A method and system for intelligent monitoring of multiple parameters of a growth environment of a quercus tree

Abstract

The application provides a copper tree growth environment multi-parameter intelligent monitoring method and system, the method comprises the following steps: collecting a multi-modal biophysical perception data set, constructing a copper digital twin model; integrating photosynthesis, water transport and nutrient absorption efficiency data to generate a copper growth state three-dimensional dynamic visualization atlas; using the biological clock synchronization and multi-scale risk prediction network to generate a copper growth risk spatio-temporal thermal map; based on this, generate adaptive warning threshold parameters, trigger a multi-source data fusion warning mechanism, real-time compare the deviation of environmental parameters and threshold value, and automatically generate a warning report if it is out of range. Based on the copper digital twin model, the three-dimensional grid spatio-temporal alignment is realized, the three-dimensional space interpolation is completed by combining the interpolation method, the three-dimensional dynamic visualization atlas is generated by mapping and superimposition fusion, and the growth state law is reflected; at the same time, the warning threshold is dynamically adjusted based on the biological clock rhythm, the adaptive parameters are formed, and the warning accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of monitoring technology for the growth environment of ancient cypress trees, and in particular to a multi-parameter intelligent monitoring method and system for the growth environment of ancient cypress trees. Background Technology

[0002] Currently, the field of monitoring the growth environment of ancient cypress trees has formed a traditional monitoring system based on single-parameter sensors. Examples include stress wave detection of internal trunk damage, analysis of tree physiological activity using crown electromagnetic fields, and assessment of root vitality using root acoustic vibrations. Some studies have attempted to integrate multi-parameter sensors to achieve preliminary data fusion. In recent years, with the development of digital twins, artificial intelligence, and the Internet of Things (IoT), the industry has begun to explore three-dimensional visualization monitoring methods. These include using lidar to construct three-dimensional models of ancient cypress trees, inverting trunk stress distribution through tomographic imaging algorithms, and using machine learning models to predict growth risks. Some cutting-edge research has involved the application of biological clock synchronization algorithms in the extraction of rhythmic features from ancient cypress trees, and attempts to use multi-scale risk prediction networks in the generation of spatiotemporal heat maps. This has formed a preliminary technical framework of "data acquisition - model construction - risk prediction," providing digital support for the protection of ancient cypress trees.

[0003] Existing technologies still have significant limitations: First, multi-parameter data fusion is mostly limited to simple superposition, lacking spatiotemporal alignment and cross-scale feature fusion mechanisms, making it difficult for growth status visualization maps to truly reflect the coupling relationship of multiple physiological parameters; Second, warning threshold parameters are mostly statically set and not dynamically adjusted in conjunction with biological clock rhythms, which easily leads to false alarms or missed alarms. Summary of the Invention

[0004] This invention aims to at least solve the technical problem of existing technologies that make it difficult to accurately reflect the coupling relationship of multiple physiological parameters, and innovatively proposes a method and system for intelligent monitoring of multiple parameters of the growth environment of ancient cypress trees.

[0005] To achieve the above-mentioned objectives of this invention, this invention provides a multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees, the method comprising: S1. Based on the ancient cypress tree, a three-dimensional biophysical field monitoring network is deployed to collect trunk stress wave data, canopy electromagnetic field data, root acoustic vibration data and microbial interaction fluorescence signals to form a multimodal biophysical sensing dataset. S2. Based on the multimodal biophysical sensing dataset, a three-dimensional model of the tree trunk stress distribution is constructed using tomographic imaging algorithm and genetic algorithm, and a digital twin model of the ancient cypress is generated using point cloud processing algorithm based on lidar scanning data. S3. Based on the aforementioned cypress digital twin model, integrate photosynthetic efficiency, water transport rate and nutrient absorption efficiency data to generate a three-dimensional dynamic visualization map of the cypress growth status. S4. Based on the three-dimensional dynamic visualization map of the growth status of the ancient cypress, a spatiotemporal heat map of the growth risk of the ancient cypress is generated using a biological clock synchronization algorithm and a multi-scale risk prediction network. S5. Generate adaptive early warning threshold parameters based on the spatiotemporal heat map of cypress growth risk; S6. Based on the adaptive early warning threshold parameter, a multi-source data fusion early warning mechanism is triggered. The deviation between the growth environment parameters of the ancient cypress tree and the threshold is compared in real time through edge computing nodes. When the deviation exceeds the preset range, an early warning report containing the risk level, the affected area and the recommended intervention measures is automatically generated.

[0006] On the other hand, the present invention also provides a multi-parameter intelligent monitoring system for the growth environment of ancient cypress trees, the system comprising: processor; Memory used to store processor-executable instructions; The processor is configured to implement the intelligent multi-parameter monitoring method for the growth environment of ancient cypress trees when executing the executable instructions.

[0007] The beneficial effects of this invention are as follows: This invention effectively solves the problems of distorted coupling relationships caused by simple superposition of multi-parameter data and false alarms and missed alarms caused by static thresholds in existing technologies by achieving spatiotemporal alignment of multi-source physiological parameter fusion and dynamic threshold adjustment of biological clock rhythms. Specifically, based on the digital twin model of Cypress, three-dimensional mesh spatiotemporal alignment of canopy leaves, trunk xylem, and root epidermal cells is achieved. Combined with Kriging interpolation and point cloud density adaptive meshing method, three-dimensional spatial interpolation of photosynthetic efficiency, water transport rate, and nutrient absorption efficiency is completed. Through multi-channel color mapping and topological superposition fusion, a three-dimensional dynamic visualization map containing the coupling relationship of multiple physiological parameters is generated, which truly reflects the spatial heterogeneity and temporal evolution law of Cypress growth status. At the same time, the warning threshold is dynamically adjusted based on the biological clock rhythm feature vector and rhythm offset index. The threshold of high-risk areas of the canopy is increased during strong synchronization periods, and the threshold of shallow root areas is decreased during desynchronization periods, forming adaptive warning threshold parameters, which greatly improves the warning accuracy and greatly reduces the false alarm rate.

[0008] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0009] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is a flowchart of a multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees according to the present invention. Detailed Implementation

[0010] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0011] Example 1 like Figure 1 As shown, a multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees is disclosed, the method comprising: S1. Based on the ancient cypress tree, a three-dimensional biophysical field monitoring network is deployed to collect trunk stress wave data, canopy electromagnetic field data, root acoustic vibration data and microbial interaction fluorescence signals to form a multimodal biophysical sensing dataset. In step S1, it is important to explain in detail that, to ensure the comprehensiveness and accuracy of data acquisition, the three-dimensional biophysical field monitoring network employs a distributed sensor node layout. Trunk stress wave data is acquired through an embedded piezoelectric sensor array, with a sampling frequency set to 1 kHz to capture subtle vibration signals. Canopy electromagnetic field data is measured non-contactly using a high-sensitivity fluxgate sensor, while simultaneously correcting for environmental noise interference using meteorological parameters. Root acoustic vibration data is collected using buried accelerometers with a frequency response range covering 20 Hz to 2 kHz to accommodate the propagation characteristics of soil media at different depths. Microbial interaction fluorescence signals are detected using a miniature fiber optic spectrometer, with targeted identification of specific metabolites achieved by adjusting the wavelength of the excitation source. All sensor nodes form a self-organizing network via the ZigBee wireless communication protocol, enabling synchronous data acquisition and transmission, effectively avoiding the complex wiring and maintenance difficulties of traditional wired monitoring systems.

[0012] S2. Based on the multimodal biophysical sensing dataset, a three-dimensional model of the tree trunk stress distribution is constructed using tomographic imaging algorithm and genetic algorithm, and a digital twin model of the ancient cypress is generated using point cloud processing algorithm based on lidar scanning data. S3. Based on the aforementioned cypress digital twin model, integrate photosynthetic efficiency, water transport rate and nutrient absorption efficiency data to generate a three-dimensional dynamic visualization map of the cypress growth status. S4. Based on the three-dimensional dynamic visualization map of the growth status of the ancient cypress, a spatiotemporal heat map of the growth risk of the ancient cypress is generated using a biological clock synchronization algorithm and a multi-scale risk prediction network. S5. Generate adaptive early warning threshold parameters based on the spatiotemporal heat map of cypress growth risk; S6. Based on the adaptive early warning threshold parameter, a multi-source data fusion early warning mechanism is triggered. The deviation between the growth environment parameters of the ancient cypress tree and the threshold is compared in real time through edge computing nodes. When the deviation exceeds the preset range, an early warning report containing the risk level, the affected area and the recommended intervention measures is automatically generated.

[0013] In step S6, it is necessary to explain in detail that the edge computing nodes adopt a distributed deployment architecture. 3-5 edge gateways are set up within a 50-meter radius of the ancient cypress tree. Each gateway integrates LoRa, NB-IoT, and WiFi communication modules to achieve low-power data interaction with various sensors in the monitoring network (stress wave sensors, chlorophyll fluorometers, stem flow probes, etc.). The real-time data acquisition frequency is dynamically adjusted according to the parameter type: trunk stress wave data is collected every 5 minutes, photosynthetic efficiency data every 15 minutes, and soil nutrient data every 30 minutes. The collected raw data first undergoes local preprocessing at the edge nodes, including outlier removal (based on the 3σ principle), data compression (using the LZ77 algorithm), and format standardization (converting to JSON-LD format) to ensure efficient and consistent data transmission.

[0014] During real-time comparison, edge nodes compare the preprocessed parameter data with the adaptive early warning threshold parameters generated in step S5 dimension by dimension. For spatially heterogeneous parameters (such as trunk stress distribution and root nutrient absorption efficiency), a spatial grid matching algorithm (such as Voronoi diagram partitioning) is used to map the parameter data to the corresponding grid cell of the Guichuang digital twin model, and then compare it with the dynamic threshold of that cell; for time-series parameters (such as the diurnal variation of water transport rate), the corresponding threshold interval is used for judgment in different time periods, combined with the circadian rhythm feature vector.

[0015] When the deviation exceeds the preset range, the edge node determines the risk level based on the degree of deviation, duration, and spatial range of the affected area, according to preset risk level classification rules (e.g., mild: deviation 10%-20% and duration <1 hour; moderate: deviation 20%-40% and duration 1-3 hours; severe: deviation >40% or duration >3 hours). Simultaneously, spatial topology analysis locates the specific affected areas (e.g., leaves on the southwest side of the canopy, the trunk at a height of 1.2 meters, and the 0-30cm shallow soil layer of the root system), and matches corresponding recommended measures based on the historical intervention case database. For example, for cases where the canopy photosynthetic efficiency is below the threshold of 25%, it is recommended to increase foliar spraying of nutrient solution; for cases where trunk stress exceeds the threshold of 30%, it is recommended to adopt support and reinforcement measures.

[0016] Finally, the early warning report is generated in the form of a structured document, which includes risk level labels, three-dimensional coordinates and visualization screenshots of the affected area, details of deviation data, recommended intervention measures and implementation priorities, and is pushed to the Gubei protection and management platform via the MQTT protocol. At the same time, relevant management personnel are notified via SMS or APP push to ensure the timely implementation of intervention measures.

[0017] The principle of a multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees in this embodiment is as follows: By integrating multi-source data and intelligent algorithms, a closed-loop monitoring and early warning system for the growth environment of ancient cypress trees was constructed. This system, centered on digital twin technology and combining biophysical field monitoring, spatiotemporal data analysis, and dynamic threshold adjustment, achieves comprehensive perception and accurate assessment of the growth status of ancient cypress trees. At the monitoring level, the system acquires high-resolution physiological and environmental data through a distributed sensor network and generates accurate digital twin models using algorithms such as tomography and point cloud processing, providing a reliable three-dimensional spatial foundation for subsequent analysis. At the analysis level, the system reveals the spatial heterogeneity and temporal evolution patterns of the ancient cypress tree's growth status through spatiotemporal interpolation and visualization mapping of key physiological parameters such as photosynthetic efficiency, water transport rate, and nutrient absorption efficiency. Simultaneously, a risk prediction model based on circadian rhythm feature vectors can capture potential anomalies in the growth process of ancient cypress trees and effectively reduce false alarm and false negative rates through an adaptive threshold mechanism. At the early warning level, real-time comparison and hierarchical response mechanisms of edge computing nodes ensure the timeliness and accuracy of early warning information. This method not only improves the intelligence level of ancient cypress tree growth monitoring.

[0018] As an optional embodiment of the present invention, generating the Cooper digital twin model may include: S201. Perform bandpass filtering, baseline correction and wavelet denoising preprocessing on the stress wave data in the multimodal biophysical sensing dataset to extract the effective stress wave dataset. In step S201, it is important to explain in detail that the preprocessing step employs a multi-level filtering strategy. First, bandpass filtering removes low-frequency environmental noise and high-frequency random interference, preserving the characteristic frequency range related to trunk stress. Second, polynomial fitting is used for baseline correction to eliminate baseline shift caused by sensor drift or temperature changes. Finally, wavelet transform is used to perform multi-scale decomposition of the signal, separating the effective stress wave components and suppressing residual noise. After the above processing, the extracted effective stress wave dataset not only has a high signal-to-noise ratio but also more realistically reflects the changing patterns of stress distribution within the tree trunk.

[0019] Based on the completed data preprocessing, a three-dimensional model of tree trunk stress distribution is further constructed by combining existing tomographic imaging algorithms and existing genetic algorithms. Specifically, the existing tomographic imaging algorithm analyzes the time difference and path difference of stress wave propagation inside the tree trunk to inversely calculate the spatial characteristics of stress distribution. The genetic algorithm is used to optimize the parameter configuration in tomographic imaging, such as the initial value of wave velocity, mesh generation accuracy, and iteration step size, thereby improving the convergence speed and computational accuracy of the model. The final generated three-dimensional stress distribution model can intuitively display the stress concentration areas and their dynamic changes in various parts of the tree trunk.

[0020] Meanwhile, in the process of generating a digital twin model of the Gouldian cypress tree based on lidar scanning data, precise geometric structural information of the canopy, trunk, and root system was extracted by denoising, registration, and segmentation operations on the original point cloud data. Based on this, the point cloud density was adaptively adjusted using the Kriging interpolation method to ensure the model's detail representation at different scales. Furthermore, the point cloud data was transformed into a regularized 3D mesh model through topology analysis and meshing.

[0021] S202. The stress wave dataset is inverted using a tomographic imaging algorithm to obtain the initial stress distribution matrix at different depths inside the tree trunk. The expression for inverting the stress wave dataset using the tomographic imaging algorithm is as follows: in, Indicates the first The pixel value vector of the three-dimensional model of trunk stress distribution in the next iteration. Indicates the first The pixel value vector of the three-dimensional model of trunk stress distribution in the next iteration. This represents the relaxation factor of the ART algorithm (range 0.1~0.5), which controls the iteration convergence speed. The transpose of the tomographic imaging system matrix is ​​constructed from the geometric relationships of the stress wave propagation paths. This represents the measured stress wave projection data vector, containing the arrival time difference data for each sensor pair. Represents the tomographic imaging system matrix. Indicates the first The unit vectors corresponding to each projection path satisfy the following condition: , Represents the row vectors of the system matrix. This represents the TV regularization coefficient (ranging from 0.01 to 0.1), which balances data fidelity and model smoothness. Represents the gradient. The stress distribution gradient field is represented by the central difference method. This represents a small constant to prevent the denominator from being zero.

[0022] In step S202, it is important to explain in detail that, in practical applications, the iterative process of the tomographic imaging algorithm needs to comprehensively consider both computational efficiency and model accuracy. To accelerate convergence, the initial model adopts an empirical distribution based on historical data, while an adaptive step-size mechanism is introduced to dynamically adjust the range of relaxation factors. During each iteration, the TV regularization coefficient is automatically optimized by monitoring the changing trend of the objective function in real time, ensuring that the model maintains spatial resolution while avoiding overfitting. Furthermore, to improve the reliability of the inversion results, the system also integrates a multi-grid method, performing rapid estimation on a coarse grid and then gradually refining it to a fine grid, thereby effectively reducing the influence of local minima and obtaining a more accurate initial stress distribution matrix.

[0023] It is worth noting that the stress distribution characteristics at different depths within the tree trunk exhibit significant anisotropy. Therefore, a directional weighting factor was specifically introduced when constructing the initial stress distribution matrix. By analyzing the directional characteristics of the stress wave propagation path, the stress values ​​at different depths were weighted and corrected. This method not only more accurately reflects the actual stress state within the tree trunk but also effectively identifies potential stress concentration areas.

[0024] S203. With minimizing the root mean square error between the stress distribution results and the actual monitoring data as the optimization objective, the genetic algorithm is used to optimize the parameters of the initial stress matrix. The population is updated through iterative operations of selection operator, crossover operator and mutation operator, and finally a three-dimensional model of the trunk stress distribution is output. In step S203, it is necessary to explain in detail that the parameter optimization process of the genetic algorithm adopts a multi-objective optimization strategy, comprehensively considering the accuracy, computational efficiency, and stability of the stress distribution model. The selection operator is sorted based on fitness function values, prioritizing the retention of individuals with smaller root mean square errors, while an elite retention mechanism is introduced to ensure that the optimal solution is not lost in each generation. The crossover operator uses a combination of single-point crossover and uniform crossover to improve convergence speed while maintaining population diversity. The mutation operator dynamically adjusts the mutation probability, promoting global search capabilities in the early stages and enhancing local fine-tuning capabilities in later stages. After multiple iterations, the output three-dimensional model of trunk stress distribution accurately reflects the stress variation law inside the trunk and possesses high spatial resolution and temporal sensitivity.

[0025] Furthermore, to verify the reliability of the model, the system also designed a cross-validation mechanism. Specifically, the collected stress wave data is divided into training and test sets according to the time series, which are used for model building and accuracy evaluation, respectively. By comparing the model's prediction results with the actual monitoring data, the correlation coefficient and relative error index are calculated to further optimize the model parameter configuration. The final generated 3D model can not only intuitively display the stress distribution characteristics of various parts of the tree trunk, but also...

[0026] S204. The point cloud data of the ancient cypress branches and canopy obtained by lidar scanning are subjected to voxel downsampling, ICP registration and Poisson meshing to construct a digital twin model of the ancient cypress containing geometric shape, spatial position and topological relationship, and the stress distribution three-dimensional model is embedded into the trunk area of ​​the ancient cypress digital twin model through coordinate mapping.

[0027] In step S204, the point cloud data processing involves several steps. First, voxel downsampling divides the point cloud data into a regular three-dimensional voxel grid. Points within each voxel are averaged or randomly selected, effectively reducing data redundancy and improving subsequent processing efficiency. Second, the ICP (Iterative Closest Point) registration algorithm is used to accurately align point cloud data acquired through multiple scans. This algorithm iteratively calculates the nearest point pairs between the source and target point clouds and minimizes the Euclidean distance between the point pairs, achieving high-precision spatial matching. To improve the robustness of registration, a weight adjustment mechanism based on normal vector consistency is introduced to avoid error accumulation caused by local noise or outliers.

[0028] Poisson meshing further transforms the registered point cloud data into a continuous 3D surface model. This method generates a surface represented by an implicit function by solving the Poisson equation for the point cloud normal field, and then constructs the final triangular mesh model using isosurface extraction techniques (such as the MarchingCubes algorithm). To ensure the model's detail representation at different scales, the system dynamically adjusts the octree depth parameter in the Poisson reconstruction, enabling the model to retain both macroscopic geometry and capture fine structural features.

[0029] After constructing the digital twin model of the cypress tree, the 3D stress distribution model was embedded into the trunk region via coordinate mapping. Specifically, the embedding process first established a global coordinate system based on LiDAR scanning data, precisely matching the node coordinates of the stress distribution model with the corresponding positions in the digital twin model. Subsequently, an interpolation algorithm was used to smoothly transition the stress values ​​in non-overlapping areas, ensuring that the embedded model is consistent both visually and numerically.

[0030] As an optional embodiment of the present invention, generating a three-dimensional dynamic visualization map of the growth status of ancient cypress trees may include: S301. Based on the coordinates of the tree crown leaves, trunk xylem, and root epidermal cell grid units of the ancient cypress digital twin model, the real-time monitoring data of the chlorophyll fluorometer, stem flow heat pulse probe, and soil nutrient probe are mapped to the corresponding grid using a timestamp alignment algorithm to generate a spatiotemporally aligned multi-source physiological parameter dataset. In step S301, it is necessary to explain in detail that the timestamp alignment algorithm first calibrates the time base of the data collected by various sensors, uniformly adopting the UTC time standard to eliminate time deviations between time zones or devices. Secondly, for sensor data with different sampling frequencies, linear interpolation or spline interpolation methods are used to fill the data gaps in the time intervals, thereby generating a continuous and synchronous sequence of physiological parameters. Based on this, spatial mapping relationships are used to map the photosynthetic efficiency detected by the chlorophyll fluorometer, the water transport rate measured by the stem flow heat pulse probe, and the nutrient absorption efficiency obtained by the soil nutrient probe to the corresponding grid cells in the Guichuang digital twin model, respectively. This process achieves spatial positioning of the data.

[0031] To further improve the accuracy of data mapping, the system introduces an adaptive weight allocation mechanism. This mechanism dynamically adjusts the weight values ​​of each data point based on the sensor's measurement accuracy and the degree of environmental interference, thereby reducing the impact of noise on the overall results. For example, under drastic changes in light conditions, the data weights from the chlorophyll fluorometer are automatically reduced to avoid misjudgments caused by external factors. Simultaneously, for key areas such as the xylem of the trunk and root epidermal cells, the system enhances data resolution by locally refining the mesh, ensuring that the physiological parameters of these parts are accurately captured and reflected in the final three-dimensional dynamic atlas.

[0032] To visually demonstrate the spatiotemporal evolution of the ancient cypress's growth status, the system employs a multi-layered visualization strategy. At the macroscopic level, photosynthetic efficiency, water transport rate, and nutrient absorption efficiency are displayed using red, blue, and green channels respectively through color gradient mapping technology, forming a comprehensive heatmap of the growth status. At the microscopic level, cross-sectional views and dynamic curves are used to display detailed parameter variation trends within specific grid cells. This multi-layered visualization design not only facilitates researchers' rapid identification of abnormal areas, but also...

[0033] S302. Based on the multi-source physiological parameter dataset, the Kriging interpolation algorithm and the point cloud density adaptive grid method are used to perform three-dimensional spatial interpolation and grid mapping on photosynthetic efficiency, water transport rate and nutrient absorption efficiency to generate a three-dimensional grid model of physiological parameters. In step S302, it is necessary to explain in detail that the Kriging interpolation algorithm estimates the physiological parameter values ​​of unknown regions by analyzing the spatial correlation of known data points and constructing a semi-variogram model. This method can not only effectively handle the interpolation problem in sparse data regions, but also dynamically adjust the interpolation accuracy according to the spatial distribution characteristics of the data. In practical applications, the system combines a point cloud density adaptive mesh method to perform non-uniform mesh division on different parts of the cypress crown, trunk, and root system. Specifically, for areas with high point cloud density, a fine-grained mesh is used to capture local details; while for areas with low point cloud density, a coarse-grained mesh is used to improve computational efficiency. This adaptive strategy ensures that the generated three-dimensional mesh model of physiological parameters has high resolution and accuracy in different regions.

[0034] Furthermore, to further optimize the interpolation results, the system introduces a multi-scale decomposition technique. By performing wavelet transform on the original data, photosynthetic efficiency, water transport rate, and nutrient absorption efficiency are decomposed into multiple frequency components, and each is interpolated independently. This method can effectively separate high-frequency noise from low-frequency trends, thereby improving the stability and robustness of the interpolation model. After interpolation, the system recombines the components through inverse transform to generate the final three-dimensional spatial distribution map.

[0035] During the grid mapping process, the system also takes into account the spatiotemporal heterogeneity of physiological parameters. For example, regarding the diurnal variation of photosynthetic efficiency, the system uses a time-sliding window mechanism to interpolate data from different time periods and employs a smooth transition algorithm to eliminate abrupt changes between time periods. Similarly, for the seasonal fluctuations in water transport rate, the system constructs a periodic trend model based on historical data and incorporates it as prior information into the interpolation process, thereby more accurately reflecting the dynamic changes in the growth status of the ancient cypress.

[0036] S303. Based on the three-dimensional mesh model of the physiological parameters, the photosynthetic efficiency, water transport rate and nutrient absorption efficiency are mapped to visualization layers of different color systems using a multi-channel color mapping rule. In step S303, it is necessary to explain in detail that the design of the multi-channel color mapping rule aims to clearly display the physiological state and dynamic changes of different parts of the ancient cypress tree through intuitive visual expression. Specifically, photosynthetic efficiency is mapped to the green series, with its hue and brightness gradually adjusted according to the numerical value to reflect the distribution of leaf activity; water transport rate is represented by the blue series, with variations in shade reflecting the spatial differences in xylem water transport capacity; nutrient absorption efficiency corresponds to the orange series, using changes in color saturation to highlight the root system's absorption of soil nutrients. This layered mapping method not only facilitates the differentiation of various physiological parameters but also enables the generation of a comprehensive growth status map through overlay display.

[0037] By continuously updating time-series data, the surface color of the 3D model changes in real time, thereby simulating the spatiotemporal evolution of the ancient cypress's growth state. For example, during periods of ample sunlight, the green brightness of the canopy area will significantly increase, while under drought conditions, the blue saturation of the trunk and root system may decrease, intuitively reflecting the impact of water stress.

[0038] S304. Overlay and fuse the visualization layers according to the spatial topological relationship of the Guber digital twin model to generate a three-dimensional static map containing the coupling relationship of multiple physiological parameters. In step S304, it is important to explain in detail that during the overlay and fusion process, the system first performs spatial alignment and coordinate calibration on each visualization layer to ensure that the distribution areas of different physiological parameters perfectly match the geometric structure of the Kupper digital twin model. To reduce visual distortion caused by differences in data resolution, the system employs adaptive filtering technology to smooth the boundary areas of each layer, thereby achieving a natural transition. Furthermore, a transparency weight adjustment mechanism is introduced during the fusion process, dynamically adjusting the display priority of each physiological parameter based on its importance. For example, in specific analytical scenarios, if it is necessary to focus on changes in water transport rate, the transparency weight of the blue layers can be enhanced, making them more prominent in the overlay result. This flexible fusion strategy improves the readability of the 3D static atlas.

[0039] S305. Based on the aforementioned three-dimensional static map, and combined with continuously acquired time-series physiological parameter data, a three-dimensional dynamic visualization map of the growth status of Cypress is generated using inter-frame linear interpolation and keyframe extraction algorithms.

[0040] In step S305, it is necessary to explain in detail that the inter-frame linear interpolation algorithm fills the data gaps in the time series by linearly estimating the changes in physiological parameters at adjacent time points, thereby generating a smooth dynamic transition effect. The keyframe extraction algorithm, based on the changing trends of physiological parameters, automatically identifies time nodes with significant characteristics and uses them as keyframes in the dynamic map. This method not only effectively reduces data redundancy but also highlights important stages of change in the growth state of the ancient cypress. In practical applications, the system combines timeline compression technology to segment data with long time spans, ensuring that the dynamic map maintains clear expressiveness at different time resolutions.

[0041] To further enhance the dynamic visualization effect, a motion fuzzy simulation mechanism is introduced into the system. This mechanism adds appropriate fuzzing effects to the dynamic map based on the rate of change of physiological parameters, thereby more realistically reflecting the continuity and gradual changes in the growth state of the ancient cypress. For example, when photosynthetic efficiency declines rapidly, the canopy area will exhibit a gradually darkening dynamic effect, intuitively demonstrating the impact of environmental stress on plant physiological activities.

[0042] As an optional embodiment of the present invention, generating a spatiotemporal heat map of cypress growth risk may include: S401. Based on the three-dimensional dynamic visualization map of the growth state of the ancient cypress, Fourier transform and cross-correlation analysis are used to extract the diurnal rhythm features of photosynthetic efficiency, stem flow rate and stomatal conductance, and generate biological clock rhythm feature vectors. In step S401, it is necessary to explain in detail that the Fourier transform extracts the periodic fluctuation components of photosynthetic efficiency, stem flow rate, and stomatal conductance through frequency domain analysis of the time series data. These components are further converted into circadian rhythm feature vectors to characterize the physiological activity patterns of the ancient cypress tree during its diurnal cycle. Cross-correlation analysis is used to assess the phase relationship and synchronicity between different physiological parameters, thereby revealing their coupling characteristics in the time dimension. For example, when photosynthetic efficiency and stomatal conductance show high synchronicity, it can be inferred that there is a strong synergistic effect between leaf transpiration and carbon fixation. In addition, to improve the accuracy of feature extraction, the system introduces an adaptive window segmentation technique, dynamically adjusting the length of the analysis window according to changes in environmental conditions to ensure that rapid changes in a short period of time and trend fluctuations on a long time scale can be captured.

[0043] S402. Based on the biological clock rhythm feature vector, the spatial data of tree crown-trunk-root system is processed using a 3D-CNN architecture to generate a spatial risk feature map. Based on the spatial risk feature map, the n=72-hour time series data is processed using a Bi-LSTM network to generate a time risk feature vector. Based on the time risk feature vector, cross-scale features are fused through an attention mechanism to generate a spatiotemporal fusion risk feature tensor. In step S402, it is necessary to explain in detail that the 3D-CNN architecture extracts risk features from local regions by performing convolutional operations on the spatial data of the tree canopy, trunk, and root system, and gradually expands to the global scope, thereby generating a spatial risk feature map. This architecture can effectively capture the spatial dependencies between different parts. Based on this, a Bi-LSTM network is introduced to process continuous n-hour time series data. Through bidirectional long short-term memory units, the system can simultaneously consider past and future contextual information, further mining potential patterns in the time dimension and generating a time risk feature vector.

[0044] To achieve more accurate risk prediction, an attention mechanism is used to fuse cross-scale features. Specifically, the system first performs multi-scale decomposition on the spatial risk feature map and the temporal risk feature vector, extracting high-frequency details and low-frequency trend information respectively. Then, by calculating the importance weights of features at each scale, the contribution ratio of each feature in the final fusion result is dynamically adjusted. This attention-based method not only enhances the expressive power of key features but also significantly reduces the impact of noise interference, thereby generating a more representative spatiotemporal fused risk feature tensor.

[0045] Furthermore, during feature fusion, the system employs adaptive normalization techniques to ensure that data from different sources are compared and integrated at a uniform scale. For example, for parameters with different dimensions, such as photosynthetic efficiency and water transport rate, the system automatically adjusts the normalization parameters based on their historical distribution characteristics, avoiding biases caused by differences in numerical ranges. This process ensures that the generated spatiotemporal fusion risk feature tensor can comprehensively reflect the overall risk level of the ancient cypress's growth state.

[0046] S403. Based on the spatiotemporal fusion risk feature tensor, the Adam optimizer and transfer learning are used to train the risk prediction model. An early stopping mechanism is used to prevent overfitting, and an optimized multi-scale risk prediction model is generated. In step S403, it is necessary to explain in detail that the Adam optimizer dynamically updates the model parameters through an adaptive learning rate adjustment strategy. Specifically, it simultaneously calculates the first and second moment estimates of the gradient, and uses the ratio of these two estimates to adjust the learning rate of each parameter. This allows the model to converge quickly in the early stages of training, and slows down the learning rate as it approaches the optimal solution to avoid oscillations. Transfer learning uses pre-trained model parameters in a relevant domain as initial values, fine-tuning only for the specific risk prediction task of this invention. This effectively reduces the training difficulty caused by insufficient data, accelerates model convergence, and improves prediction accuracy. The early stopping mechanism sets a validation set monitoring metric. When the validation set loss no longer decreases for several consecutive training cycles, the training process is automatically terminated, thereby preventing the model from overfitting to the training data and ensuring that the model has good generalization ability to new data. During actual training, the system will make targeted adjustments to the network structure of each layer of the risk prediction model based on the multi-parameter characteristics of the ancient cypress growth environment. For example, it will increase the depth of the convolutional layer to capture more complex spatial risk features, or adjust the number of LSTM units to optimize the processing capability of time series data. Finally, it will generate a multi-scale risk prediction model that can accurately identify various potential risks in the growth process of ancient cypress.

[0047] S404. Based on the multi-scale risk prediction model, a risk probability density distribution is generated using a kernel density estimation algorithm, and a spatiotemporal heat map of the growth risk of Cypress is generated through linear interpolation and rainbow color mapping rules.

[0048] In step S404, it is necessary to explain in detail that the kernel density estimation algorithm constructs a continuous risk probability distribution function by estimating the probability density of risk values ​​in the spatiotemporal fusion risk feature tensor. This function can accurately characterize the probability of risk occurrence at different spatial locations and time points. During linear interpolation, the system fills blank areas by linearly weighting the risk probability values ​​of adjacent grid cells, ensuring the continuity and smoothness of the risk distribution. The rainbow color mapping rule maps the risk probability density values ​​from low to high to a gradient of red, orange, yellow, green, blue, indigo, and violet, where red represents extremely high-risk areas and violet represents extremely low-risk areas. Through this mapping method, the spatial distribution and temporal trend of cypress growth risk are intuitively transformed into a colorful heat map, allowing researchers to clearly identify key areas and time periods where risks are concentrated. Simultaneously, the system supports dynamic timeline control of the heat map, allowing observation of the risk distribution evolution at different times by sliding the time slider, further revealing the spatiotemporal propagation patterns of risk.

[0049] As an optional embodiment of the present invention, optionally, the expression of the multi-scale risk prediction model in step S403 is: in, This represents the risk prediction output vector. This represents the activation function. This represents the weight matrix of the fully connected layer, initialized with pre-trained weights through transfer learning. This represents the attention weight vector, which dynamically calculates the contribution of spatial-temporal features through the attention mechanism. This represents the Hadamard product (element-by-element multiplication), which achieves a weighted fusion of spatial and temporal characteristics. The spatial risk feature vector is generated by processing the spatiotemporal fusion feature tensor using 3D-CNN, and includes the spatial risk distribution of the canopy, trunk, and root system. This represents a time-risk feature vector, generated by processing 72-hour time-series physiological parameter data using Bi-LSTM, and includes information on the risk evolution trend. This represents the bias vector of the fully connected layer, which is dynamically adjusted by the Adam optimizer. This represents a normalized exponential function, ensuring that the sum of the weights is 1. This represents the transpose of the spatial risk feature vector. This represents the feature dimension scaling factor, stabilizes the gradient of the dot product operation, and prevents gradient explosion.

[0050] As an optional embodiment of the present invention, generating adaptive early warning threshold parameters may include: S501. Based on the spatiotemporal heat map of the growth risk of the ancient cypress, the region growing algorithm and the contour coefficient are used to extract the spatial partitions of high / medium / low risk areas, and the high / low risk periods are identified based on time series analysis to generate spatiotemporal risk area segmentation results. In step S501, it is necessary to explain in detail that the region growing algorithm first uses grid cells with risk probability density exceeding a preset initial threshold as seed points. It then expands the region by calculating the similarity of risk values ​​between adjacent cells, merging continuous regions with similar risk characteristics into a single risk block. The silhouette coefficient verification is used to evaluate the rationality of the region segmentation. By calculating the compactness of the sample within its cluster and its separation from other clusters, the stopping condition of region growing is dynamically adjusted to ensure that the segmented high, medium, and low-risk regions have clear boundaries and internal consistency. For time series analysis, the system uses a sliding window statistical method to perform frequency statistics on risk heatmap data from different time periods. Periods where the percentage of time with risk values ​​exceeding the critical threshold exceeds 30% are identified as high-risk periods, while periods with a percentage below 5% are marked as low-risk periods. Through this spatiotemporal joint segmentation method, the generated results include both the spatial distribution range of different risk levels and the temporal regularity of risk occurrence.

[0051] S502. Based on the spatiotemporal risk region segmentation results, the biological clock rhythm feature vector and dynamic threshold adjustment algorithm are used to adjust the basic threshold through rhythm offset index mapping to generate a biological clock synchronization adaptive threshold set. In step S502, it is necessary to explain in detail that the circadian rhythm feature vector provides a physiological rhythm benchmark for threshold adjustment. The system first compares the fluctuation characteristics of physiological parameters at different time periods with the rhythm template under historical healthy conditions to calculate the rhythm deviation index. This index comprehensively considers three dimensions: phase difference, amplitude change, and periodic stability. For example, when the peak time of photosynthetic efficiency is delayed by more than 2 hours compared with the normal rhythm, or when the amplitude decay exceeds 30%, the rhythm deviation index will increase significantly. The dynamic threshold adjustment algorithm then performs nonlinear mapping adjustment of the basic threshold based on this index: in a healthy state with a low rhythm deviation index, a higher warning threshold is used to reduce false alarms; when the index exceeds the set threshold, the system automatically reduces the sensitivity of the warning threshold, and at the same time sets specific monitoring thresholds for the physiological parameters with the most significant deviations (such as stem flow rate or stomatal conductance). In addition, the algorithm also introduces a seasonal rhythm correction factor, which dynamically adjusts the threshold range according to the physiological activity differences of the ancient cypress in different growing seasons. For example, it raises the warning threshold for water stress during the vigorous growth period in summer, and lowers the triggering condition for insufficient light during the dormant period in winter, ultimately generating an adaptive threshold set that can adapt to the physiological rhythm changes of the ancient cypress.

[0052] Regarding the calculation of the rhythm deviation index, it should be noted that the system achieves quantitative calculation by constructing a multi-dimensional evaluation matrix. This matrix includes three core dimensions: phase shift, amplitude variation coefficient, and cycle stability index. The phase shift is calculated by the time difference between the peak time of the actual physiological parameter and the healthy rhythm template, in hours, with a value range of [-12, 12]. The amplitude variation coefficient is represented by the ratio of the current cycle amplitude to the historical average amplitude; an amplitude anomaly flag is triggered when this value is less than 0.7 or greater than 1.3. The cycle stability index is calculated by the ratio of the standard deviation to the mean of five consecutive cycles, reflecting the degree of fluctuation in the rhythm cycle; the threshold is set at 0.15, exceeding this value indicates cycle instability. The weight allocation of the three dimensions is determined using the analytic hierarchy process (AHP), with phase shift accounting for 40%, amplitude variation coefficient for 35%, and cycle stability index for 25%. The final rhythm deviation index is obtained by weighted summation, with a value range of [0, 1], where 0 indicates complete synchronization and 1 indicates severe disorder. During the calculation process, the system automatically excludes short-term rhythm disorder data caused by extreme weather (such as rainstorms and typhoons) and uses a moving average window (window size of 7 days) to preprocess the raw data to ensure the stability and reliability of the index calculation.

[0053] S503. Based on the biological clock synchronization adaptive threshold set, the 3D-CNN spatial weight coefficients and Bi-LSTM temporal weight coefficients are used to fuse cross-scale risk weights through Hadamard product to generate a comprehensive risk weight matrix. In step S503, it is necessary to explain in detail that the 3D-CNN spatial weight coefficients are calculated by gradient backpropagation of risk feature maps from different parts of the tree crown, trunk, and root system, reflecting the contribution of each spatial region in risk assessment. For example, when water stress occurs in the root system region, its corresponding spatial weight coefficient will increase significantly to highlight the impact of this part on the overall growth risk. The Bi-LSTM temporal weight coefficients are generated by analyzing the importance of risk feature vectors at different times in the 72-hour time series data. Higher temporal weights are assigned to risk mutation points (such as moments of sudden drop in photosynthetic efficiency) to strengthen the risk signals at key time nodes. The Hadamard product operation multiplies the spatial weight coefficient matrix and the temporal weight coefficient matrix element-wise to achieve deep fusion of spatial and temporal risk weights. During the fusion process, the system normalizes the two types of weight coefficients to ensure a uniform numerical range, and then generates a comprehensive risk weight matrix through dynamic weighted summation. This matrix not only preserves the differences in risk contribution from different spatial parts but also reflects the risk change characteristics in the time series.

[0054] Regarding the 3D-CNN spatial weight coefficients, it's worth noting that the system employs Class Activation Mapping (CAM) technology to visualize the attention distribution of convolutional layers. A spatial heatmap is generated by calculating the weighted sum of the output feature map of the last convolutional layer and the corresponding class weights. Higher pixel values ​​in this weight map indicate a greater contribution of the corresponding region to risk prediction; for example, the decay risk area at the base of an ancient cypress trunk will exhibit a high weight value. During weight calculation, a spatial attenuation factor is introduced to attenuate the weights of root regions more than 3 meters from the trunk, reducing interference from non-critical areas. Simultaneously, differentiated initial spatial weight templates are preset for ancient cypresses of different ages: young ancient cypresses emphasize the weight of the crown growth area, while older ancient cypresses strengthen the weight of areas related to trunk stability. By dynamically adjusting the initial templates in conjunction with real-time monitoring data, a 3D-CNN spatial weight coefficient matrix that reflects both species commonalities and individual differences is ultimately generated.

[0055] Regarding the Bi-LSTM time weight coefficients, it's important to note that the system calculates these coefficients using a Gated Recurrent Unit (GRU) Bi-LSTM network. This network adaptively captures long-term and short-term dependencies in time series data. Specifically, in the forward LSTM layer, the system focuses on the cumulative risk effect over the past n hours, filtering irrelevant historical data using a forgetting gate mechanism. In the backward LSTM layer, it emphasizes analyzing the risk evolution trend over the next m hours, dynamically adjusting the importance ratio of future information using an input gate. The time weight coefficients are calculated using an attention score mechanism, assigning importance scores to the hidden states at each time step. These scores are normalized using a softmax function and used as the time weight for that moment. For periods with consecutive risk warning signals, the system enhances the weight value using a time decay factor. For example, if a risk indicator exceeds a threshold for three consecutive hours, its corresponding time weight will increase exponentially. In addition, the time weight coefficient will be dynamically adjusted in combination with seasonal changes. During the critical growth period of ancient cypress (such as the spring budding period and the autumn leaf fall period), the time resolution will be increased and the time window will be refined from the usual 1 hour to 15 minutes to capture more subtle risk changes. During the period of slow physiological activity, the weight update frequency will be appropriately reduced to reduce the consumption of computing resources.

[0056] S504. Based on the comprehensive risk weight matrix, leave-one-out cross-validation and incremental learning are used to optimize the threshold parameters through error backpropagation, generating dynamically updated adaptive early warning threshold parameters.

[0057] In step S504, it is necessary to explain in detail that leave-one-out cross-validation uses a single sample from the cypress monitoring sample set as the validation set and the remaining samples as the training set, iteratively training and validating the model. This fully utilizes limited sample data to evaluate the stability of the threshold parameters. During each validation process, the system calculates the mean square error between the predicted risk value and the actual risk state. By comparing the error changes under different threshold combinations, the initial direction for threshold optimization is determined. Incremental learning, considering the dynamic changes in the cypress growth environment, adopts a sliding time window mechanism. Newly collected monitoring data is periodically added to the training set to iteratively update the threshold parameters, preventing threshold failure due to long-term environmental evolution. During error backpropagation, the system uses the deviation between the comprehensive risk weight matrix and the actual risk event as the loss function, adjusting the threshold parameters through gradient descent, enabling the warning threshold to adaptively track changes in risk characteristics. Specifically, when a certain area experiences consecutive false alarms, the system automatically reduces the threshold weight corresponding to that area; while for missed events, the threshold sensitivity of relevant parameters is increased. Meanwhile, the system sets upper and lower limits for threshold adjustment to prevent excessive threshold deviation from causing the early warning system to fail, and ultimately generates adaptive early warning threshold parameters that can dynamically adapt to changes in the growth environment of ancient cypress.

[0058] Example 2 A multi-parameter intelligent monitoring system for the growth environment of ancient cypress trees includes: processor; Memory used to store processor-executable instructions; The processor is configured to implement a multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees when executing executable instructions.

[0059] It should be noted that the computer device includes a processor, a memory, and may also include one or more of a multimedia component, an input / output (I / O) interface, and a communication component.

[0060] The processor controls the overall operation of the computer device to complete all or part of the steps in the above-mentioned intelligent monitoring method for multiple parameters of the ancient cypress tree's growth environment.

[0061] Memory is used to store various types of data to support the operation of the computer device. This data may include, for example, instructions for any application or method used to operate on the computer device, as well as application-related data. Memory can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0062] The multimedia component may include a screen and an audio component, wherein the screen may be, for example, a touch screen, and the audio component is used to output and / or input audio signals; for example, the audio component may include a microphone for receiving external audio signals, the received audio signals may be further stored in memory or transmitted via a communication component; the audio component may also include at least one speaker for outputting audio signals.

[0063] I / O interfaces provide interfaces between the processor and other interface modules, such as keyboards, mice, buttons, etc.; these buttons can be virtual buttons or physical buttons.

[0064] The communication component is used for wired or wireless communication between the computer device and other devices; wireless communication, such as Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G or 5G, or one or more combinations thereof, and the corresponding communication component may include: Wi-Fi module, Bluetooth module, NFC module, mobile communication module.

[0065] As a preferred embodiment, the computer device may be implemented by one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to execute the above-described intelligent monitoring method for multiple parameters of the ancient cypress tree's growth environment.

[0066] Although embodiments of the invention have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the claims and their equivalents.

Claims

1. A multi-parameter intelligent monitoring method for the growth environment of ancient cypress trees, characterized in that, The method includes: Based on the ancient cypress tree, a three-dimensional biophysical field monitoring network was deployed to collect trunk stress wave data, canopy electromagnetic field data, root acoustic vibration data, and microbial interaction fluorescence signals, forming a multimodal biophysical sensing dataset. Based on the multimodal biophysical sensing dataset, a three-dimensional model of tree trunk stress distribution was constructed using tomographic imaging and genetic algorithms, and a digital twin model of ancient cypress was generated using point cloud processing algorithms based on lidar scanning data. Based on the aforementioned digital twin model of the ancient cypress, data on photosynthetic efficiency, water transport rate, and nutrient absorption efficiency are integrated to generate a three-dimensional dynamic visualization map of the growth status of the ancient cypress. Based on the three-dimensional dynamic visualization map of the growth status of ancient cypress, a spatiotemporal heat map of growth risk of ancient cypress is generated by using a biological clock synchronization algorithm and a multi-scale risk prediction network. Generate adaptive early warning threshold parameters based on the spatiotemporal heat map of ancient cypress growth risk; Based on the adaptive early warning threshold parameter, a multi-source data fusion early warning mechanism is triggered. The deviation between the growth environment parameters of the ancient cypress tree and the threshold is compared in real time through edge computing nodes. When the deviation exceeds the preset range, an early warning report containing the risk level, the affected area and the recommended intervention measures is automatically generated.

2. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 1, characterized in that, The generation of the Cooper digital twin model includes: Bandpass filtering, baseline correction, and wavelet denoising preprocessing are performed on the stress wave data in the multimodal biophysical sensing dataset to extract the effective stress wave dataset. The stress wave dataset was inverted using a tomographic imaging algorithm to obtain the initial stress distribution matrix at different depths inside the tree trunk; With the goal of minimizing the root mean square error between the stress distribution results and the actual monitoring data, a genetic algorithm is used to optimize the parameters of the initial stress matrix. The population is updated through iterative operations of selection operator, crossover operator and mutation operator, and finally a three-dimensional model of the stress distribution of the tree trunk is output. Voxel downsampling, ICP registration, and Poisson meshing were performed on the point cloud data of the ancient cypress branches and canopy obtained by lidar scanning to construct a digital twin model of the ancient cypress containing geometric shape, spatial position, and topological relationship. The stress distribution three-dimensional model was then embedded into the trunk region of the ancient cypress digital twin model through coordinate mapping.

3. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 2, characterized in that, The expression for inverting the stress wave dataset using the tomographic imaging algorithm is as follows: in, Indicates the first The pixel value vector of the three-dimensional model of trunk stress distribution in the next iteration. Indicates the first The pixel value vector of the three-dimensional model of trunk stress distribution in the next iteration. Represents the relaxation factor of the ART algorithm. This represents the transpose of the tomographic imaging system matrix. This represents the measured stress wave projection data vector. Represents the tomographic imaging system matrix. Indicates the first The unit vector corresponding to each projection path Represents the TV regularization coefficient. Represents the gradient. Represents the stress distribution gradient field. This represents a small constant to prevent the denominator from being zero.

4. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 1, characterized in that, The generation of a 3D dynamic visualization map of the growth status of ancient cypress includes: Based on the coordinates of the tree crown leaves, trunk xylem, and root epidermal cell grid units of the ancient cypress digital twin model, the real-time monitoring data of the chlorophyll fluorometer, stem flow heat pulse probe, and soil nutrient probe are mapped to the corresponding grid using a timestamp alignment algorithm to generate a spatiotemporally aligned multi-source physiological parameter dataset. Based on the multi-source physiological parameter dataset, the Kriging interpolation algorithm and the point cloud density adaptive grid method are used to perform three-dimensional spatial interpolation and grid mapping on photosynthetic efficiency, water transport rate and nutrient absorption efficiency to generate a three-dimensional grid model of physiological parameters. Based on the three-dimensional mesh model of the physiological parameters, a multi-channel color mapping rule is used to map photosynthetic efficiency, water transport rate and nutrient absorption efficiency into visualization layers of different color systems. The visualization layers are superimposed and fused according to the spatial topological relationship of the Guber digital twin model to generate a three-dimensional static map containing the coupling relationship of multiple physiological parameters. Based on the aforementioned three-dimensional static map, combined with continuously acquired time-series physiological parameter data, a three-dimensional dynamic visualization map of the growth status of Cypress cypress is generated using inter-frame linear interpolation and keyframe extraction algorithms.

5. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 1, characterized in that, The generation of the spatiotemporal heat map of the growth risk of ancient cypress includes: Based on the three-dimensional dynamic visualization map of the growth status of the ancient cypress, Fourier transform and cross-correlation analysis were used to extract the diurnal rhythm features of photosynthetic efficiency, stem flow rate and stomatal conductance, and generate biological clock rhythm feature vectors. Based on the biological clock rhythm feature vector, a 3D-CNN architecture is used to process the canopy-trunk-root spatial data to generate a spatial risk feature map. Based on the spatial risk feature map, a Bi-LSTM network is combined to process n-hour time series data to generate a temporal risk feature vector. Based on the temporal risk feature vector, cross-scale features are fused through an attention mechanism to generate a spatiotemporal fusion risk feature tensor. Based on the spatiotemporal fusion risk feature tensor, the Adam optimizer and transfer learning are used to train the risk prediction model. An early stopping mechanism is used to prevent overfitting, and an optimized multi-scale risk prediction model is generated. Based on the multi-scale risk prediction model, a risk probability density distribution is generated using a kernel density estimation algorithm, and a spatiotemporal heat map of cypress growth risk is generated through linear interpolation and rainbow color mapping rules.

6. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 5, characterized in that, The expression for the multi-scale risk prediction model is: in, This represents the risk prediction output vector. This represents the activation function. This represents the weight matrix of the fully connected layer. Represents the attention weight vector. It represents the Hadamah accumulation. Represents the spatial risk feature vector. Represents the time risk feature vector. This represents the bias vector of the fully connected layer. Represents the normalized exponential function, This represents the transpose of the spatial risk feature vector. This represents the feature dimension scaling factor.

7. The intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in claim 1, characterized in that, The parameters for generating adaptive warning thresholds include: Based on the spatiotemporal heat map of the growth risk of the ancient cypress, the region growing algorithm and the contour coefficient are used to extract the spatial partitions of high / medium / low risk regions, and the high / low risk periods are identified based on time series analysis to generate spatiotemporal risk region segmentation results. Based on the spatiotemporal risk region segmentation results, the biological clock rhythm feature vector and dynamic threshold adjustment algorithm are used to adjust the basic threshold through rhythm offset exponential mapping to generate a biological clock synchronization adaptive threshold set. Based on the biological clock synchronization adaptive threshold set, the spatial weight coefficients of 3D-CNN and the temporal weight coefficients of Bi-LSTM are used to fuse cross-scale risk weights through Hadamard product to generate a comprehensive risk weight matrix. Based on the comprehensive risk weight matrix, leave-one-out cross-validation and incremental learning are used to optimize the threshold parameters through error backpropagation, generating dynamically updated adaptive early warning threshold parameters.

8. A multi-parameter intelligent monitoring system for the growth environment of ancient cypress trees, characterized in that, The system includes: processor; Memory used to store processor-executable instructions; The processor is configured to implement the intelligent monitoring method for multiple parameters of the growth environment of ancient cypress trees as described in any one of claims 1 to 7 when executing the executable instructions.

Images