Omnibearing ancient tree health monitoring method and system based on Internet of Things and AI technology

Abstract

The invention provides an all-around ancient tree health monitoring method and system based on the Internet of Things and the AI technology, and the method comprises the steps: collecting and preprocessing the physiological data of an ancient tree through a sensor, and extracting a disturbance index from the data through edge calculation based on millimeter wave radar point cloud frame data, and representing the behavior disturbance intensity of a current time step; initializing the health state variable of the ancient tree as the health state at the current moment, and inputting the health state into a pre-constructed dynamic recursion model to obtain a predicted health state variable of the ancient tree; sampling the millimeter wave radar point cloud frame data once per second, performing pre-clustering processing, constructing a graph neural network based on the target point cloud set, performing coding to extract a disturbance activation degree representing a current behavior risk, correcting a disturbance index according to the disturbance activation degree, and outputting a disturbance quantity; and calculating a comprehensive risk score, and dynamically generating a corresponding health response action based on the comprehensive risk score.

Description

Technical Field

[0001] The present invention belongs to the field of all-round ancient tree health monitoring, and in particular relates to an all-round ancient tree health monitoring method and system based on the Internet of Things and AI technology. Background Art

[0002] Ancient trees, as key components of ecosystems, possess immense biological, cultural, and environmental value. However, due to their long growth cycles, fragile structures, and slow physiological regulation, they are highly susceptible to irreversible health damage following external disturbances or environmental fluctuations. Current monitoring of ancient trees relies primarily on manual inspections and single-point environmental sensing devices, such as temperature and humidity recorders and image monitoring. These methods not only suffer from sparse sampling and delayed response, but also struggle to characterize physiological changes within the tree, failing to capture hidden signs of decline, such as decreased water conduction efficiency, impeded transpiration, or changes in tissue conductivity. Furthermore, in unmanned forest environments, illegal activities such as approaching, climbing, bark picking, and felling are difficult to identify and intervene in a timely manner. Traditional monitoring systems, which rely primarily on video recognition, are limited by insufficient understory light and unstable power supplies, hindering large-scale deployment. More critically, existing systems fail to link tree health modeling with behavioral impacts. This results in a lack of causal links between health changes and external disturbances, hindering predictive judgment and responsive decision-making, and limiting the ability to transition ancient tree management from reactive to proactive protection.

[0003] Therefore, there is an urgent need for an integrated intelligent monitoring system that can adapt to the complex environment under the forest, continuously sense multiple physiological parameters, identify abnormal behaviors, have predictive capabilities and automatically respond, so as to meet the core needs of early identification, accurate judgment and rapid response in ancient tree protection. Summary of the Invention

[0004] The purpose of this invention is to propose a comprehensive ancient tree health monitoring method and system based on the Internet of Things and AI technology. By deploying multi-source sensors at different structural levels of the tree to collect key physiological and environmental parameters such as temperature, transpiration humidity, conductivity and light, combined with the time series structure modeling mechanism, the system can predict the evolution trend of its health status before it deteriorates significantly.

[0005] To achieve the above objectives, a first aspect of the present invention provides a comprehensive ancient tree health monitoring method based on the Internet of Things and AI technology, the method comprising the following steps:

[0006] S1. Collect ancient tree physiological data through sensors and preprocess them to construct a four-dimensional time-synchronized structured input vector, while also collecting millimeter-wave radar point cloud frame data;

[0007] Based on the millimeter-wave radar point cloud frame data, a disturbance index is extracted from it through edge computing to indicate the behavioral disturbance intensity at the current time step;

[0008] S2. Initializing the old tree health state variable as the health state at the current moment, and combining the disturbance index with the pre-built dynamic recursive model to obtain the predicted old tree health state variable;

[0009] S3. Sampling the millimeter-wave radar point cloud frame data once per second and performing pre-clustering processing to obtain a target point cloud set; constructing a graph neural network based on the target point cloud set to encode and extract a disturbance activation degree representing the current behavior risk, and correcting the disturbance index according to the disturbance activation degree to output a disturbance amount;

[0010] S4. Calculate a comprehensive risk score based on the predicted ancient tree health status variables, disturbance amount, and current health status, and dynamically generate a corresponding response action based on the comprehensive risk score.

[0011] Furthermore, the collecting of physiological data of ancient trees by sensors includes:

[0012] Thermistor patches are attached to the inner bark of the trunk of the ancient tree to collect bark temperature; transpiration humidity is measured by attaching a moisture-sensitive polymer sensor to the trunk; bark conductivity is collected using a wrap-around microelectrode array; and PAR sensors in the canopy layer collect light intensity.

[0013] Wherein, the preprocessing includes:

[0014] The physiological data of the ancient trees are unified into a standard time step using linear interpolation, and the aligned vectors are generated as input vectors for four-dimensional time synchronization structuring.

[0015] Furthermore, the S2 specifically includes:

[0016] The current health status variable of the ancient tree is defined as a scalar, which represents the overall health level and has a value range of [0,1], where the smallest value indicates the worst health.

[0017] Combining the current old tree health state variable, the structured input vector, and the disturbance index to construct an update function of the dynamic recursive model to obtain a predicted old tree health state variable;

[0018] The dynamic recursive model is used to perform a time recursion mechanism on the current ancient tree health state variable, the predicted structured input vector and the disturbance index to obtain the predicted health state result for the next τ steps.

[0019] Furthermore, the predicted structured input vector and disturbance index are obtained by environmental trend extrapolation.

[0020] Furthermore, the millimeter-wave radar point cloud frame data is sampled once per second and pre-clustered to obtain a target point cloud set, specifically including:

[0021] The millimeter wave radar point cloud frame data is pre-clustered on the basis of sampling once per second to obtain the target point cloud set. Each of these A set of spatially distributed points representing a moving object.

[0022] Furthermore, a graph neural network is constructed based on the target point cloud set to encode and extract the disturbance activation representing the current behavior risk, and the disturbance index is corrected according to the disturbance activation to output the disturbance amount, which specifically includes:

[0023] Constructing a point cloud space-time graph neural network G t =(V t ,E t ), where node V t The target point cloud set O t The target point cluster centroid in E t Indicates the trajectory similarity or motion continuity between different time frames to capture behavioral trends;

[0024] Introducing the graph convolution unit GCN(·) to G t Encoding is performed to extract a perturbation activation representing the current behavioral risk; wherein the perturbation activation is a non-normalized perturbation score;

[0025] A health-state-sensitive adjustment function is introduced to map and normalize the disturbance activation degree, and finally a disturbance amount is output; wherein the disturbance amount is obtained by dynamic correction based on the health state at the current moment.

[0026] Furthermore, the health state sensitivity adjustment function dynamically modifies the disturbance amount according to the current health state by designing an adjustment coefficient to represent the response sensitivity of the health state to the disturbance, wherein the largest adjustment coefficient indicates the worst health state.

[0027] Furthermore, the S4 specifically includes:

[0028] Get the cumulative historical response, which represents the integral result of the triggered responses in the past H hours;

[0029] Based on the current health state and the predicted health state variables of the old trees, the downward trend amplitude of the health state is determined to represent the trend risk; based on the disturbance amount, the disturbance risk is determined; based on the predicted health state variables of the old trees, the severity of the future state is determined to represent the severity of the predicted health state; based on the accumulated historical responses, the response suppression penalty is determined to prevent repeated responses in a short period of time from causing system overload and misjudgment;

[0030] determining a composite risk score based on the trend risk, disturbance risk, future state severity, and response inhibition penalty;

[0031] Performing vulnerability-weighted correction on the comprehensive risk score according to the current health status to obtain a final comprehensive risk score;

[0032] Trigger response actions based on the final composite risk score.

[0033] Furthermore, triggering a response action based on the final comprehensive risk score specifically includes:

[0034] Designing a first risk grading threshold and a second risk grading threshold;

[0035] If the final comprehensive risk score is less than the first risk classification threshold, logging is triggered;

[0036] If the final comprehensive risk score is greater than the first risk classification threshold and less than the second risk classification threshold, the platform is connected via LoRa or cellular network and a structured risk report is uploaded, and a low-power infrared camera is used for directional photography;

[0037] If the final comprehensive risk score is greater than the second risk classification threshold, the buzzer is controlled and the high-brightness LED driver is started. At the same time, the sensor sampling period is changed from 5 minutes to 1 minute, and the situation is reported in real time.

[0038] In a second aspect of the present invention, a comprehensive ancient tree health monitoring system based on the Internet of Things and AI technology is provided, the system comprising:

[0039] The signal acquisition unit is used to collect ancient tree physiological data through sensors and perform preprocessing to construct a four-dimensional time-synchronized structured input vector, while also collecting millimeter-wave radar point cloud frame data;

[0040] Based on the millimeter-wave radar point cloud frame data, a disturbance index is extracted from it through edge computing to indicate the behavioral disturbance intensity at the current time step;

[0041] A health prediction unit is used to initialize the health state variable of the ancient tree as the health state at the current moment, and input the pre-built dynamic recursive model into the disturbance index to obtain the predicted health state variable of the ancient tree;

[0042] a health correction unit, configured to sample the millimeter-wave radar point cloud frame data once per second and perform pre-clustering processing to obtain a target point cloud set; construct a graph neural network based on the target point cloud set to encode and extract a disturbance activation representing the current behavior risk; and correct the disturbance index according to the disturbance activation to output a disturbance amount;

[0043] The health warning unit is used to calculate a comprehensive risk score based on the predicted health status variables, disturbance amount, and current health status of the ancient tree, and dynamically generate a corresponding response action based on the comprehensive risk score.

[0044] The beneficial technical effects of the present invention are at least as follows:

[0045] This paper has developed an intelligent ancient tree monitoring system based on predictive health modeling, driven by behavioral disturbance identification, and a closed-loop dynamic risk assessment and response system. By deploying multi-source sensors at different structural levels within the tree to collect key physiological and environmental parameters such as temperature, transpiration, humidity, conductivity, and light, and combining them with temporal structure modeling, the system can predict the evolution of health before significant degradation occurs.

[0046] Furthermore, the system innovatively incorporates a millimeter-wave point cloud behavior perception module to extract potentially destructive behaviors such as approaching and staying, converting them into disturbance variables for injection into the health model. This allows external behaviors to quantifiably influence the state judgment process, establishing a response channel between tree decline and human behavior. When health trends decline or behavioral disturbances intensify, the system calculates a risk score based on the current state and predicted results. Through vulnerability adjustment and response suppression mechanisms, it implements differentiated and controllable linkage responses, such as triggering image acquisition, audio and visual alarms, and remote reporting, thus achieving a complete closed loop from perception to judgment to action.

[0047] This invention opens up the channels between data collection, modeling, identification and response, significantly improving the sensitivity, interpretability and foresight of the ancient tree monitoring system, and solving the systematic problems of difficult state prediction, difficult behavior identification and difficult response classification in traditional methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] The present invention is further described with reference to the accompanying drawings. However, the embodiments in the accompanying drawings do not constitute any limitation to the present invention. A person skilled in the art can obtain other drawings based on the following drawings without creative effort.

[0049] Figure 1 This is a flow chart of the comprehensive ancient tree health monitoring method based on the Internet of Things and AI technology of the present invention. DETAILED DESCRIPTION

[0050] The following describes embodiments of the present invention in detail. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended only to explain the present invention and are not to be construed as limiting the present invention.

[0051] like Figure 1As shown, an embodiment of the present invention provides a comprehensive ancient tree health monitoring method based on the Internet of Things and AI technology, the method comprising:

[0052] S1. Collect ancient tree physiological data through sensors and preprocess them to construct a four-dimensional time-synchronized structured input vector, while also collecting millimeter-wave radar point cloud frame data;

[0053] Based on the millimeter-wave radar point cloud frame data, a disturbance index is extracted from it through edge computing to represent the behavioral disturbance intensity at the current time step.

[0054] Specifically, this step constructs a basic perception layer for ancient tree health modeling. The goal is to design a collection solution that can operate stably in wild forest environments and provide modelable physiological data at a high frequency, starting from the perspective of actual deployment. As objects, ancient trees have physiological structural characteristics such as slow response, tall trees, complex bark structure, and difficult-to-reach roots. Conventional single-point environmental collection methods are difficult to reflect the true physiological changes of trees, especially internal signals such as transpiration activity, water conduction, and conductivity. Therefore, this solution starts from two key parts, the trunk and the crown, to collect representative physiological and driving data respectively. At the same time, considering the subsequent modeling needs, all sensor data need to be organized into time step data vectors in a unified time sequence. And it has precise spatial deployment logic and synchronization mechanism.

[0055] Furthermore, the data collection scheme of the present invention is developed around the main body of ancient trees, and the following four key indicators are selected to construct the time step input vector:

[0056]

[0057] Where t represents the unified time step in minutes. All indicators need to be collected every 5 minutes, and the error tolerance does not exceed ±30 seconds to ensure synchronization.

[0058] T t Bark temperature is measured using a thermistor patch, attached to the inner bark of the tree 1.3 meters from the trunk. Installation is done by gently peeling the bark and then attaching the patch. The temperature measurement range is -20°C to 60°C, with an accuracy of ±0.3°C (in degrees Celsius).

[0059] H t : Transpiration humidity, measured by attaching a humidity-sensitive polymer sensor 1.5 meters north of the trunk, unit: %RH. The data sampling frequency is 1Hz, and the average value within 5 minutes is calculated to generate H t .

[0060] C t Bark electrical conductivity, collected using a wraparound microelectrode array (EMA) using the AC method, is expressed in μS / cm. The measurement period is 5 minutes, and the output is a stable mean.

[0061] I t : Light intensity, collected by a PAR sensor installed in the canopy layer, in μmol / m 2 / s. Used to reflect external transpiration driving force.

[0062] Furthermore, the collected results are unified into a standard time step t using linear interpolation to generate the aligned vector:

[0063]

[0064] where d i ∈{T,H,C,I}, Δt = 2 minutes is the time tolerance. Interpolation is used to compensate for the time step misalignment caused by differences in sensor sampling periods.

[0065] For example, when the system needs to generate If the latest data of a sensor is 10:04 and the next data is 10:06, the system will fill in the value of the sensor at 10:05 through linear interpolation.

[0066] In addition, the point cloud data P required for behavior recognition t The data is collected in real time by millimeter wave radar, which is placed 1 meter outside the base of the old tree, covering a 60° viewing angle and a sampling frequency of 1Hz. t Contains the number, distance, speed, and spatial distribution characteristics of target reflection points in the form of a multidimensional point cloud matrix, which is standardized for subsequent processing.

[0067] Output: Four-dimensional time-synchronized structured input vector for modeling; P t : Millimeter-wave radar point cloud frame data for behavior recognition.

[0068] S2. Initialize the health state variable of the old tree as the health state at the current moment, and input the pre-built dynamic recursive model into it in combination with the disturbance index to obtain the predicted health state variable of the old tree.

[0069] Specifically, the core task of this step is to use the structured input vector obtained in step 1 And point cloud data P t The behavioral perturbation factor δ extracted from t , building a modeling system for the health status of ancient trees that is interpretable, physiologically plausible, and responsive to external disturbances. Changes in the health status of ancient trees are characterized by slow changes, lags, and invisibility. Traditional methods, such as time series classifiers or simple anomaly detection models, are unable to meet practical conservation needs such as advance prediction, causal judgment, and continuous modeling.

[0070] Furthermore, in order to solve the above problems, the present invention introduces a low-dimensional state variable xt (representing the comprehensive health status of ancient trees) as an intermediate representation, and establish a and external disturbance δ t The dynamic recursive model M, that is, In addition, considering the data sparsity and suddenness of human interference in the ancient tree scenario, the present invention innovatively adds a state prediction mechanism with behavioral sensitive adjustment terms and physiological fluctuation regularization terms to the model structure, thereby achieving the ability to both judge the downward trend of the health of ancient trees in advance and cope with state mutations caused by abnormal disturbances.

[0071] This step further obtains P t , is the millimeter wave radar point cloud data, from which the disturbance index δ is extracted through edge computing t , whose value is δ t ∈[0,1], represents the behavioral disturbance intensity of the current time step (such as the weighted index formed by the frequency and duration of behaviors such as approaching, staying, and sawing).

[0072] Furthermore, the core of ancient tree health modeling is a continuous latent state variable x t Modeling and prediction are performed to reflect real physiological dynamics and be predictive. The present invention uses a recursive neural network structure to achieve this goal and makes the following design:

[0073] (1) Define the ancient tree health status variable x t A scalar representing the overall health level, ranging from 0 to 1, with smaller values ​​indicating less health. The initial value can be normalized by mapping historical expert scores.

[0074] (2) Introduce an update function M that combines environmental changes, disturbance risks, and health inertia. The recursive form is as follows:

[0075]

[0076] Among them, x t : Current health status; Structured input data, after inputting a sequence of a time window, enters the GRU (Gated Recurrent Unit) and outputs the hidden state; GRU(·): 1-layer GRU network, input is a sequence of length T=12 (i.e. the past 1 hour) Sequence, output 8-dimensional compressed vector; δ t : Perturbation factor, given by P t Calculation (e.g., a continuous stationary radar target existence time > 30 seconds is considered high disturbance); W1, W2, W3: linear weights learned during training; σ(·): Sigmoid function used for normalization; is the historical state mean; λ is the fluctuation penalty coefficient, which controls the repeated fluctuations of the state and reflects the slow-changing inertia physiological characteristics that the health of ancient trees should have.

[0077] It should be noted that the creative point of this formula is:

[0078] Introducing W3δ t Item, to achieve dynamic modeling of disturbance response driven state change, that is, when the behavior disturbance becomes more severe, the system allows x t A faster downlink response occurs;

[0079] Increase Regularization term, to avoid frequent state reversals caused by slight environmental fluctuations, ensuring that state changes conform to the physiological slow-change mechanism, and solving the problem of allergic false positives in ancient tree health assessments;

[0080] The past state x t Driven by the current integrated environment Joint modeling reflects the evolution trend of the system over time.

[0081] (3) State prediction structure:

[0082] Based on x t The time recursion mechanism, the system can construct the predicted value x t+τ :

[0083]

[0084] Among them, M (τ) represents the recursive calculation of the model for τ future time steps, and δ [t:t+τ] It is the estimated input of the future series (which can be obtained by extrapolating the environmental trend).

[0085] Output: x t : Estimation result of the health status of ancient trees at the current time step, normalized index; x t+τ : Predict the health status results in the next τ steps (such as 3 hours later) for risk scoring and response strategy calling.

[0086] S3. The millimeter-wave radar point cloud frame data is sampled once per second and pre-clustered to obtain a target point cloud set. A graph neural network is constructed based on the target point cloud set to encode and extract the disturbance activation degree representing the current behavior risk, and the disturbance index is corrected according to the disturbance activation degree to output the disturbance amount.

[0087] Specifically, this step aims to obtain the millimeter wave radar point cloud data P collected in step 1. t Identify and quantify the disturbance factor δ that reflects the intensity of human interference behavior t, as input to the health status modeling model in step 2 Unlike traditional camera recognition methods, ancient tree monitoring is often carried out in forest areas without electricity, light, or shelter, making it difficult to deploy visual sensors or to maintain long-term stable operation. This step uses point cloud data as the core input, combining spatial structure, target motion state, and behavioral temporal characteristics to extract the disturbance amount representing the risk of human behavior. The system needs to identify potential destructive behaviors such as approaching, staying for a long time, felling, shaking, and watching under unattended conditions, and output a continuous value δ t ∈[0,1], which is used as the external dynamic disturbance input in the health model. To adapt to the uncertainty of point cloud data and the ambiguity of behavior, this step designs a disturbance modeling method based on point cloud dynamic graph convolutional network and constructs a disturbance integration mechanism in combination with temporal continuity.

[0088] Furthermore, the point cloud data P t Perform pre-clustering processing based on sampling once per second to obtain the target point cloud set Each of these Represents a set of spatially distributed points of a moving object. The present invention constructs a point cloud space-time graph structure G t =(V t ,E t ), where node V t O t The target point cluster centroid in E t It represents the trajectory similarity or motion continuity between different time frames to capture behavioral trends.

[0089] This paper introduces the graph convolution unit GCN(·) to G t Encode and extract the perturbation activation s representing the current behavior risk t :

[0090] s t =ReLU(W g ·GCN(G t-29:t )+b)

[0091] Among them: GCN (G t-29:t ) is the feature vector after two-layer graph convolution processing of the 30-second behavior graph to extract spatial and temporal coherence; W g and b are weight and bias parameters, which are learned from the training set; ReLU ensures that the output is non-negative and represents the intensity of the disturbance; s t is the non-normalized perturbation fraction (real value, undefined range).

[0092] Then, considering that behavioral disturbances may have different effects on old trees with different health states (e.g. old trees with poor health are more vulnerable to the same behavior), the present invention introduces a health-state sensitive adjustment function f(x t ), the disturbance value is mapped and normalized, and the final output disturbance value is:

[0093] δ t =σ(α·s t ·f(x t ))

[0094] Among them, σ(·) is the Sigmoid function, which maps the disturbance to [0,1]; f(x t )=1+γ(1-x t ), which indicates the sensitivity of health status to disturbance, γ is the adjustment coefficient, the worse the health, the greater the amplification coefficient; α is the disturbance standardization coefficient; x t From step 2, used to dynamically correct s t .

[0095] Understandably, the creativity of this structure is reflected in two aspects: first, behavior recognition does not use explicit classification, but instead uses a point cloud structure to capture changes in behavior patterns, which is suitable for unstructured scenarios; second, the disturbance calculation is integrated with health status variables, so that the disturbance feedback has the ability to respond to the individual's health background, solving the problem of decoupling disturbance and risk in the existing system.

[0096] This step is based on the non-visual, low-power, and time-series discontinuous conditions in ancient tree protection, and designs a behavioral disturbance modeling mechanism specifically for millimeter-wave point clouds. Graph convolution is used to model point cloud behavior trends, replacing traditional classification-based recognition, which is more structurally suitable for actual forest scenarios. The health status related function f(x t ) adjusts the disturbance intensity and introduces a dynamic correction mechanism for individual sensitivity responses.

[0097] S4. Calculate a comprehensive risk score based on the predicted ancient tree health status variables, disturbance amount, and current health status, and dynamically generate a corresponding response action based on the comprehensive risk score.

[0098] Specifically, this step is based on the health status prediction result x obtained in the previous stage. t+τ and the disturbance factor δ t, building a response mechanism with predictive foresight, behavioral background understanding, dynamic suppression and differentiated linkage capabilities. Different from the threshold trigger + SMS alarm method in traditional forestry monitoring, this patent emphasizes prediction-driven response + multi-source cross-verification + local rapid intervention. Specifically, the present invention designs a structure of a dynamic scoring model + response cumulative penalty term + health vulnerability weighting function, which is used to accurately divide the risk level and automatically trigger different levels of response actions when the downward trend of the health of ancient trees is superimposed on high interference behavior. Input:

[0099] Input for this step:

[0100] x t+τ :Model in step 2 Output the predicted health status value at a future moment (e.g., 3 hours later);

[0101] x t : The current health status, which comes from the internal state output of the model;

[0102] δ t : The disturbance amount from step 3, indicating the risk level of human behavior at the current moment, with a value range of [0,1], where a larger value represents a stronger disturbance;

[0103] h t : Historical response accumulation, indicating the integral result of responses triggered in the past H hours, maintained by the system local buffer. The initial value is 0 and it increases (e.g., +1.0) each time a high-level response is triggered.

[0104] Furthermore, the key to this step is to construct a comprehensive risk score R t And generate response action A t First, define the comprehensive risk scoring model as follows:

[0105] R t =w1·(x t -x t+τ ) + +w2·δ t +w3·(1-x t+τ ) 2 -λ1·log(1+h t )

[0106] Among them, w1·(x t -x t+τ ) + is the trend risk; w2·δ t is the disturbance risk; w3·(1-x t+τ ) 2 is the severity of the future state; λ1·log(1+h t ) is the response inhibition penalty; (x t-x t+τ ) + Indicates the downward trend of health status, indicating the intensity of impending deterioration; δ t Indicates the current behavior disturbance intensity; (1-x t+τ ) 2 Indicates the severity of the predicted health status (using square to amplify low value sensitivity); log(1+h t ) is the response history cumulative penalty item, which prevents repeated responses in a short period of time from causing system overload and misjudgment; w1, w2, and w3 are the scoring item weights, with values ​​such as w1 = 1.2, w2 = 1.0, and w3 = 1.5; λ1 is the response frequency suppression item weight, which is generally 0.3 to 0.5 and can be adjusted through self-learning.

[0107] In order to further adapt to the characteristics of the ancient tree scene in which the health state changes slowly but is fragile, the present invention adds a health-sensitive weighting function to R t Perform vulnerability weighting correction:

[0108]

[0109] Among them, γ is the health sensitivity amplification factor (e.g., γ = 1.5), which ensures that when the current state of the ancient tree is low, the system response threshold is reduced, thereby improving the protection sensitivity; (1-x t ) 2 is the square of the health vulnerability of the current state, reflecting the fact that low health status is more sensitive to external risks.

[0110] For example: when x t =0.4,x t+τ =0.2,δ t =0.6,h t =2, then the original risk score is:

[0111] Trending items:

[0112] Perturbation term: w2·δ t =0.6;

[0113] Future severity items:

[0114] Cumulative response penalty items:

[0115] Get R t =0.24+0.6+0.96-0.33=1.47;

[0116] Vulnerability weighting factor: (1-0.4) 2 =0.36,γ·0.36=0.54;

[0117] have to

[0118] Furthermore, the system Trigger response action A t , define the response rules as follows:

[0119]

[0120] Among them, θ1=1.2, θ2=2.0 are risk classification thresholds; remote notification connects to the platform via LoRa or cellular network and uploads structured risk reports (including x t ,x t+τ ,δ t , ); Image acquisition uses a low-power infrared camera for directional shooting; Sound and light alarms control the buzzer + high-brightness LED driver; The sampling frequency is doubled to change the sensor sampling period from 5 minutes to 1 minute, which is maintained for 15 minutes before recovery.

[0121] Further, output: A t : The response action control instruction of the system at the current time step. The structured action can call the system's built-in devices for execution; The final weighted risk score serves as a reference for response decisions and the platform record value.

[0122] The embodiment of the present invention further provides a comprehensive ancient tree health monitoring system based on the Internet of Things and AI technology, the system comprising:

[0123] A health graph construction module is used to obtain a historical drone behavior record dataset, including access frequency, charging time, and mission intensity, as well as the cabinet's resource usage records, and construct a behavioral health graph. The module calculates the cabinet health score based on the behavior record dataset and establishes the behavioral dependency weight between the drone and the cabinet.

[0124] A dynamic topology generation module is used to obtain the spatial coordinates of the cabinet and drone and the cabinet's historical anomaly rate. Combined with the behavioral health map, it generates a dynamic spatiotemporal graph containing multiple time slices. Each time slice corresponds to a calculation of edge weights for the dynamic spatiotemporal graph. The edge weights are updated in real time to reflect cabinet resource pressure and health risks. The dynamic spatiotemporal graph is used to support multi-period prediction.

[0125] A resource prediction module, which predicts the number of vacancies and health status change trends of cabinets in future time periods based on the dynamic spatiotemporal graph;

[0126] The intelligent recommendation module generates a cabinet access recommendation priority based on the number of cabinet vacancies and health status trends in the future period, combined with the current location of the drone, and sends it to the drone for execution;

[0127] The feedback optimization module optimizes the behavioral health map based on access result feedback and dynamically adjusts the health score and behavioral dependency weight; wherein the access result feedback includes access success rate, actual cabinet health deviation and access behavior deviation.

[0128] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a division of logical functions. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0130] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, server or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0131] Although the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the claims and their equivalents.

Claims

1. A comprehensive ancient tree health monitoring method based on the Internet of Things and AI technology, characterized by: The method comprises the following steps: S1. Collect ancient tree physiological data through sensors and preprocess them to construct a four-dimensional time-synchronized structured input vector, while also collecting millimeter-wave radar point cloud frame data; Based on the millimeter-wave radar point cloud frame data, a disturbance index is extracted from it through edge computing to indicate the behavioral disturbance intensity at the current time step; S2. Initializing the old tree health state variable as the health state at the current moment, and combining the disturbance index with the pre-built dynamic recursive model to obtain the predicted old tree health state variable; S3. Sampling the millimeter-wave radar point cloud frame data once per second and performing pre-clustering processing to obtain a target point cloud set; constructing a graph neural network based on the target point cloud set to encode and extract a disturbance activation degree representing the current behavior risk, and correcting the disturbance index according to the disturbance activation degree to output a disturbance amount; S4. Calculate a comprehensive risk score based on the predicted ancient tree health status variables, disturbance amount, and current health status, and dynamically generate a corresponding response action based on the comprehensive risk score.

2. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 1 is characterized in that: The method of collecting physiological data of ancient trees through sensors includes: Thermistor patches are attached to the inner bark of the trunk of the ancient tree to collect bark temperature; transpiration humidity is measured by attaching a moisture-sensitive polymer sensor to the trunk; bark conductivity is collected using a wrap-around microelectrode array; and PAR sensors in the canopy layer collect light intensity. Wherein, the preprocessing includes: The physiological data of the ancient trees are unified into a standard time step using linear interpolation, and the aligned vectors are generated as input vectors for four-dimensional time synchronization structuring.

3. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 1 is characterized in that: Said S2 specifically includes: The current health status variable of the ancient tree is defined as a scalar, which represents the overall health level and has a value range of [0,1], where the smallest value indicates the worst health. Combining the current old tree health state variable, the structured input vector, and the disturbance index to construct an update function of the dynamic recursive model to obtain a predicted old tree health state variable; The dynamic recursive model is used to perform a time recursion mechanism on the current ancient tree health state variable, the predicted structured input vector and the disturbance index to obtain the predicted health state result for the next τ steps.

4. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 3 is characterized in that: The predicted structured input vector and disturbance index are obtained by environmental trend extrapolation.

5. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 1 is characterized in that: The millimeter wave radar point cloud frame data is sampled once per second and pre-clustered to obtain a target point cloud set, specifically including: The millimeter wave radar point cloud frame data is pre-clustered on the basis of sampling once per second to obtain the target point cloud set. Each of these A set of spatially distributed points representing a moving object.

6. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 5 is characterized in that: A graph neural network is constructed based on the target point cloud set to encode and extract the disturbance activation representing the current behavior risk, and the disturbance index is corrected according to the disturbance activation to output the disturbance amount, specifically including: Constructing a point cloud space-time graph neural network G t =(V t ,E t ), where node V t The target point cloud set O t The target point cluster centroid in E t Indicates the trajectory similarity or motion continuity between different time frames to capture behavioral trends; Introducing the graph convolution unit GCN(·) to G t Encoding is performed to extract a perturbation activation representing the current behavioral risk; wherein the perturbation activation is a non-normalized perturbation score; A health-state-sensitive adjustment function is introduced to map and normalize the disturbance activation degree, and finally a disturbance amount is output; wherein the disturbance amount is obtained by dynamic correction based on the health state at the current moment.

7. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 6 is characterized in that: The health state sensitivity adjustment function dynamically modifies the disturbance amount according to the current health state by designing an adjustment coefficient to represent the response sensitivity of the health state to the disturbance, wherein the largest adjustment coefficient indicates the worst health state.

8. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 1 is characterized in that: Said S4 specifically includes: Get the cumulative historical response, which represents the integral result of the triggered responses in the past H hours; Based on the current health state and the predicted health state variables of the old trees, the downward trend amplitude of the health state is determined to represent the trend risk; based on the disturbance amount, the disturbance risk is determined; based on the predicted health state variables of the old trees, the severity of the future state is determined to represent the severity of the predicted health state; based on the accumulated historical responses, the response suppression penalty is determined to prevent repeated responses in a short period of time from causing system overload and misjudgment; determining a composite risk score based on the trend risk, disturbance risk, future state severity, and response inhibition penalty; Performing vulnerability-weighted correction on the comprehensive risk score according to the current health status to obtain a final comprehensive risk score; Trigger response actions based on the final composite risk score.

9. The all-round ancient tree health monitoring method based on the Internet of Things and AI technology according to claim 8 is characterized in that: The response action triggered according to the final comprehensive risk score specifically includes: Designing a first risk grading threshold and a second risk grading threshold; If the final comprehensive risk score is less than the first risk classification threshold, logging is triggered; If the final comprehensive risk score is greater than the first risk classification threshold and less than the second risk classification threshold, the platform is connected via LoRa or cellular network and a structured risk report is uploaded, and a low-power infrared camera is used for directional photography; If the final comprehensive risk score is greater than the second risk classification threshold, the buzzer is controlled and the high-brightness LED driver is started. At the same time, the sensor sampling period is changed from 5 minutes to 1 minute, and the situation is reported in real time.

10. A comprehensive ancient tree health monitoring system based on the Internet of Things and AI technology, characterized by: The system comprises: The signal acquisition unit is used to collect ancient tree physiological data through sensors and perform preprocessing to construct a four-dimensional time-synchronized structured input vector, while also collecting millimeter-wave radar point cloud frame data; Based on the millimeter-wave radar point cloud frame data, a disturbance index is extracted from it through edge computing to indicate the behavioral disturbance intensity at the current time step; A health prediction unit is used to initialize the health state variable of the ancient tree as the health state at the current moment, and input the pre-built dynamic recursive model into the disturbance index to obtain the predicted health state variable of the ancient tree; a health correction unit, configured to sample the millimeter-wave radar point cloud frame data once per second and perform pre-clustering processing to obtain a target point cloud set; construct a graph neural network based on the target point cloud set to encode and extract a disturbance activation representing the current behavior risk; and correct the disturbance index according to the disturbance activation to output a disturbance amount; The health warning unit is used to calculate a comprehensive risk score based on the predicted health status variables, disturbance amount, and current health status of the ancient tree, and dynamically generate a corresponding response action based on the comprehensive risk score.

Images