Air quality monitoring data processing transmission method and device

Abstract

The application discloses an air quality monitoring data processing transmission method and device, relates to the technical field of data transmission, and comprises the following steps: collecting multi-modal environmental parameters through a distributed sensor array, calculating the pollutant concentration covariance matrix of adjacent monitoring nodes, and determining abnormal monitoring events according to the condition number; adopting a compression sensing algorithm driven by space-time correlation to perform hierarchical compression; calculating a channel comprehensive quality index, combining a deep learning model to predict a channel state, dynamically selecting a transmission protocol combination, and obtaining a multi-protocol data stream; performing space-time interpolation reconstruction, calculating reconstruction hash and writing into a blockchain, and performing integrity verification; updating the weight of the deep learning model, optimizing the compression sensing algorithm, optimizing energy distribution according to the transmission load, and performing iterative optimization of the deep learning model. The application adopts the compression sensing algorithm driven by space-time correlation to perform data compression, thereby reducing the data transmission burden.

Description

A method and apparatus for processing and transmitting air quality monitoring data Technical Field

[0001] This invention relates to the field of data transmission technology, and in particular to a method and apparatus for processing and transmitting air quality monitoring data. Background Technology

[0002] In recent years, environmental monitoring technology has made significant progress, especially in the field of air quality monitoring. Traditional air quality monitoring methods typically rely on single-point monitoring stations, using fixed sensors to collect real-time data on pollutant concentrations in the air. However, these methods often face problems such as limited sensor deployment, insufficient data coverage, and sensor interference from environmental factors. To address these shortcomings, monitoring schemes based on distributed sensor arrays have emerged in recent years. This scheme achieves wide-area coverage and improves the timeliness and accuracy of data by deploying multiple sensor nodes to collect data simultaneously in different areas. Furthermore, by combining Internet of Things (IoT) technology, cloud computing, and big data analytics, air quality monitoring systems have gradually acquired real-time data processing and remote control capabilities, greatly improving the efficiency of air pollutant monitoring and control.

[0003] Despite this, existing air quality monitoring systems still face several challenges. First, effectively processing massive amounts of environmental data when distributed sensor arrays are deployed on a large scale becomes a critical issue. Data transmission between sensor nodes is often affected by channel quality fluctuations and network bandwidth limitations, leading to data loss or delays. Second, mutual interference between sensor nodes, especially when pollutant concentrations fluctuate significantly, makes accurately identifying abnormal monitoring events a major challenge. While existing monitoring data compression algorithms can reduce data transmission burden to some extent, they lack sufficient spatiotemporal correlation analysis, resulting in low compression efficiency and an inability to optimize energy consumption and bandwidth usage while maintaining data quality. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides an air quality monitoring data processing and transmission method to solve the problem of not being able to optimize energy consumption and bandwidth usage while ensuring data quality.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides an air quality monitoring data processing and transmission method, comprising: collecting multimodal environmental parameters through a distributed sensor array; calculating the pollutant concentration covariance matrix of adjacent monitoring nodes; determining abnormal monitoring events based on condition numbers; if an abnormal monitoring event is detected, storing all data; otherwise, performing layered compression using a spatiotemporal correlation-driven compressed sensing algorithm to output monitoring data packets; calculating a channel comprehensive quality index based on the monitoring data packets; predicting channel status using a deep learning model; dynamically selecting a combination of transmission protocols to obtain multi-protocol data streams; uploading the multi-protocol data streams to the cloud; performing spatiotemporal interpolation reconstruction; calculating the reconstruction hash and writing it into the blockchain for integrity verification; updating the weights of the deep learning model; optimizing the compressed sensing algorithm; optimizing energy allocation according to the transmission load; and performing iterative optimization of the deep learning model.

[0008] As a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, the multimodal environmental parameters include PM2.5 concentration, volatile organic compound concentration, temperature and humidity.

[0009] In a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, the steps of collecting multimodal environmental parameters through a distributed sensor array, calculating the pollutant concentration covariance matrix of adjacent monitoring nodes, and determining abnormal monitoring events based on condition numbers are as follows:

[0010] The detection area is divided into grids, and the sensor nodes in each grid form adjacent node groups;

[0011] Multimodal environmental data is collected through sensor nodes, and preprocessed using Kalman filtering and data normalization to construct multimodal data vectors.

[0012] For each adjacent node group, calculate the pollutant concentration covariance matrix based on the multimodal data vector;

[0013] Calculate the condition number of the covariance matrix, set the anomaly detection threshold, and perform anomaly event judgment.

[0014] As a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, if an abnormal monitoring event is detected, full data storage is performed; otherwise, a spatiotemporal correlation-driven compressed sensing algorithm is used for layered compression to output monitoring data packets. The specific steps are as follows:

[0015] For abnormal monitoring events, the original sampled values ​​of the multimodal environmental parameters of all nodes within the current time window are fully stored, along with the node location coordinates, timestamps, covariance matrix, and condition number, and then encoded in binary.

[0016] For non-abnormal detection events, the autocorrelation coefficient and spatial correlation coefficient are calculated based on the constructed time series of the same node's continuous time window to determine the correlation.

[0017] Prioritization was implemented, with PM2.5 concentration and volatile organic compound concentration as the first-level key parameters, and temperature and humidity as the second-level auxiliary parameters.

[0018] A spatiotemporal joint sparse transform is performed on the first layer of data to generate a partial Fourier observation matrix. Compressed measurement values ​​are obtained through linear projection, and monitoring data packets are output.

[0019] The sparse coefficients are recovered in the cloud, and the error is reconstructed based on compressed sensing theory.

[0020] As a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, the steps of calculating the channel comprehensive quality index based on monitoring data packets, predicting the channel state using a deep learning model, dynamically selecting transmission protocol combinations, and obtaining multi-protocol data streams are as follows:

[0021] Obtain physical layer and network layer metrics, and calculate the weighted overall channel quality index;

[0022] Based on the channel integrated quality index, multi-dimensional fused feature vectors are extracted, a bidirectional LSTM network model is constructed, and cross-entropy loss is used for training to obtain the probability distribution of future channel states.

[0023] Based on the channel state probability distribution, the transmission protocol combination is selected and load balanced. The detection data packets are encapsulated into multi-protocol data streams and transmitted in parallel via multi-threading.

[0024] As a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, the steps of uploading multi-protocol data streams to the cloud, performing spatiotemporal interpolation reconstruction, calculating the reconstruction hash and writing it to the blockchain, and performing integrity verification are as follows:

[0025] The system parses multi-protocol data streams in the cloud, verifies transmission results through cyclic redundancy check, and performs spatiotemporal tag alignment.

[0026] Kriging interpolation was used to spatially interpolate the missing data, cubic spline interpolation was used to fill the time gaps, and cross-validation was performed.

[0027] The data hash value is generated using the SHA-256 hash algorithm, packaged into a transaction, written to a private blockchain, and its integrity is verified by comparing the hash values.

[0028] In a preferred embodiment of the air quality monitoring data processing and transmission method of the present invention, the steps of updating the deep learning model weights, optimizing the compressed sensing algorithm, optimizing energy allocation according to the transmission load, and iteratively optimizing the deep learning model are as follows:

[0029] The real-time channel state prediction error is collected, and the weights of the bidirectional LSTM network model are updated using a stochastic gradient descent algorithm with a driving term. At the same time, degradation is detected and rolled back.

[0030] To minimize the weighted sum of compressed sensing reconstruction error and energy consumption, a genetic algorithm is used to search for the optimal observation matrix;

[0031] Construct a node energy consumption model to maximize network throughput and use the Lagrange multiplier method to solve for optimal data allocation.

[0032] In a second aspect, the present invention provides an air quality monitoring data transmission system, including a data acquisition module, a data compression module, a protocol selection module, a block writing module, and a closed-loop feedback module;

[0033] The data acquisition module is used to collect multimodal environmental parameters through a distributed sensor array, calculate the pollutant concentration covariance matrix of adjacent monitoring nodes, and determine abnormal monitoring events based on the condition number.

[0034] The data compression module is used to store the full data if it is an abnormal monitoring event, and otherwise use a spatiotemporal correlation-driven compression sensing algorithm to perform layered compression and output monitoring data packets.

[0035] The protocol selection module is used to calculate the channel comprehensive quality index based on monitoring data packets, predict the channel state by combining a deep learning model, dynamically select the combination of transmission protocols, and obtain multi-protocol data streams.

[0036] The block writing module is used to upload multi-protocol data streams to the cloud, perform spatiotemporal interpolation reconstruction, calculate the reconstruction hash and write it to the blockchain, and perform integrity verification.

[0037] The closed-loop feedback module is used to update the weights of the deep learning model, optimize the compressed sensing algorithm, optimize energy allocation according to the transmission load, and perform iterative optimization of the deep learning model.

[0038] Thirdly, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein when the computer program is executed by the processor, it implements any step of the air quality monitoring data processing and transmission method as described in the first aspect of the present invention.

[0039] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the air quality monitoring data processing and transmission method as described in the first aspect of the present invention.

[0040] The beneficial effects of this invention are as follows: By collecting multimodal environmental parameters through a distributed sensor array and utilizing spatiotemporal grid partitioning and covariance matrix analysis, efficient anomaly event detection and improved data accuracy are achieved. For non-anomaly events, a spatiotemporal correlation-driven compressed sensing algorithm is used for data compression to reduce transmission load, and data transmission efficiency is optimized by dynamically selecting transmission protocol combinations. Data security and reliability are ensured by uploading to the cloud for spatiotemporal interpolation reconstruction and blockchain integrity verification. Continuous updates of the deep learning model and optimization of the compressed sensing algorithm, through closed-loop feedback control, further enhance the method's adaptability and energy efficiency, enabling the entire monitoring unit to operate efficiently and stably in complex network environments. Attached Figure Description

[0041] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0042] Figure 1 is a flowchart of the air quality monitoring data processing and transmission method in Example 1.

[0043] Figure 2 is a schematic diagram of the air quality monitoring data transmission system in Example 1. Detailed Implementation

[0044] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0045] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0046] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0047] Example 1, referring to Figures 1 and 2, is the first embodiment of the present invention. This embodiment provides an air quality monitoring data processing and transmission method, including the following steps:

[0048] S1: Collect multimodal environmental parameters through a distributed sensor array, calculate the pollutant concentration covariance matrix of adjacent monitoring nodes, and determine abnormal monitoring events based on the condition number.

[0049] Specifically, it includes the following steps:

[0050] S1.1: Multimodal environmental parameters include PM2.5 concentration, volatile organic compound (VOCs) concentration, temperature, and humidity.

[0051] It should be noted that the unit for PM2.5 concentration is μg / m³. 3 VOCs concentration is expressed in ppm, temperature in °C, and humidity in %.

[0052] S1.2: Distribute sensor nodes evenly according to geographical location, with each node integrating a PM2.5 sensor, a VOCs sensor, a temperature sensor, and a humidity sensor.

[0053] Specifically, PM2.5 sensors refer to laser scattering sensors, VOCs sensors refer to metal oxide semiconductor sensors, temperature sensors refer to thermistor sensors, and humidity sensors refer to capacitive sensors.

[0054] S1.3: Divide the detection area into grids, and the sensor nodes in each grid form adjacent node groups.

[0055] Specifically, the monitoring area is divided into square grids with fixed side lengths (e.g., 100 meters), and each grid contains a certain number (greater than or equal to 3 groups) of sensor nodes. Data from the nodes is integrated according to time windows (e.g., 5 minutes).

[0056] The nodes in each grid form an "adjacent node group". If the number of sensor nodes in a grid is less than 3 groups, the nodes of adjacent grids are merged until the number of sensor nodes is greater than or equal to 3 groups.

[0057] S1.4: Collect multimodal environmental data through sensor nodes, preprocess it using Kalman filtering and data normalization, and construct a multimodal data vector.

[0058] It should be understood that Kalman filtering is used to remove noise from the collected multimodal environmental data, and normalization is used to eliminate dimensional differences.

[0059] The constructed multimodal data vector includes PM2.5 concentration and VOCs concentration. Temperature and humidity are used as covariates for subsequent anomaly analysis and are not included in the covariance calculation.

[0060] S1.5: For each adjacent node group, calculate the pollutant concentration covariance matrix based on the multimodal data vector.

[0061] Specifically, the expression for calculating the pollutant concentration covariance matrix is ​​as follows:

[0062]

[0063] Where ∑(t) is the 2×2 pollutant concentration covariance matrix, t is the time window (5 min), N is the number of grid nodes, i is the grid node index, and v i (t) represents the multimodal data vector of the i-th node within time window t. Let T be the mean of the multimodal data vectors of all nodes in the grid, and let T be the matrix transpose operation.

[0064] The spatial consistency of pollutant concentrations between adjacent nodes is quantified by calculating the pollutant concentration covariance matrix. Under normal conditions, the pollutant concentration covariance matrix should be nearly diagonalized (with consistent concentration change trends), while under abnormal conditions, the off-diagonal elements deviate significantly.

[0065] S1.6: Calculate the condition number of the covariance matrix, set the anomaly detection threshold, and perform anomaly event judgment.

[0066] Specifically, the expression for calculating the condition number of the covariance matrix is:

[0067]

[0068] Where κ(t) is the condition number of the pollutant concentration covariance matrix ∑(t), and σ max (∑(t)) represents the largest singular value of the pollutant concentration covariance matrix ∑(t), σ min (∑(t)) represents the minimum singular value of the pollutant concentration covariance matrix ∑(t), σ max and σ min Obtained through singular value decomposition (SVD).

[0069] An anomaly detection threshold κ is set based on historical data statistics. th (e.g. κ) th =10 3 If κ(t) > κ th If it is positive, it is determined to be a preliminary abnormal monitoring event; otherwise, it is determined to be a non-abnormal monitoring event.

[0070] Supplementary verification is performed on preliminary abnormal monitoring events to check for abrupt changes in temperature and humidity (e.g., temperature changes greater than 5°C or humidity changes greater than 20% within a continuous time window) to rule out environmental interference. If an abrupt change occurs, the preliminary abnormal monitoring event is classified as a normal monitoring event; otherwise, it is classified as an abnormal monitoring event.

[0071] Ideally, multimodal environmental parameters are collected using a distributed sensor array, combined with spatiotemporal grid partitioning, to ensure the spatial uniformity and multidimensionality of the monitoring data, thereby improving data representativeness. Kalman filtering for denoising and normalization enhances data stability, and the spatial consistency of adjacent nodes is quantified based on the pollutant concentration covariance matrix, enhancing the accuracy of anomaly detection. Furthermore, the condition number calculation of the covariance matrix, combined with singular value decomposition (SVD) and temperature and humidity abrupt change verification, effectively reduces the false alarm rate, ensuring accurate identification of abnormal events. This approach not only improves the sensitivity and reliability of pollutant monitoring but also provides high-quality input for subsequent data compression, transmission, and optimization, ultimately enhancing the intelligence and stability of air quality monitoring.

[0072] S2: If it is an abnormal monitoring event, the full data is stored; otherwise, a spatiotemporal correlation-driven compressed sensing algorithm is used for hierarchical compression, and the monitoring data packet is output.

[0073] Specifically, it includes the following steps:

[0074] S2.1: For abnormal monitoring events, store the original sampled values ​​of the multimodal environmental parameters of all nodes within the current time window, and additionally store the node location coordinates, timestamps, covariance matrices and condition numbers, and encode them in binary.

[0075] It should be understood that full storage preserves the complete spatiotemporal context of anomalous events, facilitating subsequent pollution tracing and model training. Binary encoding reduces storage volume while ensuring data parsingability.

[0076] S2.2: For non-abnormal detection events, calculate the autocorrelation coefficient and spatial correlation coefficient based on the constructed time series of the same node's continuous time window, and make a correlation judgment.

[0077] Specifically, time series s are constructed for PM2.5 and VOCs concentrations over M consecutive time windows (e.g., M=10) at the same node. i (k)=||v i Let (t)|| be the autocorrelation coefficient, expressed as follows:

[0078]

[0079] Where R(τ) is the autocorrelation coefficient, M is the number of continuous time windows, τ is the time delay (e.g., τ = 1, 2), used to quantify the strength of time correlation, k is the index of the number of time windows, and s i (k) represents the magnitude of the multimodal data vector at time window k, s i (k+τ) represents the modulus of the multimodal data vector at time window k+τ.

[0080] The spatial correlation coefficient of pollutant concentrations among the same nodes is calculated using the following expression:

[0081]

[0082] Where, ρ ij The spatial correlation coefficient of pollutant concentrations, Cov is the identifier for covariance calculation, and s j (k) represents the magnitude of the multimodal data vector from another sensor at time window k. For s i standard deviation For s j The standard deviation of , where j is the index of the number of another grid node.

[0083] Furthermore, high spatiotemporal correlation indicates that the data is sparse, making it suitable for compressed sensing. If R(τ) > 0.8 and ρ ij If the correlation is greater than 0.7, it is determined to be strongly correlated, and a high compression ratio mode is enabled with priority hierarchical implementation; otherwise, it is determined to be weakly correlated and lossless compression is performed using Huffman coding.

[0084] S2.3: Implement a priority stratification, with PM2.5 concentration and volatile organic compound concentration as the first-level key parameters, and temperature and humidity as the second-level auxiliary parameters.

[0085] It should be understood that the first layer of key parameters are PM2.5 concentration and VOCs concentration, which need to retain the original resolution to improve accuracy. The second layer of auxiliary parameters are temperature and humidity, which allow downsampling to reduce transmission pressure.

[0086] S2.4: Perform a spatiotemporal joint sparse transformation on the first layer data to generate a partial Fourier observation matrix, obtain compressed measurement values ​​through linear projection, and output monitoring data packets.

[0087] It should be noted that the spatiotemporal joint sparse transformation refers to constructing an original multimodal data vector based on the number of grid nodes, the number of continuous time windows, and PM2.5 and VOCs data, performing sparse transformation using a two-dimensional wavelet transform basis, and obtaining compressed measurement values ​​through linear projection.

[0088] Furthermore, the output detection data packet structure includes: a data packet category header, compressed measurement values, sparse basis type, observation matrix index (recording randomly selected Fourier row numbers for cloud reconstruction), and second-layer downsampling auxiliary parameters.

[0089] S2.5: Recover sparse coefficients in the cloud and reconstruct errors based on compressed sensing theory.

[0090] Specifically, the Orthogonal Matching Pursuit (OMP) algorithm is used on a cloud server to recover the sparse coefficients θ from y.

[0091] According to compressed sensing theory, the reconstruction error satisfies the following expression:

[0092]

[0093] in, To reconstruct the signal vector, x K For the best K-sparse approximation, C0 is the error coefficient of the best sparse approximation (determined by the restricted isometry property RIP, C0≈1.2~2).

[0094] Ideally, compressed sensing technology driven by spatiotemporal correlation is used to perform layered compression of environmental monitoring data, optimizing data storage and transmission efficiency. In cases of abnormal monitoring events, full storage is employed to preserve complete spatiotemporal information for pollution source tracing and model training. In cases of non-abnormal events, the spatiotemporal correlation of the data is determined by calculating autocorrelation and spatial correlation coefficients, and a high or low compression ratio mode is selected based on the correlation strength. High-resolution data for key pollutants such as PM2.5 and VOCs concentrations are prioritized for preservation, while auxiliary parameters such as temperature and humidity can be downsampled. A spatiotemporal joint sparse transformation combined with linear projection of a partial Fourier observation matrix significantly reduces the data volume. The orthogonal matching pursuit (OMP) algorithm is used in the cloud to recover sparse coefficients, and reconstruction errors are controlled through compressed sensing theory.

[0095] S3: Calculate the channel comprehensive quality index based on monitoring data packets, predict the channel state by combining deep learning models, dynamically select the combination of transmission protocols, and obtain multi-protocol data streams.

[0096] Specifically, it includes the following steps:

[0097] S3.1: Obtain physical layer and network layer indicators, and calculate the weighted overall channel quality index.

[0098] Specifically, physical layer and network layer metrics include signal-to-noise ratio, packet loss rate, latency, bandwidth availability, and link stability.

[0099] The physical layer and network layer indices are normalized, and the weight allocation is dynamically adjusted using the entropy weight method. The channel comprehensive quality index A (A∈[0,1], the larger the value, the better the channel quality) is calculated by weighted summation.

[0100] S3.2: Extract multi-dimensional fusion feature vectors based on the channel integrated quality index, construct a bidirectional LSTM network model, use cross-entropy loss for training, and obtain the probability distribution of future channel states.

[0101] Specifically, a bidirectional LSTM (Bi-LSTM) network was chosen as the deep learning model.

[0102] The multidimensional fusion feature vector includes the channel quality index, signal-to-noise ratio, packet loss rate, current ambient temperature, and current ambient humidity at historical moments.

[0103] The extracted multi-dimensional fused feature vectors are used as the dataset, and the training / validation / test sets are divided in an 8:1:1 ratio.

[0104] Using cross-entropy as the loss function, and employing the Adam optimizer, the model was trained with an initial learning rate of 0.001 and a batch size of 32 until the bidirectional LSTM network model converged.

[0105] Output the channel state probability distribution Y for future times. t = [Good, Medium, Poor], when A≥0.7, the channel state is Good, when the prediction is 0.4≤A<0.7, the channel state is Medium, and when the prediction is A<0.4, the channel state is Poor.

[0106] S3.3: Based on the channel state probability distribution, select the combination of transmission protocols and perform load balancing, encapsulate the detection data packets into multi-protocol data streams, and perform multi-threaded parallel transmission.

[0107] Specifically, if the probability distribution Poor > 0.6, select the LoRaWAN protocol + adaptive coding (redundancy 30%).

[0108] If Medium > 0.5, select NB-IoT protocol + adaptive coding (redundancy 15%).

[0109] If Good > 0.8 and the current channel overall quality index A < 0.7, select MQTT protocol + no redundancy.

[0110] If the channel overall quality index A ≥ 0.7, adaptive coding is forcibly enabled (redundancy + 10%).

[0111] The adaptive coding redundancy adjustment expression is:

[0112]

[0113] Where g is the redundancy percentage and max is the maximum value identifier.

[0114] Furthermore, the monitoring data packets are segmented according to the protocol MTU (Maximum Transmission Unit): LoRaWAN packets are ≤51 bytes each, NB-IoT packets are ≤1600 bytes each, and MQTT packets are transmitted in whole packets (≤10KB).

[0115] Each data block header includes a protocol identifier field (2 bytes), which includes: Protocol_ID: protocol type encoding (e.g., 0001 = LoRaWAN), Redundancy: redundancy percentage (0-15, corresponding to 0%-30% in 2% increments), and Sequence: data block sequence number (used for reassembly at the receiving end).

[0116] Each protocol is assigned an independent thread, and the priority is dynamically adjusted according to the value A. When A≥0.7, the MQTT thread has the highest priority, and when A<0.4, the LoRaWAN thread has the highest priority.

[0117] Ideally, the Channel Quality Index (CQI) weights physical and network layer metrics to comprehensively assess channel quality and ensure accurate reflection of the transmission environment. Combining a deep learning model (bidirectional LSTM network) to predict future channel conditions enables effective selection of appropriate transmission protocols. When channel conditions are poor, the low-power LoRaWAN protocol is chosen with increased redundancy, while when channel quality is good, the MQTT protocol is selected to improve data transmission efficiency. Dynamic adjustment of protocol redundancy further enhances transmission reliability and reduces packet loss risk. Parallel transmission and load balancing of multi-protocol data streams also avoid transmission bottlenecks, ensuring efficient data transmission. Through protocol prioritization and multi-threaded processing mechanisms, transmission strategies can be optimized based on real-time channel conditions, improving overall stability, data transmission rate, and response time.

[0118] S4: Upload multi-protocol data streams to the cloud, perform spatiotemporal interpolation reconstruction, calculate the reconstruction hash and write it to the blockchain for integrity verification.

[0119] Specifically, it includes the following steps:

[0120] S4.1: Parses multi-protocol data streams in the cloud, verifies transmission results through cyclic redundancy check, and performs spatiotemporal tag alignment.

[0121] Specifically, data blocks are separated based on the protocol identifier field. Data blocks from high-reliability protocols (such as LoRaWAN) are processed first, while data blocks from low-latency protocols (such as MQTT) are reassembled first. Data block integrity is verified using Cyclic Redundancy Check (CRC-32), and data blocks that fail the check are discarded.

[0122] Each data block is attached with a spatiotemporal label, including the latitude and longitude coordinates of the sensor node (floating-point, precision 1e-6 degrees) and the acquisition timestamp (Unix time, millisecond precision).

[0123] The parsed multi-protocol data streams and spatiotemporal labels are integrated into a contaminant matrix.

[0124] S4.2: Spatial interpolation of missing data is performed using the Kriging interpolation algorithm, and time gaps are filled using cubic spline interpolation, followed by cross-validation.

[0125] It should be understood that Kriging interpolation utilizes spatial correlation to provide the best unbiased estimate, while cubic spline interpolation ensures temporal smoothness.

[0126] Cross-validation randomly retains 10% of the node data as the validation set. The interpolation error is calculated. If the interpolation error exceeds a threshold (e.g., PM2.5 concentration error > 5 μg / m³), the result is considered invalid. 3 This triggers an error alarm and requests a retransmission.

[0127] S4.3: Use the SHA-256 hash algorithm to generate data hash values, package them into transactions, write them to a private blockchain, and perform integrity verification by comparing hash values.

[0128] Specifically, the interpolated pollutant matrix is ​​standardized by rows and columns, and the data hash value is generated using the SHA-3-256 algorithm.

[0129] Pack the hashes into transactions and write them to a private blockchain (such as Hyperledger Fabric).

[0130] After users download the data, the hash is recalculated and compared with the blockchain record to perform an integrity check.

[0131] Ideally, cyclic redundancy check (CRC-32) is used to verify the integrity of data blocks, ensuring that data is not corrupted or tampered with during transmission. Spatiotemporal tag alignment is used to ensure accurate data labeling. Kriging interpolation and cubic spline interpolation are employed to fill in missing data, providing accurate spatial and temporal estimates, and cross-validation ensures that interpolation errors are within a reasonable range. The SHA-256 hash algorithm is used to generate data hash values, and blockchain provides tamper-proof integrity verification, ensuring the security and trustworthiness of transmission. Users verify transmission integrity through hash comparison, comprehensively improving the reliability, accuracy, and security of transmission, providing a reliable foundation for applications such as environmental monitoring.

[0132] S5: Update the weights of the deep learning model, optimize the compressed sensing algorithm, optimize energy allocation according to the transmission load, and perform iterative optimization of the deep learning model.

[0133] Specifically, it includes the following steps:

[0134] S5.1: Collect real-time channel state prediction errors, update the weights of the bidirectional LSTM network model using a stochastic gradient descent algorithm with a driving term, and detect degradation and roll back the data.

[0135] It should be understood that the prediction error is calculated as the mean square error (MSE) between the model output and the actual channel state.

[0136] If the error in three consecutive time windows exceeds 1.5 times the historical average error, the bidirectional LSTM network model is considered to have degraded.

[0137] If degradation is detected, the bidirectional LSTM network model is rolled back to the version with the smallest verification error in the first 10 iterations.

[0138] S5.2: Minimize the weighted sum of compressed sensing reconstruction error and energy consumption, and use a genetic algorithm to search for the optimal observation matrix.

[0139] Specifically, the expression for minimizing the weighted sum of compressed sensing reconstruction error and energy consumption is:

[0140]

[0141] Where λ1 is the perception reconstruction error weighting coefficient (taken as 0.7), λ2 is the energy consumption weighting coefficient (taken as 0.3), and E(Φ) is the energy consumption of the observation matrix (negatively correlated with the matrix sparsity).

[0142] Furthermore, the genetic algorithm is specifically configured as follows: 100 observation matrices are randomly generated, with each matrix element following a Gaussian distribution. The reconstruction error and energy consumption of each individual are calculated as a weighted sum, and the top 20% are selected as elites after sorting. Uniform crossover is used, with parent matrices randomly swapped element-wise (crossover probability of 0.8), and noise is added to the elements with a probability of 0.1, followed by Gaussian mutation.

[0143] The maximum number of iterations is 200 or the rate of change of the fitness function is less than 10. -5 The iteration terminates when the time is right.

[0144] S5.3: Construct a node energy consumption model to maximize network throughput and use the Lagrange multiplier method to solve for optimal data allocation.

[0145] Specifically, the node energy consumption model expression is as follows:

[0146]

[0147] Where E(i) is the transmission power consumption of node i, G1 is the transmission power consumption, and G2 is the standby power consumption. For node transmission time, This refers to the node's standby time.

[0148] Transmission power consumption and standby power consumption are obtained through hardware instructions from the distributed sensor array.

[0149] With the overall network throughput as the optimization objective, the Lagrange multiplier method is used to solve for and obtain the optimal data allocation ratio, and the data transmission allocation is optimized based on the ratio.

[0150] Ideally, a stochastic gradient descent algorithm with a driving term is used to dynamically update the weights of the bidirectional LSTM model, combined with error detection and rollback mechanisms to prevent degradation of the bidirectional LSTM network model and improve the accuracy of channel state prediction. A genetic algorithm is used to optimize the observation matrix, reducing transmission energy consumption while ensuring data recovery quality and improving the efficiency of compressed sensing. By constructing a node energy consumption model and using the Lagrange multiplier method to optimize data allocation, network throughput and optimal energy consumption control are maximized. This mechanism enhances the method's adaptability, reduces overall energy consumption, improves data recovery quality and transmission efficiency, and ensures long-term stable operation of the monitoring network in complex environments.

[0151] This embodiment also provides an air quality monitoring data transmission system, including: a data acquisition module, used to acquire multimodal environmental parameters through a distributed sensor array, calculate the pollutant concentration covariance matrix of adjacent monitoring nodes, and determine abnormal monitoring events based on condition numbers;

[0152] The data compression module is used to store the full data if it is an abnormal monitoring event, otherwise it uses a spatiotemporal correlation-driven compression sensing algorithm to perform layered compression and output monitoring data packets.

[0153] The protocol selection module is used to calculate the channel comprehensive quality index based on monitoring data packets, predict the channel state by combining a deep learning model, dynamically select the combination of transmission protocols, and obtain multi-protocol data streams.

[0154] The block writing module is used to upload multi-protocol data streams to the cloud, perform spatiotemporal interpolation reconstruction, calculate the reconstruction hash and write it to the blockchain, and perform integrity verification.

[0155] The closed-loop feedback module is used to update the weights of the deep learning model, optimize the compressed sensing algorithm, optimize energy allocation according to the transmission load, and perform iterative optimization of the deep learning model.

[0156] This embodiment also provides a computer device applicable to the air quality monitoring data processing and transmission method, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the air quality monitoring data processing and transmission method proposed in the above embodiment.

[0157] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0158] This embodiment also provides a storage medium storing a computer program, which, when executed by a processor, implements the air quality monitoring data processing and transmission method proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0159] In summary, this invention achieves efficient anomaly detection and improved data accuracy by: collecting multimodal environmental parameters using a distributed sensor array and utilizing spatiotemporal grid partitioning and covariance matrix analysis. For non-anomaly events, a spatiotemporal correlation-driven compressed sensing algorithm is employed for data compression to reduce transmission load, and data transmission efficiency is optimized by dynamically selecting transmission protocol combinations. Data security and reliability are ensured by uploading data to the cloud for spatiotemporal interpolation reconstruction and blockchain integrity verification. Continuous updates to the deep learning model and optimization of the compressed sensing algorithm, through closed-loop feedback control, further enhance the method's adaptability and energy efficiency, enabling the entire monitoring unit to operate efficiently and stably in complex network environments.

[0160] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for processing and transmitting air quality monitoring data, characterized in that: The process includes collecting multimodal environmental parameters through a distributed sensor array, calculating the pollutant concentration covariance matrix of adjacent monitoring nodes, and determining abnormal monitoring events based on the condition number. The specific steps are: dividing the detection area into grids, with the sensor nodes of each grid forming adjacent node groups; Multimodal environmental data is collected through sensor nodes, preprocessed using Kalman filtering and data normalization to construct multimodal data vectors; for each adjacent node group, the pollutant concentration covariance matrix is ​​calculated based on the multimodal data vectors; the condition number of the covariance matrix is ​​calculated, anomaly judgment thresholds are set, and anomaly events are judged; if it is an abnormal monitoring event, the full data is stored; otherwise, a spatiotemporal correlation-driven compressed sensing algorithm is used for hierarchical compression to output monitoring data packets. Specifically, for abnormal monitoring events, the original sampled values ​​of multimodal environmental parameters of all nodes within the current time window are fully stored, along with the node location coordinates, timestamps, covariance matrix, and condition number, and binary encoding is performed. For non-anomaly detection events, the autocorrelation coefficient and spatial correlation coefficient are calculated based on the constructed time series of continuous time windows of the same node to determine the correlation; the expression for calculating the autocorrelation coefficient is: ;in, The autocorrelation coefficient is... The number of consecutive time windows. This is a time delay used to quantify the strength of time correlation. Indexed by the number of time windows. In the time window Multimodal data vector magnitude at time In the time window The magnitude of the multimodal data vector at time; the expression for the spatial correlation coefficient is: ;in, The spatial correlation coefficient of pollutant concentration. The identifier for covariance calculation. In the time window The magnitude of the multimodal data vector from another sensor. for standard deviation for standard deviation The system indexes the number of grid nodes; implements priority hierarchical management, using PM2.5 concentration and volatile organic compound concentration as the first-layer key parameters, and temperature and humidity as the second-layer auxiliary parameters; performs a spatiotemporal joint sparse transformation on the first-layer data to generate a partial Fourier observation matrix, obtains compressed measurement values ​​through linear projection, and outputs monitoring data packets; the spatiotemporal joint sparse transformation refers to constructing an original multimodal data vector based on the number of grid nodes, the number of continuous time windows, and PM2.5 and VOCs data, performing a sparse transformation using a two-dimensional wavelet transform basis, and obtaining compressed measurement values ​​through linear projection; recovers the sparse coefficients in the cloud, reconstructs the error according to compressed sensing theory; calculates the channel comprehensive quality index based on the monitoring data packets, predicts the channel state using a deep learning model, dynamically selects the transmission protocol combination, and obtains multi-protocol data streams; uploads the multi-protocol data streams to the cloud, performs spatiotemporal interpolation reconstruction, calculates the reconstruction hash and writes it into the blockchain for integrity verification; updates the weights of the deep learning model, optimizes the compressed sensing algorithm, optimizes energy allocation according to the transmission load, and performs iterative optimization of the deep learning model.

2. The air quality monitoring data processing and transmission method as described in claim 1, characterized in that: The multimodal environmental parameters include PM2.5 concentration, volatile organic compound concentration, temperature, and humidity.

3. The air quality monitoring data processing and transmission method as described in claim 2, characterized in that: The process involves calculating the channel comprehensive quality index based on monitoring data packets, predicting channel state using a deep learning model, dynamically selecting transmission protocol combinations, and acquiring multi-protocol data streams. Specifically, this process includes: acquiring physical layer and network layer metrics and weighting them to calculate the channel comprehensive quality index; extracting multi-dimensional fusion feature vectors based on the channel comprehensive quality index, constructing a bidirectional LSTM network model, training it using cross-entropy loss, and obtaining the probability distribution of future channel states; selecting transmission protocol combinations based on the channel state probability distribution and performing load balancing; encapsulating the detection data packets into multi-protocol data streams and performing multi-threaded parallel transmission.

4. The air quality monitoring data processing and transmission method as described in claim 3, characterized in that: The process of uploading multi-protocol data streams to the cloud, performing spatiotemporal interpolation reconstruction, calculating reconstructed hashes and writing them to the blockchain, and verifying integrity involves the following steps: parsing the multi-protocol data streams in the cloud, verifying the transmission results through cyclic redundancy check, and aligning spatiotemporal tags; using the Kriging interpolation algorithm to perform spatial interpolation on missing data, using cubic spline interpolation to fill time gaps, and performing cross-validation; using the SHA-256 hash algorithm to generate data hash values, packaging them into transactions, writing them to the private blockchain, and verifying integrity by comparing hash values.

5. The air quality monitoring data processing and transmission method as described in claim 4, characterized in that: The process of updating the deep learning model weights, optimizing the compressed sensing algorithm, optimizing energy allocation based on transmission load, and iteratively optimizing the deep learning model involves the following steps: collecting real-time channel state prediction errors, updating the weights of the bidirectional LSTM network model using a stochastic gradient descent algorithm with a driving term, while detecting degradation and performing rollback; minimizing the weighted sum of compressed sensing reconstruction errors and energy consumption, using a genetic algorithm to search for the optimal observation matrix; constructing a node energy consumption model, maximizing network throughput, and using the Lagrange multiplier method to solve for the optimal data allocation.

6. An air quality monitoring data transmission system, based on the air quality monitoring data processing and transmission method according to any one of claims 1 to 5, characterized in that: It includes a data acquisition module, a data compression module, a protocol selection module, a block writing module, and a closed-loop feedback module; the data acquisition module is used to collect multimodal environmental parameters through a distributed sensor array, calculate the pollutant concentration covariance matrix of adjacent monitoring nodes, and determine abnormal monitoring events based on condition numbers; The data compression module is used to store the full data if it is an abnormal monitoring event, and otherwise use a spatiotemporal correlation-driven compression sensing algorithm to perform layered compression and output monitoring data packets. The protocol selection module is used to calculate the channel comprehensive quality index based on the monitoring data packets, predict the channel state by combining a deep learning model, dynamically select the combination of transmission protocols, and obtain multi-protocol data streams; the block writing module is used to upload the multi-protocol data streams to the cloud, perform spatiotemporal interpolation reconstruction, calculate the reconstruction hash and write it to the blockchain for integrity verification. The closed-loop feedback module is used to update the weights of the deep learning model, optimize the compressed sensing algorithm, optimize energy allocation according to the transmission load, and perform iterative optimization of the deep learning model.

7. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the air quality monitoring data processing and transmission method according to any one of claims 1 to 5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the air quality monitoring data processing and transmission method according to any one of claims 1 to 5.

Images