Intelligent sensing-oriented urban greenway environment quality dynamic monitoring and optimization method

Abstract

The present application relates to the technical field of urban environment monitoring, in particular to a city greenway environment quality dynamic monitoring and optimization method for intelligent sensing, comprising the following steps: multi-source heterogeneous sensor layout, space-time feature alignment, environment quality dynamic index construction and multi-objective optimization strategy generation; through the construction of a multi-modal environment feature matrix with space-time coding, the fusion perception of multiple elements such as air quality, sound environment, vegetation physiology and crowd density is realized; a dynamic attenuation factor and weight adjustment mechanism are used to generate an environment quality dynamic index; a multi-objective optimization model is constructed and an executable strategy is generated by combining historical trends and mutation recognition; feedback loop and strategy adaptive adjustment are realized based on digital twinning and reinforcement learning; the present application can improve the continuity of city greenway environment monitoring, the timeliness of evaluation and the intelligence of regulation.

Description

Technical Field

[0001] This invention relates to the field of urban environmental monitoring technology, and in particular to a method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing. Background Technology

[0002] With the continuous advancement of urban green infrastructure construction, urban greenways, as multi-functional open spaces integrating ecological protection, leisure and fitness, and low-carbon travel, are gradually becoming an important part of the urban ecosystem. Greenways are highly open, have diverse users, and involve complex coupling of environmental elements, thus not only providing ecological services but also directly impacting residents' travel experience and physical and mental health. In recent years, with the development of the Internet of Things and intelligent sensing technologies, more and more cities have begun to deploy environmental monitoring equipment in greenway scenarios to collect data such as air quality, noise, and vegetation status.

[0003] Existing greenway environmental monitoring methods generally suffer from the following technical bottlenecks: First, the sensor types are limited and the deployment rules lack specificity, making it difficult to cover key environmental dimensions in real-world usage scenarios, resulting in fragmented and asynchronous data collection in time and space. Second, current environmental assessment models mostly use fixed weights and static evaluation indicators, failing to dynamically reflect the complex coupling relationships and temporal changes among environmental elements, and thus providing unreliable decision-making basis. Third, existing optimization control strategies are mostly driven by preset rules, lacking the ability to respond to emergencies and the self-learning mechanism based on feedback, failing to meet the high-frequency and high-complexity dynamic management needs of urban greenways. Summary of the Invention

[0004] Based on the above objectives, this invention provides a method for dynamic monitoring and optimization of urban greenway environmental quality oriented towards intelligent sensing. It constructs a closed-loop monitoring and control method that integrates multimodal sensing, dynamic evaluation and intelligent optimization to achieve real-time sensing, accurate assessment and efficient management of greenway environmental quality.

[0005] A method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing includes the following steps: S1 deploys a multi-source heterogeneous sensor array to synchronously collect raw sensing data of the greenway space. The raw sensing data includes air quality parameters, sound environment parameters, vegetation physiological parameters, and pedestrian density parameters. S2, performs spatiotemporal reference alignment processing on the raw sensing data to generate a multimodal environment feature matrix with spatiotemporal coding; S3, construct an environmental quality scoring model based on dynamic decay factor, and input the multimodal environmental feature matrix into the environmental quality scoring model to calculate the dynamic index of environmental quality; S4 combines the dynamic index of environmental quality with historical data trends to generate a set of multi-objective optimization strategies; S5 executes the optimal strategy from the multi-objective optimization strategy set and monitors feedback in real time, adjusting strategy parameters through closed-loop control.

[0006] Optionally, S1 includes: S11, Define the sensor types and sensor array layout rules for multi-source heterogeneous sensor arrays based on the needs of greenway environmental monitoring; S12, Generate a sensor deployment location matrix based on the defined sensor type and greenway topology; S13. Based on the generated deployment location matrix, a multi-source sensor synchronous acquisition mechanism is established to synchronously acquire the raw perception data of the greenway space and encapsulate it into a data frame with a unified timestamp. S14, perform local preprocessing of raw sensing data based on unified timestamp data frames to generate structured datasets; S15, Generate raw perceptual data packets with topological identifiers based on the generated structured dataset.

[0007] Optionally, S2 includes: S21, receive the output raw sensing data packet with topology identifier, perform spatiotemporal reference alignment preprocessing, and extract the timestamp, topology code and geographic location information in the data packet; S22, Based on the extracted topological coding and geographic location information, perform spatial reference alignment to generate spatial alignment data; S23, perform time base alignment based on the extracted timestamps to generate time-aligned data.

[0008] Optionally, S2 further includes: S24, merge spatially aligned data and temporally aligned data to generate a multimodal fusion dataset; S25. Based on the generated multimodal fusion dataset, a multimodal environmental feature matrix with spatiotemporal encoding is generated through unified format reconstruction and spatiotemporal index encoding.

[0009] Optionally, S3 includes: S31, Receive the spatiotemporally encoded multimodal environmental feature matrix output by S25, and initialize the environmental quality scoring model parameters; S32, After the model initialization is completed, a dynamic adjustment mechanism for environmental parameters is constructed, namely the dynamic decay factor calculation function; S33, based on the constructed dynamic adjustment mechanism, completes the weight calculation of multimodal parameters and generates a dynamic weight matrix; S34. Based on the completion of the calculation of the dynamic weight matrix and the dynamic attenuation factor, the score of each spatiotemporal unit is defined as the dynamic index of environmental quality at that location, and spatiotemporal slice calculation is performed.

[0010] Optionally, the S3 further includes: S35, based on the calculated dynamic environmental quality index Reconstruct its dynamic environmental quality index field across the entire region; S36, Based on the generation of a dynamic index field, dynamic index normalization and classification are performed to enhance interpretability and intervention guidance; S37, based on the dynamic index field that has been completed and normalized, finally outputs the structured environmental quality assessment results.

[0011] Optionally, S4 includes: S41, based on the structured environmental quality assessment results output by S37, extracts historical data and performs time-series decomposition analysis to perceive trends and abrupt changes; S42, after completing trend identification, construct a multi-objective optimization model as the basis for strategy generation; S43. Based on the constructed multi-objective optimization model and the extracted trend information, a set of optimization candidate strategies is generated.

[0012] Optionally, S4 further includes: S44. To ensure that the candidate strategy has the expected performance, a simulation environment is constructed to perform a digital pre-evaluation. S45. After completing the simulation evaluation, the candidate strategies are optimized, ranked, and conflict resolved to generate a comparable ranking score. S46. Based on the optimal ranking results, generate the final executable multi-objective optimization strategy set and output it to the execution module.

[0013] Optionally, S5 includes: S51, based on the generated executable multi-objective optimization strategy set, parses the structured instructions and converts them into control sequences that the device can recognize; S52, based on the control sequence generated in S51, schedules various types of execution devices to implement optimization strategies; After the strategy execution is initiated, S53 relies on a multi-source heterogeneous sensor array to carry out real-time feedback monitoring and acquire feedback data.

[0014] S54. After obtaining the feedback data, compare the actual effect with the predicted effect and calculate the strategy deviation rate. S55 uses reinforcement learning to dynamically adjust policy parameters based on the deviation rate analysis results, and constructs a closed-loop optimization mechanism.

[0015] The beneficial effects of this invention are: This invention, by deploying a multi-source heterogeneous sensor array, integrates four types of sensing parameters: air quality, acoustic environment, vegetation physiology, and pedestrian density. Combined with the multi-stage spatiotemporal alignment, spatial compensation, time delay correction, and standardization processing methods in S2, it constructs a high-dimensional, multi-channel, spatiotemporally encoded multimodal environmental feature matrix. This significantly improves the completeness, synchronicity, and spatial resolution of urban greenway environmental information acquisition, providing a solid data foundation for subsequent intelligent analysis.

[0016] This invention proposes an index scoring model based on dynamic attenuation factors and directional corrections. It also constructs a dynamic weight matrix by combining the analytic hierarchy process (AHP) and entropy weighting, effectively adapting to the heterogeneous contributions of different parameters to environmental quality. Furthermore, the score of each spatiotemporal unit is directly defined as a dynamic environmental quality index, making the assessment model significantly superior to static evaluation methods in terms of timeliness and ability to express local differences, thus achieving a precise depiction of the evolution of greenway environments.

[0017] This invention constructs a multi-objective optimization model in S4, integrating four objectives: air quality, pedestrian density, acoustic environment, and vegetation health. It utilizes NSGA-III and fuzzy comprehensive evaluation to generate an executable set of strategy instructions. In S5, a digital twin pre-evaluation mechanism is established based on real-time feedback, and reinforcement learning is introduced to dynamically adjust key control parameters, achieving closed-loop control and adaptive enhancement of the optimization strategy. This mechanism can quickly respond and automatically correct the strategy after abrupt events, effectively improving the intelligence and dynamism of urban greenway environmental quality management. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only for this invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention; Figure 2 This is a schematic diagram of the S2 process in an embodiment of the present invention. Detailed Implementation

[0020] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments. It should also be noted that, to make the embodiments more comprehensive, the following embodiments are the best and preferred embodiments, and those skilled in the art can use other alternative methods to implement some well-known technologies; moreover, the accompanying drawings are only for more specific description of the embodiments and are not intended to specifically limit the present invention.

[0021] It should be noted that the use of terms such as "an embodiment," "an embodiment," "an exemplary embodiment," and "some embodiments" in the specification indicates that the described embodiment may include a specific feature, structure, or characteristic, but not every embodiment necessarily includes that specific feature, structure, or characteristic. Furthermore, when a specific feature, structure, or characteristic is described in connection with an embodiment, implementing such a feature, structure, or characteristic in conjunction with other embodiments (whether explicitly described or not) should be within the knowledge of those skilled in the art.

[0022] Generally, terms can be understood at least partly from their use in context. For example, depending at least partly on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in a singular sense, or a combination of features, structures, or characteristics in a plural sense. Additionally, the term "based on" can be understood not necessarily to convey an exclusive set of factors, but rather, alternatively, depending at least partly on the context, to allow for the presence of other factors that are not necessarily explicitly described.

[0023] like Figures 1-2 As shown, the method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing includes the following steps: S1 deploys a multi-source heterogeneous sensor array to synchronously collect raw sensing data of the greenway space. The raw sensing data includes air quality parameters, sound environment parameters, vegetation physiological parameters, and pedestrian density parameters. S2, performs spatiotemporal reference alignment processing on the raw sensing data to generate a multimodal environment feature matrix with spatiotemporal coding; S3, construct an environmental quality scoring model based on dynamic decay factor, and input the multimodal environmental feature matrix into the environmental quality scoring model to calculate the dynamic index of environmental quality; S4 combines the dynamic index of environmental quality with historical data trends to generate a set of multi-objective optimization strategies; S5 executes the optimal strategy from the multi-objective optimization strategy set and monitors feedback in real time, adjusting strategy parameters through closed-loop control.

[0024] S1 includes: S11, Based on the needs of greenway environmental monitoring, define the sensor types and sensor array deployment rules for multi-source heterogeneous sensor arrays, as follows: The air quality parameter acquisition unit includes a PM2.5 laser scattering sensor, a VOCs electrochemical sensor, and a negative oxygen ion concentration detection probe, which are used to simultaneously acquire particulate matter, volatile organic compounds, and air freshness indicators. The acoustic environment parameter acquisition unit adopts a parallel layout of an A-weighted sound pressure level sensor and a 1 / 3 octave band spectrum analysis sensor to capture the total sound pressure level and frequency band energy distribution, respectively. The vegetation physiological parameter acquisition unit consists of an infrared thermal imager and a stem flow meter, which simultaneously acquire leaf surface temperature and transpiration rate to quantify vegetation heat stress and water transport status. A visible light camera and an infrared thermal imager were added to work together. The visible light camera was installed at a 30° downward angle and took an aerial view of the greenway every 15 minutes. The vegetation pixel ratio was extracted using the U-Net image segmentation algorithm to calculate the real-time vegetation coverage. , ,in, This represents the number of pixels in the segmented vegetation area. This represents the total number of pixels in the image. The crowd density parameter acquisition unit is equipped with a binocular stereo vision camera and a WiFi probe array. It uses human posture recognition and MAC address deduplication to calculate real-time crowd density. S12, Generate the sensor deployment location matrix based on the defined sensor types and greenway topology, as follows: A monitoring node is set up every 50 meters along the main path of the greenway. Each node contains three types of sensors: air quality, sound environment, and vegetation physiology, forming a linear and continuous monitoring chain. With the greenway intersection as the center, a binocular camera is deployed to cover a radius of 200 meters to detect the flow of people. WiFi probes are added to the rest area to monitor the spatial thermal distribution. The PM2.5 sensor is installed 1.5 meters below the tree canopy to simulate the human breathing zone, the sound environment sensor is 1.2 meters above the ground corresponding to the height of a pedestrian's ear, and the vegetation physiological sensor is 20 centimeters away from the target leaf surface to ensure non-contact measurement accuracy. S13. Based on the generated deployment location matrix, a multi-source sensor synchronous acquisition mechanism is established to synchronously acquire the raw perception data of the greenway space and encapsulate it into data frames with a unified timestamp, as follows: All sensor clocks are synchronized via the PTP precise time protocol, with a time deviation of less than 1 millisecond, ensuring cross-modal data timing consistency. By deploying a multi-source heterogeneous sensor array, raw perception data of the greenway space is collected synchronously, including air quality parameters, acoustic environment parameters, vegetation physiological parameters, and pedestrian density parameters. Air quality parameters, acoustic environment parameters, vegetation physiological parameters, and human density parameters are encapsulated into data frames with a unified timestamp and uploaded to the edge computing gateway via the LoRaWAN protocol. The data frame contains three elements: sensor ID, acquisition time, and numerical value. S14, perform local preprocessing of the raw sensor data based on the uploaded unified timestamp data frame to generate a structured dataset, as follows: The sliding quartile method is used to remove transient outliers from PM2.5 data. If three consecutive sampling values ​​exceed 500 μg / m³, a hardware self-test command is triggered. Apply a Butterworth bandpass filter (20Hz-20kHz) to the acoustic environment data to eliminate interference in non-human-sensitive frequency bands and preserve effective acoustic characteristics; Wavelet thresholding was applied to the vegetation leaf surface temperature data, with a decomposition scale of 5 levels and the sym4 wavelet basis function selected to eliminate environmental radiation noise. Kalman filtering was used to reduce noise in the crowd density data, and the state equation parameters were calibrated using historical crowd migration patterns to suppress WiFi signal fluctuation interference. The preprocessed data is stored as a structured dataset containing timestamps, sensor IDs, and data types, with each data field corresponding to an S13 data frame. S15, Generate raw perceptual data packets with topological identifiers based on the generated structured dataset, as follows: Each monitoring node is assigned a unique topology code, with the coding format being "Greenway Number-Node Sequence Number-Sensor Type", where the node sequence number is bound to the node coordinates in the S12 deployment location matrix; Air quality parameters, acoustic environment parameters, vegetation physiological parameters and human flow density parameters are aligned by timestamp and packaged together, with the time series length of each parameter in the data package being consistent. The data packet header includes a CRC-32 checksum and node geographic location information (longitude, latitude, and elevation). The checksum is used to verify the integrity of data transmission. Output the complete data packet to the spatiotemporal reference alignment processing module of S2. The data packet format is a JSON nested array structure, and the key names strictly match the field names of the S14 structured dataset.

[0025] S2 includes: S21: Receive the output raw sensing data packet with topology identifier, perform spatiotemporal reference alignment preprocessing, and extract the timestamp, topology code, and geographic location information from the data packet, as follows: Parse the CRC-32 checksum in the header of the data packet. If the checksum fails, send a data retransmission request to the corresponding monitoring node. Calculate the data acquisition time interval based on the timestamp difference between two consecutive data frames. The formula is: ,in, The UTC timestamp of the current data frame. The timestamp of the previous data frame; Extract the timestamp, topology code (greenway number - node number - sensor type) and geographic location information (longitude, latitude, and elevation) from the data packet; The raw sensing data was split into four independent subsets according to sensor type: air quality parameter set, sound environment parameter set, vegetation physiological parameter set, and human flow density parameter set. S22, Based on the extracted topological encoding and geographic location information, spatial benchmark alignment is performed to generate spatially aligned data, as follows: The longitude and latitude of each monitoring node are converted into three-dimensional spatial coordinates (X,Y,Z) in the WGS84 coordinate system, and the elevation value Z is used as the vertical reference. Construct a spatial relationship mapping table, where each record contains: topology code, spatial coordinates (X,Y,Z), sensor type, and greenway number; Spatial attenuation compensation is calculated for the acoustic environment parameter set, and the compensation formula is as follows: ; in To measure the sound pressure level, The actual distance between the sensor and the sound source. Meters are used as a reference distance; S23, Perform time base alignment based on the extracted timestamps to generate time-aligned data, as detailed below: All sensor data is uniformly converted to UTC time format with a time resolution of 100 milliseconds; For data segments with acquisition time discrepancies, cubic spline interpolation is used to resample them to a unified time axis; The vegetation physiological parameter set was corrected for photosynthetically active radiation (PAR) time lag, with the time delay being... The calculation is as follows: .

[0026] S2 also includes: S24, fuse spatially aligned data and temporally aligned data to generate a multimodal fused dataset, as detailed below: The air quality parameters, sound environment parameters, and vegetation physiological parameters are correlated by Cartesian product with the timestamp according to the spatial coordinates (X,Y,Z); Spatial interpolation of pedestrian density parameters was performed, and a 200-meter gridded density distribution map was generated based on the inverse distance weighting method (IDW), which was then matched with the spatial coordinates of acoustic environment parameters. The missing data was filled using spatiotemporal kriging interpolation, and the semivariogram model was selected as exponential, with the maximum search radius set to 100 meters. Z-score normalization is applied to all parameters, using the following formula: ,in, The historical average values ​​of each parameter are... Standard deviation; S25. Based on the generated multimodal fusion dataset, a multimodal environment feature matrix with spatiotemporal encoding is generated through unified format reconstruction and spatiotemporal index encoding, as follows: Define the matrix dimension as The number of parameter channels is fixed at 12 (PM2.5, VOCs, negative oxygen ions, A-weighted sound level, 1 / 3 octave band × 10 bands, leaf surface temperature, transpiration rate, and human flow density). A spatiotemporal encoded label is attached to each matrix element. The label format is as follows: ,in, / / The coordinates in S22 are three-dimensional spatial coordinates, with precision retained to 4 decimal places; The matrix is ​​stored in blocks, each containing 60 consecutive minutes of data, with the data within each block arranged in ascending order of timestamps. Output the multimodal environmental feature matrix to the S3 environmental quality scoring model. The matrix data format is HDF5 hierarchical structure, and the metadata includes standardized parameters. And a spatial relationship mapping table.

[0027] S3 includes: S31: Receive the spatiotemporally encoded multimodal environmental feature matrix output from S25, and initialize the environmental quality scoring model parameters as follows: Extract Z-Score normalization parameters from matrix metadata Establish a parameter denormalization calculation channel; Load the dynamic decay factor calculation module, whose inputs are the vegetation coverage defined in S11 and the timestamp interval in S21; The model input layer dimension is defined to be consistent with the dimension of the multimodal environment feature matrix, containing a spatiotemporally encoded data block with 12 parameter channels; S32, after the model initialization is complete, a dynamic adjustment mechanism for environmental parameters is constructed, namely the dynamic attenuation factor calculation function, as follows: Define dynamic decay factor The calculation formula is: ,in, The environmental parameter attenuation base rate is 0.15 for PM2.5, 0.22 for VOCs, and 0.08 for noise. The data collection time interval Vegetation coverage; Apply directional correction factors to acoustic environment parameters , is represented as: ,in, The angle between the direction of the sound source and the axis of the sensor; S33, based on the constructed dynamic adjustment mechanism, completes the weight calculation of multimodal parameters and generates a dynamic weight matrix, as detailed below: The initial weight set is determined using the analytic hierarchy process (AHP). Among them, air quality has a weight of 0.35, sound environment 0.25, vegetation physiology 0.25, and human density 0.15; The superimposed entropy weight method dynamically adjusts the weights and calculates the information entropy of each parameter. : is represented as: ,in, For the first The parameter in the first... The proportion of standardized values ​​for each timestamp. The length of the time window (default 60 minutes); Generate dynamic weight matrix , is represented as: ; The weight update frequency is synchronized with the data block partitioning of S25; S34, based on the calculation of the dynamic weight matrix and dynamic attenuation factor, defines the score of each spatiotemporal unit as the dynamic index of environmental quality at that location, and performs spatiotemporal slice-by-slice calculation, as follows: The score for each spatiotemporal unit is calculated on a time-slice basis for the multimodal environment feature matrix. The calculation formula is: ,in, To standardize parameter values, and For Z-Score parameters; The directional correction for the superposition of acoustic environment parameters is expressed as: ,like When >90°, Forced to 0.1 to eliminate back noise interference.

[0028] The feature is that S3 further includes: S35, based on the calculated dynamic environmental quality index The dynamic index field of environmental quality across the entire region is reconstructed as follows: Score each spatiotemporal unit Reorganize according to the matrix dimensions of S25 to form a three-dimensional dynamic exponential field; The exponential field is Gaussian smoothed, and the convolution kernel size is 3×3×3 (time×X×Y), with a standard deviation of [missing value]. =1.5; Weighted average along the elevation direction Z, with weights Calculated according to the sensor deployment height specified in S12, it is expressed as follows: ,in The values ​​are: PM2.5 sensor 1.5 meters, sound environment sensor 1.2 meters, vegetation physiological sensor 0.2 meters; S36, Based on the generation of a dynamic index field, to enhance interpretability and intervention guidance, dynamic index normalization and grading are performed, as follows: The range method is used to map the exponent to the interval [0, 100], and the formula is as follows: ; in, , Extract the extreme values ​​of the previous 24 hours from the historical data stored in S25; The thresholds for classifying quality levels are as follows: excellent: ≥ 80; Good: 60≤ <80; Medium: 40≤ <60; Difference: <40; S37, based on the completed hierarchical normalized dynamic index field, finally outputs the structured environmental quality assessment results, as follows: Each spatiotemporal unit outputs a data structure containing: Spatiotemporal encoded labels (strictly consistent with the S25 definition format); Raw score and normalized value ; Quality grade label (Excellent / Good / Average / Poor); Dominant Influencing Factors (names and contributions of the two parameters with the highest weights) The data is stored in a spatiotemporal cube structure with a temporal resolution of 100 milliseconds and a spatial resolution of 50 meters × 50 meters × 1.5 meters (X × Y × Z).

[0029] S4 includes: S41, based on the structured environmental quality assessment results output by S37, extracts historical data and performs time-series decomposition analysis to perceive trends and abrupt changes, as detailed below: First, the quality level labels, dominant influencing factors, and normalized values ​​are analyzed from the spatiotemporal cube structure. The time range covers the 24 hours preceding the current moment; Next, seasonal trend decomposition is performed on the historical data to separate the trend components. Seasonal portion With residual components The window length is set to 24 hours; Finally, based on residual volatility detection of trend abrupt changes, if And if it lasts for more than 5 minutes, it is marked as a mutation event, among which The standard deviation of the residuals; S42, after completing trend identification, construct a multi-objective optimization model as the basis for policy generation, as follows: Based on the multi-dimensional objectives of greenway management, a set of optimization objective functions is defined, including: Air quality optimization goals: ; Sound environment optimization goals: ; Human density balance target: ; Vegetation health maintenance goals: ; At the same time, the following constraints are set: the real-time pedestrian density must not exceed the maximum carrying capacity (5 people / square meter); The frequency of dust suppression spraying is constrained by VOCs sensor data (disabled when VOCs > 200 ppb); S43, Based on the constructed multi-objective optimization model and the extracted trend information, a set of optimization candidate strategies is generated, as follows: Based on the established objective function and constraints, the NSGA-III algorithm is used to solve for the Pareto optimal solution set. The input parameters include: Spatiotemporal coding labels for the current dynamic index field of environmental quality; Dynamic weight matrix ; Trend components obtained from STL decomposition ; For each set of solutions, a corresponding control strategy is generated, which includes the following: Configuration of spray intensity and effective radius of the intelligent spray system; Sound pressure level attenuation and frequency band selection for directional sound barriers; The update frequency and directional planning of electronic signs in pedestrian guidance strategies; In vegetation irrigation strategies, the amount of irrigation is adjusted based on leaf surface temperature.

[0030] S4 also includes: S44. To ensure the candidate strategy has the expected performance, a simulation environment is constructed to perform a digital pre-evaluation, as follows: First, construct a digital twin model: import the multimodal environment feature matrix generated in step S25 into the Anylogic simulation platform, keeping the spatial and temporal resolution consistent; Then, based on the simulation of the dynamic index field of environmental quality for the next hour according to each candidate strategy, the improvement rate of each optimization objective function is calculated: , ∈1,2,3,4; For air quality improvement rate, To improve the sound environment optimization effect, The improvement rate of human flow density balancing effect Improvement rate of vegetation health maintenance effect; Finally, set the retention condition: when the air quality improvement rate Furthermore, the improvement rate of pedestrian density balancing effect When the screening strategy is in progress, it proceeds to the next round. S45. After completing the simulation evaluation, the candidate strategies are optimized, ranked, and conflict resolved to generate a comparability ranking score, as follows: First, a fuzzy comprehensive evaluation model is constructed to transform the effects of candidate strategies into an evaluation matrix. Rows represent strategies, columns represent objectives, and the membership function is: ,in, For candidate strategies in the target ( Membership values ​​on (∈1,2,3,4), For candidate strategies in optimization objectives Improvement rate , For all candidate strategies in the optimization objective The maximum / minimum improvement rate on the data is used for normalization. Next, based on the defined initial weights Calculate the overall score of the strategy : ; Conflict resolution rule: When the score difference between strategies is less than 5%, the strategy with lower energy consumption is preferred. S46, Based on the optimal ranking results, generate the final executable multi-objective optimization strategy set and output it to the execution module, as follows: Based on the ranking results, the strategy with the highest score is selected and encoded to form a JSON format control instruction package, which includes: Spatiotemporal scope (referencing the spatiotemporal coding label of S25); Control equipment and its parameter settings (such as spray intensity, sound barrier frequency band, etc.); The effective time window is delayed by 5 seconds to compensate for simulation time (S44). Finally, the structured strategy set is output to the execution module of S5.

[0031] S5 includes: S51, based on the generated executable multi-objective optimization strategy set, parses the structured instructions and converts them into control sequences that the device can recognize, as follows: First, extract the control field information contained in the policy set, including: Spatiotemporal scope: Strictly matches the spatiotemporal coding label of S25, with the following format: ; Execution parameters: such as spray intensity and sound barrier attenuation. Effective time window: Start time is delayed by 5 seconds to compensate for the set simulation calculation delay; Then, based on the overall score of the strategy The control commands are prioritized and a complete sequence of device control commands is generated. S52, based on the control sequence generated in S51, schedules various types of execution devices to implement optimization strategies, as follows: Intelligent spray system: Based on the spray intensity in the command, it controls the opening of the pre-deployed nozzle solenoid valves, with an effective radius of 5 meters; Directional sound barrier system: Control the reverse sound wave emission phase of the 1 / 3 octave band sensor, control the error of the target attenuation within ±0.5dB, and the frequency band setting should correspond exactly to the frequency band field in the strategy command (e.g., 315–400Hz); Pedestrian guidance system: Electronic signs are updated according to policy instructions, and the guidance direction is bound to the guidance direction field in the policy instructions. The guidance frequency ranges from 1 to 5 times per minute. After the strategy execution is initiated, S53 relies on a multi-source heterogeneous sensor array to carry out real-time feedback monitoring and acquire feedback data.

[0032] S54, after obtaining the feedback data, compare the actual results with the predicted results and calculate the policy deviation rate, as follows: First, based on the obtained feedback data, the actual improvement rate is recalculated. ; Then, compare the estimated values ​​of the strategy. Calculate the deviation rate : ; Set dynamic adjustment trigger conditions: when any indicator If the duration is ≥3 minutes, it is marked as a deviation from the expected effect and enters the strategy self-adjustment stage; S55, based on the deviation rate analysis results, uses reinforcement learning methods to dynamically adjust policy parameters and construct a closed-loop optimization mechanism, as detailed below: To construct a Q-learning model, the following elements are defined: State space: includes real-time environmental state variables such as current strategy number, PM2.5 concentration, sound pressure level, and crowd density; Action space: includes the spray intensity adjustment range and the sound barrier attenuation adjustment range; Reward function: With the goal of improving performance and balancing energy consumption, the reward value is calculated as follows: ; in, This is the immediate reward value after policy execution, used to guide the reinforcement learning model in selecting actions. A larger value indicates a more effective policy adjustment. The deviation rate from the air quality target is calculated as the percentage difference between the actual improvement rate and the predicted value. The improvement deviation rate for the goal of balancing pedestrian density is also based on the deviation between the actual and predicted values. It represents the control effect of the strategy in guiding pedestrian flow. This represents the increase in device energy consumption resulting from the current strategy compared to the previous execution. The optimized strategy parameters are output and then sent back for execution, forming a closed-loop control chain.

[0033] This invention encompasses any substitutions, modifications, equivalent methods, and solutions made within the spirit and scope of this invention. To provide the public with a thorough understanding of this invention, specific details are described in detail in the following preferred embodiments; however, those skilled in the art will fully understand the invention even without these details. Furthermore, to avoid unnecessary misunderstanding of the essence of this invention, well-known methods, processes, procedures, components, and circuits are not described in detail.

[0034] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing, characterized in that, Includes the following steps: S1 deploys a multi-source heterogeneous sensor array to synchronously collect raw sensing data of the greenway space. The raw sensing data includes air quality parameters, sound environment parameters, vegetation physiological parameters, and pedestrian density parameters. S2, performs spatiotemporal reference alignment processing on the raw sensing data to generate a multimodal environmental feature matrix with spatiotemporal coding; S3, construct an environmental quality scoring model based on dynamic decay factor, and input the multimodal environmental feature matrix into the environmental quality scoring model to calculate the dynamic index of environmental quality; S4 combines the dynamic index of environmental quality with historical data trends to generate a set of multi-objective optimization strategies; S5 executes the optimal strategy from the multi-objective optimization strategy set and monitors feedback in real time, adjusting strategy parameters through closed-loop control.

2. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing as described in claim 1, characterized in that, S1 includes: S11, Define the sensor types and sensor array layout rules for multi-source heterogeneous sensor arrays based on the needs of greenway environmental monitoring; S12, Generate a sensor deployment location matrix based on the defined sensor type and greenway topology; S13. Based on the generated deployment location matrix, a multi-source sensor synchronous acquisition mechanism is established to synchronously acquire the raw perception data of the greenway space and encapsulate it into a data frame with a unified timestamp. S14, perform local preprocessing of raw sensing data based on unified timestamp data frames to generate structured datasets; S15, Generate raw perceptual data packets with topological identifiers based on the generated structured dataset.

3. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing as described in claim 2, characterized in that, S2 includes: S21, receive the output raw sensing data packet with topology identifier, perform spatiotemporal reference alignment preprocessing, and extract the timestamp, topology code and geographic location information in the data packet; S22, Based on the extracted topological coding and geographic location information, perform spatial reference alignment to generate spatial alignment data; S23, perform time base alignment based on the extracted timestamps to generate time-aligned data.

4. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 3, characterized in that, S2 further includes: S24, merge spatially aligned data and temporally aligned data to generate a multimodal fusion dataset; S25. Based on the generated multimodal fusion dataset, a multimodal environmental feature matrix with spatiotemporal encoding is generated through unified format reconstruction and spatiotemporal index encoding.

5. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 4, characterized in that, S3 includes: S31, Receive the spatiotemporally encoded multimodal environmental feature matrix output by S25, and initialize the environmental quality scoring model parameters; S32, After the model initialization is completed, a dynamic adjustment mechanism for environmental parameters is constructed, namely the dynamic decay factor calculation function; S33, based on the constructed dynamic adjustment mechanism, completes the weight calculation of multimodal parameters and generates a dynamic weight matrix; S34. Based on the completion of the calculation of the dynamic weight matrix and the dynamic attenuation factor, the score of each spatiotemporal unit is defined as the dynamic index of environmental quality at that location, and spatiotemporal slice calculation is performed.

6. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 5, characterized in that, S3 further includes: S35, based on the calculated dynamic environmental quality index, reconstructs its dynamic environmental quality index field across the entire region; S36, Based on the generation of dynamic index fields, dynamic index normalization and classification are performed to enhance interpretability and intervention guidance; S37, based on the dynamic index field that has been completed and normalized, finally outputs the structured environmental quality assessment results.

7. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 6, characterized in that, S4 includes: S41, based on the structured environmental quality assessment results output by S37, extracts historical data and performs time-series decomposition analysis to perceive trends and abrupt changes; S42, after completing trend identification, construct a multi-objective optimization model as the basis for strategy generation; S43. Based on the constructed multi-objective optimization model and the extracted trend information, a set of optimization candidate strategies is generated.

8. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 7, characterized in that, S4 further includes: S44. To ensure that the candidate strategy has the expected performance, a simulation environment is constructed to perform a digital pre-evaluation. S45. After completing the simulation evaluation, the candidate strategies are optimized, ranked, and conflict resolved to generate a comparable ranking score. S46. Based on the optimal ranking results, generate the final executable multi-objective optimization strategy set and output it to the execution module.

9. The method for dynamic monitoring and optimization of urban greenway environmental quality based on intelligent sensing according to claim 8, characterized in that, S5 includes: S51, based on the generated executable multi-objective optimization strategy set, parses the structured instructions and converts them into control sequences that the device can recognize; S52, based on the control sequence generated in S51, schedules various types of execution devices to implement optimization strategies; S53, after the strategy execution is started, relies on a multi-source heterogeneous sensor array to carry out real-time feedback monitoring and obtain feedback data; S54. After obtaining the feedback data, compare the actual effect with the predicted effect and calculate the strategy deviation rate. S55 uses reinforcement learning to dynamically adjust policy parameters based on the deviation rate analysis results, and constructs a closed-loop optimization mechanism.

Images