Analysis Method and System for the Growth Law of Tea Trees Based on UAV Monitoring

Abstract

The present invention relates to the technical field of image processing, and discloses a method and system for analyzing the growth law of tea trees based on unmanned aerial vehicle (UAV) monitoring, which are used to improve the accuracy of analyzing the growth law of tea trees based on UAV monitoring. The method includes: collecting multi-temporal images of the tea tree area corresponding to the position of the sample tea tree to obtain a time-series image set including visible light and near-infrared bands; performing geometric correction, radiometric correction and orthoimage generation processing on the time-series image set to obtain a processed image set, extracting the characteristics of single tea trees from the processed image set to obtain a growth index data set; performing time series decomposition and non-linear trend analysis on the growth index data set to obtain the dynamic growth characteristics of tea trees; performing multivariate statistical analysis and principal component analysis on the dynamic growth characteristics and the pre-acquired environmental factors to obtain quantitative description data of environmental impacts; performing spatio-temporal coupling analysis and law extraction on the quantitative description data of environmental impacts to obtain the growth law of tea trees.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular, to a method and system for analyzing the growth law of tea trees based on unmanned aerial vehicle (UAV) monitoring. Background Art

[0002] Existing methods for analyzing the growth law of tea trees mainly rely on traditional ground surveys and manual observations. These methods usually use means such as regular on-site sampling, measuring the morphological parameters of tea trees, and analyzing soil and climate data to study the growth characteristics of tea trees and environmental impacts. At the same time, some studies have begun to use satellite remote sensing technology to monitor the growth status of large-scale tea gardens, and evaluate the growth trend and yield prediction of tea trees by analyzing multi-spectral images.

[0003] However, these traditional methods have some limitations. Ground surveys are time-consuming and laborious, and it is difficult to achieve large-scale and high-frequency monitoring; although satellite remote sensing has a wide coverage, its spatial resolution is limited, and it is difficult to accurately capture the growth changes of individual tea trees. In addition, existing methods often have difficulty in effectively integrating multi-source data and lack in-depth analysis of the complex interaction relationships between tea tree growth and environmental factors. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a method and system for analyzing the growth law of tea trees based on UAV monitoring, so as to improve the accuracy of analyzing the growth law of tea trees based on UAV monitoring.

[0005] The present invention provides a method for analyzing the growth law of tea trees based on UAV monitoring, including: performing high-precision geocoding and multi-factor stratified sampling on preset tea tree distribution data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree positions and sample tea tree species; based on the sample tea tree distribution information, using a preset UAV cluster to perform multi-temporal image acquisition on the tea tree area corresponding to the sample tea tree positions to obtain a time series image set including visible light and near-infrared bands; performing geometric correction, radiometric correction and orthoimage generation processing on the time series image set to obtain a processed image set, and performing single tea tree feature extraction on the processed image set through an image segmentation algorithm to obtain a growth index data set; performing time series decomposition and non-linear trend analysis on the growth index data set to obtain tea tree growth dynamic characteristics; performing multivariate statistical analysis and principal component analysis on the growth dynamic characteristics and pre-acquired environmental factors to obtain quantitatively described environmental impact data; performing spatio-temporal coupling analysis and law extraction on the quantitatively described environmental impact data to obtain the growth law of tea trees, where the growth law of tea trees includes annual growth cycle, radial growth rate, key growth period time nodes and the degree of environmental factor influence.

[0006] In the present invention, the steps of performing high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree locations and sample tea tree species, include: performing spatial interpolation and density analysis on the tea tree distribution data to obtain a tea tree distribution heat map; performing terrain overlay and slope aspect analysis on the tea tree distribution heat map to obtain a terrain-weighted tea tree distribution map; based on the terrain-weighted tea tree distribution map, performing spatial clustering on tea trees through the K-means clustering algorithm to obtain tea tree spatial distribution clusters; performing boundary extraction and shape analysis on the tea tree spatial distribution clusters to obtain tea garden contour data; based on the tea garden contour data, performing zonal sampling on the tea garden through the stratified random sampling algorithm to obtain an initial sample point set; performing spatial autocorrelation analysis and anomaly detection on the initial sample point set to obtain an optimized sample point set; based on the optimized sample point set, inferring the spatial distribution of tea tree species through the nearest neighbor interpolation algorithm to obtain a tea tree species distribution map; performing spatial frequency analysis and boundary enhancement on the tea tree species distribution map to obtain a high-precision tea tree species distribution map; based on the high-precision tea tree species distribution map, performing on-site calibration of sample points through differential GPS positioning to obtain calibrated sample tea tree location data; performing attribute association and data fusion on the calibrated sample tea tree location data to obtain the sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree locations and sample tea tree species.

[0007] In the present invention, the steps of, based on the sample tea tree distribution information, performing multi-temporal image acquisition on the tea tree area corresponding to the sample tea tree locations through a preset unmanned aerial vehicle (UAV) cluster to obtain a time-series image set including visible light and near-infrared bands, include: performing spatial clustering analysis on the sample tea tree distribution information to obtain tea tree area sub-blocks; based on the tea tree area sub-blocks, optimizing and planning the UAV flight paths through the ant colony algorithm to obtain optimized flight paths; performing time window division and task allocation on the optimized flight paths to obtain a UAV cluster collaborative operation plan; based on the UAV cluster collaborative operation plan, performing dynamic scheduling and task distribution on the UAV cluster to obtain real-time flight state data; performing attitude solution and position correction on the real-time flight state data to obtain accurate flight trajectories; based on the accurate flight trajectories, performing real-time adjustment of imaging parameters through an adaptive exposure algorithm to obtain spectrally balanced images; performing multi-scale fusion and denoising processing on the spectrally balanced images to obtain enhanced images; based on the enhanced images, performing semantic segmentation on the image content through a deep learning algorithm to obtain tea tree area masks; performing morphological processing and boundary refinement on the tea tree area masks to obtain accurate tea tree contours; based on the accurate tea tree contours, performing registration and time-series stacking on the multi-temporal images to obtain a time-series image set including visible light and near-infrared bands.

[0008] In the present invention, the steps of performing geometric correction, radiometric correction, and orthophoto generation processing on the temporal image set to obtain a processed image set, and extracting single tea plant feature from the processed image set by an image segmentation algorithm to obtain a growth index data set include: performing multi-point control point matching and geometric transformation on the temporal image set to obtain a preliminary corrected image; based on the preliminary corrected image, performing radiometric correction on the image by an adaptive histogram equalization algorithm to obtain a radiometrically balanced image; performing digital elevation model fusion and orthographic projection on the radiometrically balanced image to obtain an orthophoto; based on the orthophoto, performing preliminary segmentation on the image by a multi-scale watershed algorithm to obtain a candidate area of the tea plant contour; performing morphological operations and boundary optimization on the candidate area of the tea plant contour to obtain an accurate tea plant contour; based on the accurate tea plant contour, segmenting a single tea plant by a region growing algorithm to obtain a mask of the single tea plant; extracting texture features and performing statistical analysis on the mask of the single tea plant to obtain a descriptor of the tea plant canopy characteristics; based on the descriptor of the tea plant canopy characteristics, classifying the health status of the tea plant by a support vector machine algorithm to obtain a tea plant health index; performing time series analysis and anomaly detection on the tea plant health index to obtain a growth anomaly marker; based on the growth anomaly marker, performing multi-dimensional calculation and normalization processing on the tea plant growth index to obtain a growth index data set.

[0009] In the present invention, the steps of performing time series decomposition and non-linear trend analysis on the growth index data set to obtain the dynamic characteristics of tea plant growth include: performing missing value imputation and outlier detection on the growth index data set to obtain a preprocessed data set; based on the preprocessed data set, performing multi-scale decomposition on the time series by a wavelet transform algorithm to obtain a trend component, a seasonal component, and a residual component; performing non-parametric regression fitting on the trend component to obtain a long-term growth trend curve; based on the long-term growth trend curve, identifying growth stage transition points by a change point detection algorithm to obtain a growth stage division result; performing periodic analysis and spectral density estimation on the seasonal component to obtain a seasonal growth pattern; based on the seasonal growth pattern, modeling short-term growth fluctuations by an autoregressive integrated moving average model to obtain a short-term growth prediction model; performing non-linear time series analysis on the residual component to obtain non-linear characteristics of tea plant growth; based on the non-linear characteristics, constructing a dynamic growth model by a recursive neural network algorithm to obtain a dynamic prediction result of tea plant growth; performing sensitivity analysis and uncertainty quantification on the dynamic prediction result of tea plant growth to obtain a growth prediction confidence interval; based on the growth prediction confidence interval, performing multi-dimensional comprehensive evaluation and visualization on the dynamic characteristics of tea plant growth to obtain the dynamic characteristics of tea plant growth.

[0010] In the present invention, the step of performing multivariate statistical analysis and principal component analysis on the growth dynamic characteristics and the pre-acquired environmental factors to obtain quantitatively descriptive data of environmental impact includes: performing spatio-temporal registration and scale unification on the growth dynamic characteristics and the environmental factor data to obtain a collaborative analysis data set; based on the collaborative analysis data set, calculating a correlation matrix between environmental factors through the Pearson correlation coefficient to obtain a factor correlation evaluation result; performing hierarchical clustering analysis on the factor correlation evaluation result to obtain environmental factor groups; based on the environmental factor groups, performing multicollinearity diagnosis through the variance inflation factor method to obtain a screened environmental factor set; performing partial correlation analysis on the screened environmental factor set and the growth dynamic characteristics to obtain a ranking of direct impact factors; based on the ranking of direct impact factors, extracting main environmental impact components through the principal component analysis method to obtain a reduced-dimensional environmental feature space; performing orthogonal rotation optimization on the reduced-dimensional environmental feature space to obtain an environmental feature with enhanced interpretability; based on the environmental feature with enhanced interpretability, constructing an environment-growth relationship model through multiple linear regression to obtain environmental impact coefficients; performing Bootstrap resampling and confidence interval estimation on the environmental impact coefficients to obtain a robust evaluation result of environmental impact; based on the robust evaluation result of environmental impact, quantifying the influence degree of environmental factors through the fuzzy comprehensive evaluation method to obtain the quantitatively descriptive data of environmental impact.

[0011] In the present invention, the step of performing spatio-temporal coupling analysis and law extraction on the quantitatively described environmental impact data to obtain the tea tree growth law, where the tea tree growth law includes the annual growth cycle, radial growth rate, key growth period time nodes, and the degree of influence of environmental factors, includes: performing spatio-temporal interpolation on the quantitatively described environmental impact data to obtain a continuous environmental impact field; based on the continuous environmental impact field, calculating the local Moran index through spatial autocorrelation analysis to obtain the environmental impact spatial aggregation pattern; performing geographically weighted regression analysis on the environmental impact spatial aggregation pattern to obtain the spatial heterogeneity characteristics; based on the spatial heterogeneity characteristics, extracting the seasonal and periodic components of the environmental impact through time series decomposition to obtain the time variation pattern; performing wavelet coherence analysis on the time variation pattern to obtain the multi-scale spatio-temporal coupling characteristics; based on the multi-scale spatio-temporal coupling characteristics, performing temporal alignment of the tea tree growth and environmental factors through the dynamic time warping algorithm to obtain the synchronous change pattern; performing threshold segmentation and key time point identification on the synchronous change pattern to obtain the key growth period of the tea tree; based on the key growth period of the tea tree, constructing a non-linear mapping relationship through a radial basis function network to obtain the data on the influence relationship of environmental factors on growth; performing sensitivity analysis and uncertainty quantification on the influence relationship data to obtain the ranking of the degree of influence of environmental factors; based on the ranking of the degree of influence of environmental factors, constructing a tea tree growth-environment interaction network through the fuzzy cognitive map method to obtain the tea tree growth law, including the annual growth cycle, radial growth rate, key growth period time nodes, and the degree of influence of environmental factors.

[0012] The present invention also provides a tea tree growth law analysis system based on unmanned aerial vehicle monitoring, including:

[0013] A sampling module, configured to perform high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: the positions of the sample tea trees and the types of the sample tea trees;

[0014] An acquisition module, configured to, based on the sample tea tree distribution information, perform multi-temporal image acquisition on the tea tree area corresponding to the positions of the sample tea trees through a preset unmanned aerial vehicle cluster to obtain a time series image set including visible light and near-infrared bands;

[0015] An extraction module, configured to perform geometric correction, radiometric correction, and orthoimage generation processing on the time series image set to obtain a processed image set, and perform single tea tree feature extraction on the processed image set through an image segmentation algorithm to obtain a growth index data set;

[0016] An analysis module, configured to perform time series decomposition and non-linear trend analysis on the growth index data set to obtain the dynamic characteristics of tea tree growth;

[0017] A statistical module for performing multivariate statistical analysis and principal component analysis on the growth dynamic characteristics and pre-acquired environmental factors to obtain quantitative description data of environmental impacts;

[0018] A coupling module for performing spatio-temporal coupling analysis and law extraction on the quantitative description data of environmental impacts to obtain the growth law of tea trees, where the growth law of tea trees includes the annual growth cycle, radial growth rate, key growth period time nodes, and the degree of environmental factor influence.

[0019] In the technical solution provided by the present invention, through high-precision geocoding and multi-factor stratified sampling techniques, the representativeness of the sample tea trees and the balance of spatial distribution are ensured; the use of an unmanned aerial vehicle (UAV) cluster for multi-temporal and multi-spectral image acquisition greatly improves the efficiency and accuracy of data acquisition, and can capture the minute changes of individual tea trees; the adoption of advanced image processing techniques, including geometric correction, radiometric correction, and orthophoto generation, significantly improves the image quality and usability; through image segmentation algorithms, the accurate identification and feature extraction of individual tea trees are realized, providing high-quality input data for the quantitative analysis of growth indicators; by using time series decomposition and non-linear trend analysis methods, the dynamic characteristics of tea tree growth are deeply explored, revealing complex growth patterns; combining multivariate statistical analysis and principal component analysis, the impacts of environmental factors on tea tree growth are comprehensively evaluated, providing a quantitative description of environmental impacts; through innovative spatio-temporal coupling analysis and law extraction techniques, the key cycles, rate changes, and environmental response patterns of tea tree growth are successfully identified. From sample selection to final law extraction, by integrating multi-source data and multiple analysis techniques, a comprehensive tea tree growth-environment interaction network is successfully constructed. Based on the method of UAVs and advanced data analysis, the labor cost and time investment are greatly reduced, while the monitoring frequency and accuracy are improved. Description of the Drawings

[0020] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the description of the specific embodiments or the prior art. Obviously, the following drawings are some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0021] Figure 1 It is a flowchart of a method for analyzing the growth law of tea trees based on UAV monitoring in an embodiment of the present invention.

[0022] Figure 2 It is a flowchart of multi-temporal image acquisition of the tea tree area corresponding to the position of the sample tea trees by a pre-set UAV cluster in an embodiment of the present invention.

[0023] Figure 3Schematic diagram of a tea tree growth law analysis system based on UAV monitoring in an embodiment of the present invention.

[0024] Reference numerals:

[0025] 301, sampling module; 302, acquisition module; 303, extraction module; 304, analysis module; 305, statistics module; 306, coupling module. Detailed implementation manners

[0026] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings. Apparently, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0027] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings, and is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation of the present invention. In addition, the terms "first", "second", and "third" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance.

[0028] In addition, the technical features involved in different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0029] For ease of understanding, the specific process of the embodiment of the present invention will be described below. Please refer to Figure 1 , Figure 1 is a flowchart of a method for analyzing the growth law of tea trees based on UAV monitoring in an embodiment of the present invention. As shown in Figure 1 , it includes the following steps:

[0030] S101. Perform high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree positions and sample tea tree species;

[0031] S102. Based on the sample tea tree distribution information, use the preset UAV cluster to perform multi-temporal image acquisition on the tea tree area corresponding to the sample tea tree positions to obtain a time series image set including visible light and near-infrared bands;

[0032] S103. Perform geometric correction, radiometric correction, and orthoimage generation on the time-series image set to obtain a processed image set, and extract the characteristics of individual tea plants from the processed image set through an image segmentation algorithm to obtain a growth index data set;

[0033] S104. Perform time series decomposition and non-linear trend analysis on the growth index data set to obtain the dynamic characteristics of tea plant growth;

[0034] S105. Perform multivariate statistical analysis and principal component analysis on the dynamic growth characteristics and pre-acquired environmental factors to obtain quantitatively descriptive data on environmental impacts;

[0035] S106. Perform spatio-temporal coupling analysis and rule extraction on the quantitatively descriptive data on environmental impacts to obtain the growth rules of tea plants. The growth rules of tea plants include the annual growth cycle, radial growth rate, time nodes of key growth periods, and the degree of influence of environmental factors.

[0036] It should be noted that high-precision geocoding and multi-factor stratified sampling are performed on the pre-set tea plant distribution data. High-precision geocoding is the process of converting tea plant location information into precise geographical coordinates, while multi-factor stratified sampling takes into account multiple factors such as tea plant species and growth environment to ensure the representativeness of the samples. Through this step, distribution data containing the location and species information of sample tea plants is obtained. Based on these distribution information, a pre-set drone swarm is used to collect multi-temporal images of the selected tea plant areas. A drone swarm is a system composed of multiple drones working in coordination, which can efficiently cover large areas of tea gardens. Multi-temporal acquisition means repeating the shooting of the same area at different time points to capture the dynamic changes in tea plant growth. The acquired images include visible light and near-infrared bands, which are particularly important for vegetation analysis.

[0037] The obtained time-series image set is then subjected to a series of processes. Geometric correction is used to correct image distortion caused by lens distortion or flight attitude; radiometric correction adjusts the brightness and contrast of the image to eliminate the influence of atmospheric and lighting conditions; orthoimage generation converts the oblique aerial images into vertical overhead map projections. These steps ensure the geometric accuracy and radiometric quality of the images. Then, through an image segmentation algorithm, such as a deep learning-based semantic segmentation method, the characteristics of individual tea plants are extracted from the processed images to generate a data set containing indicators such as crown area and vegetation index. Time series decomposition and non-linear trend analysis are performed on the generated growth index data set. Time series decomposition splits the data into trend, seasonal, and random components, while non-linear trend analysis can capture the complex patterns in tea plant growth. This step reveals the dynamic characteristics of tea plant growth, such as the changing trend and periodic fluctuations of the growth rate.

[0038] Subsequently, the obtained growth dynamic characteristics are subjected to multivariate statistical analysis and principal component analysis together with the pre-acquired environmental factor data (such as temperature, precipitation, soil pH value, etc.). Multivariate statistical analysis evaluates the influence degree of each environmental factor on the growth of tea trees, while principal component analysis reduces the data dimension and highlights the most critical environmental influencing factors. This step generates quantitative description data of environmental impacts. Spatiotemporal coupling analysis and law extraction are performed on the quantitative description data of environmental impacts. Spatiotemporal coupling analysis considers how the changes of environmental factors in time and space affect the growth of tea trees, while law extraction summarizes the key characteristics of tea tree growth. Through this step, comprehensive tea tree growth laws are obtained, including annual growth cycle, radial growth rate, time nodes of key growth periods, and the influence degree of environmental factors.

[0039] For example, first, 5000 tea trees were precisely geocoded, and stratified sampling was carried out based on altitude, slope aspect, and tea tree variety, and 500 representative samples were selected. Using a cluster of 5 drones, multi-temporal shooting of the sample tea trees was carried out in the spring, summer, and autumn seasons, and 1500 high-resolution images were obtained. After image processing and segmentation, 10 growth indicators such as the crown area and leaf density of each tea tree were extracted. Time series analysis shows that tea trees grow most rapidly from April to June, and the average radial growth rate reaches 0.5 mm / week. Multivariate analysis indicates that temperature and soil moisture are the two most significant environmental factors affecting the growth of tea trees. Among them, for every 1°C increase in temperature, the growth rate of tea trees increases by about 5%.

[0040] By implementing the above steps, through high-precision geocoding and multi-factor stratified sampling techniques, the representativeness of the sample tea trees and the balance of spatial distribution are ensured; the use of a drone cluster for multi-temporal and multi-spectral image acquisition greatly improves the efficiency and accuracy of data acquisition and can capture the subtle changes of individual tea trees; the adoption of advanced image processing techniques, including geometric correction, radiometric correction, and orthophoto generation, significantly improves the image quality and usability; through image segmentation algorithms, the precise identification and feature extraction of individual tea trees are realized, providing high-quality input data for the quantitative analysis of growth indicators; the use of time series decomposition and non-linear trend analysis methods deeply explores the dynamic characteristics of tea tree growth and reveals complex growth patterns; combined with multivariate statistical analysis and principal component analysis, the influence of environmental factors on tea tree growth is comprehensively evaluated, providing a quantitative description of environmental impacts; through innovative spatiotemporal coupling analysis and law extraction techniques, the key cycles, rate changes, and environmental response patterns of tea tree growth are successfully identified. From sample selection to final law extraction, by integrating multi-source data and various analysis techniques, a comprehensive tea tree growth-environment interaction network is successfully constructed. Based on the methods of drones and advanced data analysis, the labor cost and time investment are greatly reduced, while the monitoring frequency and accuracy are improved.

[0041] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0042] (1) Perform spatial interpolation and density analysis on the tea tree distribution data to obtain a tea tree distribution heat map;

[0043] (2) Perform terrain overlay and slope aspect analysis on the tea tree distribution heat map to obtain a terrain-weighted tea tree distribution map;

[0044] (3) Based on the terrain-weighted tea tree distribution map, perform spatial clustering on the tea trees through the K-means clustering algorithm to obtain tea tree spatial distribution clusters;

[0045] (4) Perform boundary extraction and shape analysis on the tea tree spatial distribution clusters to obtain tea garden contour data;

[0046] (5) Based on the tea garden contour data, perform stratified random sampling on the tea garden to obtain an initial sample point set;

[0047] (6) Perform spatial autocorrelation analysis and outlier detection on the initial sample point set to obtain an optimized sample point set;

[0048] (7) Based on the optimized sample point set, perform spatial distribution inference on the tea tree species through the nearest neighbor interpolation algorithm to obtain a tea tree species distribution map;

[0049] (8) Perform spatial frequency analysis and boundary enhancement on the tea tree species distribution map to obtain a high-precision tea tree species distribution map;

[0050] (9) Based on the high-precision tea tree species distribution map, perform on-site calibration of the sample points through differential GPS positioning to obtain calibrated sample tea tree position data;

[0051] (10) Perform attribute association and data fusion on the calibrated sample tea tree position data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree positions and sample tea tree species.

[0052] Specifically, perform spatial interpolation and density analysis on the tea tree distribution data to generate a tea tree distribution heat map. Spatial interpolation uses the data of known points to estimate the values of unknown points, and common methods include inverse distance weighting method and Kriging method. Density analysis calculates the number or distribution density of tea trees per unit area. Through these two methods, the discrete tea tree position data is converted into a continuous distribution heat map, visually showing the tea tree dense areas. Overlay the tea tree distribution heat map with terrain data and perform slope aspect analysis. Terrain overlay considers the influence of factors such as altitude and slope on tea tree growth, while slope aspect analysis evaluates the differences in tea tree distribution on slopes in different directions. This step generates a terrain-weighted tea tree distribution map, which more accurately reflects the actual distribution of tea trees in complex terrains.

[0053] Based on the terrain-weighted tea tree distribution map, the K-means clustering algorithm is used to perform spatial clustering on tea trees. The K-means algorithm iteratively calculates, assigns data points to the nearest cluster center, and continuously updates the cluster center until convergence. This process divides tea trees into several spatially distributed clusters, and each cluster represents a relatively independent tea tree growth area. Boundary extraction and shape analysis are performed on the obtained tea tree spatial distribution clusters to obtain tea garden contour data. Edge detection algorithms such as the Canny algorithm are used for boundary extraction to identify the outer contour of the cluster. Shape analysis calculates features such as the area, perimeter, and circularity of the cluster to describe the geometric characteristics of the tea garden.

[0054] Based on the tea garden contour data, the tea garden is partitioned and sampled through the stratified random sampling algorithm. Stratified random sampling first divides the population into multiple layers and then performs simple random sampling within each layer. This method ensures the uniform distribution of samples in space and takes into account the characteristics of different regions to obtain the initial sample point set. Spatial autocorrelation analysis and outlier detection are performed on the initial sample point set. Methods such as Moran's I index are used for spatial autocorrelation analysis to evaluate the spatial dependence of sample points. Outlier detection identifies and removes outliers, such as points whose positions deviate significantly from the main distribution area. This step optimizes the sample point set and improves the representativeness.

[0055] Based on the optimized sample point set, the nearest neighbor interpolation algorithm is used to infer the spatial distribution of tea tree species. Nearest neighbor interpolation assigns the attribute value of the nearest known point to each unknown point and is suitable for spatial inference of discrete data such as tea tree species. This step generates a preliminary tea tree species distribution map. Spatial frequency analysis and boundary enhancement are performed on the tea tree species distribution map. Spatial frequency analysis evaluates the spatial distribution patterns of different tea tree species, while boundary enhancement uses image processing techniques such as the Sobel operator to highlight the boundaries between different tea tree species. This process generates a high-precision tea tree species distribution map.

[0056] Based on the high-precision tea tree species distribution map, the sample points are calibrated in the field through differential GPS positioning. Differential GPS provides centimeter-level positioning accuracy by using a combination of a reference station and a rover station. This step corrects the precise positions of the sample tea trees to obtain calibrated sample tea tree position data. Attribute association and data fusion are performed on the calibrated sample tea tree position data to obtain the final sample tea tree distribution information. Attribute association correlates the position data with attribute information such as tea tree species and age, and data fusion integrates multi-source data, such as remote sensing images and field survey data, to form complete sample tea tree distribution information.

[0057] For example, in a study of a certain tea garden, spatial interpolation and density analysis were first performed on 1,000 hectares of the tea garden to generate a heat map of tea tree distribution. The heat map shows that the tea tree density is the highest between 800 and 1,000 meters above sea level, reaching 300 plants per hectare. By overlaying the heat map with the digital elevation model, it was found that the tea tree density on the north-facing slope is 20% higher than that on the south-facing slope. The K-means algorithm (K = 5) was used to divide the tea garden into 5 main growth regions. Boundary extraction was performed on these 5 regions to obtain a tea garden contour with a total perimeter of 15 kilometers. Through stratified random sampling, 50 sample points were selected in each region, for a total of 250 initial sample points. Spatial autocorrelation analysis shows that the Moran's I index is 0.75, indicating that the sample points have strong spatial clustering. After removing 10 outliers, the nearest neighbor interpolation was used to infer the tea tree species distribution of the entire tea garden, and 3 main tea tree varieties were identified. After boundary enhancement processing, the generated high-precision distribution map clearly shows a 1-meter-wide transition zone at the variety junction. Through differential GPS field calibration, the positions of the sample points were accurate to the centimeter level. The final sample tea tree distribution information includes the precise coordinates, species, estimated age, etc. of each sample tea tree.

[0058] In a specific embodiment, as Figure 2 shown, the process of executing step S102 may specifically include the following steps:

[0059] S2001. Perform spatial clustering analysis on the sample tea tree distribution information to obtain tea tree regional sub-blocks;

[0060] S2002. Based on the tea tree regional sub-blocks, optimize and plan the UAV flight path through the ant colony algorithm to obtain an optimized flight path;

[0061] S2003. Perform time window division and task allocation on the optimized flight path to obtain a UAV cluster collaborative operation plan;

[0062] S2004. Based on the UAV cluster collaborative operation plan, perform dynamic scheduling and task distribution on the UAV cluster to obtain real-time flight status data;

[0063] S2005. Perform attitude solution and position correction on the real-time flight status data to obtain an accurate flight trajectory;

[0064] S2006. Based on the accurate flight trajectory, adjust the imaging parameters in real time through the adaptive exposure algorithm to obtain a spectral equilibrium image;

[0065] S2007. Perform multi-scale fusion and denoising processing on the spectral equilibrium image to obtain an enhanced image;

[0066] S2008. Based on the enhanced image, perform semantic segmentation on the image content through the deep learning algorithm to obtain a tea tree region mask;

[0067] S2009. Morphologically process the tea tree area mask and refine the boundary to obtain an accurate tea tree contour;

[0068] S2010. Based on the accurate tea tree contour, register and temporally stack multi-temporal images to obtain a set of temporal images containing visible and near-infrared bands.

[0069] Specifically, this solution first performs spatial clustering analysis on the sample tea tree distribution information to obtain sub-blocks of the tea tree area. The spatial clustering analysis uses the DBSCAN (Density-Based Spatial Clustering of Applications with Noise) algorithm, which divides data into different clusters based on the density of points and is suitable for processing spatially distributed data with irregular shapes. Through this step, the tea garden is divided into multiple sub-regions, facilitating subsequent UAV mission planning. Based on the obtained sub-blocks of the tea tree area, the ant colony algorithm is used to optimize the flight path of the UAV. The ant colony algorithm simulates the foraging behavior of ants and finds the optimal path through the accumulation and evaporation of pheromones. During this process, each sub-block is regarded as a node, and the algorithm iteratively optimizes the path for the UAV to visit all sub-blocks, finally obtaining an optimized flight path.

[0070] Perform time window partitioning and task allocation for the optimized flight path. Time window partitioning takes into account factors such as the battery life of the UAV and lighting conditions, and decomposes the entire flight mission into multiple time periods. Task allocation reasonably distributes the shooting tasks of sub-regions according to the performance parameters of each UAV, forming a cooperative operation plan for the UAV cluster. Based on the cooperative operation plan, dynamic scheduling and task distribution are performed for the UAV cluster. Dynamic scheduling uses the greedy algorithm to adjust UAV tasks according to real-time situations such as battery power and weather changes. Task distribution transmits specific flight instructions to each UAV through wireless communication and simultaneously receives the returned real-time flight status data.

[0071] Perform attitude solution and position correction on the received real-time flight status data. Attitude solution uses the Kalman filter algorithm to fuse sensor data such as gyroscopes and accelerometers to accurately calculate the pitch, roll, and yaw angles of the UAV. Position correction combines GPS and inertial navigation system data to correct the flight trajectory and obtain an accurate flight trajectory. Based on the accurate flight trajectory, the imaging parameters are adjusted in real time through the adaptive exposure algorithm. The adaptive exposure algorithm dynamically adjusts the exposure time, ISO sensitivity, and aperture size of the camera according to the scene brightness and contrast to ensure spectral balanced images under different lighting conditions.

[0072] Perform multi-scale fusion and denoising processing on the obtained spectral balanced images. Multi-scale fusion uses wavelet transform to effectively combine image information at different resolutions. Denoising processing adopts the non-local means algorithm to effectively reduce noise while retaining image details, obtaining enhanced images.

[0073] Based on the enhanced image, semantic segmentation of the image content is performed through deep learning algorithms. Using the U-Net convolutional neural network model, this model can accurately segment the tea tree area in the image through an encoder-decoder structure and skip connections, generating a tea tree area mask.

[0074] Morphological processing and boundary refinement are performed on the obtained tea tree area mask. Morphological processing includes opening and closing operations, which are used to remove noise and fill small holes. Boundary refinement uses an active contour model to accurately locate the tea tree contour boundary, obtaining an accurate tea tree contour.

[0075] Finally, based on the accurate tea tree contour, multi-temporal images are registered and temporally stacked. Image registration uses the SIFT feature matching algorithm to align images of different time phases to the same coordinate system. Temporal stacking then organizes the registered images in chronological order to form a temporal image set containing visible and near-infrared bands.

[0076] For example, in a 500-hectare tea garden, first, the tea tree distribution is divided into 20 sub-regions by the DBSCAN algorithm. The ant colony algorithm is used to optimize the paths of these 20 sub-regions. After 1000 iterations, an optimal path with a total flight distance of 50 kilometers is obtained. Considering the average flight time of 90 minutes for the unmanned aerial vehicle (UAV), the task is divided into 5 two-hour time windows and assigned to 5 UAVs for execution. During the flight, through dynamic scheduling, the situation where 2 UAVs returned early due to insufficient battery power was successfully handled. Attitude solution and position correction improved the flight trajectory accuracy to the centimeter level. The adaptive exposure algorithm maintained the consistency of image brightness under different lighting conditions, and the exposure time was dynamically adjusted between 1 / 1000 second and 1 / 60 second. Multi-scale fusion and denoising increased the signal-to-noise ratio of the image by 30%. The accuracy of deep learning semantic segmentation reached 95%, and approximately 1 million tea trees were identified. Morphological processing and boundary refinement improved the tea tree contour positioning accuracy to the pixel level. Finally, 5 time-phase images were registered through the SIFT algorithm, generating a temporal image set containing a total of 4 bands (RGB and NIR) and with a time span of one growing season.

[0077] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0078] (1) Perform multi-point control point matching and geometric transformation on the temporal image set to obtain a preliminary corrected image;

[0079] (2) Based on the preliminary corrected image, perform radiometric correction on the image through the adaptive histogram equalization algorithm to obtain a radiometrically balanced image;

[0080] (3) Perform digital elevation model fusion and orthographic projection on the radiometrically balanced image to obtain an orthoimage;

[0081] (4) Based on the orthoimage, the image is preliminarily segmented by the multi-scale watershed algorithm to obtain the candidate area of the tea tree contour;

[0082] (5) Morphological operations and boundary optimization are performed on the candidate area of the tea tree contour to obtain the accurate tea tree contour;

[0083] (6) Based on the accurate tea tree contour, the single tea tree is segmented by the region growing algorithm to obtain the single tea tree mask;

[0084] (7) Texture features are extracted and statistically analyzed from the single tea tree mask to obtain the tea tree canopy feature descriptor;

[0085] (8) Based on the tea tree canopy feature descriptor, the health status of the tea tree is classified by the support vector machine algorithm to obtain the tea tree health index;

[0086] (9) Time series analysis and anomaly detection are performed on the tea tree health index to obtain the growth anomaly mark;

[0087] (10) Based on the growth anomaly mark, multi-dimensional calculation and normalization processing are performed on the tea tree growth indicators to obtain the growth indicator data set.

[0088] Specifically, multi-point control point matching and geometric transformation are performed on the time series image set to obtain the preliminary corrected image. The multi-point control point matching uses the SIFT (Scale-Invariant Feature Transform) algorithm, which can find corresponding feature points in images of different time phases. The geometric transformation adopts the affine transformation to align the matching points and eliminate the position deviation and deformation between images. Based on the preliminary corrected image, the image is radiometrically corrected by the adaptive histogram equalization algorithm to obtain the radiometrically balanced image. The adaptive histogram equalization algorithm divides the image into multiple small blocks, performs histogram equalization on each small block, and then uses bilinear interpolation to merge the results, effectively enhancing the image contrast and maintaining local details.

[0089] Digital elevation model fusion and orthographic projection are performed on the radiometrically balanced image to obtain the orthoimage. The digital elevation model fusion combines the terrain elevation information with the image, and the orthographic projection eliminates the image deformation caused by terrain undulation and camera tilt, generating an orthoimage with a vertical view. Based on the orthoimage, the image is preliminarily segmented by the multi-scale watershed algorithm to obtain the candidate area of the tea tree contour. The multi-scale watershed algorithm calculates the image gradient at different scales and then simulates the process of water flow "overflowing" from low-gradient areas to high-gradient areas, effectively segmenting the general contour of the tea tree.

[0090] Morphological operations and boundary optimization are performed on the candidate area of the tea tree contour to obtain the precise tea tree contour. The morphological operations include opening and closing operations, which are used to remove noise and fill small holes. The boundary optimization uses the active contour model to accurately locate the tea tree boundary based on the principle of energy minimization. Based on the precise tea tree contour, the single tea trees are segmented by the region growing algorithm to obtain the single tea tree mask. The region growing algorithm starts from the predefined seed points and gradually merges similar pixels into the growing region until the stopping condition is met, effectively separating individual tea trees. Texture feature extraction and statistical analysis are performed on the single tea tree mask to obtain the tea tree canopy feature descriptor. The texture feature extraction uses the gray-level co-occurrence matrix method to calculate features such as energy, contrast, and homogeneity. The statistical analysis calculates the mean, variance, and other statistics of these features to form a feature vector describing the characteristics of the tea tree canopy.

[0091] Based on the tea tree canopy feature descriptor, the health status of the tea trees is classified by the support vector machine algorithm to obtain the tea tree health index. The support vector machine constructs the optimal separation hyperplane in the high-dimensional feature space, classifies the tea trees into different health levels, and outputs a numerical index representing the health degree. Time series analysis and anomaly detection are performed on the tea tree health index to obtain the growth anomaly markers. The time series analysis uses the ARIMA (Autoregressive Integrated Moving Average) model to predict the change trend of the tea tree health index. The anomaly detection marks the abnormal growth points based on the deviation between the predicted value and the actual value.

[0092] Finally, based on the growth anomaly markers, multi-dimensional calculations and normalization processing are performed on the tea tree growth indicators to obtain the growth indicator dataset. The multi-dimensional calculations include calculating indicators such as the growth rate and leaf area index, and the normalization processing unifies the indicators of different scales into the 0-1 interval for comprehensive analysis.

[0093] For example, in a tea garden with an area of 200 hectares, first, 1000 time-series images collected are processed. Approximately 5000 feature points are extracted from each image through the SIFT algorithm. After geometric transformation, the image registration error is controlled within 2 pixels. Adaptive histogram equalization increases the image contrast by 40% while maintaining local details. After fusing a digital elevation model with a resolution of 10 meters, an orthophoto with a resolution of 0.1 meters is generated. The multi-scale watershed algorithm is performed for segmentation at 3 scales, and approximately 500,000 candidate areas of tea tree contours are initially identified. Morphological operations and boundary optimization improve the accuracy of tea tree contours to the pixel level. The region growing algorithm successfully segments 480,000 individual tea trees. By calculating 12 texture features and their statistics, a 36-dimensional feature vector of each tea tree is obtained. The support vector machine algorithm classifies the health status of tea trees into 5 grades, with an accuracy rate reaching 92%. Time series analysis finds that 2% of tea trees have abnormal growth patterns. Finally, 10 growth indicators including tree height, crown width, growth rate, etc. are calculated, forming a growth indicator dataset containing 4.8 million data points.

[0094] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0095] (1) Perform missing value imputation and outlier detection on the growth indicator dataset to obtain a preprocessed dataset;

[0096] (2) Based on the preprocessed dataset, perform multi-scale decomposition on the time series through the wavelet transform algorithm to obtain a trend component, a seasonal component, and a residual component;

[0097] (3) Perform non-parametric regression fitting on the trend component to obtain a long-term growth trend curve;

[0098] (4) Based on the long-term growth trend curve, identify growth stage transition points through a change point detection algorithm to obtain a growth stage division result;

[0099] (5) Perform periodic analysis and spectral density estimation on the seasonal component to obtain a seasonal growth pattern;

[0100] (6) Based on the seasonal growth pattern, model short-term growth fluctuations through an autoregressive integrated moving average model to obtain a short-term growth prediction model;

[0101] (7) Perform non-linear time series analysis on the residual component to obtain non-linear characteristics of tea tree growth;

[0102] (8) Based on the non-linear characteristics, construct a dynamic growth model through a recursive neural network algorithm to obtain a dynamic prediction result of tea tree growth;

[0103] (9) Conduct a sensitivity analysis and uncertainty quantification on the predicted results of the tea tree growth dynamics to obtain the growth prediction confidence interval;

[0104] (10) Based on the growth prediction confidence interval, conduct a multi-dimensional comprehensive evaluation and visualization of the tea tree growth dynamics characteristics to obtain the tea tree growth dynamics characteristics.

[0105] Specifically, perform missing value imputation and outlier detection on the growth index dataset to obtain a preprocessed dataset. Multiple imputation method is used for missing value imputation, which reflects the uncertainty of imputation by creating multiple possible filling values. Outlier detection uses the method based on the interquartile range. Data points beyond 1.5 times the interquartile range are marked as potential outliers and further verified.

[0106] Based on the preprocessed dataset, perform multi-scale decomposition on the time series through the wavelet transform algorithm to obtain the trend component, seasonal component, and residual component. The wavelet transform uses the discrete wavelet transform (DWT) method to decompose the time series into subsequences of different frequencies, effectively separating the long-term trend, periodic changes, and random fluctuations. Perform non-parametric regression fitting on the trend component to obtain the long-term growth trend curve. Non-parametric regression uses the local polynomial regression method, which does not assume a global function form but fits a local polynomial around each data point to capture complex non-linear trends. Based on the long-term growth trend curve, identify the growth stage transition points through the change point detection algorithm to obtain the growth stage division result. The change point detection uses the PELT (Pruned Exact Linear Time) algorithm, which finds multiple change points in the time series in linear time through dynamic programming to accurately divide different growth stages.

[0107] Conduct a periodic analysis and spectral density estimation on the seasonal component to obtain the seasonal growth pattern. The periodic analysis uses the autocorrelation function (ACF) and partial autocorrelation function (PACF) to identify potential periodicities. The spectral density estimation uses the Welch method to reveal the intensity of different frequency components by calculating the power spectral density of the time series. Based on the seasonal growth pattern, model the short-term growth fluctuations through the autoregressive integrated moving average (ARIMA) model to obtain the short-term growth prediction model. The ARIMA model combines three components: autoregressive (AR), differencing (I), and moving average (MA), and can effectively capture the autocorrelation, trend, and periodic characteristics of the time series.

[0108] Perform nonlinear time series analysis on the residual components to obtain the nonlinear characteristics of tea tree growth. The nonlinear analysis uses phase space reconstruction technology to reconstruct the dynamic characteristics of the system through the embedding dimension and time delay parameters, revealing potential chaotic behavior or complex dynamics. Based on the nonlinear characteristics, a dynamic growth model is constructed through the Recurrent Neural Network (RNN) algorithm to obtain the dynamic prediction results of tea tree growth. The RNN model, especially the Long Short-Term Memory (LSTM) network, can capture long-term dependencies and is suitable for processing growth data with complex time dynamics.

[0109] Perform sensitivity analysis and uncertainty quantification on the dynamic prediction results of tea tree growth to obtain the growth prediction confidence interval. The sensitivity analysis uses the Morris method to evaluate the impact of different input parameters on the prediction results. Uncertainty quantification uses Monte Carlo simulation to generate the probability distribution of the prediction results through multiple random samplings. Finally, based on the growth prediction confidence interval, a multi-dimensional comprehensive evaluation and visualization of the dynamic characteristics of tea tree growth are carried out to obtain the dynamic characteristics of tea tree growth. The multi-dimensional evaluation includes calculating indicators such as growth rate, growth cycle, and environmental response. Visualization uses interactive charts to display the spatio-temporal changes of these characteristics.

[0110] For example, in a study involving 100,000 tea trees, the growth data for 5 years was first preprocessed. 2% of the missing values were filled using multiple imputation methods, and 0.5% of the outliers were identified. Wavelet transform decomposed the time series into 3 scales, extracting the long-term trend, annual cycle, and short-term fluctuations. Nonparametric regression fitting showed that the tea trees grew rapidly in the first 3 years and then the growth rate slowed down. The PELT algorithm detected 2 main growth stage transition points, at the end of the 3rd year and the 4th year respectively. Spectral density analysis revealed an obvious annual cycle and a weak quarterly cycle. The ARIMA(2,1,1) model was selected as the best short-term prediction model with an average prediction error of 3%. Nonlinear analysis indicated that tea tree growth had low-dimensional chaotic characteristics with an embedding dimension of 4. The LSTM network achieved a prediction accuracy of 95% on 80% of the training data. Sensitivity analysis showed that temperature and precipitation were the two most significant factors affecting growth. Monte Carlo simulation generated growth predictions with a 95% confidence interval. Finally, through multi-dimensional evaluation and visualization, a description of the dynamic characteristics of tea tree growth including annual average growth rate, growth cycle length, environmental response curve, etc. was obtained.

[0111] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0112] (1) Perform spatio-temporal registration and scale unification on the growth dynamic characteristics and environmental factor data to obtain a collaborative analysis data set;

[0113] (2) Based on the collaborative analysis dataset, calculate the correlation matrix among environmental factors through the Pearson correlation coefficient to obtain the factor correlation evaluation result;

[0114] (3) Conduct hierarchical clustering analysis on the factor correlation evaluation result to obtain environmental factor groups;

[0115] (4) Based on the environmental factor groups, conduct multicollinearity diagnosis through the variance inflation factor method to obtain the screened environmental factor set;

[0116] (5) Conduct partial correlation analysis on the screened environmental factor set and the growth dynamic characteristics to obtain the ranking of direct influencing factors;

[0117] (6) Based on the ranking of direct influencing factors, extract the main environmental impact components through the principal component analysis method to obtain the reduced-dimensional environmental feature space;

[0118] (7) Optimize the reduced-dimensional environmental feature space through orthogonal rotation to obtain the enhanced interpretability environmental features;

[0119] (8) Based on the enhanced interpretability environmental features, construct an environment-growth relationship model through multiple linear regression to obtain the environmental impact coefficients;

[0120] (9) Conduct Bootstrap resampling and confidence interval estimation on the environmental impact coefficients to obtain the environmental impact robustness evaluation result;

[0121] (10) Based on the environmental impact robustness evaluation result, quantify the influence degree of environmental factors through the fuzzy comprehensive evaluation method to obtain the quantitative description data of environmental impact.

[0122] Specifically, perform spatio-temporal registration and scale unification on the growth dynamic characteristics and environmental factor data to obtain the collaborative analysis dataset. Spatio-temporal registration aligns data from different sources to the same time point and spatial location, while scale unification makes data with different dimensions comparable through standardization or normalization processing. Based on the collaborative analysis dataset, calculate the correlation matrix among environmental factors through the Pearson correlation coefficient to obtain the factor correlation evaluation result. The Pearson correlation coefficient measures the linear correlation degree between two variables, with a value range from -1 to 1, and the larger the absolute value, the stronger the correlation. Conduct hierarchical clustering analysis on the factor correlation evaluation result to obtain environmental factor groups. Hierarchical clustering starts with each factor as a separate cluster and gradually merges the most similar clusters until a predetermined number of clusters or similarity threshold is reached. Based on the environmental factor groups, conduct multicollinearity diagnosis through the variance inflation factor method to obtain the screened environmental factor set. The variance inflation factor (VIF) measures the linear correlation degree between one independent variable and other independent variables. Generally, when VIF is greater than 10, it indicates severe multicollinearity.

[0123] Partial correlation analysis was performed on the screened environmental factor set and growth dynamic characteristics to obtain the ranking of direct influencing factors. Partial correlation analysis calculates the correlation degree between two variables while controlling the influence of other variables, so as to identify the environmental factors that have a direct impact on growth characteristics. Based on the ranking of direct influencing factors, the main environmental impact components were extracted by principal component analysis to obtain a reduced-dimensional environmental feature space. Principal component analysis (PCA) transforms the original variables into linearly independent new variables (principal components) through orthogonal transformation, reducing the data dimension while retaining most of the information. The reduced-dimensional environmental feature space was optimized by orthogonal rotation to obtain an enhanced interpretability environmental feature. Orthogonal rotation (such as varimax rotation) aims to make each principal component highly correlated with only a few original variables, improving the interpretability of the results. Based on the enhanced interpretability environmental feature, an environment-growth relationship model was constructed by multiple linear regression to obtain environmental impact coefficients. Multiple linear regression assumes that the dependent variable (growth characteristic) is a linear combination of independent variables (environmental factors), and the regression coefficients are estimated by the least squares method.

[0124] Bootstrap resampling and confidence interval estimation were performed on the environmental impact coefficients to obtain the results of environmental impact robustness assessment. Bootstrap resampling samples from the original data with replacement and repeats the regression analysis to estimate the distribution and confidence interval of the regression coefficients. Finally, based on the results of environmental impact robustness assessment, the influence degree of environmental factors was quantified by the fuzzy comprehensive evaluation method to obtain the quantitative description data of environmental impact. Fuzzy comprehensive evaluation converts qualitative evaluation into quantitative evaluation, considering the comprehensive influence of multiple evaluation indicators.

[0125] For example, in a study involving 50,000 tea plants, first, the growth data of 10 years and 15 environmental factors were spatially and temporally registered, and all data were aligned to a monthly sampling frequency and a 100-meter spatial resolution. By calculating the Pearson correlation coefficient, a 15×15 correlation matrix was obtained, and it was found that the correlation coefficient between temperature and precipitation was -0.75, showing a strong negative correlation. Hierarchical clustering analysis grouped the 15 factors into 4 groups, representing climate, soil, terrain, and human intervention respectively. Variance inflation factor analysis removed 3 factors with VIF greater than 10, and 12 key environmental factors were retained. Partial correlation analysis showed that temperature, soil pH value, and altitude were the three most direct factors affecting the growth of tea plants. Principal component analysis reduced the 12 factors to 4 principal components, explaining 85% of the total variance. After varimax rotation, each principal component had a clear physical meaning. For example, the first principal component mainly represented climate conditions. The determination coefficient R of the multiple linear regression model 2It reached 0.82, indicating that environmental factors could explain 82% of the growth variation. Bootstrap analysis (repeated 1000 times) showed that the influence coefficient of temperature on growth was 0.45 ± 0.05 within the 95% confidence interval. Finally, through fuzzy comprehensive evaluation, the influence weights of temperature, precipitation, and soil pH value on the growth of tea plants were quantified as 0.4, 0.3, and 0.2 respectively.

[0126] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0127] (1) Perform spatio-temporal interpolation on the quantitatively described data of environmental impacts to obtain a continuous environmental impact field;

[0128] (2) Based on the continuous environmental impact field, calculate the local Moran index through spatial autocorrelation analysis to obtain the spatial aggregation pattern of environmental impacts;

[0129] (3) Conduct geographically weighted regression analysis on the spatial aggregation pattern of environmental impacts to obtain spatial heterogeneity characteristics;

[0130] (4) Based on the spatial heterogeneity characteristics, extract the seasonal and periodic components of environmental impacts through time series decomposition to obtain the time variation pattern;

[0131] (5) Conduct wavelet coherence analysis on the time variation pattern to obtain multi-scale spatio-temporal coupling characteristics;

[0132] (6) Based on the multi-scale spatio-temporal coupling characteristics, perform temporal alignment of tea plant growth and environmental factors through the dynamic time warping algorithm to obtain the synchronous change pattern;

[0133] (7) Conduct threshold segmentation and key time point identification on the synchronous change pattern to obtain the critical growth period of tea plants;

[0134] (8) Based on the critical growth period of tea plants, construct a non-linear mapping relationship through a radial basis function network to obtain the data of the influence relationship between environmental factors and growth;

[0135] (9) Conduct sensitivity analysis and uncertainty quantification on the influence relationship data to obtain the ranking of the influence degree of environmental factors;

[0136] (10) Based on the ranking of the influence degree of environmental factors, construct a tea plant growth - environment interaction network through the fuzzy cognitive map method to obtain the growth rules of tea plants, including the annual growth cycle, radial growth rate, time nodes of the critical growth period, and the influence degree of environmental factors.

[0137] It should be noted that spatio-temporal interpolation is performed on the quantitatively described environmental impact data to obtain a continuous environmental impact field. Kriging method is used for spatio-temporal interpolation. This method takes into account the spatial autocorrelation of the data and can generate a smooth continuous surface. Based on the continuous environmental impact field, the local Moran's index is calculated through spatial autocorrelation analysis to obtain the spatial aggregation pattern of environmental impacts. The local Moran's index measures the similarity between each location and its neighboring regions, which helps to identify hot and cold spot areas. Geographically weighted regression analysis is performed on the spatial aggregation pattern of environmental impacts to obtain the spatial heterogeneity characteristics. Geographically weighted regression takes into account the influence of spatial location on the regression relationship and reveals spatial heterogeneity through local regression models. Based on the spatial heterogeneity characteristics, the seasonal and periodic components of environmental impacts are extracted through time series decomposition to obtain the time variation pattern. The X-12-ARIMA method is used for time series decomposition, which decomposes the time series into trend, seasonal and irregular components.

[0138] Wavelet coherence analysis is performed on the time variation pattern to obtain multi-scale spatio-temporal coupling characteristics. Wavelet coherence analysis reveals the correlation between two time series at different time scales and is suitable for analyzing non-stationary time series. Based on the multi-scale spatio-temporal coupling characteristics, the dynamic time warping algorithm is used to align the time series of tea tree growth and environmental factors to obtain the synchronous change pattern. The dynamic time warping algorithm finds the best match between two time series by non-linearly stretching the time axis. Threshold segmentation and key time point identification are performed on the synchronous change pattern to obtain the critical growth period of tea trees. Otsu method is used for threshold segmentation to automatically select the optimal threshold, while key time point identification is based on the mutation detection of the change rate. Based on the critical growth period of tea trees, a non-linear mapping relationship is constructed through a radial basis function network to obtain the data on the influence relationship of environmental factors on growth. The radial basis function network is a feed-forward neural network that can effectively capture the non-linear relationship between the input (environmental factors) and the output (growth characteristics).

[0139] Sensitivity analysis and uncertainty quantification are performed on the influence relationship data to obtain the ranking of the influence degrees of environmental factors. The Sobol method is used for sensitivity analysis to calculate the total effect index to evaluate the importance of each environmental factor. Uncertainty quantification uses Monte Carlo simulation to generate the probability distribution of the influence degree. Finally, based on the ranking of the influence degrees of environmental factors, a tea tree growth-environment interaction network is constructed through the fuzzy cognitive map method to obtain the growth rules of tea trees. The fuzzy cognitive map is a directed graph model, where nodes represent concepts (such as environmental factors, growth characteristics) and edges represent the causal relationships between concepts.

[0140] For example, in a tea garden study with an area of 1000 hectares, first, Kriging interpolation was performed on 10 key environmental factors to generate a continuous environmental impact field with a resolution of 100 meters. Local Moran's I analysis showed that soil moisture exhibited significant spatial aggregation, with an I value of 0.75 (p<0.01). Geographically weighted regression analysis revealed that the impact of temperature on growth was more significant in areas with higher elevations, and the local R 2 ranged from 0.6 to 0.9. Time series decomposition found that precipitation had an obvious seasonal pattern, with an amplitude of 30% of the annual average value. Wavelet coherence analysis showed that the coherence between temperature and growth rate was highest at the 6-month scale, reaching 0.85. The dynamic time warping algorithm aligned the environmental factors with the growth curve, reducing the average alignment error by 40%. Three key growth periods were identified by the Otsu method: the budding period, the rapid growth period, and the maturity period. The prediction accuracy of the radial basis function network model reached 92%, revealing the non-linear effects of temperature, precipitation, and soil pH on growth. Sobol sensitivity analysis showed that the total effect index of temperature was the highest, at 0.45. A fuzzy cognitive map constructed a network containing 15 nodes (10 environmental factors and 5 growth characteristics), revealing complex causal relationships. The finally obtained growth pattern showed that the annual growth cycle of tea trees was 300±15 days, the radial growth rate reached a peak of 0.5 mm / week during the rapid growth period, the key growth periods were in early March, mid-May, and late September respectively, and the order of the influence degrees of temperature, precipitation, and soil pH was 1, 2, 3.

[0141] An embodiment of the present invention also provides a tea tree growth pattern analysis system based on unmanned aerial vehicle monitoring, as Figure 3 shown. The tea tree growth pattern analysis system based on unmanned aerial vehicle monitoring specifically includes:

[0142] A sampling module 301, configured to perform high-precision geocoding and multi-factor stratified sampling on preset tea tree distribution data to obtain sample tea tree distribution information, where the sample tea tree distribution information includes: sample tea tree locations and sample tea tree species;

[0143] An acquisition module 302, configured to, based on the sample tea tree distribution information, perform multi-temporal image acquisition on the tea tree area corresponding to the sample tea tree locations through a preset unmanned aerial vehicle cluster to obtain a time series image set including visible light and near-infrared bands;

[0144] An extraction module 303, configured to perform geometric correction, radiometric correction, and orthoimage generation processing on the time series image set to obtain a processed image set, and perform single tea tree feature extraction on the processed image set through an image segmentation algorithm to obtain a growth index data set;

[0145] An analysis module 304, configured to perform time series decomposition and non-linear trend analysis on the growth index data set to obtain the dynamic growth characteristics of tea trees;

[0146] A statistics module 305, configured to perform multivariate statistical analysis and principal component analysis on the dynamic growth characteristics and pre-acquired environmental factors to obtain quantitative description data of environmental impacts;

[0147] A coupling module 306, configured to perform spatio-temporal coupling analysis and rule extraction on the quantitative description data of environmental impacts to obtain the growth rules of tea trees, wherein the growth rules of tea trees include annual growth cycles, radial growth rates, time nodes of key growth periods, and the influence degrees of environmental factors.

[0148] Through the collaborative work of the above-mentioned various modules, through high-precision geocoding and multi-factor stratified sampling techniques, the representativeness of the sample tea trees and the balance of spatial distribution are ensured; the use of an unmanned aerial vehicle (UAV) cluster for multi-temporal and multi-spectral image acquisition greatly improves the efficiency and accuracy of data acquisition and can capture the minute changes of individual tea trees; the adoption of advanced image processing techniques, including geometric correction, radiometric correction, and orthophoto generation, significantly improves the image quality and usability; through image segmentation algorithms, the precise identification and feature extraction of individual tea trees are realized, providing high-quality input data for the quantitative analysis of growth indicators; by using time series decomposition and non-linear trend analysis methods, the dynamic growth characteristics of tea trees are deeply explored, revealing complex growth patterns; combined with multivariate statistical analysis and principal component analysis, the impacts of environmental factors on the growth of tea trees are comprehensively evaluated, providing quantitative descriptions of environmental impacts; through innovative spatio-temporal coupling analysis and rule extraction techniques, the key cycles, rate changes, and environmental response patterns of tea tree growth are successfully identified. From sample selection to final rule extraction, by integrating multi-source data and various analysis techniques, a comprehensive tea tree growth-environment interaction network is successfully constructed. Based on the UAV and advanced data analysis methods, the labor cost and time investment are greatly reduced, while the monitoring frequency and accuracy are improved.

[0149] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the embodiments, those of ordinary skill in the art should understand that: the specific implementation manners of the present invention can still be modified or equivalently replaced, and any modification or equivalent replacement that does not depart from the spirit and scope of the present invention shall be covered by the scope of the claims of the present invention.

Claims

1. A method for analyzing tea tree growth patterns based on drone monitoring, characterized in that: include: Perform high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, wherein the sample tea tree distribution information includes: sample tea tree location and sample tea tree type; Based on the distribution information of the sample tea trees, a preset drone cluster is used to collect multi-temporal images of the tea tree area corresponding to the location of the sample tea trees, so as to obtain a time-series image set including visible light and near-infrared bands; Performing geometric correction, radiation correction and orthophoto generation processing on the time series image set to obtain a processed image set, and extracting single tea tree features from the processed image set through an image segmentation algorithm to obtain a growth index data set; Performing time series decomposition and nonlinear trend analysis on the growth index data set to obtain dynamic growth characteristics of tea trees; The growth dynamic characteristics and pre-acquired environmental factors are subjected to multivariate statistical analysis and principal component analysis to obtain quantitative description data of environmental impact, including: performing spatiotemporal registration and scale unification on the growth dynamic characteristics and environmental factor data to obtain a collaborative analysis data set; based on the collaborative analysis data set, calculating the correlation matrix between environmental factors by the Pearson correlation coefficient to obtain a factor correlation evaluation result; performing hierarchical clustering analysis on the factor correlation evaluation result to obtain an environmental factor group; based on the environmental factor group, performing multicollinearity diagnosis by the variance inflation factor method to obtain a screening environmental factor set; performing partial correlation analysis on the screening environmental factor set and the growth dynamic characteristics, Obtain the ranking of direct influencing factors; based on the ranking of direct influencing factors, extract the main environmental impact components through principal component analysis to obtain a reduced-dimensional environmental feature space; perform orthogonal rotation optimization on the reduced-dimensional environmental feature space to obtain explanatory enhanced environmental features; based on the explanatory enhanced environmental features, construct an environment-growth relationship model through multivariate linear regression to obtain an environmental impact coefficient; perform Bootstrap resampling and confidence interval estimation on the environmental impact coefficient to obtain an environmental impact robustness assessment result; based on the environmental impact robustness assessment result, quantify the degree of influence of environmental factors through a fuzzy comprehensive evaluation method to obtain the environmental impact quantitative description data; The quantitative description data of environmental impacts are subjected to spatiotemporal coupling analysis and law extraction to obtain the growth law of tea trees, wherein the growth law of tea trees includes the annual growth cycle, radial growth rate, key growth period time nodes and the degree of influence of environmental factors.

2. The tea tree growth law analysis method based on drone monitoring according to claim 1, characterized in that: The step of performing high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, wherein the sample tea tree distribution information includes: sample tea tree locations and sample tea tree types, includes: Performing spatial interpolation and density analysis on the tea tree distribution data to obtain a tea tree distribution heat map; Performing terrain superposition and slope aspect analysis on the tea tree distribution heat map to obtain a terrain-weighted tea tree distribution map; Based on the terrain-weighted tea tree distribution map, the tea trees are spatially clustered by a K-means clustering algorithm to obtain tea tree spatial distribution clusters; Boundary extraction and shape analysis are performed on the tea tree spatial distribution cluster to obtain tea garden contour data; based on the tea garden contour data, partition sampling is performed on the tea garden through a stratified random sampling algorithm to obtain an initial sample point set; Performing spatial autocorrelation analysis and anomaly detection on the initial sample point set to obtain an optimized sample point set; Based on the optimized sample point set, the spatial distribution of tea tree species is inferred by a nearest neighbor interpolation algorithm to obtain a tea tree species distribution map; Performing spatial frequency analysis and boundary enhancement on the tea tree species distribution map to obtain a high-precision tea tree species distribution map; Based on the high-precision tea tree species distribution map, sample points are calibrated on the spot through differential GPS positioning to obtain calibration sample tea tree location data; Attribute association and data fusion are performed on the calibration sample tea tree location data to obtain the sample tea tree distribution information, wherein the sample tea tree distribution information includes: sample tea tree location and sample tea tree type.

3. The tea tree growth law analysis method based on drone monitoring according to claim 1, characterized in that: The step of collecting multi-temporal images of the tea tree area corresponding to the position of the sample tea tree by using a preset drone cluster based on the distribution information of the sample tea tree to obtain a time-series image set containing visible light and near-infrared bands includes: Performing spatial cluster analysis on the distribution information of the sample tea trees to obtain tea tree area sub-blocks; optimizing and planning the flight path of the UAV through an ant colony algorithm based on the tea tree area sub-blocks to obtain an optimized flight path; Divide the optimized flight path into time windows and assign tasks to obtain a collaborative operation plan for the drone cluster; Based on the UAV cluster collaborative operation scheme, the UAV cluster is dynamically scheduled and tasks are issued to obtain real-time flight status data; Performing attitude calculation and position correction on the real-time flight status data to obtain an accurate flight trajectory; based on the accurate flight trajectory, adjusting imaging parameters in real time through an adaptive exposure algorithm to obtain a spectrally balanced image; Performing multi-scale fusion and denoising processing on the spectrally balanced image to obtain an enhanced image; based on the enhanced image, performing semantic segmentation on the image content through a deep learning algorithm to obtain a tea tree area mask; The tea tree area mask is subjected to morphological processing and boundary refinement to obtain an accurate tea tree outline; based on the accurate tea tree outline, multi-temporal images are registered and time-series stacked to obtain a time-series image set containing visible light and near-infrared bands.

4. The tea tree growth law analysis method based on drone monitoring according to claim 1, characterized in that: The step of performing geometric correction, radiation correction and orthophoto generation processing on the time series image set to obtain a processed image set, and extracting single tea tree features from the processed image set through an image segmentation algorithm to obtain a growth index data set includes: Performing multi-point control point matching and geometric transformation on the time series image set to obtain a preliminary corrected image; based on the preliminary corrected image, performing radiation correction on the image by an adaptive histogram equalization algorithm to obtain a radiation balanced image; Performing digital elevation model fusion and ortho projection on the radiation balanced image to obtain an orthophoto; based on the orthophoto, performing preliminary segmentation on the image by a multi-scale watershed algorithm to obtain a candidate area for the outline of the tea tree; Performing morphological operations and boundary optimization on the candidate tea tree contour area to obtain an accurate tea tree contour; based on the accurate tea tree contour, segmenting the individual tea trees by a region growing algorithm to obtain an individual tea tree mask; Extracting and statistically analyzing the texture features of the single tea tree mask to obtain a tea tree canopy feature descriptor; Based on the tea tree canopy feature descriptor, the health status of the tea tree is classified by a support vector machine algorithm to obtain a tea tree health index; The tea tree health index is subjected to time series analysis and anomaly detection to obtain growth anomaly markers; based on the growth anomaly markers, the tea tree growth index is subjected to multi-dimensional calculation and normalization processing to obtain a growth index data set.

5. The tea tree growth law analysis method based on drone monitoring according to claim 1, characterized in that: The step of performing time series decomposition and nonlinear trend analysis on the growth index data set to obtain dynamic characteristics of tea tree growth includes: Performing missing value interpolation and outlier detection on the growth indicator data set to obtain a preprocessed data set; based on the preprocessed data set, performing multi-scale decomposition on the time series through a wavelet transform algorithm to obtain a trend component, a seasonal component and a residual component; Performing non-parametric regression fitting on the trend component to obtain a long-term growth trend curve; based on the long-term growth trend curve, identifying the growth stage transition point through a change point detection algorithm to obtain a growth stage division result; Performing periodic analysis and spectral density estimation on the seasonal component to obtain a seasonal growth pattern; based on the seasonal growth pattern, modeling the short-term growth fluctuations through an autoregressive integrated moving average model to obtain a short-term growth prediction model; Performing nonlinear time series analysis on the residual components to obtain nonlinear characteristics of tea tree growth; Based on the nonlinear characteristics, a dynamic growth model is constructed by a recursive neural network algorithm to obtain a dynamic prediction result of tea tree growth; Performing sensitivity analysis and uncertainty quantification on the tea tree growth dynamics prediction results to obtain a growth prediction confidence interval; Based on the growth prediction confidence interval, the dynamic growth characteristics of the tea tree are comprehensively evaluated and visualized in multiple dimensions to obtain the dynamic growth characteristics of the tea tree.

6. The method for analyzing tea tree growth rules based on drone monitoring according to claim 1, characterized in that: The step of performing spatiotemporal coupling analysis and law extraction on the environmental impact quantitative description data to obtain the tea tree growth law, wherein the tea tree growth law includes the annual growth cycle, radial growth rate, key growth period time nodes and environmental factor influence degree, comprises: Performing spatiotemporal interpolation on the environmental impact quantitative description data to obtain a continuous environmental impact field; based on the continuous environmental impact field, calculating the local Moran index through spatial autocorrelation analysis to obtain a spatial aggregation pattern of environmental impact; Conducting geographically weighted regression analysis on the spatial aggregation pattern of the environmental impacts to obtain spatial heterogeneity characteristics; Based on the spatial heterogeneity characteristics, the seasonal and periodic components of environmental impacts are extracted through time series decomposition to obtain the temporal variation pattern; Performing wavelet coherence analysis on the time variation pattern to obtain multi-scale spatiotemporal coupling characteristics; Based on the multi-scale spatiotemporal coupling characteristics, the tea tree growth and environmental factors are aligned in time series through a dynamic time warping algorithm to obtain a synchronous change pattern; Performing threshold segmentation and key time point identification on the synchronous change pattern to obtain the key growth period of the tea tree; Based on the critical growth period of the tea tree, a nonlinear mapping relationship is constructed through a radial basis function network to obtain the relationship data of the influence of environmental factors on growth; Conduct sensitivity analysis and uncertainty quantification on the impact relationship data to obtain a ranking of the impact levels of environmental factors; Based on the ranking of the influence degree of the environmental factors, a tea tree growth-environment interaction network was constructed through the fuzzy cognitive map method to obtain the growth law of the tea tree, including the annual growth cycle, radial growth rate, key growth period time nodes and the influence degree of environmental factors.

7. A tea tree growth law analysis system based on drone monitoring, used to execute the tea tree growth law analysis method based on drone monitoring as claimed in any one of claims 1 to 6, characterized in that: include: A sampling module is used to perform high-precision geocoding and multi-factor stratified sampling on the preset tea tree distribution data to obtain sample tea tree distribution information, wherein the sample tea tree distribution information includes: sample tea tree location and sample tea tree type; A collection module is used to collect multi-temporal images of the tea tree area corresponding to the location of the sample tea tree through a preset drone cluster based on the distribution information of the sample tea tree, so as to obtain a time-series image set containing visible light and near-infrared bands; An extraction module is used to perform geometric correction, radiation correction and orthophoto generation processing on the time series image set to obtain a processed image set, and extract the features of individual tea trees from the processed image set through an image segmentation algorithm to obtain a growth index data set; An analysis module, used for performing time series decomposition and nonlinear trend analysis on the growth index data set to obtain dynamic growth characteristics of tea trees; The statistical module is used to perform multivariate statistical analysis and principal component analysis on the growth dynamic characteristics and pre-acquired environmental factors to obtain quantitative description data of environmental impact, including: performing spatiotemporal registration and scale unification on the growth dynamic characteristics and environmental factor data to obtain a collaborative analysis data set; based on the collaborative analysis data set, calculating the correlation matrix between environmental factors by the Pearson correlation coefficient to obtain factor correlation evaluation results; performing hierarchical clustering analysis on the factor correlation evaluation results to obtain an environmental factor group; based on the environmental factor group, performing multicollinearity diagnosis by the variance inflation factor method to obtain a screening environmental factor set; performing partial correlation between the screening environmental factor set and the growth dynamic characteristics Analyze and obtain the ranking of direct influencing factors; based on the ranking of direct influencing factors, extract the main environmental impact components through principal component analysis to obtain a reduced-dimensional environmental feature space; perform orthogonal rotation optimization on the reduced-dimensional environmental feature space to obtain explanatory enhanced environmental features; based on the explanatory enhanced environmental features, construct an environment-growth relationship model through multivariate linear regression to obtain an environmental impact coefficient; perform Bootstrap resampling and confidence interval estimation on the environmental impact coefficient to obtain an environmental impact robustness assessment result; based on the environmental impact robustness assessment result, quantify the degree of influence of environmental factors through a fuzzy comprehensive evaluation method to obtain the environmental impact quantitative description data; The coupling module is used to perform spatiotemporal coupling analysis and law extraction on the environmental impact quantitative description data to obtain the growth law of tea trees, wherein the growth law of tea trees includes the annual growth cycle, radial growth rate, key growth period time nodes and the degree of influence of environmental factors.

Images