters through the Euler algorithm: Wherein, X(t+h) represents the predicted atmospheric parameters at the future time point t+h, X(t) represents the actual atmospheric parameters at the time point t, Represents the atmospheric parameter vector [T,H,P,…] at time t n The rate of change of , and h represents the time step.

[0146] The specific implementation methods described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation method of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. An early warning method based on ecological data processing, It is characterized in that The following steps are involved: S1. Acquire multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm; S2. Define spatial data points and their spatial positions, and fill in missing values ​​in the data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the value of unknown points for blank areas in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the value of missing points for missing data points in a time series through known time series data according to cubic spline interpolation; S3. Extract features from the data after time data interpolation, calculate the air quality index through atmospheric parameters as environmental variable data; establish an ecological niche suitability model through environmental variable data, calculate the ecological niche width and predict the ecological niche suitability through the ecological niche suitability model; S4. Establish an atmospheric dynamic model based on atmospheric parameter data and predict atmospheric change trends through the atmospheric dynamic model; S5. Based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model as features, the features are enhanced through generative adversarial networks, and the atmospheric change patterns are identified through deep neural networks.

2. An early warning method based on ecological data processing as claimed in claim 1, It is characterized in that The step S1 specifically includes the following sub-steps: S101. Adaptively crop the remote sensing image data by using GIS; S102. Reducing spatial deviation by adjusting multi-phase images; S103. Correcting the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data by an atmospheric correction algorithm; S104. Detect and process clouds and shadows using a pixel reflectance value threshold method.

3. An early warning method based on ecological data processing as claimed in claim 1, It is characterized in that The step S2 specifically includes the following sub-steps: S201. Define spatial data point Z = {z 1 , z 2 …z n } and its spatial position S = {s 1 ,s 2 …s n }, wherein Z represents a spatial data point, n represents the total number of known data points, and S represents the spatial position of the spatial data point; S202. Identify missing areas in spatial data and perform spatial interpolation using Kriging: Among them, the Z(s 0 ) represents the predicted value at the unknown position, the Z(s i ) represents the predicted value at the known position, the λ i represents the weight calculated according to the spatial correlation, and n represents the total number of known data points; S203. Identify missing points in the time series and perform time interpolation by cubic spline interpolation: Wherein, S(t) represents the interpolation function at time point t, t represents the time point, and a i represents the coefficient of the polynomial, i ranges from 0, 1, 2, and 3, corresponding to the constant term, the first term, the second term, and the third term, respectively. j Represents the spline basis function B 3，j (t), the j represents the weight of the basis function, the B 3，j (t) represents the cubic B-spline basis function, which is used to create a smooth curve between data points.

4. The early warning method based on ecological data processing as claimed in claim 1, It is characterized in that In the step S3, feature extraction is performed on the data after the time data interpolation processing, and the air quality index is calculated by the atmospheric parameters as the environmental variable data, which specifically includes the following sub-steps: S3011. Calculate the air quality index through atmospheric parameters. The process is as follows: AQI=max(I p ); Among them, the I p represents a sub-index of pollutant p, where p represents a pollutant that affects air quality, and C p represents the actual concentration of pollutant p, the B high and B low Respectively represent the breakpoint value of pollutant concentration, the I high and I low Respectively represent the corresponding B high and B low Air quality index; S3012. Wavelet transform to extract features of atmospheric parameter time series: Among them, the W f (a, b) represents the wavelet transform coefficients, the ψ * represents the complex conjugate of the wavelet function, a represents a scaling parameter, b represents a translation parameter, f(t) represents a signal, and t represents a time point.

5. The early warning method based on ecological data processing as claimed in claim 1, It is characterized in that In step S3, an ecological niche suitability model is established through environmental variable data, and ecological niche suitability prediction is performed through the ecological niche suitability model, which specifically includes the following sub-steps: S3021. Estimating niche breadth from the standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche fitness model; S3022. Estimate the parameters of the logistic regression model based on maximum likelihood, and establish a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x 1 、x 2 , …, x p }, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β 0 , β 1 …β p represents the model parameters, the x 1 、x 2 …x p represents a subset of the niche breadth X.

6. The early warning method based on ecological data processing as claimed in claim 1, It is characterized in that The step S4 specifically includes the following sub-steps: S401. Through the Lotka-Volterra equation, a mathematical model describing atmospheric dynamics is established: Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state; S402. Predict the future trend of atmospheric parameters using the Euler algorithm: Wherein, X(t+k) represents the predicted atmospheric parameters at the future time point t+h, and N i (t) represents the actual atmospheric parameters at time point t, represents the atmospheric parameter vector [T, H, P, ...] at time t n The rate of change of , and h represents the time step.

7. The early warning method based on ecological data processing as claimed in claim 1, It is characterized in that The step S5 specifically includes: S501. Obtain the niche width prediction results of the niche suitability model and the atmospheric change trend prediction results of the atmospheric dynamics model, and construct a training set; S502. Perform data enhancement on the features in the training set by minimizing the loss between the generator and the discriminator through the generative adversarial network: Wherein, G represents the generator network, D represents the discriminator network, and P data Represents the probability distribution of real data, the P z represents the probability distribution of random noise, V(D, G) represents the value function, x represents the real sample, Represents the real data distribution P data (x), the expectation of the sample x under Represents the potential space distribution z~P z (z) is the expectation of sample x under G z represents the generator output, and z represents the distribution from the latent space P z (z) samples drawn from; S503. Build and train a deep learning model through the training set processed by data enhancement, and identify atmospheric change patterns.

8. An early warning system based on ecological data processing, the system is implemented based on an early warning method based on ecological data processing according to any one of claims 1 to 7, It is characterized in that include: The data acquisition module is used to obtain multi-temporal and multi-spectral remote sensing image data through earth observation satellites, wherein the remote sensing image data is atmospheric remote sensing images, and obtain atmospheric parameter data, and correct the cloud layer, atmospheric scattering and absorption that affect the remote sensing image data through an atmospheric correction algorithm; A data preprocessing module defines spatial data points and their spatial positions, and fills in missing values ​​in a data set through spatiotemporal data interpolation, wherein the spatiotemporal data interpolation includes spatial data interpolation and temporal data interpolation. The spatial data interpolation is to predict the value of an unknown point for a blank area in geographic spatial data through known data points according to the Kriging method, and the temporal data interpolation is to predict the value of a missing point for a missing data point in a time series through known time series data according to cubic spline interpolation; The feature extraction module is used to extract features from the data after time data interpolation processing, and calculate the atmospheric quality index as environmental variable data through atmospheric parameters; Model building module, used to build ecological niche suitability model and atmospheric dynamics prediction model respectively; The atmospheric change pattern prediction module is used to enhance the features through generative adversarial networks based on the niche width prediction results of the niche fitness model and the atmospheric change trend prediction results of the atmospheric dynamics model, and to identify the atmospheric change pattern through deep neural networks; The model building module specifically includes: The niche fitness prediction unit is used to establish the niche fitness model through environmental variable data, calculate the niche width and predict the niche fitness through the niche fitness model; The atmospheric dynamic model prediction unit is used to establish an atmospheric dynamic model through the niche width prediction results and environmental variable data, and to predict the atmospheric change trend through the atmospheric dynamic model.

9. An early warning system based on ecological data processing as claimed in claim 8, It is characterized in that The ecological niche suitability prediction unit specifically includes: The niche breadth prediction subunit is used to estimate the niche breadth using the standard deviation of environmental variable data: Wherein, X represents the niche width of atmospheric parameter i, represents the standard deviation of the atmospheric environmental variable i, and the niche width is used as an input parameter of the niche fitness model; The model building subunit is used to estimate the parameters of the logistic regression model based on maximum likelihood and to build a logistic regression model to predict the ecological niche suitability of the atmosphere: X={x 1 、x 2 , …, x p }, where P(x) represents the probability of suitable atmosphere under given environmental variable x, and β 0 , β 1 …β p represents the model parameters, the x 1 、x 2 …x p represents a subset of the niche breadth X.

10. An early warning system based on ecological data processing as claimed in claim 8, It is characterized in that The atmospheric dynamic model prediction unit specifically includes: The atmospheric dynamics model building subunit is used to build a mathematical model describing the atmospheric dynamics through the Lotka-Volterra equation: Wherein, dT, dH and dP represent the change rates of temperature T, humidity H and atmospheric pressure P respectively, and F T 、F H and F P Respectively represent the influence functions of temperature, humidity and atmospheric pressure on the atmospheric state; The change trend prediction subunit is used to predict the change trend of future atmospheric parameters through the Euler algorithm: Wherein, X(t+h) represents the predicted atmospheric parameters at the future time point t+h, X(t) represents the actual atmospheric parameters at the time point t, represents the atmospheric parameter vector [T, H, P, ...] at time t n The rate of change of , and h represents the time step.

Images