Real-time monitoring and early warning method for diseases and insect pests of ancient and famous trees based on multi-source environment sensing

Abstract

The invention relates to the field of intelligent monitoring, and provides a real-time monitoring and early warning method for diseases and insect pests of ancient and famous trees based on multi-source environmental sensing, which comprises the following steps: deploying a plurality of sensing nodes, continuously collecting environmental data around and on trunks of to-be-monitored ancient trees, preprocessing the environmental data and extracting features to obtain feature data; and inputting the feature data into a pre-trained disease and pest risk assessment model, outputting an ancient tree risk coefficient, and sending out an early warning signal in real time according to the change of the ancient tree risk coefficient. According to the method, early warning and early intervention can be carried out on the disease and pest phenomena of the ancient and famous trees, the growth environment of the ancient and famous trees is continuously monitored, multi-source environment data of the ancient and famous trees are collected, the risk value of the ancient and famous trees is evaluated through a model, and the defect that early diseases and pests are difficult to find is effectively overcome in a mode of pushing early warning signals in real time; the maintenance quality and survival rate of the famous and ancient tree group are fully improved, and high-precision and low-misinformation rapid identification of the risk ancient trees is realized.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring, and in particular to a method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing. Background Technology

[0002] Due to their advanced age and declining physiological functions, ancient and famous trees are susceptible to continuous threats from pests and diseases. Once an attack occurs, treatment is often difficult and recovery is slow. Currently, pest and disease monitoring in the field of ancient tree protection relies heavily on traditional periodic manual patrols. By the time external symptoms are visible to the naked eye, such as large-scale yellowing of leaves, obvious borers or gumming on the trunk, the pests and diseases have often entered the middle or late stages, missing the optimal intervention window. Later rescue measures usually cause serious damage to the ancient tree itself.

[0003] In recent years, intelligent monitoring has become a key measure for intervening in the early stages of pest and disease damage to ancient and famous trees. Its core lies in achieving early identification and warning of pests and diseases through sensing devices and data analysis technology. Document CN119622552B discloses a forest cultivation pest and disease early warning method combining multi-source data analysis. Specifically, it uses various types of sensors to comprehensively monitor pests and diseases from multiple dimensions, and integrates sensor data to accurately determine whether pest and disease problems exist in the area. However, early intervention for pests and diseases in ancient and famous trees requires high-frequency, continuous dynamic monitoring to achieve higher identification accuracy and faster real-time response speed, thereby promptly capturing early, weak signals of pest and disease occurrence. Therefore, a pest and disease monitoring and early warning method based on multi-source environmental sensing, combined with the spatiotemporal correlation and potential characteristics between sensor data, is an effective way to provide scientific protection and management for ancient and famous trees. Summary of the Invention

[0004] The purpose of this invention is to propose a method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing, so as to solve one or more technical problems existing in the prior art, and at least provide a beneficial option or create conditions.

[0005] This invention provides a method for real-time monitoring and early warning of pests and diseases in ancient and famous trees based on multi-source environmental sensing. By deploying multiple sensor nodes, it continuously collects environmental data around and on the trunk of the ancient tree to be monitored. The environmental data is preprocessed and features are extracted to obtain feature data. This feature data is input into a pre-trained pest and disease risk assessment model, which outputs a risk coefficient for the ancient tree. Based on changes in the risk coefficient, an early warning signal is issued in real time. This method can provide early warning and intervention for pests and diseases in ancient and famous trees. By continuously monitoring the growth environment of ancient and famous trees, collecting multi-source environmental data, and assessing their risk values ​​through a model, it effectively overcomes the difficulty in detecting early pests and diseases by sending real-time early warning signals. This significantly improves the maintenance quality and survival rate of ancient and famous tree groups, and achieves rapid identification of at-risk ancient trees with high accuracy and low false alarms.

[0006] To achieve the above objectives, according to one aspect of the present invention, a method for real-time monitoring and early warning of diseases and pests on ancient and famous trees based on multi-source environmental sensing is provided, the method comprising the following steps: Multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored, and environmental data is continuously collected through the sensor nodes. Environmental data is preprocessed and features are extracted to obtain feature data; Input the feature data into the pre-trained pest and disease risk assessment model to obtain the risk coefficient of ancient trees; Early warning signals are issued in real time based on changes in the risk coefficient of ancient trees.

[0007] Furthermore, multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored. The sensor nodes include at least image sensors, meteorological sensors, soil sensors, and sound and vibration sensors. Each sensor node collects its corresponding data and transmits the collected data as environmental data to a cloud platform or edge computing gateway. All sensor nodes form a sensor network.

[0008] Preferably, each sensing node collects its corresponding data, including image sensors collecting high-definition images and video sequences of the tree crown, leaves, and trunk appearance; meteorological sensors collecting ambient temperature, air humidity, light intensity, and rainfall; soil sensors collecting soil temperature, soil humidity, and soil pH; and sound and vibration sensors collecting sound waves and vibration signals from inside the tree trunk.

[0009] Furthermore, the method for preprocessing environmental data and extracting features to obtain feature data is as follows: various types of data in the environmental data are cleaned and standardized, and feature vectors related to pests and diseases are extracted through feature engineering. All feature vectors related to pests and diseases are fused to obtain feature data. Furthermore, the method for inputting feature data into a pre-trained pest and disease risk assessment model to obtain the risk coefficient of ancient trees is as follows: load a pre-trained pest and disease risk assessment model and denote it as the first model, input feature data into the first model, and output the risk coefficient of ancient trees from the output layer of the first model. The risk coefficient of ancient trees is a continuous scalar value between [0,1].

[0010] Furthermore, the pest and disease risk assessment model can be a multilayer perceptron model or a fully connected neural network model, or a hybrid neural network model constructed based on multiple models from recurrent neural network models, convolutional neural network models, or graph neural network models as the pest and disease risk assessment model. Furthermore, the method for issuing real-time early warning signals based on changes in the risk coefficient of ancient trees is as follows: Let N be the number of ancient trees to be monitored. Maintain a risk array for each of the N ancient trees to store their corresponding risk coefficients. Update the risk coefficients of the N ancient trees continuously at preset intervals. Each time the risk coefficient is updated, record the updated value and add it to the risk array. Let Tr(i) represent the i-th tree among N ancient trees, and Atr(i) represent the risk array of Tr(i). Then, N ancient trees correspond to N risk arrays Atr(1), Atr(2), ..., Atr(N). The average value M(i) of all values ​​in Atr(i) is taken as the first risk value of the i-th ancient tree Tr(i). Then, N ancient trees correspond to N first risk values. Identify and mark outliers among N first-risk values, and mark the ancient trees corresponding to the first-risk values ​​that are marked as outliers as risky ancient trees and send early warning signals; Among these, the methods used to identify outliers among N first-risk values ​​are the Isolation Forest method or the DBSCAN clustering method, or, among the N first-risk values, those higher than M... N The value of ×h0 is considered an outlier, M N h0 represents the average of N first risk values, and h0 takes values ​​within the interval [1,2].

[0011] The beneficial effects of this step are as follows: when potential pests and diseases appear in ancient and famous trees, continuous environmental monitoring and rapid early warning response help to detect and intervene in a timely manner. This step continuously updates the risk coefficient of ancient trees, takes the average value of the risk array as the first risk value to smooth short-term environmental fluctuations, and improves the accuracy and timeliness of pest and disease identification by identifying outliers, which is helpful for the breeding of trees in ancient and famous tree groups.

[0012] Preferably, the interval for updating the risk coefficient of ancient trees is set to [5, 24] hours, and the length of the risk array is set to [10, 60] (that is, the risk array is composed of the latest [10, 60] ancient tree risk coefficients). Since environmental data on ancient trees exhibits spatiotemporal correlation, considering the mutual influence of risks on neighboring trees can better improve the accuracy of early warnings; preferably, the method for identifying at-risk ancient trees can also be: Calculate the risk excess and risk average for each ancient tree. Store the absolute values ​​of the N risk excesses corresponding to the N ancient trees in an array Rk. Record the maximum and minimum values ​​in the array Rk as R0 and R1 respectively. Record the value obtained by dividing R0 by R1 as Rf0. For any ancient tree Tr(x), if there exists any ancient tree among the R0 closest ancient trees that has a risk average greater than rm(x)×Rf0 or less than rm(x)÷Rf0, then the ancient tree Tr(x) is marked as a risky ancient tree; rm(x) is the risk average of Tr(x). The calculation methods for the risk excess and risk average of ancient trees are as follows: Let rm(i) represent the risk average of ancient tree Tr(i). In the risk array Atr(i) of Tr(i), take out all values ​​greater than or equal to M(i) to form array B0(i) and take out all values ​​less than M(i) to form array B1(i). Record the lengths of arrays B0(i) and B1(i) as L0(i) and L1(i) respectively. Record the risk excess C0(i) as the value obtained by subtracting L1(i) from L0(i). The sum of all values ​​in array B0(i) multiplied by the absolute value of C0(i) is used as the first sub-value, the sum of all values ​​in array B1(i) is used as the second sub-value, and the sign(C0(i)) power of the first sub-value divided by the second sub-value is used as the risk average rm(i).

[0013] The beneficial effects of this step are as follows: For ancient trees requiring special maintenance, their tree quality is easily affected by the growth environment. When the environmental data of multiple ancient trees within a local area are similar, these ancient trees can be considered to be in a normal state. However, when the scale is expanded to the entire densely distributed area of ​​ancient trees, if the environmental data of these ancient trees deviates from the environmental baseline of the entire ancient tree group, the risk ancient tree identification method based on global statistics is prone to false alarms. The method in this step uses a collaborative assessment method combining single trees and neighbors. By setting the risk mean range of [rm(x)÷Rf0,rm(x)×Rf0], and using the dimensionless scaling value rm(i) instead of the global statistics to reflect the current risk probability of the ancient tree, it comprehensively characterizes the distribution pattern and stability of the ancient tree risk array. By calculating the ratio of the sum of high-risk values ​​to the sum of low-risk values, and introducing the absolute value of the risk excess as an amplification factor to form a weighted risk ratio, it can be seen that this ratio is sensitive to both the magnitude of the risk value and the duration of its occurrence. At the same time, the sign function sign(C) is introduced. o(i) As an index, it realizes the directional judgment of risk bias. When high-risk data points dominate, the calculation result is greater than 1, indicating that it is in a continuous high-risk state. When low-risk data points dominate, the calculation result is less than 1, reflecting the degree of deviation between the abnormal fluctuations that have occurred in history and its low-stability normal state. Through the two composite indicators of risk excess and risk average, the joint judgment process of risk correlation with neighboring trees is realized. When the ancient tree group in the whole area experiences an overall environmental data shift due to seasonal changes or the influence of the climate, the target risk ancient tree can be quickly identified.

[0014] The beneficial effects of the present invention are as follows: the method can provide early warning and early intervention for pests and diseases of ancient and famous trees. By continuously monitoring the growth environment of ancient and famous trees, collecting multi-source environmental data and evaluating their risk value through models, the method effectively overcomes the defect that early pests and diseases are difficult to detect by pushing early warning signals in real time, thus fully improving the maintenance quality and survival rate of ancient and famous tree groups, and achieving rapid identification of risky ancient trees with high accuracy and low false alarm. Attached Figure Description

[0015] Figure 1 The diagram shows a flowchart of a method for real-time monitoring and early warning of diseases and pests in ancient and famous trees based on multi-source environmental sensing. Detailed Implementation

[0016] The following will provide a clear and complete description of the concept, specific structure, and technical effects of the present invention in conjunction with the embodiments and accompanying drawings, so as to fully understand the purpose, solution, and effects of the present invention. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0017] In the description of this invention, "several" means one or more, "more than" means two or more, "greater than," "less than," and "exceeding" are understood to exclude the stated number, while "above," "below," and "within" are understood to include the stated number. The use of "first" and "second" in the description is merely for distinguishing technical features and should not be construed as indicating or implying relative importance, or implicitly indicating the number of indicated technical features, or implicitly indicating the order of the indicated technical features.

[0018] like Figure 1 The diagram shown is a flowchart of a real-time monitoring and early warning method for diseases and pests of ancient and famous trees based on multi-source environmental sensing according to the present invention. The following is a combination of... Figure 1 This paper describes a method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing, according to an embodiment of the present invention.

[0019] This invention proposes a method for real-time monitoring and early warning of diseases and pests on ancient and famous trees based on multi-source environmental sensing. The method includes the following steps: Multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored, and environmental data is continuously collected through the sensor nodes. Environmental data is preprocessed and features are extracted to obtain feature data; Input the feature data into the pre-trained pest and disease risk assessment model to obtain the risk coefficient of ancient trees; Early warning signals are issued in real time based on changes in the risk coefficient of ancient trees.

[0020] Furthermore, multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored. The sensor nodes include at least image sensors, meteorological sensors, soil sensors, and sound and vibration sensors. Each sensor node collects its corresponding data and transmits the collected data as environmental data to a cloud platform or edge computing gateway. All sensor nodes form a sensor network.

[0021] The specific deployment locations of the sensor nodes are determined based on expert experience or generated using a grid coverage method.

[0022] Specifically, the sensor nodes transmit data via LoRa, NB-IoT, or 4G / 5G wireless transmission.

[0023] Furthermore, the method for preprocessing environmental data and extracting features to obtain feature data is as follows: data cleaning and standardization are performed on various types of data in the environmental data, and feature vectors related to pests and diseases are extracted through feature engineering. All feature vectors related to pests and diseases are fused to obtain feature data. Feature data is a multi-dimensional fused feature vector. Various types of data in the environmental data refer to different sensor data collected by different types of sensors.

[0024] Specifically, the feature vectors related to pests and diseases include leaf texture features, leaf color features, leaf morphological features, air temperature and humidity features, light and rainfall features, and soil temperature and humidity features.

[0025] Furthermore, the method for inputting feature data into a pre-trained pest and disease risk assessment model to obtain the risk coefficient of ancient trees is as follows: load a pre-trained pest and disease risk assessment model and denote it as the first model, input feature data into the first model, and output the risk coefficient of ancient trees from the output layer of the first model. The risk coefficient of ancient trees is a continuous scalar value between [0,1].

[0026] Furthermore, the pest and disease risk assessment model is a multilayer perceptron model or a fully connected neural network model, or a hybrid multilayer neural network model constructed based on multiple models among recurrent neural network models, convolutional neural network models or graph neural network models as the pest and disease risk assessment model; the pre-trained pest and disease risk assessment model is obtained by using historical environmental data and its corresponding ancient tree health status labels as training samples and supervised learning training.

[0027] Specifically, the first model performs a linear transformation on the input feature data through the first fully connected layer to generate a new feature representation. Then, the new feature representation is non-linearly processed by the ReLU activation function and input into the subsequent hidden layer. The hidden layer continues to iterate the linear transformation and ReLU activation of the features. Finally, the network output layer of the model converts the final feature representation into a continuous scalar value in the interval [0,1] through the Sigmoid activation function, which is the ancient tree risk coefficient.

[0028] Furthermore, the method for issuing real-time early warning signals based on changes in the risk coefficient of ancient trees is as follows: Let N be the number of ancient trees to be monitored. Maintain a risk array for each of the N ancient trees to store their corresponding risk coefficients. Update the risk coefficients of the N ancient trees continuously at preset intervals. Each time the risk coefficient is updated, record the updated value and add it to the risk array. The data acquisition, processing and model inference process is triggered periodically according to a preset cycle, and the risk coefficients of all ancient trees are calculated and updated in batches based on the latest multi-source environmental data. Let Tr(i) represent the i-th tree among N ancient trees, where i is the index, i=1,2,…,N. Let Atr(i) represent the risk array of Tr(i). Then, N ancient trees correspond to N risk arrays Atr(1), Atr(2),…,Atr(N). The average value M(i) of all values ​​in Atr(i) is taken as the first risk value of the i-th ancient tree Tr(i). Then, N ancient trees correspond to N first risk values. Identify and mark outliers among N first-risk values, mark the ancient trees corresponding to the first-risk values ​​that are marked as outliers as risky ancient trees and send an early warning signal. The early warning signal includes relevant tree information of the risky ancient trees. Among these, the methods used to identify outliers among N first-risk values ​​are the Isolation Forest method or the DBSCAN clustering method, or, among the N first-risk values, those higher than M... N The value of ×h0 is considered an outlier, M N h0 represents the average of N first risk values, and h0 takes values ​​within the interval [1,2].

[0029] Specifically, if the interval is set to 5 hours and the length of the risk array is set to 60, then the 60 data points will cover approximately 12.5 days, meaning the warning signal will be updated every 12.5 days.

[0030] Specifically, the warning signal includes the ancient tree's unique identifier, geographical location, environmental data snapshot, and risk level. The risk level is determined based on the magnitude of the outlier in the first risk value, and the threshold for risk level is set by expert experience.

[0031] Because environmental data on ancient trees exhibits spatiotemporal correlation, considering the mutual influence of risks from neighboring trees can significantly improve the accuracy of early warnings. Specifically, other methods for identifying at-risk ancient trees include: Calculate the risk excess and risk average for each ancient tree. Store the absolute values ​​of the N risk excesses corresponding to the N ancient trees in an array Rk. Record the maximum and minimum values ​​in the array Rk as R0 and R1 respectively. Record the value obtained by dividing R0 by R1 as Rf0. For any ancient tree Tr(x), if any of the R0 nearest ancient trees to Tr(x) has a risk average greater than rm(x)×Rf0 or less than rm(x)÷Rf0, then ancient tree Tr(x) is marked as a risky ancient tree; rm(x) is the risk average of Tr(x); the information of neighboring ancient trees of Tr(x) is pre-stored in the database; The calculation methods for the risk excess and risk average of ancient trees are as follows: Let rm(i) represent the risk average of ancient tree Tr(i). In the risk array Atr(i) of Tr(i), take out all values ​​greater than or equal to M(i) to form array B0(i) and take out all values ​​less than M(i) to form array B1(i). Record the lengths of arrays B0(i) and B1(i) as L0(i) and L1(i) respectively. Record the risk excess C0(i) as the value obtained by subtracting L1(i) from L0(i). The sum of all values ​​in array B0(i) multiplied by the absolute value of C0(i) is used as the first sub-value, the sum of all values ​​in array B1(i) is used as the second sub-value, and the sign(C0(i)) power of the first sub-value divided by the second sub-value is used as the risk average rm(i). The mathematical expression for rm(i) is:

[0032] In the formula, sum(B0(i)) and sum(B1(i)) represent the sum of all values ​​in array B0(i) and the sum of all values ​​in array B1(i), respectively. sign(C0(i)) represents the sign function. When C0(i)>0, sign(C0(i)) equals 1, and when C0(i)<0, sign(C0(i)) equals -1.

[0033] This invention provides a method for real-time monitoring and early warning of pests and diseases in ancient and famous trees based on multi-source environmental sensing. By deploying multiple sensor nodes, it continuously collects environmental data around and on the trunk of the ancient tree to be monitored. The environmental data is preprocessed and features are extracted to obtain feature data. This feature data is input into a pre-trained pest and disease risk assessment model, which outputs a risk coefficient for the ancient tree. Based on changes in the risk coefficient, an early warning signal is issued in real time. This method can provide early warning and intervention for pests and diseases in ancient and famous trees. By continuously monitoring the growth environment of ancient and famous trees, collecting multi-source environmental data, and assessing their risk values ​​through a model, it effectively overcomes the difficulty in detecting early pests and diseases by sending real-time early warning signals. This significantly improves the maintenance quality and survival rate of ancient and famous tree groups, and achieves rapid identification of risky ancient trees with high accuracy and low false alarms. Although the description of this invention has been quite detailed and several embodiments have been described in particular, it is not intended to limit itself to any of these details or embodiments or any particular embodiments, thereby effectively covering the intended scope of the invention. Furthermore, the invention has been described above with respect to embodiments foreseeable by the inventors, in order to provide a useful description, while non-substantial modifications to the invention that have not yet been foreseen may still represent equivalent modifications to the invention.

Claims

1. A method for real-time monitoring and early warning of diseases and pests on ancient and famous trees based on multi-source environmental sensing, characterized in that, The method includes the following steps: Multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored, and environmental data is continuously collected through the sensor nodes. Environmental data is preprocessed and features are extracted to obtain feature data; Input the feature data into the pre-trained pest and disease risk assessment model to obtain the risk coefficient of ancient trees; Early warning signals are issued in real time based on changes in the risk coefficient of ancient trees.

2. The method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing according to claim 1, characterized in that, Multiple sensor nodes are deployed around and on the trunk of the ancient tree to be monitored. The sensor nodes include at least image sensors, meteorological sensors, soil sensors, and sound and vibration sensors. Each sensor node collects its corresponding data and transmits the collected data as environmental data to a cloud platform or edge computing gateway. All sensor nodes form a sensor network.

3. The method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing according to claim 1, characterized in that, The method for preprocessing environmental data and extracting features to obtain feature data is as follows: data cleaning and standardization are performed on various types of data in the environmental data, and feature vectors related to pests and diseases are extracted through feature engineering. All feature vectors related to pests and diseases are then fused to obtain feature data.

4. The method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing according to claim 1, characterized in that, The method for obtaining the risk coefficient of ancient trees by inputting feature data into a pre-trained pest and disease risk assessment model is as follows: load a pre-trained pest and disease risk assessment model and denote it as the first model, input feature data into the first model, and output the risk coefficient of ancient trees from the output layer of the first model. The risk coefficient of ancient trees is a continuous scalar value between [0,1].

5. A method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing, as described in claim 4, is characterized in that... The pest and disease risk assessment model is either a multilayer perceptron model or a fully connected neural network model.

6. A method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing, as described in claim 4, is characterized in that... The method for issuing real-time early warning signals based on changes in the risk coefficient of ancient trees is as follows: Let N be the number of ancient trees to be monitored. Maintain a risk array for each of the N ancient trees to store their corresponding risk coefficients. Update the risk coefficients of the N ancient trees continuously at preset intervals. Each time the risk coefficient is updated, record the updated value and add it to the risk array. Let Tr(i) represent the i-th tree among N ancient trees, and Atr(i) represent the risk array of Tr(i). Then, N ancient trees correspond to N risk arrays Atr(1), Atr(2), ..., Atr(N). The average value M(i) of all values ​​in Atr(i) is taken as the first risk value of the i-th ancient tree Tr(i). Then, N ancient trees correspond to N first risk values. Outliers are identified and marked among N first-risk values. Ancient trees corresponding to the first-risk values ​​marked as outliers are marked as risky ancient trees and warning signals are sent.

7. A method for real-time monitoring and early warning of diseases and pests of ancient and famous trees based on multi-source environmental sensing, as described in claim 6, is characterized in that... Methods used to identify outliers among N first-risk values ​​include the Isolation Forest method or the DBSCAN clustering method.

Images