Monitoring method and system for measuring particulate matters by using near-infrared laser

Abstract

The invention provides a monitoring method and system for measuring particulate matters by using near-infrared laser, and relates to the technical field of monitoring for measuring particulate matters by using near-infrared laser, and the method comprises the steps: obtaining monitoring videos and regional meteorological data of a plurality of discharge channels of factory particulate matters; determining a plurality of supplementary placement points based on the laser radar data of each preliminary placement point; acquiring laser radar data of each supplementary placement point; determining a corrected monitoring area based on the laser radar data of each preliminary placement point, the laser radar data of each supplementary placement point, the monitoring videos of a plurality of emission channels of the factory particulate matters, and the regional meteorological data; determining a plurality of target monitoring points based on the corrected monitoring area; and placing the plurality of near-infrared laser radars at the plurality of target monitoring points to monitor the particulate matters. The method can accurately identify the high-concentration diffusion area of the particulate matters and determine the optimal dynamic layout scheme of the near-infrared laser radars.

Description

Technical Field

[0001] This invention relates to the field of monitoring technology for particulate matter measurement using near-infrared lasers, and specifically to a monitoring method and system for measuring particulate matter using near-infrared lasers. Background Technology

[0002] Factory particulate matter monitoring is a crucial aspect of industrial environmental regulation, aiming to achieve compliant control and precise treatment of industrial emission systems by real-time tracking and quantification of particulate matter concentration distribution from emission channels. Currently, particulate matter monitoring based on near-infrared lidar typically employs fixed-point deployment or manual point selection by technicians. This involves either pre-setting long-term monitoring points within the factory area or having monitoring personnel determine detection locations based on experience during specific tasks. However, this method has significant limitations in balancing monitoring costs and capturing high-value pollution targets. Fixed deployment struggles to respond to the real-time migration paths of particulate matter driven by instantaneous weather and emission conditions, resulting in the equipment's field of view often being in low-concentration areas, leading to ineffective use of monitoring resources. Manual point selection, on the other hand, is limited by subjective perception biases, making it difficult to accurately pinpoint critical high-concentration core areas during complex diffusion processes. This limitation not only leads to significant resource waste and maintenance costs when monitoring point deployment is redundant but also results in the inability to cover core pollution areas when points are sparse. Furthermore, insufficient targeting of the monitoring layout directly weakens the data support for subsequent analysis, ultimately reducing the accuracy of monitoring results in reflecting the actual pollution situation.

[0003] Therefore, accurately identifying areas of high particulate matter diffusion and determining the optimal dynamic deployment scheme for near-infrared lidar are urgent problems to be solved. Summary of the Invention

[0004] The main technical problem solved by this invention is how to accurately identify areas of high particulate matter diffusion and determine the optimal dynamic deployment scheme of near-infrared lidar.

[0005] According to a first aspect, the present invention provides a method for monitoring particulate matter using near-infrared lasers, comprising: acquiring monitoring videos of multiple emission channels of particulate matter from a factory and regional meteorological data; determining a preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on the monitoring videos of the multiple emission channels of particulate matter from the factory and the regional meteorological data; placing multiple near-infrared laser radars at the multiple preliminary placement points within the preliminary monitoring area and acquiring laser radar data at each preliminary placement point; determining multiple supplementary placement points based on the laser radar data at each preliminary placement point; placing multiple near-infrared laser radars at the supplementary placement points and acquiring laser radar data at each supplementary placement point; determining a corrected monitoring area based on the laser radar data at each preliminary placement point, the laser radar data at each supplementary placement point, the monitoring videos of the multiple emission channels of particulate matter from the factory, and the regional meteorological data; determining multiple target monitoring points based on the corrected monitoring area; and placing multiple near-infrared laser radars at the multiple target monitoring points to monitor particulate matter.

[0006] In one possible implementation, the multiple target monitoring points include multiple first target monitoring points, multiple second target monitoring points, and multiple third target monitoring points. Determining the multiple target monitoring points based on the modified monitoring area includes: determining multiple first target monitoring points based on lidar data from each initial placement point, lidar data from each supplementary placement point, the modified monitoring area, multiple high-concentration area information, and the monitoring importance of each high-concentration area; the first target monitoring points are selected from the initial placement points and supplementary placement points; determining multiple second target monitoring points based on the multiple high-concentration area information and the monitoring importance of each high-concentration area; determining multiple third candidate target monitoring points based on the multiple first target monitoring points and the multiple second target monitoring points; generating simulated lidar data for each third candidate target monitoring point based on monitoring videos of multiple emission channels of particulate matter from the factory, regional meteorological data, the multiple third candidate target monitoring points, lidar data from each initial placement point, and lidar data from each supplementary placement point; and determining multiple third target monitoring points based on the simulated lidar data of each third candidate target monitoring point.

[0007] In one possible implementation, determining multiple second target monitoring points based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area includes: constructing a particulate matter concentration map, which includes multiple high-concentration area nodes and multiple edges between the multiple high-concentration areas, wherein the node features of the high-concentration area nodes include high-concentration area information and the monitoring importance of the high-concentration area; and processing the particulate matter concentration map based on the second target monitoring point determination model to obtain multiple second target monitoring points.

[0008] In one possible implementation, the second target monitoring point determination model is a graph neural network model.

[0009] According to a second aspect, the present invention provides a monitoring system for measuring particulate matter using near-infrared lasers, comprising: a data acquisition module for acquiring monitoring videos of multiple emission channels of particulate matter from a factory and regional meteorological data; a preliminary planning module for determining a preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on the monitoring videos of the multiple emission channels of particulate matter from the factory and the regional meteorological data; a preliminary deployment module for placing multiple near-infrared laser radars at the multiple preliminary placement points within the preliminary monitoring area and acquiring laser radar data at each preliminary placement point; a supplementary planning module for determining multiple supplementary placement points based on the laser radar data at each preliminary placement point; a supplementary deployment module for placing multiple near-infrared laser radars at the supplementary placement points and acquiring laser radar data at each supplementary placement point; a region correction module for determining a corrected monitoring area based on the laser radar data at each preliminary placement point, the laser radar data at each supplementary placement point, the monitoring videos of the multiple emission channels of particulate matter from the factory, and the regional meteorological data; a target determination module for determining multiple target monitoring points based on the corrected monitoring area; and a monitoring execution module for placing multiple near-infrared laser radars at the multiple target monitoring points to monitor particulate matter.

[0010] In one possible implementation, the multiple target monitoring points include multiple first target monitoring points, multiple second target monitoring points, and multiple third target monitoring points. The target determination module is further configured to: determine multiple first target monitoring points, multiple high-concentration area information, and the monitoring importance of each high-concentration area based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the corrected monitoring area, and the first target monitoring points selected from the initial placement points and supplementary placement points; determine multiple second target monitoring points based on the multiple high-concentration area information and the monitoring importance of each high-concentration area; determine multiple third candidate target monitoring points based on the multiple first target monitoring points and the multiple second target monitoring points; generate simulated lidar data for each third candidate target monitoring point based on the monitoring videos of multiple emission channels of particulate matter from the factory, the regional meteorological data, the multiple third candidate target monitoring points, the lidar data of each initial placement point, and the lidar data of each supplementary placement point; and determine multiple third target monitoring points based on the simulated lidar data of each third candidate target monitoring point.

[0011] In one possible implementation, determining multiple second target monitoring points based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area includes: constructing a particulate matter concentration map, which includes multiple high-concentration area nodes and multiple edges between the multiple high-concentration areas, wherein the node features of the high-concentration area nodes include high-concentration area information and the monitoring importance of the high-concentration area; and processing the particulate matter concentration map based on the second target monitoring point determination model to obtain multiple second target monitoring points.

[0012] In one possible implementation, the second target monitoring point determination model is a graph neural network model.

[0013] According to a third aspect, embodiments of the present invention provide an electronic device, including: a processor; a memory; and a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the method as described above, the method including: acquiring monitoring videos of multiple emission channels of particulate matter from a factory and regional meteorological data; determining a preliminary monitoring area and multiple preliminary placement points in the preliminary monitoring area based on the monitoring videos of the multiple emission channels of particulate matter from the factory and the regional meteorological data; placing multiple near-infrared lidars at the multiple preliminary placement points in the preliminary monitoring area and acquiring lidar data at each preliminary placement point; determining multiple supplementary placement points based on the lidar data at each preliminary placement point; placing multiple near-infrared lidars at the supplementary placement points and acquiring lidar data at each supplementary placement point; determining a corrected monitoring area based on the lidar data at each preliminary placement point, the lidar data at each supplementary placement point, the monitoring videos of the multiple emission channels of particulate matter from the factory, and the regional meteorological data; determining multiple target monitoring points based on the corrected monitoring area; and placing multiple near-infrared lidars at the multiple target monitoring points to monitor particulate matter.

[0014] According to a fourth aspect, this embodiment provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aforementioned monitoring method for measuring particulate matter using near-infrared lasers. The method includes: acquiring monitoring videos of multiple emission channels of particulate matter from a factory and regional meteorological data; determining a preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on the monitoring videos of the multiple emission channels and the regional meteorological data; placing multiple near-infrared laser radars at the multiple preliminary placement points within the preliminary monitoring area and acquiring laser radar data at each preliminary placement point; determining multiple supplementary placement points based on the laser radar data at each preliminary placement point; placing multiple near-infrared laser radars at the supplementary placement points and acquiring laser radar data at each supplementary placement point; determining a corrected monitoring area based on the laser radar data at each preliminary placement point, the laser radar data at each supplementary placement point, the monitoring videos of the multiple emission channels of particulate matter from the factory, and the regional meteorological data; determining multiple target monitoring points based on the corrected monitoring area; and placing multiple near-infrared laser radars at the multiple target monitoring points to monitor particulate matter.

[0015] This invention provides a method and system for monitoring particulate matter using near-infrared lasers. The method includes acquiring monitoring videos of multiple emission channels of particulate matter from a factory and regional meteorological data; determining a preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on the monitoring videos of the multiple emission channels and the regional meteorological data; placing multiple near-infrared laser radars at the multiple preliminary placement points within the preliminary monitoring area and acquiring laser radar data at each preliminary placement point; determining multiple supplementary placement points based on the laser radar data at each preliminary placement point; placing multiple near-infrared laser radars at the supplementary placement points and acquiring laser radar data at each supplementary placement point; determining a corrected monitoring area based on the laser radar data at each preliminary placement point, the laser radar data at each supplementary placement point, the monitoring videos of the multiple emission channels of particulate matter from the factory, and the regional meteorological data; determining multiple target monitoring points based on the corrected monitoring area; and placing multiple near-infrared laser radars at the multiple target monitoring points to monitor particulate matter. This method can accurately identify areas of high-concentration particulate matter diffusion and determine the optimal dynamic deployment scheme for near-infrared laser radars. Attached Figure Description

[0016] Figure 1 A schematic flowchart of a method for monitoring particulate matter using near-infrared laser measurement provided in an embodiment of the present invention;

[0017] Figure 2 A schematic diagram illustrating the distribution of multiple particulate matter emission channels in a factory, provided as an embodiment of the present invention;

[0018] Figure 3 A schematic diagram of a near-infrared lidar provided in an embodiment of the present invention;

[0019] Figure 4 This is a schematic diagram of a process for determining multiple target monitoring points provided in an embodiment of the present invention;

[0020] Figure 5 This is a schematic diagram of a process for determining multiple second target monitoring points, provided by an embodiment of the present invention.

[0021] Figure 6 This is a schematic diagram of a monitoring system for measuring particulate matter using near-infrared laser, provided as an embodiment of the present invention. Detailed Implementation

[0022] The present invention will now be described in further detail with reference to specific embodiments and accompanying drawings. Similar elements in different embodiments are referred to by associated similar element reference numerals. In the following embodiments, many details are described to facilitate a better understanding of the invention. However, those skilled in the art will readily recognize that some features may be omitted in different situations, or may be replaced by other elements, materials, or methods. In some cases, certain operations related to the present invention are not shown or described in the specification. This is to avoid obscuring the core parts of the invention with excessive description. For those skilled in the art, detailed description of these related operations is not necessary; they can fully understand the related operations based on the description in the specification and general technical knowledge in the art.

[0023] In this embodiment of the invention, the following are provided: Figure 1 The method for monitoring particulate matter using near-infrared laser measurement, as shown, includes steps S1 to S8:

[0024] Step S1: Obtain monitoring videos and regional meteorological data from multiple emission channels of particulate matter from the factory.

[0025] The monitoring videos of multiple particulate matter emission channels in a factory are real-time video data collected by camera equipment installed near each particulate matter emission channel in the factory. Figure 2 This is a schematic diagram illustrating the distribution of multiple particulate matter emission channels in a factory, as provided in an embodiment of the present invention.

[0026] Monitoring videos of multiple emission channels for particulate matter from a factory can be used to visually record the color, concentration, and emission direction of the gas emitted from the emission channels.

[0027] Regional meteorological data are real-time environmental parameters collected by meteorological stations in the area where the factory is located. Regional meteorological data include wind speed, wind direction, ambient temperature, relative humidity, and atmospheric pressure.

[0028] Step S2: Based on the monitoring videos of multiple emission channels of particulate matter from the factory and the local meteorological data, determine the preliminary monitoring area and multiple preliminary placement points in the preliminary monitoring area.

[0029] In some embodiments, a preliminary regional planning model can be used to determine a preliminary monitoring area and multiple preliminary placement points within that area. The preliminary regional planning model is a Transformer model. The inputs to the preliminary regional planning model are monitoring videos of multiple particulate matter emission channels from the factory and regional meteorological data. The outputs of the preliminary regional planning model are the preliminary monitoring area and multiple preliminary placement points within that area.

[0030] The Transformer model consists of an encoder and a decoder. The encoder captures global dependencies in the input sequence through a self-attention mechanism and transforms them into a high-dimensional representation. The decoder generates the target sequence using the encoder's output and a mask multi-head attention mechanism. The Transformer model can process data in parallel and has significant advantages when dealing with long-range temporal information.

[0031] The preliminary monitoring area is a geographical area that needs to be monitored in advance by lidar, defined by a preliminary regional planning model based on the factory's particulate matter emission characteristics and meteorological conditions.

[0032] The initial placement points in the preliminary monitoring area were determined through the preliminary area planning model, which specifies the locations for placing near-infrared lidar within the preliminary monitoring area.

[0033] Determining the initial monitoring area and multiple initial placement points within that area provides precise spatial guidance for particulate matter capture. By defining the initial monitoring area, the broad factory environment can be narrowed down to a specific geographical range where particulate matter has a high probability of spreading, thus achieving initial screening of monitoring targets. Planning multiple initial placement points discretizes the continuous diffusion area into specific spatial coordinates, thereby constructing a basic spatial sampling network targeting the pollution plume morphology, ensuring that the initial spatial layout effectively covers core pollution characteristics.

[0034] Monitoring videos from multiple emission channels of particulate matter at the factory recorded the initial state and instantaneous flow direction of particulate matter emissions, while regional meteorological data provided dynamic environmental parameters affecting the diffusion velocity and direction of particulate matter. These monitoring videos and regional meteorological data provided the model with the physical environmental context of the entire process from source emission to atmospheric transport, enabling the model to extrapolate the spatial distribution trend of particulate matter based on current emission intensities and meteorological conditions.

[0035] The Transformer model, through its self-attention mechanism, can capture changes in plume morphology and temporal series characteristics of meteorological data within a sequence of monitored video frames. It can utilize location encoding to understand the temporal evolution and spatial distribution of data, and establish a nonlinear mapping between environmental parameters and particulate matter dispersion range. This allows for the prediction of preliminary monitoring areas covering major pollution plumes and the planning of multiple preliminary placement points that can effectively grid the area.

[0036] In some embodiments, determining the preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on monitoring videos of multiple emission channels of particulate matter from the factory and regional meteorological data includes steps S21-S23:

[0037] Step S21: Based on the monitoring videos of multiple emission channels of particulate matter from the factory and the regional meteorological data, determine the injection direction of each emission channel, the emission intensity level of each emission channel, the geometric morphology characteristics of the plume, and the meteorological condition adaptability score.

[0038] In some embodiments, the Transformer model can be used to determine the injection direction of each emission channel, the emission intensity level of each emission channel, the plume geometry, and the meteorological condition suitability score.

[0039] The injection direction of each emission channel is determined by the Transformer model, which provides azimuth information on the direction of airflow movement in the emission channel. The injection direction of each emission channel includes both horizontal and vertical angle values.

[0040] The emission intensity level for each emission channel is output by the Transformer model and is a grading index that quantifies the intensity of particulate matter emissions from each emission channel. A higher emission intensity level indicates that the emission channel emits more particulate matter and at a higher concentration.

[0041] The plume geometry features are information about the geometric shape of the emitted flue gas in space, output by the Transformer model. These features include plume rise height, plume visual width, plume diffusion angle, and plume edge sharpness.

[0042] The meteorological condition suitability score, output by the Transformer model, is a numerical indicator that quantifies the impact of meteorological conditions on emission dispersion. A higher meteorological condition suitability score indicates that the current wind speed and direction conditions are more favorable for pollutant dispersion and transport along the emission pathway.

[0043] Meteorological condition suitability score is used to measure the degree of fit between current wind speed and direction conditions and emission characteristics.

[0044] The Transformer model, utilizing its self-attention mechanism, can process frame sequences of surveillance video to capture the dynamic and continuous changes in plume emission, thereby accurately determining the emission direction and calculating emission intensity. Simultaneously, by combining the time-series characteristics of meteorological data, the model can analyze the real-time reshaping effect of wind on plume morphology, establishing a nonlinear mapping between visual morphology and physical geometric features and the degree of meteorological influence. Ultimately, it accurately determines the plume's geometric morphology and its suitability score for meteorological conditions.

[0045] Step S22: Based on the injection direction of each emission channel, the emission intensity level of each emission channel, the geometric morphology of the plume, and the meteorological condition adaptability score, determine the high probability diffusion area and diffusion core area of ​​particulate matter.

[0046] In some embodiments, deep neural networks can be used to determine the high-probability diffusion region and the diffusion core region of particulate matter.

[0047] A deep neural network is a nonlinear computational model consisting of an input layer, multiple hidden layers, and an output layer. Each layer in a deep neural network contains multiple neurons, which can perform linear transformations on the input data through weight matrices and bias terms. Deep neural networks can process the transformation results using nonlinear activation functions, thereby enabling them to fit extremely complex function mappings.

[0048] High-probability particulate matter (PPM) diffusion areas are geographical regions with a high probability of PPM diffusion, predicted by deep neural networks based on emission characteristics and meteorological conditions. These PPM diffusion areas represent the maximum potential impact range of pollutants.

[0049] The diffusion core area, identified through a deep neural network, is the central geographical region with the highest pollutant concentration and most severe impact within the high-probability particulate matter diffusion area. Located at the center of the high-probability particulate matter diffusion area, the diffusion core area exhibits significantly higher pollutant concentrations than the surrounding areas.

[0050] Deep neural networks can learn the complex nonlinear relationships between different emission patterns and downstream spatial distribution through feature fusion. They can simulate the statistical laws governing atmospheric turbulent diffusion processes and, based on current jet momentum and meteorological influencing factors, deduce the transport trajectory of particulate matter in three-dimensional space, thereby delineating broad-coverage high-probability diffusion areas and concentration-rich diffusion core regions.

[0051] Step S23: Based on the high-probability diffusion area of ​​particulate matter and the diffusion core area, determine the preliminary monitoring area and multiple preliminary placement points in the preliminary monitoring area.

[0052] In some embodiments, a deep neural network may be used to determine the initial monitoring area and multiple initial placement points within the initial monitoring area.

[0053] Deep neural networks can analyze the spatial shape and size of areas with a high probability of particulate matter diffusion. The model can learn the optimal monitoring area delineation strategy by combining the location characteristics of the diffusion core area. By evaluating the monitoring effectiveness of different area delineation schemes, deep neural networks can determine the initial monitoring area boundaries that can both cover areas with a high probability of particulate matter diffusion and focus on the diffusion core area. At the same time, based on the geometric features and monitoring needs of the initial monitoring area, deep neural networks can optimize the spatial location of multiple initial placement points within the area, so that the monitoring network achieves the best balance between coverage and monitoring density.

[0054] Step S3: Place multiple near-infrared lidars at multiple initial placement points in the initial monitoring area, and acquire lidar data for each initial placement point.

[0055] The lidar data for each initial placement point is measurement data reflecting the characteristics of particulate matter around that location, collected by near-infrared lidar devices deployed at multiple initial placement points in the initial monitoring area. Figure 3 This is a schematic diagram of a near-infrared lidar provided in an embodiment of the present invention.

[0056] The lidar data at the initial placement point includes the echo signal intensity sequence, particulate matter extinction coefficient distribution, and particulate matter concentration distribution at different heights.

[0057] The echo signal intensity sequence is the signal intensity scattered back by atmospheric particles at different distances after the laser pulse is emitted.

[0058] The extinction coefficient distribution of particulate matter refers to the degree of energy attenuation of laser light due to absorption and scattering of particulate matter along the propagation path.

[0059] The numerical values ​​for particulate matter concentration distribution at different altitudes represent the spatial distribution of particulate matter concentration at different altitude levels.

[0060] Step S4: Determine multiple supplementary placement points based on the lidar data of each initial placement point.

[0061] In some embodiments, a supplementary monitoring and determination model can be used to determine multiple supplementary placement points. The supplementary monitoring and determination model is a deep neural network. The input to the supplementary monitoring and determination model is the LiDAR data of each initial placement point, and the output of the supplementary monitoring and determination model is the multiple supplementary placement points.

[0062] Multiple supplementary placement points were determined through a supplementary monitoring model to cover the blind spots of the initial placement point monitoring and refine the deployment coordinates of near-infrared lidar equipment.

[0063] The lidar data from each initial placement point reflects the actual particulate matter distribution intensity and spatial uncertainty within the initially planned area. This data provides experimental evidence for the model to identify monitoring blind spots. Deep neural networks, through nonlinear transformations of multiple layers of neurons, can extract deep features from the lidar data at each initial placement point, thereby identifying edge areas with drastic concentration changes or insufficient signal coverage. The deep neural network can maximize the target based on monitoring completeness, and then locate the areas requiring enhanced observation within the gaps of the initial monitoring grid, ultimately determining multiple supplementary placement points.

[0064] Step S5: Place multiple near-infrared lidars at the supplementary placement points and acquire lidar data for each supplementary placement point.

[0065] The lidar data for each replenishment point is measurement data reflecting the characteristics of particulate matter in the surrounding area, collected by near-infrared lidar equipment deployed at the replenishment point. The lidar data for each replenishment point includes the echo signal intensity sequence, particulate matter extinction coefficient distribution, and particulate matter concentration distribution at different heights.

[0066] Step S6: Determine the corrected monitoring area based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the monitoring video of multiple emission channels of particulate matter from the factory, and the regional meteorological data.

[0067] In some embodiments, a correction area determination model can be used to determine the correction monitoring area. The correction area determination model is a Transformer model. The inputs to the correction area determination model are the lidar data from each initial placement point, the lidar data from each supplementary placement point, the monitoring video from multiple emission channels of particulate matter from the factory, and the regional meteorological data. The output of the correction area determination model is the correction monitoring area.

[0068] The revised monitoring area is determined by the revised area determination model, which is a precise geographical range for monitoring the core of particulate matter emissions and diffusion after the initial monitoring area has been calibrated, optimized, and updated.

[0069] Modifying the monitoring area allows for more precise targeting of the core areas of particulate matter emissions and diffusion, while excluding irrelevant or less influential areas, thus making the monitoring more targeted and efficient.

[0070] LiDAR data from each initial placement point and each supplementary placement point provides the measured concentration distribution field across the entire area. Monitoring videos from multiple emission channels of particulate matter from the factory, along with regional meteorological data, provide the boundary conditions for pollution source emissions and environmental transport. The Transformer model possesses powerful sequence modeling capabilities, enabling feature fusion of multi-point LiDAR observation data with visual flow data and meteorological sequences. By comparing the predicted diffusion field with the actual observation field, the Transformer model can automatically correct deviations in the initial planning and eliminate redundant areas with low contribution, ultimately determining a corrected monitoring area that better reflects the actual pollutant diffusion pattern.

[0071] Step S7: Determine multiple target monitoring points based on the corrected monitoring area.

[0072] In some embodiments, Figure 4 This is a schematic diagram of a process for determining multiple target monitoring points according to an embodiment of the present invention. The determination of multiple target monitoring points includes steps S31 to S35:

[0073] Step S31: Based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the modified monitoring area, multiple first target monitoring points, multiple high-concentration area information, and the monitoring importance of each high-concentration area are determined. The first target monitoring points are selected from the initial placement points and supplementary placement points.

[0074] In some embodiments, a monitoring point optimization model can be used to determine multiple first target monitoring points, multiple high-concentration area information, and the monitoring importance of each high-concentration area. The monitoring point optimization model is a deep neural network. The inputs to the monitoring point optimization model are the lidar data of each initial placement point, the lidar data of each supplementary placement point, and the corrected monitoring area. The outputs of the monitoring point optimization model are the multiple first target monitoring points, the multiple high-concentration area information, and the monitoring importance of each high-concentration area.

[0075] Multiple primary target monitoring points were determined through a monitoring point optimization model, selected from the core monitoring locations of the initial placement points and supplementary placement points.

[0076] Information on multiple high-concentration areas was determined through a monitoring point optimization model, which corrected the specific information of areas with high particulate matter concentrations within the monitoring area. This information includes the geographical location of each high-concentration area, the average particulate matter concentration within the area, the peak concentration, and the duration of the high concentration.

[0077] The monitoring importance of each high-concentration area is determined by the monitoring point optimization model, which is used to quantify the contribution of each high-concentration area to the overall monitoring task.

[0078] The importance of monitoring high-concentration areas can be used to guide the priority allocation of monitoring resources; the higher the value, the more important the high-concentration area needs to be monitored.

[0079] The lidar data from each initial placement point and each supplementary placement point provides complete information on the particulate matter concentration distribution within the corrected monitoring area. Using this data, the model can accurately identify the location and extent of high-concentration areas. The corrected monitoring area defines an optimized monitoring geographic scope and provides spatial constraints for point location optimization.

[0080] Deep neural networks can learn the spatial distribution characteristics of particulate matter concentration from lidar data at each initial placement point and each supplementary placement point, thereby identifying the distribution patterns of high-concentration areas. By analyzing and correcting the concentration distribution patterns within the monitoring area, deep neural networks can select the points that contribute the most to the overall monitoring network from the initial and supplementary placement points as multiple primary target monitoring points. Furthermore, deep neural networks can evaluate the monitoring value of different high-concentration areas, thereby identifying multiple high-concentration area information and quantifying the monitoring importance of each high-concentration area.

[0081] Step S32: Determine multiple second target monitoring points based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area.

[0082] In some embodiments, Figure 5 This is a schematic flowchart illustrating the process of determining multiple second target monitoring points according to an embodiment of the present invention. The determination of multiple second target monitoring points includes steps S41-S42:

[0083] Step S41: Construct a particulate matter concentration map, which includes multiple high-concentration region nodes and multiple edges between the high-concentration regions. The node features of the high-concentration region nodes include high-concentration region information and the monitoring importance of the high-concentration region.

[0084] A particulate matter concentration map is a structured data format that represents the relationships between high-concentration regions in a topological manner. A particulate matter concentration map consists of multiple high-concentration region nodes and multiple edges connecting these regions. Each high-concentration region node represents a high-concentration region. The node characteristics of each high-concentration region node include information about the high-concentration region and its monitoring importance.

[0085] Each edge in the multiple edges represents the positional relationship between two adjacent high-concentration region nodes. The positional relationship includes the relative direction and physical distance between the center points of the two high-concentration regions.

[0086] Physical distance is used to represent the geographical proximity of different high-concentration areas, while relative direction is used to represent the spatial orientation of high-concentration areas driven by meteorological fields.

[0087] Edges can quantify the topological connections in the spatial distribution of different high-concentration areas, thereby enabling the description of the potential diffusion paths or spatial coupling of pollutants within the modified monitoring area.

[0088] Step S42: Based on the second target monitoring point determination model, the particulate matter concentration spectrum is processed to obtain multiple second target monitoring points.

[0089] The second target monitoring point determination model is a graph neural network model. The input of the second target monitoring point determination model is the particulate matter concentration spectrum, and the output of the second target monitoring point determination model is multiple second target monitoring points.

[0090] Graph neural network models consist of graph neural networks (GNNs) and fully connected layers. A graph neural network is a deep learning architecture capable of processing graphs. Through message passing mechanisms, GNNs can aggregate information between nodes and their neighbors to capture the complex topology and node relationships within the entire graph.

[0091] Multiple secondary target monitoring points are determined by the secondary target monitoring point determination model and are used to strengthen the monitoring of key areas in high-concentration regions.

[0092] Secondary target monitoring points can be deployed within or at the boundary of high-concentration areas to supplement monitoring coverage.

[0093] By constructing particulate matter concentration maps, the spatial evolution network of each pollution core within the monitoring area can be topologically reshaped and corrected. This introduction of spatial correlation is crucial for improving monitoring efficiency because the spatiotemporal evolution of particulate matter is not isolated but significantly constrained by the dynamic interactions and spatial coupling effects between adjacent high-concentration areas. By mapping high-concentration area information and their monitoring importance to node features, and simultaneously defining the relative direction and physical distance between areas as edges, a deep digital expression of the physical evolution constraints within the monitoring area can be achieved. This data construction method can drive graph neural networks to analyze the transport weights and response relationships between high-concentration areas at a higher spatial scale, thereby significantly improving the scientific rigor and coverage accuracy of monitoring point deployment.

[0094] Utilizing graph neural network models for deep processing of particulate matter concentration maps overcomes the limitations of traditional point-based algorithms in capturing nonlinear spatial dependencies and effectively learns the complex interaction logic and information propagation characteristics between nodes. Graph neural network models possess powerful topological feature extraction capabilities when processing such non-Euclidean spatial data and can accurately identify key monitoring locations along diffusion paths, thereby determining multiple secondary target monitoring points with global collaborative monitoring capabilities.

[0095] Graph neural networks (Graph Neural Networks) aggregate information from adjacent high-concentration areas by passing messages across particulate matter concentration maps and learn the contextual representation of each high-concentration area. Based on the monitoring importance of high-concentration areas and the correlations between areas, Graph Neural Networks can infer the optimal distribution of monitoring points within high-concentration areas. Through graph convolution operations, Graph Neural Network models can capture pollutant transport paths and diffusion patterns between high-concentration areas, thereby identifying multiple secondary target monitoring points at key locations to achieve effective monitoring coverage of key areas.

[0096] Step S33: Determine multiple third candidate target monitoring points based on the multiple first target monitoring points and the multiple second target monitoring points.

[0097] In some embodiments, a candidate monitoring point determination model can be used to determine multiple third candidate target monitoring points. The candidate monitoring point determination model is a deep neural network. The input to the candidate monitoring point determination model is the multiple first target monitoring points and the multiple second target monitoring points, and the output of the candidate monitoring point determination model is the multiple third candidate target monitoring points.

[0098] Multiple third-candidate target monitoring points are candidate position sequences determined by the candidate monitoring point determination model after spatial fusion and layout optimization of multiple first-target monitoring points and multiple second-target monitoring points. These sequences are used to simulate and verify the monitoring effectiveness.

[0099] The first target monitoring point provides representative monitoring site information selected based on measured data, while the second target monitoring point provides key enhanced monitoring site information determined based on the spatial correlation characteristics of high-concentration areas. By analyzing the spatial relationship between multiple first and second target monitoring points within the corrected monitoring area, the model can assess the coverage overlap of different coordinate combinations, eliminate redundant points, and optimize the spatial layout, thereby identifying multiple third candidate target monitoring points.

[0100] Deep neural networks can comprehensively analyze the spatial distribution characteristics of multiple primary and secondary target monitoring points and evaluate the overall monitoring effectiveness under different spatial coordinate arrangements. By learning the spatial relationships between points and the complementarity of monitoring coverage, deep neural networks can spatially fuse and optimize the layout of multiple primary and secondary target monitoring points, thereby deducing and determining multiple third candidate target monitoring points with better spatial representativeness within the corrected monitoring area. Deep neural networks can balance the coverage and monitoring sensitivity of the monitoring network, thus determining the set of candidate monitoring points that can maximize the capture of particulate matter diffusion dynamics.

[0101] Step S34: Based on the monitoring videos of multiple emission channels of particulate matter from the factory, the regional meteorological data, the multiple third candidate target monitoring points, the lidar data of each initial placement point, and the lidar data of each supplementary placement point, simulated lidar data for each third candidate target monitoring point is generated.

[0102] In some embodiments, a monitoring simulation model can be used to generate simulated lidar data for each third candidate target monitoring point. The monitoring simulation model is a generative adversarial network (GAN). The inputs to the monitoring simulation model are monitoring videos of multiple emission channels of particulate matter from the factory, regional meteorological data, the multiple third candidate target monitoring points, lidar data for each initial placement point, and lidar data for each supplementary placement point. The output of the monitoring simulation model is simulated lidar data for each third candidate target monitoring point.

[0103] Generative Adversarial Networks (GANs) consist of a generator and a discriminator. The generator and discriminator learn the distribution characteristics of the data through adversarial training. The generator is responsible for generating samples similar to real data, while the discriminator is responsible for distinguishing between real and generated samples. During training, they compete against each other, ultimately enabling the generator to produce high-quality simulated data.

[0104] The simulated lidar data for each third candidate target monitoring point is generated by a monitoring simulation model, and is the virtual lidar monitoring data at each third candidate target monitoring point.

[0105] The simulated lidar data for each third candidate target monitoring point includes the simulated echo signal intensity sequence, simulated particulate matter extinction coefficient distribution, and simulated particulate matter concentration distribution at different heights at each third candidate target monitoring point.

[0106] Simulated lidar data for each third candidate target monitoring point is used to evaluate the monitoring effectiveness of the candidate monitoring points before actual deployment.

[0107] Monitoring videos from multiple emission channels of the factory particulate matter recorded the initial emission intensity of the pollution source and the real-time visual diffusion morphology of the plume, providing physical boundary features for the model to simulate the extinction characteristics of particulate matter. Regional meteorological data, including dynamic parameters such as wind speed and direction that affect the drift trajectory of pollutants, can determine the spatial offset trend of particulate matter concentration in the simulation data. LiDAR data from each initial placement point and each supplementary placement point provided echo signal characteristics and concentration distribution samples under real-world conditions, enabling the model to learn the energy attenuation law and background noise distribution of real laser light during atmospheric transmission, ensuring that the simulated LiDAR data for each third candidate target monitoring point conforms to the measured physical characteristics.

[0108] Generative Adversarial Networks (GANs) are trained through a game-like interaction between a generator and a discriminator to achieve deep simulation of complex environmental data. The generator uses LiDAR data from each initial placement point and each supplementary placement point as training samples, learning the physical statistical distribution pattern of real laser echo signal attenuation with distance. The model can combine emission intensity recorded from monitoring videos of multiple particulate matter emission channels at the factory with diffusion parameters provided by regional meteorological data to establish a mapping relationship between pollution source emissions and meteorologically driven concentration fields. The discriminator receives the simulated signal generated by the generator and compares its features with the LiDAR data from each initial and supplementary placement point, identifying minor differences in physical consistency and signal-noise distribution. Based on feedback from the discriminator, the generator continuously optimizes its internal weights to ensure that the simulated LiDAR data for each third candidate target monitoring point maintains a high degree of consistency with real measurement characteristics in dimensions such as simulated echo signal intensity sequence, simulated particulate matter extinction coefficient distribution, and simulated particulate matter concentration distribution at different heights.

[0109] Step S35: Determine multiple third target monitoring points based on the simulated lidar data of each third candidate target monitoring point.

[0110] In some embodiments, a simulation monitoring and analysis model can be used to determine multiple third target monitoring points. The simulation monitoring and analysis model is a deep neural network. The input to the simulation monitoring and analysis model is the simulated LiDAR data for each of the third candidate target monitoring points, and the output of the simulation monitoring and analysis model is the multiple third target monitoring points.

[0111] The optimal monitoring point is determined by a simulation monitoring analysis model from multiple candidate third target monitoring points.

[0112] The multiple third-target monitoring points are final target locations that have been verified through model simulations and proven to have stable monitoring performance under various simulated operating conditions. These multiple third-target monitoring points, together with the multiple first-target monitoring points and the multiple second-target monitoring points, constitute a complete set of target monitoring points.

[0113] The simulated lidar data for each third candidate target monitoring point provides a virtual experimental sample for evaluating the detection effectiveness of each candidate site under simulated emission conditions. The simulated echo signal intensity sequence included in the simulated lidar data for each third candidate target monitoring point reflects the detection sensitivity of the candidate site to distant targets; the simulated particulate matter extinction coefficient distribution reflects the site's ability to identify changes in atmospheric transparency; and the simulated particulate matter concentration distribution at different altitudes demonstrates the site's accuracy in capturing the vertical spatial distribution of particulate matter. By performing multi-dimensional analysis of these quantitative indicators, the simulation monitoring and analysis model can compare the accuracy and stability of different candidate sites in capturing the dynamics of pollution plumes, thereby identifying the coordinate location with the best monitoring performance among multiple third candidate target monitoring points.

[0114] Deep neural networks can extract evaluation features from simulated lidar data of each third candidate target monitoring point and establish a nonlinear mapping model between simulated signal features and the quality of monitoring sites. By performing deep feature analysis on the simulated echo signal intensity sequence, simulated particulate extinction coefficient distribution, and simulated particulate concentration distribution at different heights in the simulated lidar data of each third candidate target monitoring point, deep neural networks can evaluate the detection sensitivity, spatial resolution, and stability of plume diffusion capture of each site under simulated conditions. Utilizing its powerful pattern recognition capabilities, deep neural networks can analyze the signal quality and concentration response accuracy in the simulated data of each candidate site, thereby quantitatively evaluating the comprehensive monitoring value of each third candidate target monitoring point. Based on the evaluation results, deep neural networks can rank and optimize multiple third candidate target monitoring points based on their effectiveness, thereby identifying the multiple third target monitoring points with the best monitoring effectiveness.

[0115] Step S8: Place multiple near-infrared lidars at the multiple target monitoring points to monitor particulate matter.

[0116] Once the multiple target monitoring points are identified, multiple near-infrared lidars are actually deployed on multiple target monitoring points, which together consist of multiple first target monitoring points, multiple second target monitoring points, and multiple third target monitoring points, to carry out actual monitoring of particulate matter in the factory.

[0117] Based on the same inventive concept Figure 6 This is a schematic diagram of a monitoring system for measuring particulate matter using near-infrared laser, provided by an embodiment of the present invention. The monitoring system for measuring particulate matter using near-infrared laser includes:

[0118] The data acquisition module 51 is used to acquire monitoring videos of multiple emission channels of particulate matter from the factory and regional meteorological data;

[0119] The preliminary planning module 52 is used to determine the preliminary monitoring area and multiple preliminary placement points in the preliminary monitoring area based on the monitoring videos of multiple emission channels of particulate matter from the factory and the meteorological data of the region.

[0120] The initial deployment module 53 is used to place multiple near-infrared lidars at multiple initial placement points in the initial monitoring area and acquire lidar data at each initial placement point.

[0121] Supplementary planning module 54 is used to determine multiple supplementary placement points based on the lidar data of each initial placement point;

[0122] The supplementary deployment module 55 is used to place multiple near-infrared lidars at the supplementary placement points and acquire lidar data for each supplementary placement point;

[0123] The area correction module 56 is used to determine the correction monitoring area based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the monitoring video of multiple emission channels of particulate matter in the factory, and the regional meteorological data.

[0124] Target determination module 57 is used to determine multiple target monitoring points based on the modified monitoring area;

[0125] The monitoring execution module 58 is used to place multiple near-infrared lidars at the multiple target monitoring points to monitor particulate matter.

[0126] It should be noted that, in order to simplify the descriptions disclosed herein and thus aid in the understanding of one or more embodiments of the invention, the foregoing description of embodiments of this specification may sometimes combine multiple features into a single embodiment, drawing, or description thereof. However, this method of disclosure does not imply that the subject matter of this specification requires more features than those mentioned in the claims. In fact, the embodiments contain fewer features than all the features of a single embodiment disclosed above.

[0127] Finally, it should be understood that the embodiments described in this specification are merely illustrative of the principles of the embodiments described herein. Other variations may also fall within the scope of this specification. Therefore, alternative configurations of the embodiments described herein are intended to be illustrative rather than limiting, and should be considered consistent with the teachings of this specification. Accordingly, the embodiments described herein are not limited to those explicitly introduced and described herein.

Claims

1. A method for monitoring particulate matter using near-infrared laser, characterized in that, include: Acquire monitoring videos and regional meteorological data from multiple emission channels of particulate matter from the factory; Based on monitoring videos of multiple particulate matter emission channels from the factory and meteorological data of the region, a preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area were determined. Multiple near-infrared lidars are placed at multiple initial placement points in the initial monitoring area, and lidar data is acquired at each initial placement point; Based on the lidar data of each initial placement point, multiple supplementary placement points are determined; Multiple near-infrared lidars are placed at the supplementary placement points, and lidar data for each supplementary placement point is acquired; The corrected monitoring area is determined based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the monitoring video of multiple emission channels of particulate matter in the factory, and the regional meteorological data. Multiple target monitoring points are determined based on the modified monitoring area; Multiple near-infrared lidar sensors are placed at the multiple target monitoring points to monitor particulate matter.

2. The monitoring method for particulate matter using near-infrared laser as described in claim 1, characterized in that, Multiple target monitoring points include multiple first target monitoring points, multiple second target monitoring points, and multiple third target monitoring points. Determining multiple target monitoring points based on the corrected monitoring area includes: Based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the modified monitoring area, multiple first target monitoring points, multiple high-concentration area information, and the monitoring importance of each high-concentration area are determined. The first target monitoring points are selected from the initial placement points and supplementary placement points. Based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area, multiple second target monitoring points are determined; Based on the plurality of first target monitoring points and the plurality of second target monitoring points, a plurality of third candidate target monitoring points are determined; Based on the monitoring videos of multiple particulate matter emission channels of the factory, the regional meteorological data, the multiple third candidate target monitoring points, the lidar data of each initial placement point, and the lidar data of each supplementary placement point, simulated lidar data for each third candidate target monitoring point is generated. Multiple third target monitoring points are determined based on the simulated lidar data of each third candidate target monitoring point.

3. The monitoring method for particulate matter using near-infrared laser as described in claim 2, characterized in that, The determination of multiple second target monitoring points based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area includes: Construct a particulate matter concentration map, which includes multiple high-concentration region nodes and multiple edges between the high-concentration regions. The node characteristics of the high-concentration region nodes include high-concentration region information and the monitoring importance of the high-concentration region. The particulate matter concentration spectrum is processed based on the second target monitoring point determination model to obtain multiple second target monitoring points.

4. The monitoring method for particulate matter using near-infrared laser as described in claim 3, characterized in that, The second target monitoring point determination model is a graph neural network model.

5. A monitoring system for measuring particulate matter using near-infrared laser, characterized in that, include: The data acquisition module is used to acquire monitoring videos of multiple particulate matter emission channels from the factory and regional meteorological data; The preliminary planning module is used to determine the preliminary monitoring area and multiple preliminary placement points within the preliminary monitoring area based on monitoring videos of multiple particulate matter emission channels of the factory and meteorological data of the region. The initial deployment module is used to place multiple near-infrared lidars at multiple initial placement points in the initial monitoring area and acquire lidar data at each initial placement point; A supplementary planning module is used to determine multiple supplementary placement points based on the lidar data of each initial placement point; The supplementary deployment module is used to place multiple near-infrared lidars at the supplementary placement points and acquire lidar data for each supplementary placement point. The area correction module is used to determine the correction monitoring area based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the monitoring video of multiple emission channels of particulate matter in the factory, and the regional meteorological data. The target determination module is used to determine multiple target monitoring points based on the modified monitoring area; The monitoring execution module is used to place multiple near-infrared lidars at the multiple target monitoring points to monitor particulate matter.

6. The monitoring system for measuring particulate matter using near-infrared laser as described in claim 5, characterized in that, The multiple target monitoring points include multiple first target monitoring points, multiple second target monitoring points, and multiple third target monitoring points. The target determination module is further used for: Based on the lidar data of each initial placement point, the lidar data of each supplementary placement point, the modified monitoring area, multiple first target monitoring points, multiple high-concentration area information, and the monitoring importance of each high-concentration area are determined. The first target monitoring points are selected from the initial placement points and supplementary placement points. Based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area, multiple second target monitoring points are determined; Based on the plurality of first target monitoring points and the plurality of second target monitoring points, a plurality of third candidate target monitoring points are determined; Based on the monitoring videos of multiple particulate matter emission channels of the factory, the regional meteorological data, the multiple third candidate target monitoring points, the lidar data of each initial placement point, and the lidar data of each supplementary placement point, simulated lidar data for each third candidate target monitoring point is generated. Multiple third target monitoring points are determined based on the simulated lidar data of each third candidate target monitoring point.

7. The monitoring system for measuring particulate matter using near-infrared laser as described in claim 6, characterized in that, The determination of multiple second target monitoring points based on the information of the multiple high-concentration areas and the monitoring importance of each high-concentration area includes: Construct a particulate matter concentration map, which includes multiple high-concentration region nodes and multiple edges between the high-concentration regions. The node characteristics of the high-concentration region nodes include high-concentration region information and the monitoring importance of the high-concentration region. The particulate matter concentration spectrum is processed based on the second target monitoring point determination model to obtain multiple second target monitoring points.

8. The monitoring system for measuring particulate matter using near-infrared laser as described in claim 7, characterized in that, The second target monitoring point determination model is a graph neural network model.

9. An electronic device, characterized in that, include: processor; Memory; And a computer program; wherein the computer program is stored in the memory and configured to be executed by the processor to implement the monitoring method for measuring particulate matter using near-infrared laser as described in any one of claims 1 to 4.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the monitoring method for measuring particulate matter using near-infrared laser as described in any one of claims 1 to 4.

Images