Lidar-based tea garden shading condition quantification method and shading tree species screening method

Abstract

The application discloses a kind of tea garden shade condition quantification method and shade tree species screening method based on laser radar, the method obtains the three-dimensional point cloud data of tea garden shade tree and tea tree by handheld laser radar, extracts tree leaf area index, porosity, tree height, breast height diameter, crown diameter and other structure parameters after pretreatment;Based on principal component analysis method, the parameter dimension is reduced to construct shade index, realize the quantitative evaluation of shade effect;And with tree structure parameters as input, tea tree growth condition index as output, train machine learning regression model to predict tea tree growth condition.Furthermore, by comparing the shade index or predicted growth index corresponding to different tree species, the optimal shade tree species suitable for tea garden is screened out.The present application effectively overcomes the defects of strong subjectivity of traditional manual evaluation and lack of targeted general remote sensing index, and provides an efficient and quantifiable technical means for precise configuration and intelligent management of shade forest in northern tea garden.

Description

Technical Field

[0001] This invention relates to the field of agricultural remote sensing and smart tea garden management technology, specifically to a method for constructing a tea garden shading index and selecting shading tree species based on handheld lidar. Background Technology

[0002] The shading conditions in tea gardens directly affect the stability of the canopy light environment and microclimate, significantly impacting tea quality. Real-time, accurate monitoring and quantification of shading levels are crucial for optimizing the tea tree growing environment and improving tea garden management efficiency and decision-making accuracy. Appropriate shading helps alleviate strong light stress, stabilize canopy leaf temperature, increase amino acid content in tea leaves, and optimize the phenol-to-amino acid ratio, which is of significant practical importance for promoting high-quality, efficient, and ecological tea production.

[0003] Traditional tea garden shading monitoring mainly relies on manual visual inspection, fixed-point measurement with lux meters, estimation using canopy analyzers, and multispectral remote sensing inversion. These methods generally suffer from problems such as over-reliance on manual labor, low efficiency, limited accuracy, and strong subjectivity. Therefore, more and more scholars are beginning to use remote sensing indices to monitor changes in regional ecosystems. For example, Wu Jiaming's team proposed the Shrub Structure Index (SSI) based on UAV lidar data, successfully achieving high-precision mapping of individual shrub biomass; Manon Collard et al. constructed the Index of Biodiversity Potential (IBP) using airborne lidar to assess forest biodiversity; and Li Zhipeng et al. established the Nitrogen Nutrition Index (NNI) for tea trees using UAV multi-sensor imaging technology, achieving accurate diagnosis of nitrogen nutrition status in tea gardens.

[0004] However, current tea garden shading monitoring still faces the following problems: Existing remote sensing indices are mostly general indicators, failing to be specifically optimized for the unique "tree-tea tree" dual-canopy structure of tea gardens and the shading needs under northern temperate climate conditions. For example, the Shrub Structure Index (SSI) focuses on assessing the biomass of individual shrubs, the Biodiversity Potential Index (IBP) is geared towards forest biodiversity assessment, and the Tea Tree Nitrogen Nutrition Index (NNI) focuses on the nutritional status of the tea tree itself. None of these indices involve the quantitative relationship between the three-dimensional canopy structure of shading trees and the growth response of tea trees under the forest canopy, making it difficult to directly quantify the shading effect of tea gardens. Furthermore, traditional shading assessments rely on manual experience and lack quantifiable and reproducible objective evaluation standards, leading to significant subjectivity and uncertainty in the selection and configuration of shading tree species.

[0005] Therefore, there is an urgent need to develop a technical method that can accurately acquire the three-dimensional canopy structure of tea gardens, efficiently quantify shading conditions, and support the scientific selection of shading tree species, so as to make up for the shortcomings of existing technologies and improve the efficiency and precision of tea garden management. Summary of the Invention

[0006] In view of this, the present invention provides a method for quantifying the shading status of tea gardens based on lidar, specifically by constructing a shading index as a quantitative evaluation indicator and establishing a method for screening shading tree species in tea gardens based on the shading index.

[0007] Firstly, a method for quantifying the shading status of tea gardens based on lidar includes the following steps:

[0008] Data acquisition steps: The tea garden area containing shade trees and tea trees is scanned using a lidar scanning system to obtain three-dimensional point cloud data of the tea garden area;

[0009] Data processing steps: The three-dimensional point cloud data is preprocessed, and based on the preprocessed point cloud data, multiple preset structural parameters of the shade trees are extracted;

[0010] Shading index construction steps: Based on principal component analysis, multiple preset structural parameters of the extracted shading trees are subjected to dimensionality reduction processing to generate a comprehensive shading index. The shading index is used to quantitatively evaluate the shading effect of the shading trees on the tea garden. The principal component analysis is based on the correlation between the multiple preset structural parameters and the indicators characterizing the growth status of the tea trees to determine the weight contribution of each preset structural parameter in the shading index.

[0011] Evaluation model establishment steps: Using multiple preset structural parameters of the shade trees as input features and indicators representing the growth status of tea trees as output, a machine learning regression model is trained and established to predict the growth status of tea trees based on the structural parameters of the shade trees.

[0012] Specifically, the preset structural parameters of the shade tree include: tree leaf area index, tree porosity, tree height, diameter at breast height, crown diameter, north-south crown diameter, east-west crown diameter, crown area, crown volume, and height under branches, or a combination thereof.

[0013] Specifically, the shading index construction step further includes:

[0014] The extracted pre-defined structural parameters of the shade trees are subjected to data standardization processing.

[0015] Principal component analysis was performed on the standardized data to obtain multiple principal components and their corresponding eigenvalues, variance contribution rates, and loading coefficients.

[0016] Select the first few principal components with eigenvalues ​​greater than 1 and / or whose cumulative variance contribution rate reaches a preset threshold to construct the shading index;

[0017] Based on the loading coefficients and eigenvalues ​​of the selected principal components, the linear combination coefficients of each preset structural parameter are calculated, and the comprehensive score coefficient is further calculated by combining the variance contribution rate.

[0018] Based on the comprehensive score coefficient, the standardized values ​​of each preset structural parameter are weighted and summed to generate the shading index.

[0019] Specifically, the shading index is composed of a weighted sum of multiple preset structural parameter variables of the shading tree and their corresponding weighting coefficients, and its expression is as follows:

[0020] Shading index = A1X1 + A2X2 + A3X3 + ... + A n X n

[0021] Among them, X1, X2, X3...X n A1, A2, A3...A1 represent multiple preset structural parameter variables of the shade tree. n These are the weighting coefficients that correspond one-to-one with each structural parameter variable.

[0022] Specifically, the weighting coefficients A1, A2, A3...A n The comprehensive score coefficient is obtained by normalizing the coefficient.

[0023] Specifically, the machine learning regression model is one of LightGBM, Xgboost, Random Forest, Gaussian Process Regression, or Support Vector Machine.

[0024] Specifically, the evaluation model establishment steps also include:

[0025] The Optuna hyperparameter optimization framework is used to tune the hyperparameters of the selected machine learning regression model in order to obtain the optimal combination of model parameters; wherein, the hyperparameter tuning aims to minimize the prediction error of the validation set.

[0026] Specifically, in the evaluation model establishment step, the coefficient of determination (R²) is used. 2 The predictive performance of the machine learning regression model is evaluated by at least one of the following metrics: mean absolute error (MAE), mean squared error (MSE), and root mean square error (RMSE).

[0027] Specifically, the index characterizing the growth status of tea trees is represented by the difference between the tea tree leaf area index and the tea tree porosity obtained after standardizing the tea tree point cloud data.

[0028] Secondly, a method for selecting shade tree species for tea gardens is also provided, including:

[0029] For multiple shade tree species to be screened, the quantification method described above is used to obtain the shade index of each shade tree species under the same or standardized conditions, and / or the machine learning regression model described above is used to predict the corresponding tea tree growth status index.

[0030] By comparing the shading index and / or predicted tea tree growth status indicators corresponding to each shading tree species to be screened, the tree species that result in the optimal tea tree growth status indicators or the shading index falling within the preset optimal range are determined as the preferred shading tree species.

[0031] The beneficial effects of this invention are:

[0032] First, addressing the limitation of existing general remote sensing indices in quantifying the shading effect of the "tree-tea tree" double canopy in tea gardens, this invention proposes a method for constructing a shading index based on the multi-dimensional three-dimensional structural parameters of shading trees. This method uses principal component analysis to uncover the inherent statistical correlation between parameters such as tree leaf area index, porosity, and canopy volume and the growth status of tea trees, compressing high-dimensional structural information into a single comprehensive index. This achieves efficient and objective quantification of shading intensity in tea gardens, filling the technical gap in shading assessment indicators for specific tea garden scenarios.

[0033] Secondly, this invention combines tree structure parameters obtained from lidar point clouds with a machine learning regression model to establish a predictive model from the structural features of shading trees to the growth response of tea trees, and improves the prediction accuracy through hyperparameter optimization. This technical solution effectively overcomes the problems of strong subjectivity and low efficiency in traditional manual visual assessment, elevating the assessment of shading effect from qualitative experience judgment to quantitative model prediction, significantly improving the objectivity and reproducibility of the assessment results.

[0034] Third, based on the aforementioned shading index and prediction model, this invention further provides a scientific screening method for shading tree species in tea gardens. It can pre-evaluate and prioritize different candidate tree species based on their shading effects, providing a data-driven decision-making basis for the precise configuration of shading forests in northern ecological tea gardens. This method has significant practical value for improving the intelligent management level of tea gardens and the quality of tea. Attached Figure Description

[0035] The present invention includes the following figures:

[0036] Figure 1 This is a flowchart illustrating the quantification of shading conditions and the selection of shading tree species based on the method of this invention. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be described in further detail below with reference to the accompanying drawings and specific embodiments.

[0038] It should be noted that, in order to meet the needs of ecological planting in northern tea gardens, common trees in places such as campuses and parks can be selected as candidate tree species for shading. This method can be used to complete the suitability screening and achieve the scientific configuration of shading forests in tea gardens.

[0039] This embodiment provides a method for quantifying the shading status of tea gardens and a method for selecting shading tree species based on handheld lidar, specifically including the following steps:

[0040] Step 1: Data Collection

[0041] In an artificial ecological tea garden at a tea plantation, a LiGrip H120 handheld lidar scanning system was used to collect data, obtaining three-dimensional point cloud data of the shade trees and tea plants. Simultaneously, to screen suitable shade tree species for northern tea gardens, point cloud data were also collected for 54 common northern shade tree species in locations such as the East Campus of Liaocheng University, Fenghuangyuan Park, and Wanghu Park.

[0042] Step 2: Data Processing

[0043] Insta360 Studio 2022 (Version: 4.4) software was used to perform video stitching and frame alignment on multi-view camera data obtained from the LiGrip H120, generating panoramic images corresponding to the point cloud space. Then, LiDAR360MLS was used to decode and colorize the LiDAR data and the processed camera data. The decoded point cloud data was preprocessed using LiDAR360 software, including cropping, noise reduction, ground point classification, and ground point-based normalization.

[0044] Based on the obtained normalized point cloud, individual tree attributes of shade trees were extracted using a seed point-based tree segmentation method, specifically including: tree height, diameter at breast height (DBH), crown diameter, north-south crown diameter, east-west crown diameter, crown area, crown volume, and branch height. Simultaneously, forestry community parameters were calculated based on the grid, yielding tree porosity and leaf area index. Furthermore, point cloud data of tea trees in the corresponding regions were extracted, and tea tree porosity and leaf area index were calculated.

[0045] It should be noted that the shading effect of shade trees on tea plants is influenced by a combination of various structural parameters, with varying degrees of correlation among these parameters, and different parameters contributing differently to tea plant growth. Directly using a single parameter or a simple linear combination for evaluation is insufficient to fully reflect the comprehensive regulatory effect of the three-dimensional canopy structure of the shade trees on the understory light environment. Therefore, this embodiment employs principal component analysis (PCA) to integrate 10 tree structural parameters into a single shading index while preserving the main information of the original variables. The weight contribution of each parameter in the index is determined based on the statistical correlation between each parameter and tea plant growth indicators (tea tree leaf area index, tea tree porosity), thereby achieving an objective quantification of the shading effect.

[0046] Step 3: Constructing the Shading Index of the Tea Garden

[0047] SPSS software was used to analyze the correlation between 10 structural parameter variables of shade trees and tea tree variables (tea tree porosity and tea tree leaf area index).

[0048] Principal component analysis (PCA) was used to integrate 10 variables related to trees into a tea garden shading index. To avoid weight imbalance caused by inconsistent dimensions, the 12 remote sensing variables were standardized before PCA. PCA was then performed to obtain the eigenvalues, variance contribution rates, and loading coefficients of each principal component, as shown in Tables 1 and 2, respectively.

[0049] Table 1. Eigenvalues ​​and variance contribution rates of the correlation matrix of tree variables

[0050]

[0051] Table 2 Principal component loadings for 10 tree variables

[0052]

[0053] Table 1 shows that the eigenvalues ​​of the first six principal components (Y1, Y2, Y3, Y4, Y5, Y6) are all greater than 1, with a cumulative contribution rate as high as 94.654%, reflecting most of the information of the 10 tree variables. Therefore, Y1 to Y6 are used to construct the tea garden shading index.

[0054] Based on the data in Tables 1 and 2, the linear combination coefficient (LCC) and comprehensive score coefficient (CSC) are calculated using the following formulas to obtain the shading index.

[0055]

[0056]

[0057] In equation (1), LC ijλ represents the loading coefficients of the j-th principal components and the i-th index, and λ represents the initial eigenvalues ​​of j-th. In equation (2), n represents the number of variables participating in the principal component analysis, VC represents the variance contribution rate, and CVC represents the cumulative variance contribution rate.

[0058] Based on the data in Tables 1 and 2, the LCC is calculated using formula (1), yielding the linear composite expression for Y1, Y2, Y3, Y4, Y5, and Y6:

[0059]

[0060]

[0061]

[0062]

[0063]

[0064]

[0065] Then, using formula (2) to calculate CSC, we obtain the formula for calculating the principal component (Y):

[0066]

[0067] Finally, each coefficient of the Y equation is normalized using a percentage-based method. The final formula for the tea garden shading index is as follows:

[0068]

[0069]

[0070] Where X1 is the tree leaf area index, X2 is the tree porosity, X3 is the tree height, X4 is the diameter at breast height (DBH), X5 is the crown diameter, X6 is the north-south crown diameter, X7 is the east-west crown diameter, X8 is the crown area, X9 is the crown volume, and X... 10 It is high below the branch.

[0071] Step 4: Model Building

[0072] Data from three tea plantation surveys were selected, comprising 18 sample plots, 828 tree data points, and corresponding tea tree growth data. The tea tree growth status index was represented by the difference between the standardized tea tree leaf area index and the tea tree porosity. The shading tree dataset was used as input features, and the tea tree growth status dataset was used as output, with the training and test sets divided in a 3:1 ratio.

[0073] Five machine learning regression models—Lightweight Gradient Boosting Machine (LightGBM), Extreme Gradient Boosting (Xgboost), Random Forest (RF), Gaussian Process Regression (GPR), and Support Vector Machine (SVM)—were used to establish prediction models for tea tree growth. LightGBM, Xgboost, and RF did not require feature scaling; SVM and GPR required selecting the optimal kernel function based on data characteristics. The coefficients (R²) were then determined. 2 The mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE) were used as evaluation metrics to compare the predictive performance of the five models.

[0074] Step 5: Model Optimization

[0075] The Optuna hyperparameter optimization framework was used to systematically tune the five models mentioned above. Key hyperparameters to be optimized and their search space were defined for each model. For example, the search space for Xgboost included: learning rate 0.01–0.3, maximum depth 3–15, subsample size 0.6–1.0, column sampling 0.6–1.0, and number of estimators 100–1000; the corresponding hyperparameter ranges for other models were set similarly. Minimizing the root mean square error of the validation set was used as the optimization objective.

[0076] Using Optuna's Tree-based Parzen Estimator (TPE) sampling algorithm, 50 sequential parameter searches and evaluations are performed on each model to automatically find the approximate optimal hyperparameter combination for each model on the validation set. In this embodiment, the optimized hyperparameters of Xgboost are: learning rate 0.087, maximum depth 8, subsamples 0.85, column sampling 0.79, and number of estimators 520. The final prediction model is constructed using this optimal parameter combination.

[0077] Step Six: Model Evaluation and Tree Species Selection

[0078] Using R 2 The predictive performance of the optimized model on tea tree growth status was evaluated using four indicators: MAE, MSE, and RMSE. The results showed that the Xgboost model exhibited the best overall predictive performance, with a test set determination coefficient R0. 2 The model's accuracy of 0.787 is the highest among all models, while its MSE (0.051) and RMSE (0.226) are the lowest, indicating that this model has the strongest explanatory power for the variation in tea tree growth status and the smallest prediction bias.

[0079] For 54 common shading tree species in northern China, the structural parameters extracted from their lidar point cloud data were input into an optimized Xgboost model to predict the corresponding tea tree growth indicators under shading conditions for each species. By comparing the predicted indicators for different species, the species that produce the best tea tree growth indicators or whose shading index falls within a preset optimal range are identified as preferred shading tree species, thus achieving scientific selection of shading tree species for tea gardens. For example, the range corresponding to the top 10% of species in terms of shading index can be used as the preset optimal range, or specific thresholds can be set based on actual tea garden management experience.

[0080] The above embodiments have provided a detailed description of the technical solutions of the present invention. Obviously, the present invention is not limited to the described embodiments. Based on the embodiments of the present invention, those skilled in the art can make various changes, but any changes that are equivalent or similar to the present invention fall within the scope of protection of the present invention. Contents not described in detail in this specification are prior art known to those skilled in the art.

Claims

1. A method for quantifying the shading status of tea gardens based on lidar, characterized in that, Includes the following steps: Data acquisition steps: The tea garden area containing shade trees and tea trees is scanned using a lidar scanning system to obtain three-dimensional point cloud data of the tea garden area; Data processing steps: The three-dimensional point cloud data is preprocessed, and based on the preprocessed point cloud data, multiple preset structural parameters of the shade trees are extracted. Shading index construction steps: Based on principal component analysis, multiple preset structural parameters of the extracted shading trees are subjected to dimensionality reduction processing to generate a comprehensive shading index. The shading index is used to quantitatively evaluate the shading effect of the shading trees on the tea garden. The principal component analysis is based on the correlation between the multiple preset structural parameters and the indicators characterizing the growth status of the tea trees to determine the weight contribution of each preset structural parameter in the shading index. Evaluation model establishment steps: Using multiple preset structural parameters of the shade trees as input features and indicators representing the growth status of tea trees as output, a machine learning regression model is trained and established to predict the growth status of tea trees based on the structural parameters of the shade trees.

2. The method according to claim 1, characterized in that, The preset structural parameters of the shade tree include: tree leaf area index, tree porosity, tree height, diameter at breast height, crown diameter, north-south crown diameter, east-west crown diameter, crown area, crown volume, and height under branches, or a combination thereof.

3. The method according to claim 1, characterized in that, The step of constructing the shading index further includes: The extracted pre-defined structural parameters of the shade trees are subjected to data standardization processing. Principal component analysis was performed on the standardized data to obtain multiple principal components and their corresponding eigenvalues, variance contribution rates, and loading coefficients. Select the first few principal components with eigenvalues ​​greater than 1 and / or whose cumulative variance contribution rate reaches a preset threshold to construct the shading index; Based on the loading coefficients and eigenvalues ​​of the selected principal components, the linear combination coefficients of each preset structural parameter are calculated, and the comprehensive score coefficient is further calculated by combining the variance contribution rate. Based on the comprehensive score coefficient, the standardized values ​​of each preset structural parameter are weighted and summed to generate the shading index.

4. The method according to claim 3, characterized in that, The shading index is composed of a weighted sum of multiple preset structural parameter variables of the shading tree and their corresponding weight coefficients, and its expression is as follows: Shading index = A1X1 + A2X2 + A3X3 + ... + A n X n Among them, X1, X2, X3...X n A1, A2, A3...A1 represent multiple preset structural parameter variables of the shade tree. n These are the weighting coefficients that correspond one-to-one with each structural parameter variable.

5. The method according to claim 4, characterized in that, The weighting coefficients A1, A2, A3...A n The comprehensive score coefficient is obtained by normalizing the coefficient.

6. The method according to claim 1, characterized in that, The machine learning regression model is one of LightGBM, Xgboost, Random Forest, Gaussian process regression, or support vector machine.

7. The method according to claim 6, characterized in that, The steps for establishing the evaluation model also include: The Optuna hyperparameter optimization framework is used to tune the hyperparameters of the selected machine learning regression model in order to obtain the optimal combination of model parameters; wherein, the hyperparameter tuning aims to minimize the prediction error of the validation set.

8. The method according to claim 1, characterized in that, In the evaluation model establishment step, the coefficient of determination R is used. 2 The predictive performance of the machine learning regression model is evaluated by at least one of the following metrics: mean absolute error (MAE), mean square error (MSE), and root mean square error (RMSE).

9. The method according to claim 1, characterized in that, The index characterizing the growth status of tea trees is represented by the difference between the tea tree leaf area index and the tea tree porosity obtained after standardizing the tea tree point cloud data.

10. A method for selecting shade tree species for tea gardens, characterized in that, include: For multiple shade tree species to be screened, the method described in any one of claims 1 to 9 is used to obtain the shade index of each shade tree species under the same or standardized conditions, and / or to use a machine learning regression model to predict the corresponding tea tree growth status index. By comparing the shading index and / or predicted tea tree growth status indicators corresponding to each shading tree species to be screened, the tree species that result in the optimal tea tree growth status indicators or the shading index falling within the preset optimal range are determined as the preferred shading tree species.

Images