Particle detection system and method for preventing and controlling atmospheric pollution

Abstract

The invention provides a particle detection system and method for air pollution prevention and control. The method comprises the following steps: firstly, dividing particle concentration data into a plurality of modal particle data segments; determining the sampling support degree between the sensors of the same type according to the self-adaptive confidence distance between the sensors in each modal particle data segment, and performing concentration fusion on each modal particle data segment based on the sampling support degree between all the sensors of the same type to obtain a plurality of local fusion particle concentrations; determining the measurement weight proportion of each type of sensor according to the fluctuation characteristics of the environmental parameters of the sensor and the particle concentration data collected by the multi-source heterogeneous sensor; and determining the global fusion particle concentration according to the local fusion particle concentration and the measurement weight proportion of each type of sensor, and taking the global fusion particle concentration as a particle detection result of the current environment of the target area. According to the scheme, layered fusion detection of the atmospheric pollutant particle concentration can be realized based on multi-source heterogeneous data, so that the technical accuracy of an atmospheric pollution monitoring result is improved.

Description

Technical Field

[0001] This application relates to the field of pollution detection technology, and more specifically, to a particulate detection system and method for air pollution control. Background Technology

[0002] High-precision particulate matter detection is of great significance for formulating scientific pollution prevention and control policies, evaluating the effectiveness of air quality improvement, and preventing respiratory diseases. However, air pollution has obvious spatiotemporal dynamic characteristics, and a single sensor or isolated monitoring point is difficult to comprehensively and accurately reflect the particulate matter distribution in the target area. Therefore, the development of detection technology that can cover a wide area and reflect changes in particulate matter concentration in real time has become an important requirement for environmental monitoring.

[0003] However, existing particle concentration detection methods mainly rely on single-source or homogeneous sensors for data acquisition and processing. These methods are easily affected by individual sensor differences, sudden anomalies, and environmental factors, leading to biases or localized distortions in the monitoring results. For example, different sensors may produce significant concentration deviations under the same environment due to differences in measurement principles, ranges, or stability. Traditional simple averaging or weighted averaging fusion methods cannot adequately eliminate these deviations, making it difficult to accurately reflect the overall particle concentration of a region. Furthermore, they lack systematic correction for abnormal data and environmental adaptability processing. Therefore, how to achieve hierarchical fusion detection of atmospheric pollutant particle concentrations based on multi-source heterogeneous data, thereby improving the technical accuracy of atmospheric pollution monitoring results, has become a challenge for the industry. Summary of the Invention

[0004] This application provides a particle detection system and method for air pollution control, which can realize the hierarchical fusion detection of atmospheric pollutant particle concentration based on multi-source heterogeneous data, thereby improving the technical accuracy of air pollution monitoring results.

[0005] In a first aspect, this application provides a particulate detection method for air pollution control, comprising the following steps:

[0006] Particulate concentration data of air pollutants in the target area are collected using multi-source heterogeneous sensors;

[0007] Spatiotemporal feature constraints are applied to the particle concentration data to remove short-term abrupt changes or isolated outliers, resulting in particle concentration constraint data. The particle concentration constraint data is then divided into multiple modal particle data segments according to the sensor type.

[0008] The sampling support between sensors of the same type is determined by the adaptive confidence distance between sensors within each modal particle data segment. Then, the concentration of each modal particle data segment is fused based on the sampling support between all sensors of the same type to obtain multiple local fused particle concentrations.

[0009] The environmental parameters of the environment in which the multi-source heterogeneous sensor is located are obtained, and the measurement weight ratio of each type of sensor is determined by the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor.

[0010] The global fusion particle concentration is determined based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then the global fusion particle concentration is used as the particle detection result of the current environment of the target area.

[0011] In some embodiments, subjecting the particle concentration data to spatiotemporal feature constraints and removing short-term abrupt changes or isolated outliers to obtain particle concentration constrained data specifically includes:

[0012] Collect particle concentration data for the target area within a preset time period;

[0013] Based on the temporal continuity of the particle concentration data, instantaneous abrupt change points are detected, and the data corresponding to the abrupt change points are marked as abnormal data.

[0014] By combining the measurement results of spatially nearby sensors, isolated outliers are identified and the abnormal data is removed to obtain particle concentration constraint data.

[0015] In some embodiments, dividing the particle concentration constraint data into multiple modal particle data segments according to sensor type specifically includes:

[0016] Based on the type information of the multi-source heterogeneous sensors, the particle concentration constraint data are grouped by type;

[0017] Particle concentration constraint data corresponding to the same type of sensor are combined into independent datasets, and each dataset is assigned a unique identifier for subsequent processing and storage.

[0018] Each type of dataset is used as a corresponding modal granular data segment.

[0019] In some embodiments, determining the sampling support among sensors of the same type by using the adaptive confidence distance between sensors within each modal particle data segment specifically includes:

[0020] Extract the differences in particle concentration measurements from similar sensors over the same time period;

[0021] The adaptive confidence distance between sensors is determined based on measurement differences and the historical stability of each sensor.

[0022] The sampling consistency between sensors is determined based on the adaptive confidence distance, thereby determining the sampling support between sensors of the same type.

[0023] In some embodiments, concentration fusion of each modal particle data segment is performed based on the sampling support among all sensors of the same type to obtain multiple local fused particle concentrations, specifically including:

[0024] In each modal particle data segment, fusion weights are assigned based on the sampling support between sensors;

[0025] Weighted fusion operations are used to fuse sensor data within the same modality;

[0026] A residual detection step is introduced during the fusion process to constrain and correct abnormal residuals;

[0027] Output the local fusion particle concentration corresponding to each modality particle data segment.

[0028] In some embodiments, determining the measurement weighting of each type of sensor based on the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensors specifically includes:

[0029] Extract the temporal fluctuation characteristics of environmental parameters within the target area;

[0030] Based on the aforementioned temporal fluctuation characteristics, a factor is determined to assess the degree of influence of environmental fluctuations on the measurement results of various types of sensors.

[0031] The measurement weights of each type of sensor are determined based on various influence factors.

[0032] In some embodiments, determining the global fusion particle concentration based on the local fusion particle concentration and the measurement weighting of each type of sensor specifically includes:

[0033] The local particle concentration is weighted and combined according to the measurement weights of each type of sensor;

[0034] A cross-modal correction mechanism is introduced in the weighted combination process to adjust the initial weighting results based on the correlation between particle concentrations of different modes;

[0035] Global consistency constraints are established based on environmental parameters, and the adjusted weighted results are checked and corrected according to the global consistency constraints.

[0036] The global fusion particle concentration is determined by a weighted combination of cross-modal correction and global consistency constraints.

[0037] Secondly, this application provides a particulate detection system for air pollution control, comprising:

[0038] The data acquisition module is used to collect particulate concentration data of air pollutants in the target area through multi-source heterogeneous sensors;

[0039] The feature processing module is used to perform spatiotemporal feature constraints on the particle concentration data, remove short-term abrupt changes or isolated outliers, obtain particle concentration constraint data, and then divide the particle concentration constraint data into multiple modal particle data segments according to the sensor type.

[0040] The feature processing module is also used to determine the sampling support between sensors of the same type through the adaptive confidence distance between sensors in each modal particle data segment, and then perform concentration fusion on each modal particle data segment based on the sampling support between all sensors of the same type to obtain multiple local fused particle concentrations.

[0041] The feature processing module is also used to acquire environmental parameters of the environment in which the multi-source heterogeneous sensor is located, and to determine the measurement weight ratio of each type of sensor by means of the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor.

[0042] The fusion module is used to determine the global fusion particle concentration based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then use the global fusion particle concentration as the particle detection result of the current environment of the target area.

[0043] Thirdly, this application provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described particulate detection method for air pollution prevention and control.

[0044] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described particulate detection method for air pollution control.

[0045] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects:

[0046] In this embodiment, particulate concentration data of atmospheric pollutants in the target area are collected using multi-source heterogeneous sensors. The particulate concentration data is constrained by spatiotemporal characteristics to remove short-term abrupt changes or isolated outliers, resulting in constrained particulate concentration data. This constrained data is then divided into multiple modal particulate data segments according to sensor type. The sampling support between sensors of the same type is determined using adaptive confidence distance within each modal particulate data segment. Concentration fusion is then performed on each modal particulate data segment based on the sampling support between all sensors of the same type, resulting in multiple locally fused particulate concentrations. Environmental parameters of the environment where the multi-source heterogeneous sensors are located are acquired. The measurement weights of each type of sensor are determined based on the fluctuation characteristics of these environmental parameters and the particulate concentration data collected by the multi-source heterogeneous sensors. The global fused particulate concentration is determined based on the locally fused particulate concentration and the measurement weights of each type of sensor. This global fused particulate concentration is then used as the particulate detection result for the current environment of the target area.

[0047] Therefore, this application determines the global fusion particle concentration based on the local fusion particle concentration and the measurement weighting of each type of sensor, and uses the global fusion particle concentration as the particle detection result of the current environment in the target area. First, by constraining the original particle concentration data with spatiotemporal features, short-term mutations and isolated outliers can be effectively eliminated, making the data processed later more stable and reliable, reducing the interference of local anomalies on the overall detection result, and thus improving the quality of basic data. Second, the constrained data is divided into multiple modal particle data segments according to sensor type, and the sampling support between sensors of the same type is calculated through adaptive confidence distance within each modality. This quantifies the measurement consistency and historical stability between sensors, realizes weighted fusion of similar data, and improves the accuracy and reliability of local concentration estimation. On this basis, the local fusion result is compared with environmental parameters. By combining the fluctuation characteristics, the measurement weights of different types of sensors are determined, enabling reasonable weighting of different sensors in the global fusion process based on environmental adaptability and measurement reliability. This ensures the sensitivity and responsiveness of the fusion results to actual environmental changes. Finally, the local fused particle concentration is fused globally according to the measurement weights, while introducing cross-modal correction and global consistency constraints. This corrects the correlation between different modes and environmental consistency, eliminates local biases, and achieves accurate characterization of regional particle concentration. In summary, this application's scheme, through multi-level and multi-dimensional data processing and fusion strategies, not only improves the accuracy and reliability of particle concentration measurement but also enhances the system's adaptability to abnormal data and environmental fluctuations. It achieves hierarchical fusion detection of atmospheric pollutant particle concentration, thereby significantly improving the technical accuracy of atmospheric pollution monitoring results. Attached Figure Description

[0048] Figure 1This is an exemplary flowchart of a particulate detection method for air pollution control according to some embodiments of this application;

[0049] Figure 2 This is a schematic flowchart illustrating the process of determining sampling support according to some embodiments of this application;

[0050] Figure 3 This is a schematic flowchart illustrating the process of determining the weighting of measurement values ​​according to some embodiments of this application;

[0051] Figure 4 This is a schematic diagram of the structure of a particulate detection system for air pollution control, as shown in some embodiments of this application;

[0052] Figure 5 This is a schematic diagram of the structure of a computer device for implementing a particulate detection method for air pollution control, according to some embodiments of this application. Detailed Implementation

[0053] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] refer to Figure 1 The figure is an exemplary flowchart of a particulate detection method for air pollution control according to some embodiments of this application. The particulate detection method for air pollution control mainly includes the following steps:

[0055] In step 101, particulate concentration data of air pollutants in the target area are collected using a multi-source heterogeneous sensor.

[0056] It should be noted that the multi-source heterogeneous sensor in this application refers to multiple types of sensors with different detection principles, sensitivity ranges and measurement characteristics deployed in the same target area. They achieve multi-dimensional coverage and complementary monitoring of atmospheric particulate matter through collaborative acquisition. The particulate concentration data refers to the concentration information of suspended particulate matter in the atmosphere collected by particulate matter sensors under specific time and space conditions.

[0057] In practice, a multi-source heterogeneous sensor array is deployed in the target area to collect particulate concentration data of air pollutants. This multi-source heterogeneous sensor includes a light-scattering particle sensor, a beta-ray absorption particle monitor, a laser particle size analyzer, and a gravimetric sampler. The light-scattering particle sensor is used to monitor the mass concentrations of PM2.5 and PM10 in real time; the beta-ray absorption monitor provides high-precision long-term reference values; the laser particle size analyzer acquires particle size distribution information for fine and ultrafine particles; and the gravimetric sampler provides reference-level calibration data. In specific applications, the multi-source heterogeneous sensor array can be deployed in a fixed, mobile, or portable manner to achieve spatial coverage of the target area. During the data acquisition process, the system uses a unified data acquisition module to synchronize the sampling frequency and timestamps of different types of sensors and uses a network time protocol for clock calibration to ensure the consistency of the data's time dimension. The particulate concentration data is transmitted to the data center via wired or wireless communication networks. Simultaneously, data integrity checks are performed at the acquisition end, marking and caching missing data, data with abnormal formats, or data exceeding the measurement range for subsequent processing.

[0058] In step 102, the particle concentration data is subjected to spatiotemporal feature constraints to remove short-term abrupt changes or isolated outliers, resulting in particle concentration constraint data. Then, the particle concentration constraint data is divided into multiple modal particle data segments according to the sensor type.

[0059] In some embodiments, subjecting the particle concentration data to spatiotemporal feature constraints and removing short-term abrupt changes or isolated outliers to obtain particle concentration constrained data specifically includes:

[0060] Collect particle concentration data for the target area within a preset time period;

[0061] Based on the temporal continuity of the particle concentration data, instantaneous abrupt change points are detected, and the data corresponding to the abrupt change points are marked as abnormal data.

[0062] By combining the measurement results of spatially nearby sensors, isolated outliers are identified and the abnormal data is removed to obtain particle concentration constraint data.

[0063] It should be noted that the spatiotemporal feature constraints in this application refer to the joint screening and constraints on the collected data based on temporal continuity and spatial proximity, in order to ensure the stability and consistency of particle concentration data.

[0064] In practice, the original sampling data of each sensor within the target area within a preset time period are first resampled and aligned using a unified time reference (e.g., using a step size of 1 minute or 10 seconds). Missing samples are marked, and a short-time linear / neighbor-filling strategy is used to ensure the availability of continuous windows. Then, a short-window denoising and mutation detection process is applied to each sensor sequence in the time dimension. Specifically, this includes first removing impulse noise using a median filter or moving median, and then calculating the first-order difference sequence within a sliding window (commonly with a window length of 3 to 7 time slots) and using the historical median absolute difference (Median Absolute Difference) as the basis. Adaptive thresholds for Deviation (MAD) or historical variance are used to determine transient abrupt changes, or mature breakpoint detection algorithms (such as Bayesian online breakpoint detection) are used in parallel. By setting penalty parameters based on information criteria, the location of change points is determined. Data identified by breakpoint detection and not supported by subsequent spatial evidence are marked as temporal anomalies. Spatially, a neighborhood set is established for each sensor (using nearest neighbor k of 3-6 or radius search). At each time point, a neighborhood reference value is calculated (preferably using weighted median or inverse distance weighted mean to reduce anomaly sensitivity). The normalized residual of the sensor reading relative to the neighborhood reference is calculated. Normalization can use the neighborhood MAD or the sensor's own historical standard deviation as the denominator. When the residual exceeds a set threshold (e.g., three times the MAD or an empirical multiple) and is consistent with other sensors in the neighborhood, the reading is marked as a breakpoint. Isolated spatial anomalies are identified. The temporal and spatial judgment results are fused to construct an anomaly confidence score for each moment (the temporal anomaly flag and the spatial anomaly flag are synthesized according to preset weights or logical voting). Points with scores exceeding the threshold are removed or cleared. For short-term isolated missing data (i.e., below the preset duration threshold), neighborhood-weighted interpolation or local time-series interpolation can be used for recovery. Long-term missing data is retained as missing data and its metadata is recorded. The entire process can be implemented using a combination of readily available tools, such as using Pandas for resampling and interpolation, using ruptures or similarity libraries for breakpoint detection, using scikit-learn's neighborhood retrieval for spatial neighborhood construction, and using statistical functions to calculate MAD / IQR and manage thresholds. The detection identifier, threshold, and uncertainty information for each step are saved in the pipeline for subsequent traceability and parameter adjustment.

[0065] In some embodiments, dividing the particle concentration constraint data into multiple modal particle data segments according to sensor type specifically includes:

[0066] Based on the type information of the multi-source heterogeneous sensors, the particle concentration constraint data are grouped by type;

[0067] Particle concentration constraint data corresponding to the same type of sensor are combined into independent datasets, and each dataset is assigned a unique identifier for subsequent processing and storage.

[0068] Each type of dataset is used as a corresponding modal granular data segment.

[0069] It should be noted that the modal particle data segment in this application refers to an independent set of particle concentration data grouped by sensor type.

[0070] In practice, firstly, based on the model or type information of the multi-source heterogeneous sensors, the collected particle concentration constraint data are grouped by type. This grouping can be achieved by reading sensor attribute tables or data tag information to ensure accurate classification of data from sensors of the same type. Secondly, particle concentration constraint data corresponding to the same type of sensor are combined into independent datasets, and each dataset is assigned a unique identifier to facilitate subsequent data processing, storage, and retrieval. This step can be implemented using conventional data structures or database tables to ensure the reliability and operability of data management. Finally, each type of dataset is used as a corresponding modal particle data segment for subsequent intramodal concentration fusion and cross-modal analysis. Through these steps, hierarchical data management based on sensor type can be achieved, providing structured and reliable input data for subsequent particle concentration fusion and environmental monitoring.

[0071] In step 103, the sampling support between sensors of the same type is determined by the adaptive confidence distance between sensors within each modal particle data segment. Then, based on the sampling support between all sensors of the same type, the concentration of each modal particle data segment is fused to obtain multiple local fused particle concentrations.

[0072] In some embodiments, reference Figure 2 As shown in the figure, this is a flowchart illustrating the process of determining sampling support in some embodiments of this application. In this embodiment, determining the sampling support between sensors of the same type through the adaptive confidence distance between sensors within each modal particle data segment specifically includes:

[0073] In step 1031, the differences in particle concentration measurements from the same type of sensor within the same time period are extracted;

[0074] In step 1032, the adaptive confidence distance between sensors is determined based on the measurement differences and the historical stability of each sensor;

[0075] In step 1033, the sampling consistency between sensors is determined based on the adaptive confidence distance, thereby determining the sampling support between sensors of the same type.

[0076] It should be noted that the particle concentration measurement difference in this application refers to the difference or deviation between the measured values ​​of particulate matter concentration in the target area atmosphere by the same type of sensor within the same time period; the adaptive confidence distance in this application is a comprehensive consistency characteristic index that measures the real-time measurement consistency and historical stability of the same type of sensor; and the sampling support in this application is a characteristic index that measures the reliability and credibility of data from the same type of sensor during the fusion process.

[0077] It should also be noted that this application's solution, by introducing particle concentration measurement differences and sensor historical stability to calculate adaptive confidence distance, solves the problems of difficulty in quantifying sampling consistency and the impact of outlier data on fusion results in multi-source sensor data fusion compared to existing technologies. The adaptive confidence distance dynamically reflects the real-time consistency and long-term stability between sensors, and then maps it to sampling support, achieving precise quantification of the reliability of each sensor's data in the fusion process. The technical advantage of this method lies in its ability to reasonably allocate weights when performing weighted fusion of local modal data, significantly improving the accuracy and stability of local fused particle concentration, reducing outlier interference, and enhancing the detection reliability and data credibility of the entire particle detection system in complex environments.

[0078] In practical implementation, it is necessary to first align the data collected by sensors of the same type. During the implementation process, sensors of the same type can be selected based on factors such as the sensor's measurement principle, range, and accuracy. For example, only optical scattering PM2.5 sensors should be retained, ensuring that the range and accuracy are consistent. Devices with different measurement principles or incompatible parameters should be excluded to ensure the consistency of the benchmark for subsequent difference analysis. Then, a time window synchronization method is used to align the sampled data, setting an analysis time window, such as five consecutive sampling points. By using the sensor's built-in timestamp or network time synchronization technology, it is ensured that each sampling point corresponds to the same time, resulting in a structured dataset. Further, the particle concentration measurement difference is extracted. In the implementation process, outliers can be removed from single sensor data first. The mean and standard deviation of the concentration within the time window are calculated using the three-standard-deviation criterion. Sampling points exceeding the mean plus or minus three standard deviations are identified as outliers and filled using linear interpolation to ensure data continuity. Then, all sensors of the same type are paired up, and the absolute difference in concentration of each pair of sensors at the same time point is calculated. The average difference value within the time window is then calculated to quantify the real-time measurement between sensors. The differences are then identified. In calculating the adaptive confidence distance, concentration data from the same time window of the target sensor over a past period are extracted. The standard deviation of each window is calculated, and the historical average standard deviation is used to measure sensor stability and quantify it as a stability coefficient. The adaptive confidence distance is calculated by weighting real-time measurement differences and historical stability. Real-time differences are normalized and multiplied by the difference weight, and stability differences are multiplied by the stability weight. The sum of these two values ​​yields the adaptive confidence distance. A smaller value indicates higher overall consistency between the two sensors in real-time measurement and historical stability. Finally, in determining the sampling support based on the adaptive confidence distance, a percentile threshold method can be used to determine the confidence distance threshold, representing the upper limit of distance pairing for most normal sensors. The adaptive confidence distance is then mapped to sampling support: when the distance is less than or equal to the threshold, support is calculated through linear mapping, with smaller distances resulting in higher support; when the distance exceeds the threshold, it is set as the minimum effective support to ensure numerical rationality. Finally, the average support of each sensor with other sensors of the same type is taken to obtain the sampling support between each sensor and other sensors of the same type.

[0079] In some embodiments, concentration fusion of each modal particle data segment is performed based on the sampling support among all sensors of the same type to obtain multiple local fused particle concentrations, specifically including:

[0080] In each modal particle data segment, fusion weights are assigned based on the sampling support between sensors;

[0081] Weighted fusion operations are used to fuse sensor data within the same modality;

[0082] A residual detection step is introduced during the fusion process to constrain and correct abnormal residuals;

[0083] Output the local fusion particle concentration corresponding to each modality particle data segment.

[0084] It should be noted that the fusion weight in this application refers to an indicator used to measure the proportion of contribution of each sensor's data to the fusion result; it should also be noted that by introducing a residual detection step in the fusion process, the abnormal residuals in this application can be constrained and corrected, which can effectively suppress the interference of abnormal data from a single sensor on the fusion result, improve the accuracy and stability of the local fusion particle concentration, and enhance the reliability and data credibility of the entire particle detection system in complex environments.

[0085] In specific implementation, firstly, for each modal particle data segment, the sampling support data of sensors of the same type are acquired. The sampling support of each sensor is normalized so that the sum of the weights equals 1. The normalization method can be linear normalization, which divides the sampling support of each sensor by the sum of the sampling support of all sensors in the same modality. In this way, sensors with high reliability occupy a larger weight in subsequent fusion calculations, ensuring that the fusion result is sensitive to high-confidence data and suppresses low-confidence data, thereby improving the impact of abnormal fluctuations of a single sensor on the overall result. Secondly, the particle concentration data of all sensors in each modal data segment are weighted and summed. Specifically, for each sensor measurement value at the same time point, the corresponding normalization weight is multiplied and then accumulated to obtain the fusion concentration value at that time point. This operation is repeated for the entire time window to obtain the time series fusion concentration. This can be implemented using the weighted average method or the weighted least squares method. Among them, the weighted least squares method is susceptible to noise or slight variations in the data between sensors. For minor deviations, errors can be further reduced to improve fusion accuracy. Then, after the initial fusion results are generated, the residuals are calculated between the fused concentration at each time point and the measured values ​​of each sensor, i.e., residual = fused concentration - single sensor measured value. Using residual statistical methods, such as the three-standard-deviation principle, abnormal residuals are identified. For residuals exceeding the set threshold, linear correction or filtering methods can be used for constraint, such as Kalman filtering or weighted correction, to minimize the impact of abnormal data on the fusion results, thereby ensuring the stability and robustness of the fusion results. Finally, after completing the residual constraint correction, the fusion results of each modal particle data segment are output as local fused particle concentrations, including time series concentration values ​​and their corresponding confidence information. Each modality is output independently, which is convenient for subsequent cross-modal global fusion or further analysis. This method can effectively reduce the interference of single sensor outliers on the fusion results, improve the accuracy, reliability and stability of local concentration fusion, and meet the actual needs of atmospheric particle detection in complex environments.

[0086] In step 104, environmental parameters of the environment in which the multi-source heterogeneous sensor is located are obtained, and the measurement weight ratio of each type of sensor is determined by the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor.

[0087] It should be noted that obtaining environmental parameters of the environment in which the multi-source heterogeneous sensor is located refers to collecting external environmental information such as temperature, humidity, wind speed, and air pressure in the area where the sensor is located, which affect particulate matter measurement.

[0088] In some embodiments, reference Figure 3 As shown in the figure, this is a flowchart illustrating the process of determining the measurement weighting in some embodiments of this application. In this embodiment, determining the measurement weighting of each type of sensor based on the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensors specifically includes:

[0089] In step 1041, the temporal fluctuation characteristics of environmental parameters within the target area are extracted;

[0090] In step 1042, the influence factor of environmental fluctuations on the measurement results of various types of sensors is determined based on the time-series fluctuation characteristics;

[0091] In step 1043, the measurement weights of each type of sensor are determined based on each influence factor.

[0092] It should be noted that the temporal fluctuation characteristics in this application are indicators used to describe the magnitude and frequency characteristics of environmental parameters changing over time; the influence degree factor in this application is an indicator used to quantify the magnitude of the impact of environmental fluctuations on sensor measurement results; and the measurement weight ratio in this application is an indicator used to characterize the contribution ratio of each type of sensor in the fusion result.

[0093] In practical implementation, firstly, when extracting the fluctuation characteristics of environmental parameters, the mean and standard deviation of each environmental parameter are calculated according to a preset time window. The ratio of the standard deviation to the mean is used as the environmental volatility index to quantify the temporal fluctuation amplitude of the environmental parameters. Simultaneously, the long-term trend of the environmental parameters can be smoothed using moving average or exponentially weighted average methods to obtain stable temporal characteristics. Secondly, the environmental volatility is multiplied by the response coefficient corresponding to the sensor type to obtain a single-parameter influence factor. Then, all environmental parameter influence factors are weighted or accumulated to obtain a single environmental influence factor E_i. The value of E_i ranges from 0 to 1; a larger E_i value indicates a more significant impact of environmental fluctuations on that type of sensor. It should be further noted that the response coefficient refers to... The index for quantifying the sensitivity of a specific type of sensor to measurement errors under changing environmental parameters can be determined through experimental calibration or regression analysis of historical data. Then, based on the environmental impact factor E_i, the measurement weight W_i of each type of sensor is further determined. The calculation method is to define the weight of each sensor type as W_i=(1-E_i) / Σ(1-E_j), where Σ(1-E_j) is the sum of the environmental stability coefficients of all sensor types. Through this normalization method, it can be ensured that the sum of the weights of all sensor types is 1, so that sensors with high environmental sensitivity automatically have their weights reduced when environmental fluctuations are large, while sensors with high environmental stability receive higher weights, thereby improving the accuracy and stability of the fusion results in the subsequent global fusion process.

[0094] It should be noted that the proposed solution extracts the fluctuation characteristics of environmental parameters and calculates the environmental impact factor of each type of sensor, thereby achieving dynamic quantification of the measurement contribution of different sensors in complex environments. This solves the problems of uneven measurement accuracy of multi-source heterogeneous sensors under environmental fluctuations and the easy introduction of errors in direct fusion in the prior art. By determining the measurement weight ratio based on the impact factor, the weight of highly environmentally sensitive sensors can be automatically reduced and the weight of highly stable sensors can be increased during the fusion process. This significantly improves the accuracy and stability of the particle concentration fusion across the entire domain, enhances the reliability and data credibility of the system in variable environments, and achieves accurate, dynamic, and environmentally adaptive sensor data fusion.

[0095] In step 105, the global fusion particle concentration is determined based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then the global fusion particle concentration is used as the particle detection result of the current environment of the target area.

[0096] In some embodiments, determining the global fusion particle concentration based on the local fusion particle concentration and the measurement weighting of each type of sensor specifically includes:

[0097] The local particle concentration is weighted and combined according to the measurement weights of each type of sensor;

[0098] A cross-modal correction mechanism is introduced in the weighted combination process to adjust the initial weighting results based on the correlation between particle concentrations of different modes;

[0099] Global consistency constraints are established based on environmental parameters, and the adjusted weighted results are checked and corrected according to the global consistency constraints.

[0100] The global fusion particle concentration is determined by a weighted combination of cross-modal correction and global consistency constraints.

[0101] It should be noted that the global fusion particle concentration in this application is a comprehensive measurement indicator used to reflect the overall particulate pollution level of the target area; the cross-modal correction mechanism in this application refers to a processing method used to adjust the deviation between particle concentration data of different modes, so as to make the multimodal fusion results consistent; the global consistency constraint condition in this application refers to the rule of reasonably limiting the fusion results based on environmental parameters, and its function is to ensure that the global fusion particle concentration is consistent with the actual environmental conditions and improve the reliability of the fusion results.

[0102] It should also be noted that this scheme achieves high-precision determination of global particle concentration by combining the local fusion particle concentration with the measurement weights of various types of sensors, and introducing cross-modal correction and global consistency constraints. Compared with existing technologies, this solves the technical problems of coarse multimodal data processing, insufficient environmental adaptability, and the influence of anomalous modes on the fusion results in traditional fusion methods. Specifically, by using weighted combination to fully utilize the reliability and stability of various types of sensors, a reasonable allocation of contributions from different sensors is achieved; the cross-modal correction mechanism adjusts the preliminary weighted results based on intermodal correlation, effectively eliminating the interference of single-modal anomalies or deviations on the global concentration; the global consistency constraint uses environmental parameters to verify and correct the adjusted fusion results, ensuring that the global concentration is highly consistent with actual environmental conditions, thereby guaranteeing the reliability and accuracy of the fusion results. The overall technical effect is reflected in enhancing the accuracy, robustness, and environmental adaptability of multi-source heterogeneous sensor data fusion, improving the real-time performance and reliability of particle concentration detection, and meeting the technical requirements for accurate monitoring and data reliability in air pollution prevention and control.

[0103] In specific implementation, firstly, the local fusion particle concentration obtained from each modal particle data segment is used as the basic input. For each modal concentration C_m, a weighted processing is performed by combining the measurement weight ratio W_m of its corresponding sensor type: the preliminary global fusion concentration C_global_init=Σ(W_m×C_m) is calculated, where Σ represents the summation over all modes, and normalization is used to ensure that ΣW_m=1, so as to ensure that the contribution ratio of different types of sensors in the fusion result is consistent with its reliability and stability. This weighting method uses a well-known weighted average algorithm to ensure that the fusion result reflects the comprehensive information of each modality. Secondly, cross-modal correction is performed on the preliminary weighted result to correct the deviation between different modes. Specifically, the intermodal correlation index, such as Pearson correlation coefficient or Spearman rank correlation coefficient, is calculated for each modal concentration sequence; modes that deviate significantly from the overall trend are identified; and linear regression or least squares method is used to correct the deviating modes so that their adjusted concentrations are... The degree value is kept consistent with the relevant modes. This step ensures the synergy of multimodal data and reduces the impact of single-mode anomalies on global fusion. Then, global consistency constraints are established based on environmental parameters (temperature, humidity, air pressure, wind speed, etc.). Specifically, the mean and volatility of environmental parameters are extracted first, and the allowable fluctuation range of the global fusion concentration is defined. For example, the allowable concentration fluctuation does not exceed the threshold corresponding to the environmental volatility. The weighted result C_global_adj after cross-modal correction is checked. If the concentration value exceeds the allowable range, it is corrected to a reasonable range by truncation or weighted adjustment so that the final fusion concentration is consistent with the environmental conditions. Finally, through the above weighted combination, cross-modal correction and global consistency constraint fusion processing, the final global fusion particle concentration C_global_final is obtained. C_global_final considers the measurement contribution of each mode, as well as environmental adaptability and intermodal correlation, to achieve high accuracy and stability of global fusion.

[0104] It should be noted that using the fused particle concentration as the particle detection result of the target area's current environment means using this fused concentration value as a real-time assessment indicator of the area's particulate pollution level.

[0105] On the other hand, in some embodiments, this application provides a particulate detection system for air pollution control, with reference to... Figure 4 The figure is a schematic diagram of a particulate detection system for air pollution control according to some embodiments of this application. The particulate detection system for air pollution control includes: a data acquisition module 401, a feature processing module 402, and a fusion module 403, which are described below:

[0106] The acquisition module 401 is used to acquire particulate concentration data of air pollutants in the target area through multi-source heterogeneous sensors;

[0107] Feature processing module 402 is used to perform spatiotemporal feature constraints on the particle concentration data, remove short-term abrupt changes or isolated outliers, obtain particle concentration constraint data, and then divide the particle concentration constraint data into multiple modal particle data segments according to the sensor type.

[0108] The feature processing module 402 is further configured to determine the sampling support between sensors of the same type through the adaptive confidence distance between sensors in each modal particle data segment, and then perform concentration fusion on each modal particle data segment based on the sampling support between all sensors of the same type to obtain multiple local fused particle concentrations.

[0109] The feature processing module 402 is also used to acquire environmental parameters of the environment in which the multi-source heterogeneous sensor is located, and to determine the measurement weight ratio of each type of sensor by means of the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor.

[0110] The fusion module 403 is used to determine the global fusion particle concentration based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then use the global fusion particle concentration as the particle detection result of the current environment of the target area.

[0111] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described particulate detection method for air pollution prevention and control.

[0112] In some embodiments, reference Figure 5 The figure is a schematic diagram of the structure of a computer device for implementing a particulate detection method for air pollution control, according to some embodiments of this application. The particulate detection method for air pollution control in the above embodiments can be implemented through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0113] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0114] The communication bus 502 can be used to transmit information between the aforementioned components.

[0115] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0116] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. The particulate detection method for air pollution control in the above embodiments can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0117] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0118] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single-CPU) processor or a multi-core (multi-CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0119] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0120] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described particulate detection method for air pollution control.

[0121] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0122] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A particulate matter detection method for air pollution control, characterized in that, Includes the following steps: Particulate concentration data of air pollutants in the target area are collected using multi-source heterogeneous sensors; Spatiotemporal feature constraints are applied to the particle concentration data to remove short-term abrupt changes or isolated outliers, resulting in particle concentration constraint data. The particle concentration constraint data is then divided into multiple modal particle data segments according to the sensor type. The sampling support between sensors of the same type is determined by the adaptive confidence distance between sensors within each modal particle data segment. Then, the concentration of each modal particle data segment is fused based on the sampling support between all sensors of the same type to obtain multiple local fused particle concentrations. The environmental parameters of the environment in which the multi-source heterogeneous sensor is located are obtained, and the measurement weight ratio of each type of sensor is determined by the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor. The global fusion particle concentration is determined based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then the global fusion particle concentration is used as the particle detection result of the current environment of the target area.

2. The method as described in claim 1, characterized in that, The particle concentration data is subjected to spatiotemporal feature constraints to remove short-term abrupt changes or isolated outliers, resulting in particle concentration constraint data, specifically including: Collect particle concentration data for the target area within a preset time period; Based on the temporal continuity of the particle concentration data, instantaneous abrupt change points are detected, and the data corresponding to the abrupt change points are marked as abnormal data. By combining the measurement results of spatially nearby sensors, isolated outliers are identified and the abnormal data is removed to obtain particle concentration constraint data.

3. The method as described in claim 1, characterized in that, The particle concentration constraint data is divided into multiple modal particle data segments according to the sensor type, specifically including: Based on the type information of the multi-source heterogeneous sensors, the particle concentration constraint data are grouped by type; Particle concentration constraint data corresponding to the same type of sensor are combined into independent datasets, and each dataset is assigned a unique identifier for subsequent processing and storage. Each type of dataset is used as a corresponding modal granular data segment.

4. The method as described in claim 1, characterized in that, Determining the sampling support among sensors of the same type by using adaptive confidence distance between sensors within each modal particle data segment specifically includes: Extract the differences in particle concentration measurements from similar sensors over the same time period; The adaptive confidence distance between sensors is determined based on measurement differences and the historical stability of each sensor. The sampling consistency between sensors is determined based on the adaptive confidence distance, thereby determining the sampling support between sensors of the same type.

5. The method as described in claim 1, characterized in that, Based on the sampling support among all sensors of the same type, the concentration of each modal particle data segment is fused to obtain multiple local fused particle concentrations, specifically including: In each modal particle data segment, fusion weights are assigned based on the sampling support between sensors; Weighted fusion operations are used to fuse sensor data within the same modality; A residual detection step is introduced during the fusion process to constrain and correct abnormal residuals; Output the local fusion particle concentration corresponding to each modality particle data segment.

6. The method as described in claim 1, characterized in that, The determination of the measurement weighting of each type of sensor based on the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensors specifically includes: Extract the temporal fluctuation characteristics of environmental parameters within the target area; Based on the aforementioned temporal fluctuation characteristics, a factor is determined to assess the degree of influence of environmental fluctuations on the measurement results of various types of sensors. The measurement weights of each type of sensor are determined based on various influence factors.

7. The method as described in claim 1, characterized in that, The determination of the global fusion particle concentration based on the local fusion particle concentration and the measurement weighting of each type of sensor specifically includes: The local particle concentration is weighted and combined according to the measurement weights of each type of sensor; A cross-modal correction mechanism is introduced in the weighted combination process to adjust the initial weighting results based on the correlation between particle concentrations of different modes; Global consistency constraints are established based on environmental parameters, and the adjusted weighted results are checked and corrected according to the global consistency constraints. The global fusion particle concentration is determined by a weighted combination of cross-modal correction and global consistency constraints.

8. A particulate matter detection system for air pollution control, characterized in that, include: The data acquisition module is used to collect particulate concentration data of air pollutants in the target area through multi-source heterogeneous sensors; The feature processing module is used to perform spatiotemporal feature constraints on the particle concentration data, remove short-term abrupt changes or isolated outliers, obtain particle concentration constraint data, and then divide the particle concentration constraint data into multiple modal particle data segments according to the sensor type. The feature processing module is also used to determine the sampling support between sensors of the same type through the adaptive confidence distance between sensors in each modal particle data segment, and then perform concentration fusion on each modal particle data segment based on the sampling support between all sensors of the same type to obtain multiple local fused particle concentrations. The feature processing module is also used to acquire environmental parameters of the environment in which the multi-source heterogeneous sensor is located, and to determine the measurement weight ratio of each type of sensor by means of the fluctuation characteristics of the environmental parameters and the particle concentration data collected by the multi-source heterogeneous sensor. The fusion module is used to determine the global fusion particle concentration based on the local fusion particle concentration and the measurement weight ratio of each type of sensor, and then use the global fusion particle concentration as the particle detection result of the current environment of the target area.

9. A computer device comprising a memory and a processor, the memory storing code, characterized in that, The processor is configured to acquire the code and execute the particulate detection method for air pollution control as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the particulate detection method for air pollution control as described in any one of claims 1 to 7.

Images