Street tree crown coverage benefit evaluation and optimization method

Abstract

The invention relates to the technical field of urban landscaping, and discloses a street tree crown coverage benefit evaluation and optimization method, which comprises the following steps: S1, collecting multi-source data of street trees in a target area; s2, performing fusion processing on the multi-source data, extracting feature parameters for evaluation, inputting the feature parameters into a pre-trained crown benefit evaluation model, and outputting a comprehensive benefit score of crown coverage by the model; wherein the crown benefit evaluation model is a machine learning model obtained by training sample data with expert score labels; and S3, based on the comprehensive benefit score and a benefit restriction factor identified through model interpretive analysis, matching and generating a targeted crown optimization scheme from a pre-constructed optimization strategy knowledge base. According to the method, the street tree crown coverage benefit evaluation and optimization of the complex urban environment can be systematically completed, the evaluation accuracy is relatively high, and the method has dynamics and operability.

Description

Technical Field

[0001] This invention relates to the field of urban greening technology, specifically to a method for evaluating and optimizing the benefits of street tree canopy coverage. Background Technology

[0002] With the rapid development of urban economies, the impact of urban climate issues on public health and the ecological environment is intensifying, making the urban ecological environment increasingly fragile. The significant heat island effect, the pervasive smog and dust, the sharp rise in carbon dioxide concentration, and frequent public health problems have prompted scholars to focus on the comprehensive benefits of urban green spaces. According to the "Classification Standard for Urban Green Space" (CJJ / T 85-2017), urban green spaces are classified into four main types based on spatial heterogeneity and function: parks, protective green spaces, plaza green spaces, and ancillary green spaces. Among these, street trees, as an important component of ancillary green spaces, not only reflect the traditional urban landscape and carry historical and cultural value, but also play a crucial role in regulating urban climate, improving the ecological environment, and promoting public health.

[0003] However, with the acceleration of urbanization, problems such as the urban heat island effect, air pollution, and ecological imbalance are becoming increasingly prominent. As an important carrier of the urban ecosystem, the canopy of street trees has significant ecological and social benefits. Studies have shown that the canopy of street trees not only reduces surface temperature and alleviates the urban heat island effect through transpiration, but also purifies particulate pollutants and harmful gases in the air by physically trapping them. In addition, its three-dimensional spatial structure plays an important role in regulating the local microclimate. Although many cities have made significant progress in the selection and management of street tree species, they still face some problems and challenges. For example, unreasonable canopy structure leads to low shading efficiency, and the excessively low branch height of some tree species poses a safety hazard for pedestrians. Furthermore, the increasing threats of high temperatures and pests and diseases also affect the functions of street trees in regulating the urban heat island effect, absorbing PM2.5, and reducing noise. In the long run, this will weaken the public's sense of identity and belonging to urban green landscapes, and is not conducive to creating a beautiful and livable urban image.

[0004] Currently, research on street trees mainly focuses on the optimal selection of street tree species, adaptability analysis, and ecological function assessment. Systematic research on the comprehensive benefits of street tree canopies remains relatively lacking. Although existing studies have revealed the ecological functions of street trees in improving urban air quality, regulating microclimate, and mitigating the urban heat island effect, the multi-dimensional and systematic comprehensive benefits of tree canopies at the urban ecological and social levels have not received sufficient attention. Existing research generally suffers from single-dimensionality and fragmented methods; ecological and social benefit assessments are often isolated and fail to be integrated into a unified framework. Most assessments rely on limited field sampling and manual interpretation, which is inefficient and difficult to apply on a large scale. Although some studies have attempted to construct evaluation index systems, the index weights largely depend on expert experience, are highly subjective, and lack adaptive adjustments for complex urban environments (such as mountainous terrain and diverse street canyons). Furthermore, the lack of deep integration between data collection methods and assessment models results in assessments that are often static and descriptive, failing to reveal key constraints affecting benefits and making it difficult to directly guide precise management and optimization strategies. Summary of the Invention

[0005] The present invention aims to provide a method for evaluating and optimizing the benefits of street tree canopy cover, which can systematically complete the evaluation and optimization of the benefits of street tree canopy cover in complex urban environments, with high evaluation accuracy and dynamic and operable features.

[0006] The basic solution provided by this invention is: a method for evaluating and optimizing the benefits of street tree canopy cover, comprising the following steps: S1. Collect multi-source data of roadside trees in the target area, wherein the multi-source data includes at least remote sensing data, near-ground motion monitoring data and Internet of Things sensor data; S2. The multi-source data is fused to extract feature parameters for evaluation, and the feature parameters are input into a pre-trained canopy benefit evaluation model. The canopy benefit evaluation model outputs a comprehensive benefit score of canopy coverage. The canopy benefit evaluation model is a machine learning model trained on sample data with expert rating labels. S3. Based on the comprehensive benefit score and the benefit constraint factors identified through model interpretability analysis, a targeted canopy optimization scheme is generated by matching from the pre-constructed optimization strategy knowledge base.

[0007] The working principle and advantages of this invention are as follows: This invention provides a method for evaluating and optimizing the benefits of street tree canopy cover. It can systematically evaluate and optimize the benefits of street tree canopy cover in complex urban environments, with high accuracy and possessing both dynamism and operability. The key points are: First, this solution achieves the systematic integration and efficient utilization of multi-source heterogeneous data. Traditional street tree assessment methods typically rely on a single data source, such as manual on-site measurement or remote sensing image analysis, which suffers from limited data dimensions, low work efficiency, and difficulty in comprehensively reflecting the true state of tree canopies in complex urban environments. This solution constructs a multi-level, multi-perspective data acquisition system by simultaneously collecting and fusing remote sensing data, near-ground motion monitoring data, and IoT sensor data. Remote sensing data provides macroscopic coverage and topographic background, near-ground motion monitoring data captures refined three-dimensional tree structures and street-level visual landscapes, and IoT sensor data enables continuous in-situ monitoring of ecological processes and environmental factors. This systematic data integration solution overcomes the limitations of a single data source, providing a more comprehensive and accurate data foundation for assessment through mutual verification and supplementation of data, significantly improving the reliability and scientific rigor of the assessment results.

[0008] Secondly, this solution automates and objectifies the evaluation process by employing a machine learning model trained on expert rating labels as the core of the evaluation. This model can automatically learn the complex mapping relationship between multi-dimensional feature parameters and the overall benefits of tree canopy coverage, without requiring rigid, pre-defined calculation formulas or weight allocations. This data-driven evaluation method not only significantly improves evaluation efficiency but also uncovers potential patterns that are difficult for the human brain to intuitively recognize, making the evaluation results more accurate, consistent, and highly repeatable.

[0009] Third, most existing technologies stop at presenting evaluation results, lacking an effective way to translate evaluation conclusions into specific operational guidelines. This solution, by introducing model interpretability analysis technology, can accurately identify key constraints affecting the overall benefit score, transforming abstract scores into specific technical parameter issues. Combined with a pre-built optimization strategy knowledge base, it can automatically map the identified constraints into targeted optimization measures, forming a complete canopy optimization plan. This helps solve the problem of the disconnect between evaluation and optimization in traditional management, providing clear, explicit, and immediately executable technical guidance for landscape management departments. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of the process flow of an embodiment of the method for evaluating and optimizing the benefits of street tree canopy coverage according to the present invention. Detailed Implementation

[0011] The following detailed explanation illustrates the specific implementation methods: Example 1 The basic implementation examples are as follows: Figure 1 As shown: A method for evaluating and optimizing the benefits of street tree canopy cover, including the following steps: S1. Collect multi-source data of roadside trees within the target area. The multi-source data includes at least remote sensing data, near-ground motion monitoring data, and Internet of Things sensor data.

[0012] The remote sensing data includes lidar point cloud data and multispectral image data, which are used to extract individual tree structural parameters (such as individual tree location, tree height, and crown width), digital elevation models, and crown coverage.

[0013] Specifically, in this embodiment, a drone platform equipped with a professional-grade airborne LiDAR system such as the RIEGL VQ-1560ii is rented or deployed. Flight planning ensures a point cloud density of no less than 16 points / square meter to capture detailed canopy top structures. Hyperspectral imagery (such as CASI 1500, ≥48 bands) is acquired simultaneously to retrieve parameters such as chlorophyll content and leaf water content, serving as early indicators of tree health. Flight operations should be conducted under cloudless, windless, or light wind conditions to minimize data noise.

[0014] The near-ground mobile monitoring data is collected by ground mobile platforms (such as inspection vehicles) equipped with lidar and imaging equipment (such as hyperspectral imagers and high dynamic range cameras), including vehicle-mounted point cloud data and vehicle-mounted image data, which are used to extract the three-dimensional green volume of individual trees, canopy structure parameters (such as branch height, canopy width, canopy height, etc.) and street green view rate.

[0015] The IoT sensor data is collected through miniature weather stations and air quality sensors deployed in the target area to obtain continuous time-series data on photosynthetically active radiation, temperature and humidity gradients, and pollutant concentrations.

[0016] In this embodiment, a stratified random sampling method is used to collect IoT sensor data. Specifically, the data is first collected by dominant tree species (such as banyan, Chinese tallow tree, ginkgo, sycamore, etc.) and the width-to-height ratio of the street canyon (such as...). ,0.5-1.0, The system is divided into layers, and then 3-5 streets are randomly selected in each layer, with 1-2 monitoring nodes set up on each street.

[0017] Each monitoring node includes: 1) a canopy microclimate sensor installed in the middle of the tree canopy, in a shaded area, to measure canopy temperature, relative humidity, and photosynthetically active radiation; 2) a control weather station set up near an open, paved surface; and 3) a multi-gas analysis module installed at sidewalk height (approximately 1.5 meters) for monitoring... , , (etc.) and laser dust sensors (for monitoring PM2.5 and PM10). All sensors upload data to the cloud platform at a frequency of 1 minute per upload via 4G / 5G DTU modules.

[0018] On typical weather days (sunny and cloudless), using portable photosynthesis meters such as the LI-6800, measurements were taken on 5 standard trees of each tree species during two periods of strong photosynthetically active radiation, from 9:00 to 11:00 and from 14:00 to 16:00. Three sun-loving leaves from the upper part of the canopy of each tree were measured, and key parameters such as net photosynthetic rate, transpiration rate, and stomatal conductance were recorded.

[0019] S101, Adaptive Correction Steps for Terrain and Road Features: Obtain the digital elevation model of the target area and calculate the terrain factor of the target area; simultaneously, calculate the road structure factor of the target area based on near-ground motion monitoring data; use the terrain factor and road structure factor to perform parameter correction on the original ecological benefit indicators extracted from the multi-source data to obtain the feature parameters after terrain and road feature correction.

[0020] The primary ecological benefit indicators include the following secondary indicators: canopy cover benefits, soil and water conservation benefits, carbon sequestration and oxygen release capacity, air purification capacity, microclimate regulation function, and biodiversity support.

[0021] Each secondary indicator has a tertiary indicator; among them, carbon sequestration and oxygen release capacity includes carbon storage increment and photosynthetic intensity; microclimate regulation function includes cooling effect and humidification effect.

[0022] The topographic factors include the theoretical solar radiation calculated based on the digital elevation model using a solar radiation transmission model, and the topographic humidity index calculated based on a hydrological analysis algorithm. The parameter correction of the original ecological benefit indicators includes: The photosynthetic intensity or cooling effect index is corrected based on the ratio of the actual location to the theoretical solar radiation on the horizontal plane. The humidification effect index is corrected according to the grade coefficient determined by the topographic humidity index.

[0023] Specifically, based on a high-precision DEM, the actual theoretical solar radiation (kWh / m²) for each individual tree location is calculated using professional GIS software (such as ArcGIS Pro's "Regional Solar Radiation" tool). Then, the theoretical solar radiation for that location if it were a horizontal plane is calculated. Finally, the measured photosynthetic intensity of each individual tree is corrected: Corrected photosynthetic intensity = Measured photosynthetic intensity × (Horizontal plane radiation / Actual radiation). Similarly, the topographic humidity index for each location is calculated based on the DEM, and a grade coefficient is set (e.g., 0.9 for the top of the slope, 1.0 for the middle of the slope, and 1.1 for the bottom of the slope) to perform a reverse correction for the humidification effect (ΔRH).

[0024] S2. The multi-source data is fused and processed to extract feature parameters for evaluation. The feature parameters are then input into a pre-trained canopy benefit evaluation model, which outputs a comprehensive benefit score for canopy coverage. The canopy benefit evaluation model is a machine learning model trained on sample data with expert rating labels.

[0025] Specifically, the extraction of feature parameters for evaluation includes: From lidar point cloud data, individual trees are separated using a hierarchical clustering algorithm, and the three-dimensional green content of individual trees is calculated using a convex hull algorithm or a voxel model.

[0026] From street view image data, vegetation pixels are identified through a semantic segmentation model, and the green view rate of each road segment is automatically calculated; Environmental fluctuation characteristic parameters, including daily average, peak value, and volatility, are extracted from the time-series data of IoT sensor data.

[0027] In this embodiment, the following operation steps are specifically included: S201. Unify all multi-source data into the same high-precision geographic coordinate system (such as the CGCS2000 coordinate system). Also, assign a unified timestamp (using UTC time) to all multi-source data. For periodically collected data (such as remote sensing data), record the exact date and time of collection. For continuously monitored IoT sensor data, synchronize its data stream with a network time protocol server to ensure the consistency of time-series data.

[0028] S202, Perform multi-platform point cloud data fusion and accurate 3D green volume calculation.

[0029] Point cloud data acquired by UAV platform LiDAR (top-down view) and vehicle-mounted LiDAR (side-view view) are fused. Using the existing Progressive Profound Triangulation Filtering (PTD) algorithm, ground points and non-ground points are accurately separated from the fused point cloud. Subsequently, the existing Voxel Cloud Connectivity-Based Watershed Segmentation (VCCS) algorithm is applied to the non-ground points. This algorithm effectively solves the problem of adjacent canopy point clouds sticking together, thus achieving more accurate individual tree segmentation.

[0030] For the segmented point cloud of individual trees, the Alpha-Shape algorithm is used for shape simulation. This algorithm can construct a 3D model that accurately reflects the complex morphology of the tree canopy surface, such as its unevenness and holes, based on the actual distribution of the point cloud. The 3D spatial volume calculated based on this refined model is the refined 3D green volume of the tree.

[0031] Based on single-tree point clouds, basic tree height and crown width can be extracted, as well as crown volume density (describing the density of leaf distribution in the crown space) and vertical structure distribution profile (analyzing the distribution of leaf area at different heights) can be calculated.

[0032] Spectral data acquired by a vehicle-mounted hyperspectral imager is fused with three-dimensional spatial information provided by laser point clouds. Each point in the three-dimensional point cloud is assigned a corresponding spectral feature, forming a unified "point-spectrum" data volume. Using this fused data, and through machine learning classifiers (such as support vector machines and random forests), higher-precision tree species classification can be achieved, and early spectral anomalies that are difficult to detect with the human eye due to pests, diseases, or nutrient stress can be detected.

[0033] For street view images, a semantic segmentation deep learning model (such as SegFormer) pre-trained on a large image dataset is used to perform pixel-level classification for each image. Not only is the total green view ratio calculated, but green vegetation is further divided into tree layer, shrub layer, and grassland layer, thereby calculating the tree green view ratio specific to street trees. In addition, the proportion of sky pixels obscured by tree canopies in the image can be statistically analyzed as a visual indicator of shading effect.

[0034] S203. Under a unified spatial reference, the monitoring data of each IoT sensor is correlated with the nearest roadside tree or tree group in its geographical location. At the same time, micro-environmental parameters such as the sky's openness and building occupancy around the sensor location are extracted from the point cloud data.

[0035] In-depth time series analysis is performed on continuous time-series data (such as temperature, humidity, and pollutant concentration) acquired by IoT sensors. Daily averages are calculated, and key features are extracted, including: Cooling / humidifying efficiency: Calculate the cumulative temperature / humidity difference (i.e., area under the curve) between the shaded area and the control point during the daily high-temperature period (e.g., 12:00-16:00) to reflect the regulation effect throughout the day.

[0036] Pollutant peak reduction capacity: Analyze the daily variation curve of pollutant concentration and calculate the reduction rate of pollution concentration during the morning / evening peak by roadside trees.

[0037] Environmental volatility: Calculate the daily temperature range and the daily humidity fluctuation range to assess the contribution of the tree canopy to environmental stability.

[0038] S204. Perform environmental adaptive correction and feature standardization, including: Topographic factor correction: Using a high-precision DEM and a professional solar radiation transfer model (such as r.sun), the actual solar radiation received at each canopy location on a typical day is calculated. Based on this, the measured photosynthetic intensity and cooling effect are normalized and corrected to eliminate the assessment bias caused by topographic shading.

[0039] Street canyon effect correction: Based on the accurate 3D street model extracted by vehicle-mounted LiDAR, the width-to-height ratio (W / H) of the street is extracted. Through expert review, excellent green view rate benchmarks are set for different W / H intervals, and the humidification effect and green view rate of all road segments are standardized accordingly, so that the values ​​measured on narrow streets and wide roads are comparable.

[0040] S205. Construct multidimensional feature vectors.

[0041] For each evaluation unit (a single tree or a homogeneous street segment), all the feature values ​​processed, corrected, and extracted in the above steps are combined into a structured multidimensional feature vector.

[0042] For example, the multidimensional feature vector in this embodiment may include: Structural characteristics: three-dimensional green volume, crown width, tree height, crown volume density, and height under branches.

[0043] Spectral and visual characteristics: tree species coding, green visibility of trees, and spectral stress index.

[0044] Ecological process characteristics: corrected photosynthetic intensity, daily cumulative cooling value, and average pollutant removal rate.

[0045] Environmental background characteristics: green view rate, topographic humidity index, road width-to-height ratio.

[0046] Furthermore, the canopy benefit evaluation model is trained using a gradient boosting decision tree algorithm.

[0047] In this embodiment, 500 sample units (which can be single trees or 50-meter-long homogeneous street sections) are randomly selected from the area where data has been collected. A panel of five experts in landscaping, ecology, and urban planning is organized to review all data for each unit (including 3D models, street view images, time-series curves, and extracted feature values) and independently score the ecological, social, and comprehensive benefits of each sample unit on a 100-point scale. The average score is then used as the label value for that sample.

[0048] A feature vector is constructed for each sample unit, containing all corrected and extracted metrics (such as 3D green volume, corrected photosynthetic intensity, standardized green visibility, cooling effect, humidification effect, dust retention efficiency, etc., totaling over 50 features). The sample library is randomly divided into training and test sets in a 7:3 ratio. The XGBoost algorithm is used to train the model on the training set, and the model hyperparameters (such as learning rate, maximum tree depth, subsampling ratio, etc.) are optimized through grid search and 5-fold cross-validation. The model performance is validated on the test set. Root mean square error and coefficient of determination are used as evaluation metrics. For example, a well-trained model should have an RMSE of less than 8 for its overall benefit prediction score. It should be higher than 0.85.

[0049] For a new region to be evaluated, after completing all data processing in S1, a feature vector is generated for each evaluation unit and input into the trained XGBoost model. The model will directly output the ecological benefit score, social benefit score, and comprehensive benefit score (0-100 points) for that unit.

[0050] Furthermore, the identification of benefit constraint factors through model interpretability analysis refers to: using the SHAP analysis framework to calculate the contribution of each feature parameter to the comprehensive benefit score, and identifying the top N feature parameters with negative contribution and the largest absolute value as core constraint factors, where N is a natural number.

[0051] In this embodiment, the SHAP library in Python is used to interpret the model prediction results for each evaluation unit. The contribution value (SHAP value) of each feature (such as "canopy volume" and "tree pit permeability score") to the final score of the unit is analyzed. The top three features with negative SHAP values ​​in absolute terms are automatically selected and identified as the "core limiting factors" of the unit, generating a diagnostic report. For example, the report might state: "The main reasons for the street's low efficiency are: 1. Poor tree pit permeability (contribution -12.5 points); 2. Insufficient canopy volume (contribution -8.3 points)."

[0052] S3. Based on the comprehensive benefit score and the benefit constraint factors identified through model interpretability analysis, a targeted canopy optimization scheme is generated by matching from the pre-constructed optimization strategy knowledge base.

[0053] The optimization strategy knowledge base is a structured database constructed in the form of rules of constraint factors and optimization strategies; the identified benefit constraint factors are matched with the rules in the optimization strategy knowledge base to generate a list of optimization schemes containing one or more optimization strategies and their priorities.

[0054] In this embodiment, the optimization strategy knowledge base is constructed using a graph database (such as Neo4j), with "constraint factors", "optimization measures", "tree species" and "site conditions" as nodes, and the relationships between them (such as "applicable to", "prohibited to", "enhanced to") as edges.

[0055] Once the limiting factors are diagnosed, a graph traversal algorithm can not only match directly related strategies but also discover associated and synergistic optimization measures. For example, if "poor dust retention capacity" and "poor tree pit permeability" are diagnosed, a combined strategy of "replacing the small-leaved banyan with sycamore (high dust retention)" and "simultaneously modifying the tree pit permeability" is recommended, along with the expected benefit improvement and implementation priority of this combination.

[0056] The optimization strategy includes at least one of the following: Tree shaping and pruning strategies addressing canopy structure issues; for example, for trees with insufficient canopy coverage and green volume, develop stepped pruning or "open-center" shaping strategies to expand the canopy width and increase the leaf area index. For trees with poor shading effect (cooling effect), implement crown-raising pruning or thinning of inner branches to improve understory light and ventilation. For trees posing safety hazards (such as branches too low or touching power lines), propose structural pruning plans.

[0057] Strategies for replanting or mixed planting to address the issue of single-species tree planting: For example, in areas with weak carbon sequestration, oxygen release, and air purification capabilities, it is recommended to introduce tree species with high photosynthetic efficiency and strong pollutant absorption capacity (such as *Firmiana simplex* and *Platanus sylvestris*). In areas with insufficient biodiversity support (food and nectar sources), it is suggested to replant or mix plantings of native tree species with combined nectar and food sources, such as *Ligustrum lucidum* and *Loquat*. In areas dominated by deciduous tree species (such as *Ginkgo biloba*) with poor overall ecological benefits, it is proposed to gradually replace them with native tree species with strong resilience and more balanced ecological functions, such as *Pterocarya stenoptera* and *Koelreuteria paniculata*. Along wide roads, a combination of *Firmiana simplex* (high photosynthesis) and *Platanus sylvestris* (high dust retention) can be used to compensate for the functional deficiencies of a single tree species.

[0058] Strategies for improving the permeability of tree pits to address site conditions; for example, in areas with poor soil and water conservation benefits, it is recommended to replace impermeable tree pit paving (such as porous plastic) with environmentally friendly permeable materials such as permeable concrete.

[0059] Strategies for configuring forest understory spaces and facilities to address social benefits. For example, for areas with low social benefits, forest understory space transformation plans are proposed: installing mobile fitness facilities and sparse woodland lawns on wide roads; embedding ecological tree pits and fixed rest seats in narrow streets and alleys; and setting up time-limited street stall areas ("green economy") in commercial districts, and rationally planning parking spaces to avoid root damage.

[0060] In addition, based on the health problems (such as diseases and poor growth) identified in the assessment, develop targeted water and fertilizer management and pest and disease control plans.

[0061] S4. Dynamic Feedback and Model Optimization Steps: After implementing the canopy optimization scheme, re-execute steps S1 and S2 to obtain optimized canopy coverage benefit data. Use the data before and after optimization as new samples to incrementally learn the canopy benefit evaluation model in order to update the model parameters.

[0062] This embodiment provides a method for evaluating and optimizing the benefits of street tree canopy cover, which can systematically evaluate and optimize the benefits of street tree canopy cover in complex urban environments, with high evaluation accuracy and dynamic and operable features.

[0063] Example 2 A method for evaluating and optimizing the benefits of street tree canopy cover, based on Example 1, further includes a preliminary step: constructing a multi-dimensional comprehensive evaluation index system for canopy cover benefits.

[0064] The evaluation system is constructed using primary indicators based on two core dimensions: ecological benefits and social benefits. Under the ecological benefits dimension, six secondary indicators are established: canopy cover benefits, soil and water conservation benefits, carbon sequestration and oxygen release capacity, air purification capacity, microclimate regulation function, and biodiversity support. Under the social benefits dimension, four secondary indicators are established: mental health benefits, physical health benefits, safety and comfort benefits, and landscape aesthetic benefits.

[0065] Each secondary indicator is further decomposed to form measurable tertiary indicators. For example: The benefits of tree canopy coverage can be broken down into: green volume, tree canopy coverage rate, and green view rate.

[0066] Air purification capacity can be broken down into: SO2 absorption capacity, NO2 absorption capacity, and dust retention capacity.

[0067] Physiological health benefits correspond to the third-level indicator: forest exercise.

[0068] Next, the Delphi method (expert consultation method) was employed, inviting domain experts to assign pairwise comparisons of the relative importance of indicators at the same level. Based on the expert assignments, a series of judgment matrices were constructed. The Analytic Hierarchy Process (AHP) was used, and specialized software (such as yaahp) was employed to process the judgment matrices, calculating and determining the global weights from the first-level indicators to the third-level indicators. A consistency check was performed on the weight allocation results (requiring a consistency ratio CR < 0.1) to ensure the scientific validity and logical rationality of the weights. The final weight allocation results are shown in Table 1.

[0069] Table 1. Multi-dimensional Comprehensive Evaluation Index System and Weights for Tree Canopy Cover Benefits

[0070] In S2, during the construction of the pre-trained canopy benefit assessment model, 500 sample units (which can be single trees or 50-meter-long homogeneous street sections) are randomly selected from the area where data has been collected. A group of five experts in landscaping, ecology, and urban planning is organized to independently score the ecological, social, and comprehensive benefits of each sample unit on a 100-point scale, based on the multi-dimensional comprehensive evaluation index system for canopy cover benefits (17 tertiary indicators for ecology and society) constructed in this scheme. The final average score is taken as the label value of that sample.

[0071] A feature vector was constructed for each sample unit, containing all corrected and extracted metrics (such as 3D green volume, corrected photosynthetic intensity, standardized green visibility, cooling effect, humidification effect, dust retention efficiency, etc., totaling more than 50 features). The sample library was randomly divided into a training set (350 samples) and a test set (150 samples) in a 7:3 ratio. The XGBoost algorithm was used to train the model using the training set, and the model hyperparameters (such as learning rate, maximum tree depth, subsampling ratio, etc.) were optimized through grid search and 5-fold cross-validation.

[0072] Validate the model performance on the test set. Use root mean square error and coefficient of determination as evaluation metrics. For example, a well-trained model should have an RMSE of less than 8 and an R² of more than 0.85 for its overall benefit prediction score.

[0073] For a new region to be evaluated, after completing all data processing in the first stage, a feature vector is generated for each evaluation unit and input into the trained XGBoost model. The model will directly output the ecological benefit score, social benefit score, and comprehensive benefit score (0-100 points) for that unit.

[0074] Furthermore, in S2, the canopy benefit assessment model outputs sub-scores for both ecological and social benefits. The generation of the optimization strategy considers the balance between the overall benefit score and the ecological and social benefit sub-scores to generate an optimization scheme that promotes synergistic improvement. Specifically, this includes the following operational steps: Step 1, Benefit Status Diagnosis Step: Obtain the Ecological Benefit Sub-Score (ES) and the Social Benefit Sub-Score (SS), calculate the absolute value of the difference between them, and calculate a Benefit Balance Index (BI) based on the ratio of this absolute value to the sum of the two scores; the formula for calculating the Benefit Balance Index (BI) is as follows: , where both ES and SS are greater than zero.

[0075] Step 2, State Type Determination Step: Compare the Ecological Benefit Sub-Score (ES) and Social Benefit Sub-Score (SS) with preset thresholds to classify the evaluation unit into multiple preset benefit state types. The benefit state types include at least: Ecological Dominant Type, determined when ES is greater than the sum of SS and the first preset threshold; Social Dominant Type, determined when SS is greater than the sum of ES and the second preset threshold; Low-Score Balanced Type, determined when the comprehensive benefit score is lower than the third preset threshold and BI is higher than the fourth preset threshold.

[0076] Step 3, Collaborative Strategy Matching Step: The identified benefit constraint factors and benefit state types are used together as query conditions to match the corresponding optimization strategies from the optimization strategy knowledge base.

[0077] Step 4, Scheme Generation Step: Based on the matched optimization strategy, generate an optimization scheme that aims to improve the overall benefit score while promoting the balance between the ecological benefit sub-score and the social benefit sub-score.

[0078] Specifically, a comprehensive priority score is calculated for each matched optimization strategy. The formula for this score is as follows: Overall priority score = Basic priority × W1 + Expected improvement in ecological benefits × W2 + Expected improvement in social benefits × W3 + Expected improvement in balance × W4; Where W1, W2, W3, and W4 are preset weight coefficients; Furthermore, the values ​​of the weighting coefficients W2, W3, and W4 are dynamically adjusted based on the benefit status type of the current evaluation unit.

[0079] The specific method of dynamic adjustment is as follows: When the type is determined to be ecologically dominant, the weighting coefficient W3 for the expected improvement in social benefits and the weighting coefficient W4 for the expected improvement in balance are increased. When the model is determined to be socially dominant, the weighting coefficient W2 for the expected improvement in ecological benefits and the weighting coefficient W4 for the expected improvement in balance are increased.

[0080] This embodiment provides a method for evaluating and optimizing the benefits of street tree canopy cover. Compared to Embodiment 1, this method offers higher decision-making accuracy and can consider benefits from multiple perspectives. The key points are: The performance ceiling of machine learning models is highly dependent on the quality of the training data. This approach uses a specially constructed multi-dimensional comprehensive evaluation index system for tree canopy cover benefits. The training samples for the model are independently scored back-to-back based on all multi-source fusion data (rather than partial information), and the average value is used as the label. This process minimizes subjective bias and creates an authoritative, consistent, and reliable "gold standard" dataset. This ensures that the subsequently trained model learns relatively objective benefit judgment rules that integrate multidisciplinary wisdom.

[0081] Furthermore, traditional assessments may only focus on single dimensions such as green coverage or tree survival rate. This system, through 2 primary indicators, 10 secondary indicators, and 17 tertiary indicators, systematically covers the core values ​​of concern to all sectors of ecology and society, from carbon sequestration and oxygen release, air purification, to promoting public health and enhancing urban aesthetics. This ensures that the model's training dimensions are comprehensive and accurate, and the final assessment results can truly reflect the comprehensive value of street trees as a key urban infrastructure, rather than their single function.

[0082] Furthermore, because the model learns the complex relationship between this comprehensive indicator system and the final benefit score, its internal logic is consistent with this system when making predictions or diagnoses. This makes the model's output (whether a score or a diagnose of limiting factors) logically traceable and interpretable. For example, when the model diagnoses "green visibility rate" as a limiting factor, we can clearly find its corresponding relationship in the indicator system and the social benefit dimension, thus enabling decision-makers to understand and trust the model's judgment.

[0083] Second, this scheme specifically considers the balance between the overall benefit score and the sub-scores of ecological and social benefits. Traditional optimization might only improve the lowest-scoring items identified in the assessment. However, this scheme, by introducing "benefit status type" and "balance index," can intelligently identify the "uneven development" problem in the region. For example, it can not only identify "weak carbon sequestration capacity" but also determine that the street has "excessive social functions and insufficient ecological functions," thus prescribing a composite prescription of "maintaining existing recreational functions while specifically enhancing ecological functions," avoiding the robbing Peter to pay Paul approach during the optimization process.

[0084] The above descriptions are merely embodiments of the present invention. Commonly known structures and characteristics of the solutions are not described in detail here. Those skilled in the art are aware of all common technical knowledge in the field prior to the application date or priority date, are aware of all existing technologies in that field, and have the ability to apply conventional experimental methods prior to that date. Those skilled in the art can, under the guidance of this application, improve and implement this solution in combination with their own capabilities. Some typical known structures or methods should not be obstacles for those skilled in the art to implement this application. It should be noted that those skilled in the art can make several modifications and improvements without departing from the structure of the present invention. These should also be considered within the scope of protection of the present invention, and will not affect the effectiveness of the implementation of the present invention or the practicality of the patent.

Claims

1. A method for evaluating and optimizing the benefits of street tree canopy cover, characterized in that, Includes the following steps: S1. Collect multi-source data of roadside trees in the target area, wherein the multi-source data includes at least remote sensing data, near-ground motion monitoring data and Internet of Things sensor data; S2. The multi-source data is fused to extract feature parameters for evaluation, and the feature parameters are input into a pre-trained canopy benefit evaluation model. The canopy benefit evaluation model outputs a comprehensive benefit score of canopy coverage. The canopy benefit evaluation model is a machine learning model trained on sample data with expert rating labels. S3. Based on the comprehensive benefit score and the benefit constraint factors identified through model interpretability analysis, a targeted canopy optimization scheme is generated by matching from the pre-constructed optimization strategy knowledge base.

2. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, Between S1 and S2, it also includes: S101, Adaptive Correction Steps for Terrain and Road Features: Obtain the digital elevation model of the target area and calculate the terrain factor of the target area; simultaneously, calculate the road structure factor of the target area based on near-ground motion monitoring data; use the terrain factor and road structure factor to perform parameter correction on the original ecological benefit indicators extracted from the multi-source data to obtain the feature parameters after terrain and road feature correction. The primary ecological benefit indicators include the following secondary indicators: canopy cover benefits, soil and water conservation benefits, carbon sequestration and oxygen release capacity, air purification capacity, microclimate regulation function, and biodiversity support; Each secondary indicator has a tertiary indicator; among them, carbon sequestration and oxygen release capacity includes carbon storage increment and photosynthetic intensity; microclimate regulation function includes cooling effect and humidification effect.

3. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 2, characterized in that, The topographic factors include the theoretical solar radiation calculated based on the digital elevation model using a solar radiation transmission model, and the topographic humidity index calculated based on a hydrological analysis algorithm. The parameter correction of the original ecological benefit indicators includes: The photosynthetic intensity or cooling effect index is corrected based on the ratio of the actual location to the theoretical solar radiation on the horizontal plane. The humidification effect index is corrected according to the grade coefficient determined by the topographic humidity index.

4. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, In S1: The remote sensing data includes lidar point cloud data and multispectral image data, which are used to extract single-tree structure parameters, digital elevation models and canopy coverage. The near-ground mobile monitoring data is collected by a ground-based mobile platform equipped with lidar and imaging equipment, and is used to extract the three-dimensional green volume of individual trees, canopy structure parameters, and street green view rate. The IoT sensor data is collected through miniature weather stations and air quality sensors deployed in the target area to obtain continuous time-series data on photosynthetically active radiation, temperature and humidity gradients, and pollutant concentrations.

5. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, The extracted feature parameters for evaluation include: From lidar point cloud data, individual trees are separated using a hierarchical clustering algorithm, and the three-dimensional green content of individual trees is calculated using a convex hull algorithm or a voxelization model. From street view image data, vegetation pixels are identified through a semantic segmentation model, and the green view rate of each road segment is automatically calculated; Environmental fluctuation characteristic parameters, including daily average, peak value, and volatility, are extracted from the time-series data of IoT sensor data.

6. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, The canopy benefit evaluation model was trained using the gradient boosting decision tree algorithm. Furthermore, the identification of benefit constraint factors through model interpretability analysis refers to: using the SHAP analysis framework to calculate the contribution of each feature parameter to the comprehensive benefit score, and identifying the top N feature parameters with negative contribution and the largest absolute value as core constraint factors, where N is a natural number.

7. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, The optimization strategy knowledge base is a structured database constructed in the form of constraint factors-optimization strategy rules; In step S3, the identified benefit constraint factors are matched with the rules in the optimization strategy knowledge base to generate a list of optimization schemes containing one or more optimization strategies and their priorities.

8. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, The optimization strategy includes at least one of the following: Tree shaping and pruning strategies for addressing canopy structure problems; Strategies for replanting or mixed planting to address the problem of single-function tree species; Strategies for improving the permeability of tree pits to address site condition issues; Strategies for configuring forest understory space and facilities to address social benefits.

9. The method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, Following S3, it also includes: S4. Dynamic Feedback and Model Optimization Steps: After implementing the canopy optimization scheme, re-execute steps S1 and S2 to obtain optimized canopy coverage benefit data. Use the data before and after optimization as new samples to incrementally learn the canopy benefit evaluation model in order to update the model parameters.

10. A method for evaluating and optimizing the benefits of street tree canopy cover according to claim 1, characterized in that, In S2, the canopy benefit assessment model outputs sub-scores for two dimensions: ecological benefits and social benefits. The optimization strategy is generated by considering the balance between the comprehensive benefit score and the ecological and social benefit sub-scores, in order to generate an optimization scheme that promotes synergistic improvement.

Images