Atmospheric pollution source intelligent identification and tracking method and related equipment thereof

Abstract

The invention belongs to the technical field of atmospheric pollution source identification and tracking, and particularly provides an atmospheric pollution source intelligent identification and tracking method and related equipment thereof, and the method mainly comprises the steps: obtaining real-time atmospheric pollution data of a plurality of sensors, carrying out the preprocessing of the real-time atmospheric pollution data, and generating spatial continuous pollution data; performing feature extraction operation based on the spatial continuous pollution data, and outputting a simplified feature vector; pollution source preliminary identification is carried out on the simplified feature vector, and the position of a refining pollution source is determined; a tracking path is generated according to the position of the refining pollution source, and a smooth prediction path is obtained through smoothing processing. According to the invention, accurate identification and real-time tracking of the mobile pollution source can be realized, the identification accuracy and tracking precision are improved, the response delay is reduced, the method can adapt to a complex urban environment, and efficient support is provided for air pollution management and control.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of atmospheric pollution source identification and tracking, and particularly relates to an atmospheric pollution source intelligent identification and tracking method and a related device thereof. BACKGROUND

[0002] The current identification and tracking of atmospheric pollution sources are often carried out with the aid of a fixed sensor network, and pollution-related data are collected by arranging sensors in a specific area to provide basic information for subsequent pollution monitoring and tracing. The fixed sensor network has certain applications in the scene of urban atmospheric pollution control.

[0003] When facing a mobile pollution source, the arrangement characteristics of the fixed sensor network can cause sparsity of data collection, and various disturbances in the environment can bring noise to the collected data. In addition, pollutants from different sources are prone to mixing, which can cause real-time identification of the mobile pollution source to be difficult to achieve ideal results, the tracking accuracy is affected, and there is a delay in response speed. Especially in a complex urban environment with dense building distribution and heavy traffic flow, it is not conducive to timely control of the mobile pollution source, thereby affecting the efficiency of atmospheric pollution control. SUMMARY

[0004] The application provides an atmospheric pollution source intelligent identification and tracking method and a related device thereof, which effectively solves the problems of poor identification effect, low tracking accuracy, and response delay caused by data sparsity, noise interference, and multi-source mixing of the fixed sensor network when facing a mobile pollution source, especially in a complex urban environment. The application realizes accurate identification and real-time tracking of the mobile pollution source, improves the identification accuracy and tracking accuracy, reduces the response delay, and can adapt to a complex urban environment to provide efficient support for atmospheric pollution control.

[0005] To achieve the above purpose, the application adopts the following technical solutions:

[0006] In a first aspect, the application provides an atmospheric pollution source intelligent identification and tracking method, comprising:

[0007] Obtaining real-time atmospheric pollution data of a plurality of sensors, preprocessing the real-time atmospheric pollution data, and generating spatially continuous pollution data.

[0008] Performing a feature extraction operation based on the spatially continuous pollution data, and outputting a simplified feature vector.

[0009] Performing preliminary identification of the pollution source on the simplified feature vector, and determining a refined pollution source position.

[0010] Generating a tracking path according to the refined pollution source position, and obtaining a smoothed predicted path through smoothing processing.

[0011] The smoothed prediction path is updated and adjusted in combination with real-time atmospheric pollution data to obtain a verification path.

[0012] The verification path is output and visualized to form a final output.

[0013] Further, the real-time atmospheric pollution data is preprocessed to generate spatially continuous pollution data, including:

[0014] The real-time atmospheric pollution data collected by multiple sensors is filtered by a Gaussian filter to eliminate random errors and obtain clean pollution data.

[0015] The clean pollution data is subjected to timestamp alignment processing to unify the time reference of each sensor data and obtain time-synchronized pollution data.

[0016] The time-synchronized pollution data is subjected to spatial interpolation to generate the spatially continuous pollution data.

[0017] Further, a feature extraction operation is performed based on the spatially continuous pollution data to output a simplified feature vector, including:

[0018] The spatially continuous pollution data is analyzed using a polynomial fitting method to extract pollution trend data.

[0019] The peak values in the pollution trend data are identified by a local extreme value detection algorithm to obtain pollution peak value data.

[0020] The pollution peak value data is subjected to feature quantization using a principal component analysis method to obtain a pollution feature vector.

[0021] The pollution feature vector is subjected to dimensionality reduction processing using linear discriminant analysis to obtain the simplified feature vector.

[0022] Further, the simplified feature vector is subjected to preliminary identification of pollution sources to determine a refined pollution source location, including:

[0023] The simplified feature vector is subjected to clustering analysis to obtain pollution clustering results.

[0024] The pollution clustering results are compared with a pollution source type database by a Euclidean distance matching algorithm to obtain potential pollution source types.

[0025] The potential pollution source types are subjected to confidence evaluation to obtain confirmed pollution sources.

[0026] The position of the confirmed pollution sources is subjected to optimized calculation using a least squares method to output a refined pollution source location.

[0027] Further, a tracking path is generated according to the refined pollution source location, and a smoothed prediction path is obtained through smoothing processing, including:

[0028] Processing the refined pollution source position data at continuous time points by a gradient descent method to obtain a pollution source movement direction; based on the time sequence of the refined pollution source position, a difference calculation method is used to derive a pollution source movement speed.

[0029] Combining the pollution source movement direction and movement speed, a Kalman filtering algorithm is used for path prediction to output a predicted path.

[0030] Performing a sliding average processing on the predicted path to output a smoothed predicted path.

[0031] Further, the smoothed predicted path is updated and adjusted in combination with real-time atmospheric pollution data to obtain a verified path, including:

[0032] Comparing the smoothed predicted path with real-time sensor data through residual analysis to output path deviation data.

[0033] Using a regression analysis method to process the path deviation data to output path adjustment parameters.

[0034] Based on the adjustment parameters, the smoothed predicted path is corrected using a Gaussian process regression to output a corrected path.

[0035] Using a cross-validation method to verify the rationality of the corrected path to output a verified path.

[0036] Further, the verified path is output and visualized to form a final output, including:

[0037] Visualizing the verified path through a heat map rendering technique to obtain a visualized tracking path.

[0038] Combining the visualized tracking path, confirming the pollution source and refining the pollution source position, a pollution source tracking report is generated.

[0039] Packaging and transmitting the pollution source tracking report in JSON format to a monitoring platform to output the final result.

[0040] In a second aspect, the present application provides an atmospheric pollution source intelligent identification and tracking system, comprising:

[0041] Data processing and preprocessing module: obtaining real-time atmospheric pollution data of multiple sensors, preprocessing the real-time atmospheric pollution data to generate spatial continuous pollution data.

[0042] Feature extraction module: performing feature extraction operation based on the spatial continuous pollution data to output a simplified feature vector.

[0043] Pollution source preliminary identification: performing pollution source preliminary identification on the simplified feature vector to determine a refined pollution source position.

[0044] Path generation and smoothing module: generate tracking path according to the refined pollution source position, and obtain smoothed prediction path through smoothing processing.

[0045] Path update verification: update and adjust the smoothed prediction path in combination with real-time atmospheric pollution data to obtain a verified path.

[0046] Result output visualization module: output and visualize the verified path to form a final output.

[0047] In a third aspect, the present application provides an atmospheric pollution source intelligent identification and tracking device, comprising a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program to realize the steps of the atmospheric pollution source intelligent identification and tracking method.

[0048] In a fourth aspect, the present application provides a storage medium, which stores computer program instructions, and the computer program instructions are read and executed by a processor to perform the steps of the atmospheric pollution source intelligent identification and tracking method.

[0049] In a fifth aspect, the present application provides a computer program product, which comprises a computer program or instructions, and the computer program or instructions are executed by a processor to realize the steps of the atmospheric pollution source intelligent identification and tracking method.

[0050] Advantages of the present application:

[0051] The present application adopts the scheme of obtaining real-time sensor data and preprocessing to generate spatial continuous data, extracting features, preliminarily identifying pollution source positions, generating and smoothing tracking paths, updating and adjusting verified paths in combination with real-time data, and outputting visualization, which effectively solves the problems of poor identification effect, low tracking precision, and response delay of fixed sensor networks in the prior art when facing mobile pollution sources due to sparse data, noise interference, and mixed multiple sources, especially in complex urban environments, realizes accurate identification and real-time tracking of mobile pollution sources, improves identification accuracy and tracking precision, reduces response delay, and can adapt to complex urban environments, providing efficient support for atmospheric pollution control.

[0052] Other features and advantages of the present application will be set forth in the following description, and in part will become apparent to those skilled in the art from the description, or can be learned by practice of the present application. The objects and other advantages of the present application can be realized and obtained by the structure indicated in the specification and drawings. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative effort.

[0054] Figure 1 A flowchart of a method for intelligent identification and tracking of an atmospheric pollution source is shown.

[0055] Figure 2 A module diagram of a system for intelligent identification and tracking of an atmospheric pollution source is shown. DETAILED DESCRIPTION

[0056] In order to solve the problems raised in the background art, the present application generates spatially continuous pollution data by acquiring real-time atmospheric pollution data of multiple sensors and preprocessing, outputs a simplified feature vector through feature extraction, generates and smooths a tracking path after identifying and refining the location of the pollution source, updates and visualizes the verification path in combination with real-time data, can accurately identify mobile pollution sources, and improves tracking accuracy and reduces delay.

[0057] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the following will clearly and completely describe the technical solutions in the embodiments of the present application with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application, but not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without any creative effort are within the scope of protection of the present application.

[0058] In some embodiments, as shown in Figure 1 The present application provides a method for intelligent identification and tracking of an atmospheric pollution source, comprising:

[0059] S1. Acquire real-time atmospheric pollution data of multiple sensors, preprocess the real-time atmospheric pollution data, and generate spatially continuous pollution data.

[0060] S2. Perform feature extraction operation based on the spatially continuous pollution data, and output a simplified feature vector.

[0061] S3. Perform preliminary identification of the pollution source on the simplified feature vector, and determine the refined pollution source location.

[0062] S4. Generate a tracking path according to the refined pollution source location, and obtain a smoothed predicted path through smoothing processing.

[0063] S5. Update and adjust the smoothed predicted path in combination with real-time atmospheric pollution data, and obtain a verification path.

[0064] S6. Output and visualize the results of the verification path to form the final output.

[0065] The real-time atmospheric pollution data refers to raw monitoring data collected by a plurality of atmospheric pollution sensors deployed in the monitoring area in real time, including concentration values of one or more atmospheric pollutants (such as PM2.5, PM10, SO2, NO2, O3, CO, etc.) measured by each sensor, a device identifier unique to each sensor, a specific timestamp of data collection, and geographic location information of the latitude and longitude of each sensor.

[0066] In some embodiments, the preprocessing of the real-time atmospheric pollution data in S1 to generate spatially continuous pollution data includes:

[0067] S11. Noise filtering of the real-time atmospheric pollution data collected by the plurality of sensors by Gaussian filtering to eliminate random errors to obtain clean pollution data.

[0068] Specifically, a Gaussian convolution kernel is constructed, and the convolution kernel is convolved with the original concentration data of each sensor to obtain the denoised concentration values.

[0069] The metadata information such as the sensor device number, the data collection timestamp, and the sensor geographic coordinates is retained, and finally the clean pollution data is obtained, which includes the denoised concentration values, the sensor device number, the data collection timestamp, and the sensor geographic coordinates.

[0070] S12. Time stamp alignment processing is performed on the clean pollution data to unify the time reference of each sensor data to obtain time-synchronized pollution data.

[0071] The linear interpolation method is used to perform time alignment calculation on the concentration values to obtain the aligned concentration values under the unified time reference.

[0072] The sensor device number and the sensor geographic coordinate information are inherited to generate a unified time sequence.

[0073] The time-synchronized pollution data is finally obtained, which includes the aligned concentration values, the unified time sequence, the sensor location information, and the sensor device number.

[0074] S13. Spatial interpolation is performed on the time-synchronized pollution data to generate spatially continuous pollution data.

[0075] The Kriging interpolation method is used, the sensor geographic coordinates are taken as the known point positions, and the corresponding aligned concentration values are taken as the known point values. The entire monitoring area is subjected to gridding interpolation calculation, and the spatial grid concentration values and the interpolation uncertainty of each grid point are output.

[0076] According to the preset spatial range and resolution, a regular latitude and longitude grid coordinate covering the entire monitoring area is generated.

[0077] The final obtained spatial continuous pollution data includes spatial grid concentration values, latitude and longitude grid coordinates, interpolation uncertainty, and data validity time.

[0078] In some embodiments, the feature extraction operation based on the spatial continuous pollution data in S2 outputs a simplified feature vector, including:

[0079] S21. A polynomial fitting method is used to analyze the variation trend of the spatial continuous pollution data, and pollution trend data is extracted.

[0080] The spatial continuous pollution data is processed using a polynomial fitting method, a trend surface reflecting the regional background pollution distribution is fitted, the original data is subtracted from the trend surface, and pollution trend data is obtained, which includes trend residual values R and corresponding spatial coordinates (x, y).

[0081] S22. Peak values in the pollution trend data are identified by a local extreme value detection algorithm to obtain pollution peak data.

[0082] A sliding window method is used to scan the entire region. For a given point (x0, y0), its window Ω is a rectangular region centered on it with a width of ω. If R(x0, y0) > R(x, y), and R(x0, y0) > T, the point is identified as a peak value; where R(x0, y0) represents the trend residual value calculated at spatial coordinate point (x0, y0), and T represents a preset intensity threshold, which is determined by calculating the statistical distribution of all trend residual values R(x, y).

[0083] The final obtained pollution peak data includes a set of spatial coordinates of peak points {(x p ,y p )} and corresponding trend residual values {R(x p ,y p )}.

[0084] S23. Principal component analysis is used to quantize the pollution peak data to obtain a pollution feature vector.

[0085] The pollution peak data is constructed into an original feature matrix X with a size of m×n, where m is the number of peak points and n is the feature dimension.

[0086] The original feature matrix X is centralized to obtain a centralized feature matrix X c , representing the column mean vector of the original feature matrix X.

[0087] Calculate X c The covariance matrix Σ,

[0088] By solving the eigenvalue problem Σv j =λ j v j , thus obtaining λ j and v j , λ j v represents the variance of the data carried by the principal component corresponding to the j-th eigenvalue. j This represents the j-th eigenvector, and the direction of the vector defines the direction of the j-th principal component.

[0089] Select the first k principal components, project the original data onto the new principal component space, and obtain the principal component score matrix, i.e., the pollution feature vector Y: Y = X c ·V, where V represents the projection matrix, which is composed of the first k eigenvectors v1, v2, ..., v k The composition; where the value of k is determined by the minimum number of principal components required to satisfy the cumulative variance contribution rate to reach the preset target.

[0090] S24. Linear discriminant analysis is used to reduce the dimensionality of the pollution feature vector to obtain a simplified feature vector.

[0091] Assign a class label to each sample in the pollution feature vector Y, and calculate the within-class scatter matrix S based on the pollution feature vector Y and the class label. ω and the inter-class scatter matrix S b .

[0092] By solving the generalized eigenvalue problem S b w=γS ω w is the projection vector that maximizes the ratio of between-class divergence to within-class divergence, where γ represents the generalized eigenvalue.

[0093] The input pollution feature vector Y is projected onto the solved optimal discrimination direction, i.e., the simplified feature vector Z is calculated, Z = YW. The simplified feature vector Z is a low-dimensional feature representation rich in discrimination information; where W is the projection matrix composed of projection vectors.

[0094] In some embodiments, S3 involves preliminary identification of pollution sources from the simplified feature vector to determine the location of refined pollution sources, including:

[0095] S31. Perform cluster analysis on the simplified feature vectors to obtain the pollution clustering results.

[0096] Based on the input simplified feature vector Z, a clustering algorithm is used to calculate the feature similarity between sample points, and the sample points with similar features are merged into several clusters according to the distance in the feature space, to obtain the pollution clustering result, which contains a sample clustering label vector and a clustering center coordinate matrix C, the k-th row of the matrix C is a vector c k , representing the coordinates of the center point of the k-th cluster in the feature space.

[0097] S32. The pollution clustering result is matched with the pollution source type database by the Euclidean distance matching algorithm to obtain the potential pollution source type.

[0098] The pollution source type database contains a preset total number of pollution source categories M, a feature template vector t m of the m-th pollution source category, and a pollution source type corresponding to each template vector.

[0099] For each cluster center coordinate vector c k in the matrix C, the Euclidean distance d(c k ,t m ) between it and all template vectors t m in the pollution source type database is calculated, and the pollution source type corresponding to the template vector with the smallest Euclidean distance is preliminarily identified as the potential pollution source type represented by the cluster, and the obtained potential pollution source type includes the matched pollution source type, the matching distance d k .

[0100] S33. Confidence evaluation is performed on the potential pollution source type to obtain the confirmed pollution source.

[0101] A distance threshold D thres is set, if the matching distance d k ≤D thres , it is confirmed that the potential type passes the confidence evaluation, otherwise it is marked as "unknown" and excluded.

[0102] The obtained confirmed pollution source contains the confirmed pollution source type and the original space coordinate set {(x i ,y i )} of all peak points corresponding to the pollution source.

[0103] S34. The position of the confirmed pollution source is optimized by using the least squares method, and the refined pollution source position is output.

[0104] Based on the original space coordinate set {(x i ,y i )} of all peak points corresponding to the pollution source, an optimal center position (x0, y0) is fitted by using the least squares method, and the optimization problem min∑ i [d i ((x0, y0), (xi , y i )) 2 , to obtain the refined pollution source position, specifically the optimized pollution source center coordinates (x0, y0).

[0105] In some embodiments, S4 includes generating a tracking path according to the refined pollution source position, and obtaining a smoothed predicted path through smoothing processing, including:

[0106] S41. Process the refined pollution source position data at consecutive time points through a gradient descent method to obtain a pollution source movement direction; and based on the time series of the refined pollution source position, use a difference calculation method to derive the pollution source movement speed.

[0107] Based on the continuous refined pollution source position, use a gradient descent method to fit the local tangent direction of the movement trajectory, which is the pollution source movement direction θ t .

[0108] Use a difference calculation method to calculate the pollution source movement instantaneous speed v t , the formula is: Where (x t , y t ), (x t-1 , y t-1 ) represent the position coordinates at t, t-1 time, i.e. the position coordinates at adjacent time points, and t and t -1 are the corresponding time points.

[0109] S42. Combine the pollution source movement direction and movement speed, and use a Kalman filter algorithm for path prediction to output a predicted path.

[0110] Use the movement direction θ t and the pollution source movement instantaneous speed v t as observation values, input them into the Kalman filter algorithm, use the state equation and the observation equation to optimally estimate the movement state of the pollution source, and predict its future movement path to obtain a predicted path, which contains a predicted path point sequence, i.e. a series of predicted future time pollution source position coordinates {(x t+1 , y t+1 ), (x t+2 , y t+2 ), …}.

[0111] S43. Perform a moving average processing on the predicted path to output a smoothed predicted path.

[0112] The original coordinates of each path point and its adjacent multiple points are replaced by the average of the coordinates, so as to effectively filter out high-frequency jitter and noise in the path, make the path more smooth and reasonable, and obtain a smoothed predicted path, which contains a sequence of smoothed path points {(x' t+1 ,y′ t+1 ),(x′ t+2 ,y′ t+2 ),…}.

[0113] In some embodiments, the smoothed predicted path is updated and adjusted by combining real-time atmospheric pollution data in S5 to obtain a verified path, which includes:

[0114] S51. The smoothed predicted path is compared with real-time sensor data by residual analysis to output path deviation data.

[0115] The newly collected real-time sensor data is obtained, and the real-time sensor data is compared with the predicted path point at the corresponding time. For each sensor-equipped predicted point (x' t+i ,y′ t+i ), the Euclidean distance d i between it and the center of the adjacent multiple sensor measured concentrations is calculated, and the output path deviation data includes a sequence {(t+i, d i )}, i.e., the position deviation d i corresponding to each future time t+i.

[0116] S52. The path deviation data is processed by a regression analysis method to output path adjustment parameters.

[0117] A regression model d = a + bt is constructed with time t as the independent variable and position deviation d i or Δc i as the dependent variable. The model parameters are fitted by least squares method or the like to minimize the total error, and the path adjustment parameters, i.e., the coefficients a and b of the regression model, are obtained; wherein d represents the position path deviation data, and a and b represent the intercept and slope, respectively.

[0118] S53. The smoothed predicted path is corrected based on the adjustment parameters using Gaussian process regression to output a corrected path.

[0119] The coefficient set a and b defines a linear trend model d(t) = a + bt of the path deviation with respect to time. The linear trend model d(t) is used as the mean function m(t) of the Gaussian process regression, i.e., m(t) = a + bt, and a kernel function k(t, t ′ ) is selected, such as a radial basis function; wherein d(t) represents the deviation value between the predicted position of the pollution source at any time t and the actually observed position of the pollution source.

[0120] The original smooth prediction path is taken as observation data, and the posterior probability of the Gaussian process is calculated based on the mean function m(t) and the kernel function k(t, t ′ ), to obtain the posterior probability distribution.

[0121] From the calculated posterior probability distribution, the mean vector is extracted as the corrected path point coordinates, and the diagonal elements of the covariance matrix are extracted as the uncertainty of each point, to obtain the corrected path containing the corrected path point sequence {(x t+i , y t+i )} and the corrected uncertainty of each point .

[0122] S54. Verify the rationality of the corrected path using the cross-validation method, and output the verification path.

[0123] The generalization ability of the corrected path is evaluated using the M-fold cross-validation method. Specifically, the historical data set can be randomly divided into M subsets, and M-1 subsets are used as the training set to perform the correction process of S51-S53, and the prediction error such as the root mean square error RMSE is calculated on the remaining 1 subset. After k cycles, the comprehensive evaluation index is obtained.

[0124] The verification path obtained contains the verified path point sequence and the overall verification index; wherein the overall verification index can be the average RMSE value.

[0125] In some embodiments, the result output and visualization of the verification path in S6 form the final output, including:

[0126] S61. Visualize the verification path by heat map rendering technology to obtain the visual tracking path.

[0127] The verification path is visualized by heat map rendering, and each point on the path is regarded as a heat source, and its heat can be determined by the standard deviation σ i of the point or other functions, such as the greater the variance, the higher the uncertainty, and the color is more inclined to cool tones such as blue; the smaller the variance, the higher the confidence, and the color is more inclined to warm tones such as red, and finally the visual tracking path is obtained.

[0128] S62. Combine the visual tracking path, confirm the pollution source, and refine the pollution source location to generate a pollution source tracking report.

[0129] The visual tracking path, the confirmed pollution source, and the refined pollution source location are integrated and aggregated to automatically generate a structured report.

[0130] S63. Package the pollution source tracking report in JSON format and transmit it to the monitoring platform to output the final result.

[0131] Specifically, the structured information in the report can be standardized and packaged by using JSON (JavaScript Object Notation) to form a set of machine-readable key-value pairs, and the packaged JSON data packet is transmitted to the upper monitoring platform or decision support system through an API interface.

[0132] In some embodiments, as shown in Figure 2 The present application provides an atmospheric pollution source intelligent identification and tracking system, which comprises:

[0133] A data processing and preprocessing module is configured to acquire real-time atmospheric pollution data of a plurality of sensors, preprocess the real-time atmospheric pollution data, and generate spatially continuous pollution data.

[0134] A feature extraction module is configured to perform a feature extraction operation based on the spatially continuous pollution data and output a simplified feature vector.

[0135] A pollution source preliminary identification module is configured to perform a pollution source preliminary identification on the simplified feature vector and determine a refined pollution source position.

[0136] A path generation and smoothing module is configured to generate a tracking path according to the refined pollution source position and obtain a smoothed predicted path through smoothing processing.

[0137] A path updating and verification module is configured to update and adjust the smoothed predicted path in combination with real-time atmospheric pollution data and obtain a verified path.

[0138] A result output and visualization module is configured to perform result output and visualization on the verified path and form a final output.

[0139] In some embodiments, the present application provides an atmospheric pollution source intelligent identification and tracking device, which comprises a memory and a processor; the memory is configured to store a computer program; and the processor is configured to execute the computer program to implement the steps of the atmospheric pollution source intelligent identification and tracking method.

[0140] In some embodiments, the present application provides a storage medium, which stores computer program instructions; when the computer program instructions are read and executed by a processor, the steps of the atmospheric pollution source intelligent identification and tracking method are performed.

[0141] In some embodiments, the present application provides a computer program product, which comprises computer programs or instructions; when the computer programs or instructions are executed by a processor, the steps of the atmospheric pollution source intelligent identification and tracking method are implemented.

[0142] Any reference to storage, memory, database or other medium herein includes non-volatile and / or volatile storage. Non-volatile storage can include read-only memory (ROM), programmable ROM (PROM), electronically programmable ROM (EPROM), or electrically erasable programmable ROM (EEPROM). Volatile storage can include random-access memory (RAM), or external cache memory.

[0143] It should be noted that, in the present document, relational terms such as "first" and "second", and the like can be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Moreover, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. In other words, any of the elements listed in the description of the technology are not exhaustive of the elements of the technology.

[0144] Although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art will understand that the technology disclosed in the foregoing embodiments can be modified or equivalent substitutions can be made to some of the technical features, and such modifications or substitutions do not cause the essence of the corresponding technical solution to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for intelligent identification and tracking of air pollution sources, characterized in that, include: Acquire real-time air pollution data from multiple sensors, preprocess the real-time air pollution data, and generate spatial continuous pollution data; Perform feature extraction based on the spatially continuous pollution data and output a simplified feature vector. The simplified feature vector is used for preliminary identification of pollution sources to determine the location of refining pollution sources; A tracking path is generated based on the location of the refining pollution source, and a smoothed prediction path is obtained through smoothing processing. The smooth prediction path is updated and adjusted by combining real-time air pollution data to obtain a validation path; The verification path is then processed and visualized to form the final output.

2. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, The real-time air pollution data is preprocessed to generate spatially continuous pollution data, including: Gaussian filtering is used to filter noise from real-time air pollution data collected by multiple sensors to eliminate random errors and obtain clean pollution data. Perform timestamp alignment processing on clean pollution data to unify the time reference of data from various sensors and obtain time-synchronized pollution data; Spatial interpolation is performed on the time-synchronized contaminated data to generate the spatially continuous contaminated data.

3. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, Based on the aforementioned spatially continuous pollution data, a feature extraction operation is performed to output a simplified feature vector, including: The variation trend of the spatial continuous pollution data was analyzed using a polynomial fitting method, and pollution trend data was extracted. The pollution peak data is obtained by identifying the peak values ​​in the pollution trend data using a local extremum detection algorithm. Principal component analysis was used to quantify the pollution peak data to obtain a pollution feature vector. The simplified feature vector is obtained by using linear discriminant analysis to reduce the dimensionality of the pollution feature vector.

4. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, Preliminary identification of pollution sources is performed on the simplified feature vector to determine the location of refined pollution sources, including: Cluster analysis was performed on the simplified feature vectors to obtain the pollution clustering results; The pollution clustering results are compared with the pollution source type database using the Euclidean distance matching algorithm to obtain the potential pollution source types. A confidence assessment was conducted on the types of potential pollution sources to confirm the pollution sources. The least squares method is used to optimize the calculation of the location of the identified pollution source and output the refined pollution source location.

5. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, A tracing path is generated based on the location of the refining pollution source, and a smoothed predicted path is obtained through smoothing processing, including: The location data of refining pollution sources at continuous time intervals are processed by gradient descent to obtain the direction of movement of the pollution sources; based on the time series of the locations of the refining pollution sources, the velocity of the pollution sources is derived by differential calculation method. By combining the direction and speed of the pollution source's movement, the Kalman filter algorithm is used to predict the path and output the predicted path. Perform a moving average on the predicted path to output a smoothed predicted path.

6. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, The smoothed prediction path is updated and adjusted by combining real-time air pollution data to obtain a validation path, including: By comparing the smoothed predicted path with real-time sensor data through residual analysis, the path deviation data is output. Regression analysis is used to process path deviation data and output path adjustment parameters. Based on the adjusted parameters, Gaussian process regression is used to correct and smooth the prediction path, and the corrected path is output. Use cross-validation to verify the rationality of the modified path and output the validation path.

7. The intelligent identification and tracking method for air pollution sources according to claim 1, characterized in that, The verification path is processed to output and visualize the results, forming the final output, including: The verification path is visualized using heatmap rendering technology, resulting in a visually traceable path. By combining the described visual tracking path, confirming the location of pollution sources and refining the pollution source locations, a pollution source tracking report is generated; The pollution source tracing report is packaged in JSON format and transmitted to the monitoring platform to output the final results.

8. An intelligent identification and tracking system for air pollution sources, characterized in that, include: Data processing and preprocessing module: acquires real-time air pollution data from multiple sensors, preprocesses the real-time air pollution data, and generates spatial continuous pollution data; Feature extraction module: Performs feature extraction operation based on the spatially continuous pollution data and outputs a simplified feature vector; Preliminary pollution source identification: The simplified feature vector is used to perform preliminary pollution source identification to determine the location of refined pollution sources; Path generation and smoothing module: Generates a tracking path based on the location of the refined pollution source, and obtains a smoothed predicted path through smoothing processing; Path update verification: The smoothed prediction path is updated and adjusted by combining real-time air pollution data to obtain a verification path; The result output visualization module outputs and visualizes the verification path to form the final output.

9. An intelligent identification and tracking device for air pollution sources, characterized in that, It includes a memory and a processor; the memory is used to store computer programs; the processor is used to execute the computer programs to implement the steps of the intelligent identification and tracking method for air pollution sources.

10. A storage medium, characterized in that, The storage medium stores computer program instructions, which are read and executed by a processor to perform the steps of the intelligent identification and tracking method for air pollution sources.

Images