A method and system for tracing dust emissions in electronic greenhouse scenarios

Abstract

This invention relates to the field of dust pollution source tracing technology, and particularly to a method and system for dust emission source tracing in an electronic greenhouse setting. The method includes: acquiring dust pollution images within the electronic greenhouse and performing preprocessing such as denoising, contrast enhancement, and background suppression; extracting the diffusion vector of dust particles using a dense optical flow algorithm based on an image pyramid; simultaneously acquiring monitoring data and meteorological parameters at multiple times within a set time window; correcting the Gaussian diffusion model parameters based on the optical flow analysis results, calculating the concentration transfer coefficient between potential emission sources and monitoring points, and generating a transfer matrix; forming a source intensity inversion equation set by simultaneously solving the observation equations at multiple times, and solving for the optimal source intensity distribution matrix under constraints of temporal smoothness, spatial continuity, and non-negativity; finally, performing spatial clustering and temporal analysis on the results to generate a pollution source heat map and emission intensity change curve, thus achieving a visual display of the source tracing results.

Description

Technical Field

[0001] This invention relates to the field of dust pollution source tracing technology, and in particular to a dust emission source tracing method and system for electronic greenhouse scenarios. Background Technology

[0002] In open-air storage yards, mines, and material storage environments, loading, unloading, stacking, and wind-driven processes generate large amounts of dust particles, affecting surrounding air quality, threatening worker health, and causing material loss, posing challenges to environmental management. Currently used dust monitoring methods include laser scattering, gravity sampling, and optical sensor monitoring. These methods can obtain local or single-point dust concentrations, but they struggle to accurately identify the location of dust emission sources and the emission intensity of each source.

[0003] Furthermore, dust diffusion in open-air storage yards is significantly affected by wind speed, wind direction, and complex airflow conditions. Traditional diffusion models with fixed parameters and single-point observation methods cannot effectively reflect the actual diffusion process across multiple sources and at multiple times, resulting in low source strength identification accuracy and a tendency to generate unstable local optima. Therefore, there is an urgent need for a visualization solution that can combine visual information with physical models to achieve multi-source, multi-time dust emission tracing. Summary of the Invention

[0004] To overcome the above deficiencies, this invention provides a method and system for tracing dust emissions in electronic greenhouse scenarios. It aims to improve the accuracy and real-time performance of dust pollution tracing by integrating high-definition image optical flow analysis with Gaussian diffusion model inversion technology to achieve spatial positioning, intensity identification, and dynamic visualization of dust emission sources.

[0005] In a first aspect, the present invention provides the following technical solution: a method for tracing the source of dust emissions in an electronic greenhouse setting, comprising:

[0006] Images of dust pollution inside an electronic greenhouse were collected and preprocessed, and an image pyramid was constructed based on the preprocessed pollution images.

[0007] An optical flow algorithm is applied to the image pyramid to track the motion vector field of dust particles and calculate the diffusion vector of dust in each region in real time.

[0008] Within a set time window, data from multiple moments are collected synchronously at fixed time intervals. The data at each moment includes the measured dust concentration at gridded monitoring points, the diffusion vector obtained from optical flow analysis, and real-time meteorological parameters.

[0009] For the optical flow analysis results at each sampling time, the parameters of the Gaussian diffusion model are corrected, and the concentration transport coefficient from each gridded potential emission source to each monitoring point is calculated in combination with the meteorological conditions at the sampling time, generating a transport matrix at the sampling time.

[0010] The transfer matrices and observation data corresponding to multiple sampling times are combined to form a set of source strength inversion equations. The source strength inversion equations are solved by an optimization algorithm. The optimization objective comprehensively considers the minimization of data fitting error, source strength temporal smoothness constraint, spatial continuity constraint, and non-negativity constraint.

[0011] Based on the inversion solution results, the spatial locations of major dust emission sources are identified, the emission intensity and contribution ratio of each source are quantified, and pollution source heat maps and time series emission curves are generated to visualize the pollution source tracing results.

[0012] Preferably, the preprocessing step includes:

[0013] The acquired dust pollution images were denoised using a bilateral filtering method.

[0014] An adaptive histogram equalization method is used to enhance the contrast of the denoised image.

[0015] The frame difference method is used to perform background suppression processing on the enhanced image.

[0016] Preferably, the step of constructing the image pyramid includes:

[0017] The preprocessed dust pollution image is scaled in multiple layers to generate image layers of different resolutions;

[0018] Each scaled image is smoothed to suppress aliasing noise caused by downsampling;

[0019] The images of each scaling layer are saved sequentially to form an image pyramid sequence.

[0020] Preferably, the step of calculating the diffusion vector of dust in each region includes:

[0021] In each layer of the image pyramid, the dense optical flow method is applied to the pixel blocks of the image region to calculate the motion vector;

[0022] Multi-scale fusion of dense optical flow results from adjacent frames is performed to obtain the average motion vector for each region;

[0023] The diffusion vector of dust in each region is calculated based on the fused motion vector, including the diffusion direction and diffusion velocity.

[0024] Preferably, the Gaussian diffusion model parameter correction step includes:

[0025] Extract the diffusion vector of each grid region at the sampling time, and calculate the dominant direction and average velocity of the vector field;

[0026] By taking the dominant direction of the vector field as the actual dust diffusion direction, the wind direction parameter in the Gaussian diffusion model is corrected.

[0027] The diffusion intensity index is calculated based on the degree of dispersion of the optical flow vector, and the transverse and longitudinal diffusion coefficients in the Gaussian diffusion model are corrected.

[0028] Preferably, the step of generating the sampling time transfer matrix includes:

[0029] The greenhouse meteorological parameters were collected at the sampling time, and the theoretical concentration transport coefficients between potential emission sources and each monitoring point were calculated based on the modified Gaussian diffusion model.

[0030] Substitute wind speed, wind direction, and diffusion coefficient parameters into the Gaussian diffusion equation to solve for the concentration distribution of dust in the grid space;

[0031] The calculated concentration distribution results are normalized to form the transmission relationship between potential emission sources and monitoring points;

[0032] The transmission relationships of each grid cell are combined according to their spatial location to generate the overall transmission matrix at the sampling time.

[0033] Preferably, the step of forming the source strength inversion equation set includes:

[0034] The transmission matrix generated at each sampling time is paired with the dust concentration observation data of the monitoring point at the corresponding time.

[0035] Using the dust concentration at each monitoring point as the observation vector, the transmission matrix as the coefficient matrix, and the gridded source strength as the vector to be determined, a linear observation equation for each time point is established.

[0036] By combining the linear observation equations from multiple sampling times in a time series manner, a joint inversion equation system containing the time dimension is constructed.

[0037] Preferably, the steps for solving the source strength inversion equation system include:

[0038] An optimization model is constructed with the objective of minimizing the error between the measured and calculated dust concentration values ​​at monitoring points;

[0039] The optimization model incorporates time smoothness, spatial continuity, and non-negativity constraints to form a comprehensive constraint optimization equation.

[0040] The comprehensive constraint optimization equation is solved using an iterative optimization algorithm to obtain the optimal source strength distribution matrix.

[0041] Preferably, the steps for visualizing pollution source tracing results include:

[0042] Spatial cluster analysis is performed on the source intensity distribution matrix obtained by inversion to identify the spatial location of the main dust emission sources;

[0043] Calculate the contribution ratio of each emission source to the dust concentration at the monitoring point and its time change trend, and quantify the emission intensity of each source;

[0044] Based on the source strength distribution results, a spatial heat map of pollution sources is generated, and combined with time series data, an emission intensity change curve is generated to achieve a visual display of the pollution source tracing results.

[0045] Secondly, the present invention provides the following technical solution: a dust emission tracing system for electronic greenhouse scenarios, used to implement the aforementioned dust emission tracing method, the system comprising:

[0046] The diffusion vector calculation module is used to apply optical flow algorithms to the image pyramid, track the motion vector field of dust particles, and calculate the diffusion vector of dust in each area in real time.

[0047] The grid data acquisition module is used to synchronously acquire data from multiple moments at fixed time intervals within a set time window. The data at each moment includes the measured dust concentration of the grid monitoring points, the diffusion vector obtained from optical flow analysis, and real-time meteorological parameters.

[0048] The transmission matrix construction module is used to correct the Gaussian diffusion model parameters based on the optical flow analysis results at each sampling time, and calculate the concentration transmission coefficient from each gridded potential emission source to each monitoring point in combination with the meteorological conditions at the sampling time, thereby generating the transmission matrix at the sampling time.

[0049] The spatiotemporal joint inversion module is used to combine the transfer matrices and observation data corresponding to multiple sampling times to form a source strength inversion equation set. The source strength inversion equation set is solved by an optimization algorithm. The optimization objective comprehensively considers the minimization of data fitting error, source strength temporal smoothness constraint, spatial continuity constraint, and non-negativity constraint.

[0050] The pollution source tracing result visualization module is used to identify the spatial location of major dust emission sources based on the inversion solution results, quantify the emission intensity and contribution ratio of each source, generate pollution source heat maps and time series emission curves, and realize the visualization of pollution source tracing results.

[0051] The present invention has the following beneficial effects:

[0052] This invention combines high-definition image acquisition, dense optical flow analysis, and a Gaussian diffusion model to achieve dynamic spatial and temporal tracking of dust within an electronic greenhouse. Through joint inversion of data from multiple time points and monitoring points, it can accurately locate major dust emission sources and quantify the emission intensity and contribution ratio of each source. This method overcomes the error problems inherent in traditional single-point monitoring or empirical estimation methods under complex airflow conditions, improving the accuracy and reliability of dust source tracing and providing a precise basis for the scientific formulation of emission control strategies. Attached Figure Description

[0053] Figure 1 This is a flowchart of a dust emission source tracing method for electronic greenhouse scenarios proposed in this invention;

[0054] Figure 2 This is a framework diagram of a dust emission traceability system for electronic greenhouse scenarios proposed in this invention. Detailed Implementation

[0055] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0056] Example 1

[0057] In the first embodiment of the present invention, the present invention provides a method and system for tracing dust emissions in electronic greenhouse scenarios, such as... Figure 1 As shown, it includes the following steps:

[0058] S100: Acquires and preprocesses images of dust pollution inside the electronic greenhouse, and constructs an image pyramid based on the preprocessed pollution images;

[0059] Preferably, the preprocessing step includes:

[0060] The acquired dust pollution images were denoised using a bilateral filtering method.

[0061] An adaptive histogram equalization method is used to enhance the contrast of the denoised image.

[0062] The frame difference method is used to perform background suppression processing on the enhanced image.

[0063] Specifically, the acquired raw dust pollution images contain high-frequency noise introduced by changes in illumination and the image sensor. To remove noise while preserving the edge information of dust particles, this embodiment employs a bilateral filtering method for noise reduction.

[0064] The image after bilateral filtering suffers from limited grayscale dynamic range. To improve the distinction between dust areas and the background, this embodiment employs an adaptive histogram equalization method. During processing, the image is divided into several 8×8 pixel sub-blocks; histogram equalization is performed independently on each sub-block, and the contrast amplification factor is limited to the range of 2.0–4.0 to avoid excessive noise amplification; finally, bilinear interpolation is used to fuse the results of adjacent sub-blocks, thereby obtaining an enhanced image with global brightness balance and clear dust particle edges.

[0065] In open-air storage yard scenarios, changes in lighting, camera shake, and ground reflections can easily cause background interference. To highlight areas of moving dust, this embodiment employs frame differencing for background suppression. Let the current frame be... The previous frame was Calculate the absolute difference between two frames: Thresholding is performed on the difference image, with the threshold T dynamically set to 1.2 to 1.5 times the average gray level of the image. Isolated noise points are removed and continuous regions are connected through morphological opening and closing operations, thereby extracting the significant dust movement areas.

[0066] After the above steps, the output preprocessed image can significantly suppress background interference and highlight the dust particle area, providing an effective image for subsequent analysis.

[0067] Preferably, the step of constructing the image pyramid includes:

[0068] The preprocessed dust pollution image is scaled in multiple layers to generate image layers of different resolutions;

[0069] Each scaled image is smoothed to suppress aliasing noise caused by downsampling;

[0070] The images of each scaling layer are saved sequentially to form an image pyramid sequence.

[0071] Specifically, the preprocessed dust pollution image is downsampled layer by layer to obtain a multi-resolution image sequence, forming an image pyramid structure. Let the original preprocessed image be... This represents the 0th level of the pyramid, using a scaling factor. Perform layer-by-layer scaling to generate the first... Layer image:

[0072] ;

[0073] Generally, 3 to 5 layers of images are sufficient to meet the resolution requirements for optical flow analysis.

[0074] During downsampling, direct scaling can easily lead to aliasing and high-frequency noise distortion. Therefore, a smoothing filter is applied to the previous layer image before each scaling step. In this embodiment, a Gaussian smoothing filter is used, with a filter kernel size of 3×3 or 5×5 and a standard deviation of 0.8 to 1.2.

[0075] The scaled images from each layer are saved sequentially in descending order of resolution to construct a complete image pyramid sequence. ,in, For the highest resolution layer, The lowest resolution layer is used. The pyramid sequence is organized in a hierarchical indexing manner. Each layer of the image is used for hierarchical matching and motion estimation in the optical flow algorithm. In subsequent optical flow calculations, the global motion trend is estimated first in the low-resolution layer, and then the image is refined layer by layer to a high-resolution image, finally obtaining a high-precision dust motion vector field.

[0076] After the above steps, the generated image pyramid can effectively preserve the multi-scale features of dust particles, reduce the impact of local occlusion and noise, and provide a multi-layer input basis for subsequent dense optical flow analysis.

[0077] S200: Apply optical flow algorithm to the image pyramid to track the motion vector field of dust particles and calculate the diffusion vector of dust in each area in real time;

[0078] Preferably, the step of calculating the diffusion vector of dust in each region includes:

[0079] In each layer of the image pyramid, the dense optical flow method is applied to the pixel blocks of the image region to calculate the motion vector;

[0080] Multi-scale fusion of dense optical flow results from adjacent frames is performed to obtain the average motion vector for each region;

[0081] The diffusion vector of dust in each region is calculated based on the fused motion vector, including the diffusion direction and diffusion velocity.

[0082] Specifically, to obtain the motion features of dust in the image sequence, a dense optical flow algorithm is applied to each layer of the image pyramid to calculate the motion vector of each pixel between consecutive frames. This involves taking two adjacent frames... and The grayscale values ​​are then normalized to [0,1]. The classical optical flow constraint equation is employed. Constraints are imposed, among which , These are the spatial gradients of the image. For time gradient, , Let x and y be the optical flow components of a pixel. The Farnebäck dense optical flow algorithm is preferred. Its core idea is to model local image regions using a quadratic polynomial and obtain the pixel-level motion vector field by minimizing the matching error. The algorithm parameters are set to a window radius of 5–9 pixels and a polynomial neighborhood standard deviation of 1.2–1.5. The dense optical flow field of each image layer is calculated. This indicates the direction and speed of pixel movement.

[0083] Because dust motion exhibits velocity differences and inconsistent directions at different scales, multi-scale fusion processing is necessary to avoid noise interference in single-layer optical flow. The low-resolution layer optical flow field is upsampled to a high-resolution scale, and the weight of the k-th layer optical flow field is set as follows: The weights are dynamically set based on the layer resolution and texture sharpness, and then fused using a blending formula. Calculate the multi-scale optical flow results after fusion. .

[0084] The fused optical flow field is used to calculate the dust diffusion vector in each region. First, the image is divided into grid regions of fixed size. For example, a 10×10 or 20×20 pixel block, then calculate the mean of the optical flow component within each grid region:

[0085] ;

[0086] in, This represents the number of pixels in the region. The diffusion direction is then calculated. and diffusion rate Finally, the diffusion vector is obtained. .

[0087] Through the above steps, the present invention can achieve pixel-level dust movement capture and can stably identify the direction of dust flow even in low contrast or high noise environments.

[0088] S300: Within a set time window, data from multiple moments are collected synchronously at fixed time intervals. The data at each moment includes the measured dust concentration at gridded monitoring points, the diffusion vector obtained from optical flow analysis, and real-time meteorological parameters.

[0089] Specifically, at each sampling time The grid diffusion set is obtained through the S200 process. Each of these items is Simultaneously, the concentration vector at the monitoring point is obtained. and meteorological quantity ,in for Wind speed at any moment for The direction of the wind field at any given moment.

[0090] S400: Based on the optical flow analysis results at each sampling time, the Gaussian diffusion model parameters are corrected, and the concentration transfer coefficient from each gridded potential emission source to each monitoring point is calculated in combination with the meteorological conditions at the sampling time, generating the transfer matrix at the sampling time.

[0091] Preferably, the Gaussian diffusion model parameter correction step includes:

[0092] Extract the diffusion vector of each grid region at the sampling time, and calculate the dominant direction and average velocity of the vector field;

[0093] By taking the dominant direction of the vector field as the actual dust diffusion direction, the wind direction parameter in the Gaussian diffusion model is corrected.

[0094] The diffusion intensity index is calculated based on the degree of dispersion of the optical flow vector, and the transverse and longitudinal diffusion coefficients in the Gaussian diffusion model are corrected.

[0095] Specifically, diffusion vectors within all grids at a uniform sampling time are collected. ,in This represents the total number of grid cells, which in turn gives the average velocity of the vector field. Solve for the horizontal component of the average velocity of the vector field. and vertical components Then, the dominant direction is determined based on the components. .

[0096] In the traditional Gaussian diffusion model, meteorological wind direction Typically measured by meteorological sensors, the direction of diffusion can deviate from the actual diffusion direction in complex space environments. To improve simulation accuracy, the dominant direction obtained from optical flow analysis is used... With weather and wind direction Perform fusion correction: ,in This represents the fusion coefficient. The wind speed is also fused and corrected accordingly. .

[0097] Calculate the degree of optical flow dispersion:

[0098] ;

[0099] in, and They represent the first The average velocity in the horizontal and vertical directions of each grid is then used to correct the diffusion coefficient of the Gaussian diffusion model. The correction process is as follows:

[0100] ;

[0101] in, and To represent the initial empirical baseline value of the Gaussian diffusion model, and This represents the adjustment coefficient, with typical values ​​ranging from 0.1 to 0.5.

[0102] Preferably, the step of generating the sampling time transfer matrix includes:

[0103] The greenhouse meteorological parameters were collected at the sampling time, and the theoretical concentration transport coefficients between potential emission sources and each monitoring point were calculated based on the modified Gaussian diffusion model.

[0104] Substitute wind speed, wind direction, and diffusion coefficient parameters into the Gaussian diffusion equation to solve for the concentration distribution of dust in the grid space;

[0105] The calculated concentration distribution results are normalized to form the transmission relationship between potential emission sources and monitoring points;

[0106] The transmission relationships of each grid cell are combined according to their spatial location to generate the overall transmission matrix at the sampling time.

[0107] Specifically, at each sampling time For each potential emission source grid point j and each monitoring point s, the theoretical transmission coefficient is calculated using the modified Gaussian diffusion equation in a coordinate system aligned with the prevailing wind direction. Its form is:

[0108] ;

[0109] in, and These are the horizontal and vertical distances of monitoring point j relative to source point s, respectively (after coordinate rotation transformation).

[0110] For each source point s, the theoretical concentration contribution of each monitoring point downwind is calculated according to the Gaussian diffusion equation:

[0111] ;

[0112] in, This represents the emission rate under the assumption of unit source strength, tentatively set to 1 unit. By calculating the sum of the theoretical concentrations at all monitoring points for all source points, a dust concentration distribution map of the entire grid space can be obtained.

[0113] To facilitate solving the subsequent system of equations, for all Normalization is performed according to the direction of monitoring point j:

[0114] ;

[0115] in, This represents the total number of monitoring points. It ensures that the sum of the transmission coefficients of all potential sources at the same monitoring point is 1, thereby enhancing the numerical stability and physical interpretation of the equations. This coefficient characterizes the transmission rate at the sampling time. The concentration transfer relationship between all potential emission sources and each monitoring point.

[0116] The transmission coefficients corresponding to each grid cell are combined according to their spatial location and numbering order to form the overall transmission matrix at the sampling time. ,in For the number of monitoring points, The number of potential source grid points.

[0117] Through the above steps, a high spatiotemporal resolution transmission matrix can be generated at each sampling time. This matrix expresses the physical process of dust diffusion in the form of a linear mapping, providing a computable basis for subsequent multi-time source intensity inversion.

[0118] S500: The source strength inversion equations are formed by combining the transfer matrices and observation data corresponding to multiple sampling times. The source strength inversion equations are solved by optimization algorithm. The optimization objective comprehensively considers the minimization of data fitting error, source strength time smoothness constraint, spatial continuity constraint and non-negativity constraint.

[0119] Preferably, the step of forming the source strength inversion equation set includes:

[0120] The transmission matrix generated at each sampling time is paired with the dust concentration observation data of the monitoring point at the corresponding time.

[0121] Using the dust concentration at each monitoring point as the observation vector, the transmission matrix as the coefficient matrix, and the gridded source strength as the vector to be determined, a linear observation equation for each time point is established.

[0122] By combining the linear observation equations from multiple sampling times in a time series manner, a joint inversion equation system containing the time dimension is constructed.

[0123] Specifically, after completing the Gaussian diffusion correction calculation at each sampling time, the system obtains the transmission matrix and the observation vector of the monitoring point, respectively:

[0124] Transmission Matrix Indicates at the sampling time Below, from Potential emission sources Unit concentration transfer coefficient at each monitoring point;

[0125] Monitoring point observation vector This indicates the sampling time at each monitoring point. The measured dust concentration value. The system compares the timestamps to determine the value for each... With the corresponding Pair them one by one to form time-synchronized data pairs.

[0126] For each sampling time The gridded source strength vector to be determined For the unknowns, the following linear observation equation is established:

[0127] ;

[0128] in, The measurement error term is used to represent sensor error and model residuals. To fully utilize observation information from multiple time points and improve inversion stability and noise resistance, this invention combines the observation equations from all time points in chronological order to construct a time-extended inversion equation set, specifically in the following form:

[0129] ;

[0130] This implementation process can unify observation information and transmission patterns from multiple time points into a mathematical model, providing a physically consistent and computable basic framework for the spatial location and time series inversion of dust emission sources.

[0131] Preferably, the steps for solving the source strength inversion equation system include:

[0132] An optimization model is constructed with the objective of minimizing the error between the measured and calculated dust concentration values ​​at monitoring points;

[0133] The optimization model incorporates time smoothness, spatial continuity, and non-negativity constraints to form a comprehensive constraint optimization equation.

[0134] The comprehensive constraint optimization equation is solved using an iterative optimization algorithm to obtain the optimal source strength distribution matrix.

[0135] Specifically, after forming a joint observation equation set containing multiple sampling times, an optimized model for source strength inversion is established with the objective of minimizing the squared error between the dust concentration observation data and the model calculation values:

[0136] ;

[0137] in, The concentration vector observed at all times; The transfer matrix for all time points; The source strength vector is the gridded source strength vector for all time moments to be determined; and These are the weighting coefficients for time and space constraints. The first term of the objective function is the observation error term, reflecting the difference between the model-predicted concentration and the measured concentration.

[0138] To ensure the inversion results are continuous in the time dimension and physically reasonable, a time smoothness constraint term is constructed:

[0139] ;

[0140] in This represents the number of sampling times. This parameter is used to penalize excessive changes in source intensity between adjacent times, ensuring that dust emission intensity changes smoothly over time.

[0141] Considering that actual dust diffusion sources are usually spatially continuous, a spatial gradient penalty term is introduced:

[0142] ;

[0143] in Represents grid cells Surrounding source strength gradient, This represents all cell grids.

[0144] To conform to the physical fact that actual dust emission intensity cannot be negative, a non-negativity constraint is added to the optimization solution process:

[0145] . ;

[0146] That is, the source strength at any point and at any time is greater than 0.

[0147] To solve the aforementioned constrained optimization problem, this invention preferably employs an iterative solution strategy based on the alternating direction multiplier method or the non-negative least squares method. After the solution is obtained, the optimal source strength distribution matrix is ​​output. .

[0148] By implementing this step, a robust solution for source strength inversion can be achieved under conditions of multi-time and multi-monitoring point data. By introducing temporal smoothness, spatial continuity and non-negativity constraints, the ill-posedness problem of the inversion equation system is effectively suppressed, making the source strength distribution results continuous in time, smooth in space and physically reasonable.

[0149] S600: Based on the inversion solution results, identify the spatial location of the main dust emission sources, quantify the emission intensity and contribution ratio of each source, generate pollution source heat maps and time series emission curves, and realize the visualization of pollution source tracing results.

[0150] Preferably, the steps for visualizing pollution source tracing results include:

[0151] Spatial cluster analysis is performed on the source intensity distribution matrix obtained by inversion to identify the spatial location of the main dust emission sources;

[0152] Calculate the contribution ratio of each emission source to the dust concentration at the monitoring point and its time change trend, and quantify the emission intensity of each source;

[0153] Based on the source strength distribution results, a spatial heat map of pollution sources is generated, and combined with time series data, an emission intensity change curve is generated to achieve a visual display of the pollution source tracing results.

[0154] Spatial clustering analysis was performed on the source intensity distribution matrix obtained by inversion. By setting a source intensity threshold, regions with high source intensity values ​​in the gridded results were clustered to identify the spatial locations of representative dust emission sources.

[0155] Based on the source intensity time series data obtained from the inversion, the contribution ratio of each major emission source to the dust concentration at different monitoring points is calculated. By comparing the measured concentrations at the monitoring points with the contribution values ​​obtained from the model inversion, the trend of the contribution of each source in different time periods is obtained, thereby quantitatively describing the dynamic change characteristics of the emission intensity of pollution sources.

[0156] Subsequently, using a geographic information system or visualization software, the source intensity values ​​of each grid were mapped to a spatial coordinate system to generate a spatial heat map of dust pollution sources, with color gradients reflecting the magnitude and distribution range of the source intensity. This heat map visually displays the spatial distribution characteristics of dust emission sources within the electronic greenhouse. Finally, the time-series source intensity data of each major emission source obtained through inversion were plotted as emission intensity variation curves to demonstrate the emission change trends of each source over different time periods.

[0157] By combining heatmaps and time-series curves, the results of dust pollution source tracing are visualized, making it easy for users to intuitively analyze the spatial location, emission contribution, and temporal variation patterns of major pollution sources.

[0158] Example 2

[0159] In the process of monitoring dust emissions from electronic greenhouses, the complex airflow inside the greenhouse, the concealed distribution of pollution sources, and the difficulty of accurately tracing the source using traditional single-point monitoring methods lead to difficulties in identifying dust sources and low control efficiency. To solve these problems, this invention provides a dust emission tracing system for electronic greenhouse scenarios, the structure of which is as follows: Figure 2 As shown. The system includes an image preprocessing module, a diffusion vector calculation module, a grid data acquisition module, a transfer matrix construction module, a spatiotemporal joint inversion module, and a pollution source tracing result visualization module. Its specific execution process is as follows:

[0160] Image preprocessing module

[0161] This module first acquires real-time images of dust pollution using high-definition cameras deployed at different heights and angles inside the electronic greenhouse. The acquired images are then denoised using a bilateral filtering algorithm to eliminate random lighting interference and camera noise. Subsequently, an adaptive histogram equalization method is used to enhance the contrast of the images, increasing the brightness differences in dust particle areas. Finally, a frame difference method is used for background suppression to remove the influence of static structures, thereby obtaining an image sequence containing only the dynamic features of the dust.

[0162] Diffusion Vector Calculation Module

[0163] This module applies a dense optical flow algorithm to consecutive frames of images at each layer of the image pyramid to extract the motion information of dust particles and calculate the motion vector of each pixel. The system uses a multi-scale fusion strategy to perform a weighted average of the optical flow results from adjacent frames to obtain the average motion vector for each grid region and extract the diffusion vector.

[0164] Grid data acquisition module

[0165] This module divides the electronic greenhouse space into multiple regular grid units and deploys dust monitoring points within the grids. Through the data acquisition subsystem, it synchronously collects measured dust concentration values, meteorological parameters (including wind speed and direction), and diffusion vectors obtained from optical flow analysis at fixed intervals within a set time window, achieving temporal alignment and spatial registration of multi-source information.

[0166] Transmission Matrix Construction Module

[0167] At each sampling time, the system corrects the Gaussian diffusion model parameters based on the optical flow diffusion results, takes the dominant optical flow direction as the actual diffusion direction, and adjusts the lateral and longitudinal diffusion coefficients using the optical flow vector discretization. Then, combined with real-time meteorological parameters, the concentration transfer coefficient between potential emission sources and each monitoring point is calculated.

[0168] Pollution source tracing results visualization module

[0169] This module performs spatial cluster analysis on the source intensity distribution results obtained from the inversion, automatically identifies the locations of major dust emission sources, and calculates the contribution ratio and temporal trend of each source to the dust concentration at the monitoring points. Finally, the system generates a spatial heat map of the pollution sources and emission intensity variation curves, providing a visual representation of the source tracing results and a basis for subsequent dust control and treatment strategies.

[0170] Through the coordinated operation of this system, the precise location and quantitative analysis of dust pollution sources in electronic greenhouses can be achieved under complex airflow conditions, significantly improving the real-time performance, reliability, and scientific rigor of dust emission monitoring.

[0171] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art can still modify the technical solutions described in the foregoing embodiments or make equivalent substitutions for some of the technical features. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for tracing dust emissions in electronic greenhouse scenarios, characterized in that, include: Images of dust pollution inside an electronic greenhouse were collected and preprocessed, and an image pyramid was constructed based on the preprocessed pollution images. An optical flow algorithm is applied to the image pyramid to track the motion vector field of dust particles and calculate the diffusion vector of dust in each region in real time. Within a set time window, data from multiple moments are collected synchronously at fixed time intervals. The data at each moment includes the measured dust concentration at gridded monitoring points, the diffusion vector obtained from optical flow analysis, and real-time meteorological parameters. For the optical flow analysis results at each sampling time, the Gaussian diffusion model parameters are corrected, and the concentration transport coefficient from each gridded potential emission source to each monitoring point is calculated in combination with the meteorological conditions at the sampling time, generating a transport matrix at the sampling time. The transfer matrices and observation data corresponding to multiple sampling times are combined to form a source strength inversion equation set. The source strength inversion equation set is solved by an optimization algorithm. The optimization objective comprehensively considers the minimization of data fitting error, source strength temporal smoothness constraint, spatial continuity constraint, and non-negativity constraint. Based on the inversion solution results, the spatial location of the main dust emission sources is identified, the emission intensity and contribution ratio of each source are quantified, and a pollution source heat map and time series emission curve are generated to visualize the pollution source tracing results. The steps for generating the sampling time transfer matrix include: The greenhouse meteorological parameters were collected at the sampling time, and the theoretical concentration transport coefficients between potential emission sources and each monitoring point were calculated based on the modified Gaussian diffusion model. Substitute wind speed, wind direction, and diffusion coefficient parameters into the Gaussian diffusion equation to solve for the concentration distribution of dust in the grid space; The calculated concentration distribution results are normalized to form the transmission relationship between potential emission sources and monitoring points; The transmission relationships of each grid cell are combined according to their spatial location to generate the overall transmission matrix at the sampling time.

2. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The preprocessing steps include: The acquired dust pollution images were denoised using a bilateral filtering method. An adaptive histogram equalization method is used to enhance the contrast of the denoised image. The frame difference method is used to perform background suppression processing on the enhanced image.

3. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps involved in constructing an image pyramid include: The preprocessed dust pollution image is scaled in multiple layers to generate image layers of different resolutions; Each scaled image is smoothed to suppress aliasing noise caused by downsampling; The images of each scaling layer are saved sequentially to form an image pyramid sequence.

4. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps for calculating the dust diffusion vector in each region include: In each layer of the image pyramid, the dense optical flow method is applied to the pixel blocks of the image region to calculate the motion vector; Multi-scale fusion of dense optical flow results from adjacent frames is performed to obtain the average motion vector for each region; The diffusion vector of dust in each region is calculated based on the fused motion vector, including the diffusion direction and diffusion velocity.

5. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps for correcting the parameters of the Gaussian diffusion model include: Extract the diffusion vector of each grid region at the sampling time, and calculate the dominant direction and average velocity of the vector field; By taking the dominant direction of the vector field as the actual dust diffusion direction, the wind direction parameter in the Gaussian diffusion model is corrected. The diffusion intensity index is calculated based on the degree of dispersion of the optical flow vector, and the transverse and longitudinal diffusion coefficients in the Gaussian diffusion model are corrected.

6. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps to form the source strength inversion equation set include: The transmission matrix generated at each sampling time is paired with the dust concentration observation data of the monitoring point at the corresponding time. Using the dust concentration at each monitoring point as the observation vector, the transmission matrix as the coefficient matrix, and the gridded source strength as the vector to be determined, a linear observation equation for each time point is established. By combining the linear observation equations from multiple sampling times in a time series manner, a joint inversion equation system containing the time dimension is constructed.

7. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps to solve the source strength inversion equation system include: An optimization model is constructed with the objective of minimizing the error between the measured and calculated dust concentration values ​​at monitoring points; The optimization model incorporates time smoothness, spatial continuity, and non-negativity constraints to form a comprehensive constraint optimization equation. The comprehensive constraint optimization equation is solved using an iterative optimization algorithm to obtain the optimal source strength distribution matrix.

8. The dust emission source tracing method for electronic greenhouse scenarios according to claim 1, characterized in that, The steps for visualizing pollution source tracing results include: Spatial cluster analysis is performed on the source intensity distribution matrix obtained by inversion to identify the spatial location of the main dust emission sources; Calculate the contribution ratio of each emission source to the dust concentration at the monitoring point and its time change trend, and quantify the emission intensity of each source; Based on the source strength distribution results, a spatial heat map of pollution sources is generated, and combined with time series data, an emission intensity change curve is generated to achieve a visual display of the pollution source tracing results.

9. A dust emission traceability system for electronic greenhouse scenarios, characterized in that, The system for implementing the dust emission source tracing method according to any one of claims 1-8 includes: The image preprocessing module is used to acquire and preprocess dust pollution images inside the electronic greenhouse, and to construct an image pyramid based on the preprocessed pollution images. The diffusion vector calculation module is used to apply optical flow algorithms to the image pyramid, track the motion vector field of dust particles, and calculate the diffusion vector of dust in each area in real time. The grid data acquisition module is used to synchronously acquire data from multiple moments at fixed time intervals within a set time window. The data at each moment includes the measured dust concentration of the grid monitoring points, the diffusion vector obtained from optical flow analysis, and real-time meteorological parameters. The transmission matrix construction module is used to correct the Gaussian diffusion model parameters based on the optical flow analysis results at each sampling time, and calculate the concentration transmission coefficient from each gridded potential emission source to each monitoring point in combination with the meteorological conditions at the sampling time, thereby generating the transmission matrix at the sampling time. The spatiotemporal joint inversion module is used to combine the transfer matrices and observation data corresponding to multiple sampling times to form a source strength inversion equation set. The source strength inversion equation set is solved by an optimization algorithm. The optimization objective comprehensively considers the minimization of data fitting error, source strength temporal smoothness constraint, spatial continuity constraint, and non-negativity constraint. The pollution source tracing result visualization module is used to identify the spatial location of major dust emission sources based on the inversion solution results, quantify the emission intensity and contribution ratio of each source, generate pollution source heat maps and time series emission curves, and realize the visualization of pollution source tracing results. The steps for generating the sampling time transfer matrix include: The greenhouse meteorological parameters were collected at the sampling time, and the theoretical concentration transport coefficients between potential emission sources and each monitoring point were calculated based on the modified Gaussian diffusion model. Substitute wind speed, wind direction, and diffusion coefficient parameters into the Gaussian diffusion equation to solve for the concentration distribution of dust in the grid space; The calculated concentration distribution results are normalized to form the transmission relationship between potential emission sources and monitoring points; The transmission relationships of each grid cell are combined according to their spatial location to generate the overall transmission matrix at the sampling time.

Images