Urban Wind Speed ​​Monitoring and Assimilation Method and System Based on Tree Swaying Characteristics

Abstract

This invention provides a method and system for urban wind speed monitoring and assimilation based on tree swaying characteristics, including an image acquisition module, an image processing and feature extraction module, a model training and storage module, a wind speed inversion module, a data assimilation module, and a data storage and visualization module. The method includes the following steps: deploying cameras in the target area; acquiring tree image sequences using the cameras; extracting swaying characteristic parameters for different tree species based on the tree image sequences; constructing swaying-wind speed mapping models for different tree species based on the swaying characteristic parameters and the tree's own attributes; inverting the urban wind speed distribution based on the swaying-wind speed mapping models; and assimilating the wind speed distribution into an urban wind field model. The beneficial effects of this invention are: achieving refined urban-scale wind speed monitoring using existing camera resources, improving the spatiotemporal resolution of wind speed data, and providing strong support for urban meteorological services, disaster prevention and mitigation, etc.

Description

Technical Field

[0001] This invention belongs to the field of meteorological monitoring technology, and in particular relates to a method and system for monitoring and assimilating urban wind speed based on tree swaying characteristics. Background Technology

[0002] Urban wind speed monitoring is a crucial foundation for meteorological services, disaster early warning (such as typhoons and strong winds), and urban planning. Current urban wind speed monitoring primarily relies on equipment such as meteorological stations and wind measurement towers, which have the following technical limitations:

[0003] (1) Insufficient spatiotemporal resolution: Traditional wind measurement equipment is expensive and difficult to deploy, making it difficult to achieve large-scale, high-density coverage in densely populated urban areas. This results in low spatial resolution of wind speed data (usually at the kilometer level), making it difficult to reflect the fine differences in wind fields between blocks and building clusters.

[0004] (2) High monitoring costs: Dedicated wind measuring instruments (such as ultrasonic anemometers) are expensive, large-scale deployment requires high investment, and maintenance costs are high.

[0005] (3) Poor adaptability: Cities are full of high-rise buildings and have complex terrain. Traditional wind measurement equipment is easily affected by building obstruction, and the monitoring data is not representative enough. Some areas (such as commercial areas and residential areas) cannot install wind measurement equipment due to space limitations.

[0006] While some studies have attempted to retrieve wind speed using image recognition (such as video-based particle motion analysis), the following problems exist:

[0007] (1) Relying on manually placed markers (such as smoke or balloons) has limited applicability.

[0008] (2) The relationship between the swaying characteristics of naturally occurring "biosensors" (such as trees) in the city and wind speed was not considered.

[0009] (3) The lack of an assimilation mechanism with urban wind field patterns makes it difficult to integrate scattered monitoring data into wind field information across the entire region.

[0010] Therefore, there is an urgent need for a low-cost, easy-to-deploy, and high spatiotemporal resolution urban wind speed monitoring method that utilizes the abundant tree resources in cities and combines image recognition and data assimilation technologies to achieve refined wind speed monitoring and wind field optimization. Summary of the Invention

[0011] In view of this, the present invention aims to propose a method and system for urban wind speed monitoring and assimilation based on tree swaying characteristics, so as to solve at least one of the problems existing in the above-mentioned prior art.

[0012] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0013] In a first aspect, the present invention provides a method for urban wind speed monitoring and assimilation based on tree swaying characteristics, comprising the following steps:

[0014] S1. Deploy cameras and high-precision wind measuring instruments in the target area;

[0015] S2. Use a camera to capture tree image sequences;

[0016] S3. Based on tree image sequences, extract the swaying feature parameters of different tree species;

[0017] S4. Based on the swaying characteristic parameters of different tree species and the tree's own attributes, construct a swaying-wind speed mapping model for different tree species.

[0018] S5. Based on the swing-wind speed mapping model, invert the urban wind speed distribution;

[0019] S6. Assimilate wind speed distribution into urban wind field patterns;

[0020] In step S3, based on the tree image sequence, swing feature parameters are extracted, including:

[0021] S31. Preprocess the tree image sequence to obtain a preprocessed image;

[0022] S32. Based on the preprocessed image, use deep learning algorithms to locate the tree position, segment the tree region, and distinguish the tree species;

[0023] S33. Based on tree regions of different tree species, use optical flow or feature point matching technology to track the movement trajectory of tree canopy and branches of different tree species.

[0024] S34. Based on the tracking results of different tree species, extract the swing characteristic parameters of different tree species;

[0025] In step S4, based on the swaying characteristic parameters of different tree species and the tree's own attributes, a swaying-wind speed mapping model for different tree species is constructed, including:

[0026] S41. Match the oscillation characteristic parameters and wind speed to obtain calibration data;

[0027] S42. Based on the calibration data, a swing-wind speed mapping model is constructed using a supervised learning algorithm.

[0028] In step S6, the wind speed distribution is assimilated into the urban wind field model, including:

[0029] Based on urban wind field model data, a data assimilation algorithm is used to assimilate wind speed distribution as observation data into the wind field model, thereby correcting the initial field or boundary conditions.

[0030] Furthermore, in step S31, the tree image sequence is preprocessed to obtain a preprocessed image, including:

[0031] S311, Remove noise interference from the image;

[0032] S312. Stabilize the image;

[0033] S313. Extract the region of interest from the image.

[0034] Furthermore, in step S41, the oscillation characteristic parameters and wind speed are matched to obtain calibration data, including:

[0035] S411. Set the wind speed data acquisition time to synchronize it with the acquisition time of the oscillation characteristic parameters;

[0036] S412. Use high-precision wind measuring instruments to collect wind speed data in real time;

[0037] S413. Based on the swing characteristic parameters and wind speed data, calibration data is obtained.

[0038] Furthermore, in step S5, based on the oscillation-wind speed mapping model, the urban wind speed distribution is inverted, including:

[0039] S51. Perform steps S2 to S3 to extract the swing feature parameters;

[0040] S52. Based on the swaying characteristic parameters and the tree's own attributes, call the swaying-wind speed mapping model to calculate the instantaneous wind speed and obtain the tree wind speed data;

[0041] S53. Combining urban terrain data, using spatial interpolation algorithms, the tree wind speed data is interpolated to the wind speed distribution of the target area.

[0042] Furthermore, in step S1, the target area is covered with different underlying surfaces and different tree species;

[0043] In step S2, the camera captures data at a frequency of no less than 10 frames per second.

[0044] Furthermore, in step S3, the swing characteristic parameters include swing amplitude, swing frequency, swing intensity, and damping coefficient.

[0045] Furthermore, in step S4, the tree's own attributes include tree height, crown width, and tree species.

[0046] Furthermore, in step S4, the inputs to the sway-wind speed mapping model include sway feature parameters and tree attributes; the outputs of the sway-wind speed mapping model include wind speed data; the wind speed data includes wind speed magnitude, wind direction, and turbulence intensity.

[0047] Furthermore, in step S33, the optical flow method is either a sparse optical flow method or a dense optical flow method;

[0048] In step S33, the feature point matching technique includes a feature extraction algorithm and a matching algorithm. The feature extraction algorithm is SIFT, ORB, or SURF, and the matching algorithm is brute-force matching or KD-tree matching.

[0049] In step S42, the supervised learning algorithm is a random forest, gradient boosting tree, or neural network;

[0050] In step S53, the spatial interpolation algorithm is either Kriging interpolation or inverse distance weighted interpolation;

[0051] In step S6, the data assimilation algorithm is three-dimensional variational assimilation or ensemble Kalman filtering algorithm, and the urban wind field model is WRF-UCM or Envi-met.

[0052] Secondly, based on the concept of a unified approach, the present invention also provides an urban wind speed monitoring and assimilation system based on tree swaying characteristics, including an image acquisition module, an image processing and feature extraction module, a model training and storage module, a wind speed inversion module, a data assimilation module, and a data storage and visualization module.

[0053] The image acquisition module is used to acquire a continuous image sequence of trees within the target area;

[0054] The image processing and feature extraction module is used to process image sequences and extract swaying feature parameters of different types of trees;

[0055] The model training and storage module is used to establish and store swing-wind speed mapping models for different types of trees based on measured wind speed data and swing feature parameters through supervised learning.

[0056] The wind speed inversion module is used to calculate the wind speed distribution in the target area using the oscillation-wind speed mapping model;

[0057] The data assimilation module is used to assimilate the wind speed distribution into the urban wind field pattern;

[0058] The data storage and visualization module is used to store raw data, intermediate results, and assimilated wind field data, and provides visualization display.

[0059] Furthermore, the sway-wind speed mapping model includes an input layer, a preprocessing layer, a tree species-specific model layer, and an output layer. The input layer is used to input sway feature parameters and tree attributes. The preprocessing layer is used for feature stitching and generating calibration data. The tree species-specific model layer is used to train models specific to broad-leaved trees, coniferous trees, and shrubs. The output layer is used to output wind speed data.

[0060] Compared with existing technologies, the urban wind speed monitoring and assimilation method and system based on tree swaying characteristics described in this invention has the following beneficial effects:

[0061] (1) Low cost and easy deployment: Utilizing existing urban camera resources, there is no need to deploy dedicated wind measurement equipment on a large scale, which significantly reduces monitoring costs; trees, as natural "sensors", are widely distributed throughout the city and are suitable for complex urban environments.

[0062] (2) High spatiotemporal resolution: Through dense camera coverage and image sequence analysis, wind speed monitoring with a spatial resolution of 10-50 meters and a temporal resolution of seconds can be achieved, capturing wind field differences at the street scale (such as building flow around and funnel effect).

[0063] (3) High adaptability: It is not limited by urban terrain and can cover a variety of areas such as commercial areas, residential areas, and parks. It is especially suitable for areas where traditional wind measurement equipment is difficult to deploy.

[0064] (4) Enhanced data value: By assimilating data, scattered tree wind speed data are integrated into the urban wind field model, improving the accuracy of wind field simulation and providing a more reliable basis for urban meteorological services, typhoon disaster prevention and mitigation, and ventilation corridor planning.

[0065] (5) Tree species adaptability optimization: Establish a dedicated mapping model for the swing characteristics of different tree species to avoid the loss of accuracy of a single model in complex vegetation environment.

[0066] (6) Good integration of technologies: Combining a variety of advanced image recognition and tracking technologies improves the accuracy and reliability of tree swaying parameter extraction, thereby improving the accuracy of wind speed monitoring. Attached Figure Description

[0067] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0068] Figure 1 This is a schematic diagram of the overall process described in an embodiment of the present invention;

[0069] Figure 2 This is a schematic diagram of the swing-wind speed mapping model architecture described in an embodiment of the present invention;

[0070] Figure 3 This is a schematic diagram of the extracted tree swaying feature parameters described in an embodiment of the present invention. Detailed Implementation

[0071] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0072] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined with "first," "second," etc., may explicitly or implicitly include one or more of that feature. In the description of this invention, unless otherwise stated, "a plurality of" means two or more.

[0073] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art will understand the specific meaning of the above terms in this invention based on the specific circumstances.

[0074] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0075] like Figures 1 to 3 As shown, an urban wind speed monitoring and assimilation system based on tree swaying characteristics includes an image acquisition module, an image processing and feature extraction module, a model training and storage module, a wind speed inversion module, a data assimilation module, and a data storage and visualization module. The image acquisition module acquires continuous image sequences of trees within a target area. The image processing and feature extraction module processes the image sequences to extract swaying characteristic parameters for different tree species. The model training and storage module, based on measured wind speed data and swaying characteristic parameters, establishes and stores sway-wind speed mapping models for different tree species through supervised learning. The wind speed inversion module calculates the wind speed distribution in the target area using the sway-wind speed mapping model. The data assimilation module assimilates the wind speed distribution into an urban wind field model. The data storage and visualization module stores raw data, intermediate results, and assimilated wind field data, and provides a visual representation.

[0076] The oscillation-wind speed mapping model consists of an input layer, a preprocessing layer, a tree species-specific model layer, and an output layer. The input layer is used to input oscillation feature parameters (oscillation amplitude, oscillation frequency, oscillation intensity, and damping coefficient) and tree attributes (tree height, crown width, and tree species). The preprocessing layer is used for feature stitching (7-dimensional vector) and generating calibration data (synchronous acquisition). The tree species-specific model layer is used for normalization (dimension elimination) and supervision labels (measured wind speed data), and to train models specific to broad-leaved trees, coniferous trees, and shrubs. The output layer is used to output wind speed data, which includes wind speed magnitude, wind direction, and turbulence intensity.

[0077] A method for monitoring and assimilating urban wind speed based on tree swaying characteristics includes the following steps:

[0078] Step 1: Acquisition of tree image sequences

[0079] A1: Within the target city area, utilize existing surveillance cameras (such as road surveillance cameras and community security cameras) or add dedicated cameras to collect continuous image sequences containing trees.

[0080] A2: Camera deployment needs to cover different underlying surfaces (such as streets, parks, squares) and different tree species (such as broad-leaved trees, coniferous trees, shrubs), with a collection frequency of no less than 10 frames / second to ensure that the details of tree swaying can be captured.

[0081] Step 2: Extraction of Tree Swaying Feature Parameters

[0082] B1: Preprocessing the acquired image sequence: A series of preprocessing operations are performed on the acquired image sequence, including using algorithms such as Gaussian filtering to remove noise interference from the image. Image stabilization is performed based on feature matching techniques such as SIFT (Scale Invariant Feature Transform) to eliminate the adverse effects caused by camera shake, and the region of interest (ROI), i.e., the area where the trees are located, is accurately extracted through advanced algorithms such as semantic segmentation, preparing for subsequent accurate analysis.

[0083] B2: Tree Target Detection and Tracking: Different tree species exhibit varying swaying characteristics that correlate with wind speed. Deep learning algorithms (such as YOLO and Mask R-CNN) are needed to identify trees in preprocessed images and distinguish between different tree species. YOLO can quickly locate tree positions, while Mask R-CNN can more accurately segment tree regions, providing a more precise target range for subsequent tracking.

[0084] B3: In order to obtain the swing characteristics of trees, the movement trajectory of the tree canopy and branches is tracked by optical flow or feature point matching technology.

[0085] Among them, the optical flow method is based on the assumption that the brightness of pixels in adjacent frames remains unchanged. It obtains the instantaneous motion vector (optical flow vector) of the pixel by solving the motion equation. For tree observation, the tree canopy and branches are regarded as a set of numerous pixels. The sparse optical flow method (such as the Lucas-Kanade algorithm) is used to track the movement of a small number of key feature points (such as obvious corner points) on the tree. The computational load is small and can meet the real-time requirements.

[0086] The formula for the sparse optical flow method is as follows:

[0087] ;

[0088] In the formula, I(x,y,t) represents the pixel brightness value at image coordinates (x,y) at time t. This represents the spatial gradient of pixel brightness in the x-direction. This represents the spatial gradient of pixel brightness in the y-direction. This represents the gradient of pixel brightness over time. Let be the optical flow velocity component of the pixel in the x-direction. y is the optical flow velocity component of the pixel in the y direction;

[0089] Solve for the pixel velocity vector .

[0090] Alternatively, dense optical flow methods (such as the Farneback algorithm) can be used to calculate the motion vectors of all pixels in the tree image to obtain more comprehensive swing information, but this requires a large amount of computation.

[0091] Alternatively, feature point matching techniques can be used, employing feature extraction algorithms such as SIFT (Scale Invariant Feature Transform), ORB (Fast Feature Point), or SURF (Speed-Up Robust Feature Transform). Feature points of the trees are extracted and feature vectors are generated in the first frame. In subsequent frames, matching algorithms (such as brute-force matching or KD-tree matching) are used to find the corresponding positions, and the pixel displacement difference of the feature points is calculated. SIFT is robust to scale, rotation, and illumination changes, but computationally intensive; ORB is efficient and suitable for embedded devices; SURF is faster than SIFT and suitable for dynamic scenes.

[0092] B4: Extracting Sway Feature Parameters: Based on the tracking results, calculate the sway feature parameters for different tree species, including:

[0093] Swing amplitude: This refers to the maximum displacement of the tree canopy during the swaying process. It is calculated by measuring the maximum displacement difference between the tracked feature points during the movement, and the unit is centimeters (cm).

[0094] Oscillation frequency: refers to the number of oscillation cycles of a tree per unit time. It is accurately calculated by analyzing the periodic changes in the displacement of characteristic points over time, and the unit is Hertz (Hz).

[0095] Sway intensity: defined as the root mean square of the rate of change of displacement, reflecting the intensity of tree swaying, and measured in centimeters per second (cm / s).

[0096] Damping coefficient: Used to measure the rate at which the tree's swing decays. It is calculated by analyzing the displacement decay during the swing process and reflects the energy loss characteristics of the tree during the swing.

[0097] Step 3: Construction of the oscillation-wind speed mapping model

[0098] C1: After obtaining the swaying characteristics of the trees in the image, it is necessary to match them with the actual wind speed for data calibration. High-precision anemometers, such as ultrasonic anemometers, should be precisely deployed near carefully selected representative trees (at a distance of no more than 5 meters). While the camera is acquiring images, the anemometer simultaneously collects real-time wind speed data, including key information such as wind speed magnitude, wind direction, and turbulence intensity, ensuring strict synchronization with the acquisition time of the tree swaying characteristic parameters to establish an accurate correspondence.

[0099] C2: Model Training: Based on the calibration data, supervised learning algorithms (such as random forest, gradient boosting tree, neural network) are used to construct sway-wind speed mapping models for different tree species. The model input is sway feature parameters and tree attributes (such as tree height, crown width, tree species), and the output is the wind speed at the corresponding location.

[0100] Suppose that the swing-wind speed mapping model for a specific tree species (such as camphor tree, sycamore, etc., the document mentions covering broad-leaved trees, conifers, shrubs, etc.) is as follows: (t represents the tree species category label, such as t=1 for broad-leaved trees, t=2 for coniferous trees, etc.), then the input-output relationship formula of the model is as follows:

[0101] ;

[0102] in, This represents the measured wind speed (unit: m / s) at the location corresponding to a specific tree species t in the i-th set of calibration data, which is the supervised learning target value of the model (collected by an ultrasonic anemometer).

[0103] This represents the swaying characteristic parameter vector of a specific tree species t in the i-th set of calibration data, i.e. ,in The amplitude of the swing. The oscillation frequency, For the oscillation intensity, The damping coefficient;

[0104] For the i-th set of calibration data, the self-attribute vector of a specific tree species t is, i.e. ,in Tree height (unit: m). Crown width (unit: m). Label the tree species category;

[0105] The model prediction error for the i-th set of calibration data satisfies the expected value E[ ]=0. The sway-wind speed mapping function for a specific tree species t is obtained by training a supervised learning algorithm (random forest, gradient boosting tree, neural network) specified in the file.

[0106] C3: Model Optimization: Adjust model parameters through cross-validation to ensure prediction accuracy in different wind speed ranges (such as 0-5m / s, 5-10m / s, >10m / s), especially optimizing model stability under strong wind conditions.

[0107] Step 4: Urban wind speed distribution inversion

[0108] D1: Extract the swing feature parameters from tree images captured by all cameras within the target area through step 2.

[0109] D2: Call the sway-wind speed mapping model for the corresponding tree species to calculate the instantaneous wind speed at each tree location.

[0110] D3: Based on spatial interpolation algorithms (such as Kriging interpolation and inverse distance weighting), combined with urban terrain data (such as building height and street orientation), discrete tree wind speed data is interpolated into a refined wind speed distribution for the target area (spatial resolution can reach 10-50 meters).

[0111] Step 5: Wind speed data assimilation

[0112] E1: Collect existing urban wind field model data, such as simulation results from models like WRF-UCM (Weather Research and Forecasting-Urban Canopy Model) and Envi-met, which can provide preliminary prediction information for urban wind fields.

[0113] E2: Using data assimilation algorithms (such as three-dimensional variational assimilation and ensemble Kalman filtering), the refined wind speed distribution retrieved in step 4 is used as observation data and assimilated into the wind field model to correct the initial field or boundary conditions of the model.

[0114] The three-dimensional variational assimilation formula is as follows:

[0115] ;

[0116] in, The objective function for three-dimensional variational assimilation is to minimize this function to obtain the optimal wind field state vector. ;

[0117] Let be the optimal urban wind field state vector to be solved;

[0118] The background field vector for the wind field model;

[0119] B is the background field error covariance matrix. Its inverse matrix;

[0120] This is the wind speed observation vector for tree inversion;

[0121] For the observation operator, the wind field state vector Mapping from the model grid space to the observation point space;

[0122] R is the observation error covariance matrix. It is its inverse matrix.

[0123] E3: Outputs optimized urban wind field data, including the spatiotemporal distribution of wind speed and direction.

[0124] Example 1:

[0125] 1. Deployment of Experimental Area and Equipment

[0126] (1) Select a typical urban area (approximately 5km × 5km) including different underlying surfaces such as commercial areas, parks, and residential areas.

[0127] (2) Image acquisition: Utilize the existing 20 surveillance cameras in the area (resolution 1920×1080, frame rate 25 frames / second) and add 5 dedicated cameras in the park to cover 10 common trees (such as camphor tree, sycamore, ginkgo, privet, etc.).

[0128] (3) Calibration equipment: Ultrasonic anemometers (measurement range 0-60m / s, accuracy ±0.1m / s) were deployed next to 8 representative trees to record wind speed data and camera images simultaneously.

[0129] 2. Extraction of tree swaying parameters

[0130] (1) Preprocessing: Gaussian filtering is used for noise reduction, SIFT feature matching is used to eliminate camera shake, and semantic segmentation algorithm is used to extract tree canopy ROI.

[0131] (2) Detection and Tracking: The YOLOv5 algorithm is used to detect trees with an accuracy of 95%; the KLT optical flow method is used to track canopy feature points and calculate the displacement changes in each frame. The KLT optical flow method can efficiently obtain the movement trajectory of trees by tracking sparse feature points, which is suitable for the real-time processing requirements in this embodiment.

[0132] (3) Parameter calculation: Extract the swing amplitude (unit cm), frequency (unit Hz), intensity (cm / s) and damping coefficient, and output a set of statistical values ​​every 10 seconds.

[0133] 3. Mapping Model Construction

[0134] (1) Data set: Collect one month of synchronous data (including calm, light, and strong wind weather) and generate 100,000 samples (each sample includes tree species attributes, 4 swing parameters, and measured wind speed).

[0135] (2) Model Training: A random forest regression model was used, with the training and test sets divided in a 7:3 ratio. The input consisted of sway parameters, tree species codes, and tree height, and the output was wind speed. After optimization, the model achieved a good R-value in the wind speed range of 0-10 m / s. 2 The accuracy reached 0.91, and the root mean square error (RMSE) was 0.6 m / s.

[0136] 4. Wind speed distribution inversion

[0137] (1) Extract the sway parameters from the real-time images of 25 cameras (processed every 5 seconds) and call the model of the corresponding tree species to calculate the wind speed at a single point.

[0138] (2) Combining the regional digital elevation model (DEM) with building vector data, Kriging interpolation is used to generate a wind speed distribution map with a resolution of 50 meters, and the wind speed vector map is output.

[0139] 5. Data Assimilation and Application

[0140] (1) The WRF-UCM urban wind field model (spatial resolution 1km) was used as the background field.

[0141] (2) Using a three-dimensional variational assimilation algorithm, the inverted wind speed data with a resolution of 50 meters is assimilated into the model, and the initial field is updated 4 times a day.

[0142] (3) After assimilation, the model’s simulation error of street-scale wind speed is reduced by 40%, and it can accurately capture the local strong winds caused by the “narrowing effect” between buildings.

[0143] Advantages and benefits of this invention:

[0144] (1) Low cost and easy deployment: Utilizing existing urban camera resources, there is no need to deploy dedicated wind measurement equipment on a large scale, which significantly reduces monitoring costs; trees, as natural "sensors", are widely distributed throughout the city and are suitable for complex urban environments.

[0145] (2) High spatiotemporal resolution: Through dense camera coverage and image sequence analysis, wind speed monitoring with a spatial resolution of 10-50 meters and a temporal resolution of seconds can be achieved, capturing wind field differences at the street scale (such as building flow around and funnel effect).

[0146] (3) High adaptability: It is not limited by urban terrain and can cover a variety of areas such as commercial areas, residential areas, and parks. It is especially suitable for areas where traditional wind measurement equipment is difficult to deploy.

[0147] (4) Enhanced data value: By assimilating data, scattered tree wind speed data are integrated into the urban wind field model, improving the accuracy of wind field simulation and providing a more reliable basis for urban meteorological services, typhoon disaster prevention and mitigation, and ventilation corridor planning.

[0148] (5) Tree species adaptability optimization: Establish a dedicated mapping model for the swing characteristics of different tree species to avoid the loss of accuracy of a single model in complex vegetation environment.

[0149] (6) Good integration of technologies: Combining a variety of advanced image recognition and tracking technologies improves the accuracy and reliability of tree swaying parameter extraction, thereby improving the accuracy of wind speed monitoring.

[0150] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for monitoring and assimilating urban wind speed based on tree swaying characteristics, characterized in that: Includes the following steps: S1. Deploy cameras and high-precision wind measuring instruments in the target area; S2. Use a camera to capture tree image sequences; S3. Based on tree image sequences, extract the swaying feature parameters of different tree species; S4. Based on the swaying characteristic parameters of different tree species and the tree's own attributes, construct a swaying-wind speed mapping model for different tree species. S5. Based on the swing-wind speed mapping model, invert the urban wind speed distribution; S6. Assimilate wind speed distribution into urban wind field patterns; In step S3, based on the tree image sequence, swing feature parameters are extracted, including: S31. Preprocess the tree image sequence to obtain a preprocessed image; S32. Based on the preprocessed image, use deep learning algorithms to locate the tree position, segment the tree region, and distinguish the tree species; S33. Based on tree regions of different tree species, use optical flow or feature point matching technology to track the movement trajectory of tree canopy and branches of different tree species. S34. Based on the tracking results of different tree species, extract the swing characteristic parameters of different tree species; In step S4, based on the swaying characteristic parameters of different tree species and the tree's own attributes, a swaying-wind speed mapping model for different tree species is constructed, including: S41. Match the oscillation characteristic parameters and wind speed to obtain calibration data; S42. Based on the calibration data, a swing-wind speed mapping model is constructed using a supervised learning algorithm. In step S6, the wind speed distribution is assimilated into the urban wind field model, including: Based on urban wind field model data, a data assimilation algorithm is used to assimilate wind speed distribution as observation data into the wind field model, thereby correcting the initial field or boundary conditions.

2. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S31, the tree image sequence is preprocessed to obtain a preprocessed image, including: S311, Remove noise interference from the image; S312. Stabilize the image; S313. Extract the region of interest from the image.

3. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S41, the oscillation characteristic parameters and wind speed are matched to obtain calibration data, including: S411. Set the wind speed data acquisition time to synchronize it with the acquisition time of the oscillation characteristic parameters; S412. Use high-precision wind measuring instruments to collect wind speed data in real time; S413. Based on the swing characteristic parameters and wind speed data, calibration data is obtained.

4. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S5, based on the swing-wind speed mapping model, the urban wind speed distribution is inverted, including: S51. Perform steps S2 to S3 to extract the swing feature parameters; S52. Based on the swaying characteristic parameters and the tree's own attributes, call the swaying-wind speed mapping model to calculate the instantaneous wind speed and obtain the tree wind speed data; S53. Combining urban terrain data, using spatial interpolation algorithms, the tree wind speed data is interpolated to the wind speed distribution of the target area.

5. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S1, the target area is covered with different underlying surfaces and different tree species; In step S2, the camera captures data at a frequency of no less than 10 frames per second.

6. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S3, the swing characteristic parameters include swing amplitude, swing frequency, swing intensity, and damping coefficient; In step S4, the tree's own attributes include tree height, crown width, and tree species.

7. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 1, characterized in that: In step S4, the inputs to the sway-wind speed mapping model include sway feature parameters and tree attributes; the outputs of the sway-wind speed mapping model include wind speed data; the wind speed data includes wind speed magnitude, wind direction, and turbulence intensity.

8. The urban wind speed monitoring and assimilation method based on tree swaying characteristics according to claim 4, characterized in that: In step S33, the optical flow method is either a sparse optical flow method or a dense optical flow method; In step S33, the feature point matching technique includes a feature extraction algorithm and a matching algorithm. The feature extraction algorithm is SIFT, ORB, or SURF, and the matching algorithm is brute-force matching or KD-tree matching. In step S42, the supervised learning algorithm is a random forest, gradient boosting tree, or neural network; In step S53, the spatial interpolation algorithm is either Kriging interpolation or inverse distance weighted interpolation; In step S6, the data assimilation algorithm is three-dimensional variational assimilation or ensemble Kalman filtering algorithm, and the urban wind field model is WRF-UCM or Envi-met.

9. A city wind speed monitoring and assimilation system based on tree swaying characteristics, applied to the city wind speed monitoring and assimilation method based on tree swaying characteristics as described in any one of claims 1-8, characterized in that: It includes an image acquisition module, an image processing and feature extraction module, a model training and storage module, a wind speed inversion module, a data assimilation module, and a data storage and visualization module; The image acquisition module is used to acquire a continuous image sequence of trees within the target area; The image processing and feature extraction module is used to process image sequences and extract swaying feature parameters of different types of trees; The model training and storage module is used to establish and store swing-wind speed mapping models for different types of trees based on measured wind speed data and swing feature parameters through supervised learning. The wind speed inversion module is used to calculate the wind speed distribution in the target area using the oscillation-wind speed mapping model; The data assimilation module is used to assimilate the wind speed distribution into the urban wind field pattern; The data storage and visualization module is used to store raw data, intermediate results, and assimilated wind field data, and provides visualization display.

10. The urban wind speed monitoring and assimilation system based on tree swaying characteristics according to claim 9, characterized in that: The swing-wind speed mapping model includes an input layer, a preprocessing layer, a tree species-specific model layer, and an output layer. The input layer is used to input swing feature parameters and tree attributes. The preprocessing layer is used for feature stitching and generating calibration data. The tree species-specific model layer is used to train models specific to broad-leaved trees, coniferous trees, and shrubs. The output layer is used to output wind speed data.

Images