Forest fire monitoring method and monitoring system based on three-temperature model

Abstract

The invention relates to a forest fire monitoring method and monitoring system based on a three-temperature model, and the method comprises the steps: S1, extracting an initial surface temperature feature, an initial canopy temperature feature and an initial atmospheric temperature feature according to a radiation feature, and obtaining a standard three-temperature feature set; s2, generating a temperature weight distribution parameter based on the standard three-temperature feature set, and obtaining an evapotranspiration prediction value according to the temperature weight distribution parameter; and S3, comparing the evapotranspiration prediction value with a preset fire danger threshold interval, and outputting a visual fire danger distribution map as a forest fire monitoring result. In a forest fire monitoring scene, an existing monitoring method has significant limitation; on the basis of the scheme, the regional evapotranspiration can be accurately predicted in real time in a complex and changeable forest ecological system, so that a reliable basis is provided for fire risk assessment and early warning.

Description

Technical Field

[0001] This invention relates to the field of intelligent monitoring, and in particular to a forest fire monitoring method and system based on a three-temperature model. Background Technology

[0002] In forest fire monitoring scenarios, existing monitoring methods have significant limitations when dealing with complex terrain, variable climate, and differences in vegetation cover. In other words, current technologies cannot simultaneously meet the following requirements: ensuring the rationality of predictions through physical constraints, while adapting to the specific patterns of different forest environments; processing data with spatiotemporal scale differences from satellite remote sensing and meteorological stations, while achieving efficient inference under conditions of limited edge computing resources.

[0003] Based on this, the present invention proposes a forest fire monitoring method and system based on a three-temperature model, corresponding to a forest fire monitoring strategy based on a three-temperature model and a Tabnet fine-tuned neural network architecture for predicting evapotranspiration. It involves the intersection of forest fire monitoring, meteorological forecasting and deep learning, and solves the shortcomings of existing technologies. Summary of the Invention

[0004] The main technical problem addressed by this invention is how to accurately predict regional evapotranspiration in real time within complex and ever-changing forest ecosystems, providing a reliable basis for fire risk assessment and early warning. The core idea is to simultaneously collect three-dimensional temperature field information—surface temperature, canopy temperature, and atmospheric temperature—based on a three-temperature model. Then, using a TabNet fine-tuned neural network architecture, adaptive feature selection and deep fusion are performed on multi-source remote sensing data (such as satellite infrared and LST), ground meteorological station data (temperature, humidity, wind speed, etc.), and vegetation indices, thereby efficiently simulating the evapotranspiration process. This invention innovatively combines a physics-driven three-temperature model with a data-driven TabNet fine-tuning architecture. On the one hand, it uses the three-temperature model to capture the physical mechanisms of evapotranspiration; on the other hand, it leverages TabNet's attention mechanism to achieve dynamic weighting and interpretability analysis of influencing factors. Furthermore, this application supports deployment on edge computing devices or monitoring stations, enabling real-time online monitoring and prediction of key forest areas, improving the system's timeliness and applicability.

[0005] This invention proposes a forest fire monitoring method based on a three-temperature model. The method includes: S1, extracting initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on radiation characteristics to obtain a standard three-temperature feature set; S2, generating temperature weight allocation parameters based on the standard three-temperature feature set, and obtaining evapotranspiration prediction values ​​based on the temperature weight allocation parameters; S3, comparing the evapotranspiration prediction values ​​with a preset fire risk threshold range, and outputting a visualized fire risk distribution map as the forest fire monitoring result.

[0006] The further technical solution is as follows: S1, extracting initial surface temperature features, initial canopy temperature features and initial atmospheric temperature features based on radiation characteristics to obtain a standard three-temperature feature set, including: S31, extracting initial surface temperature features, initial canopy temperature features and initial atmospheric temperature features based on remote sensing data of the target area collected by the sensor; S32, performing data standardization processing on the initial surface temperature features, initial canopy temperature features and initial atmospheric temperature features to generate a standard three-temperature feature set.

[0007] The further technical solution is as follows: S2, generating temperature weight allocation parameters based on a standard three-temperature feature set, and obtaining evapotranspiration prediction values ​​based on the temperature weight allocation parameters, includes: S33, inputting the standard three-temperature feature set into an evapotranspiration prediction model fine-tuned based on the TabNet architecture, dynamically filtering and weighting the surface temperature, canopy temperature, and atmospheric temperature features to generate temperature weight allocation parameters; S34, performing weighted fusion processing on the standard three-temperature feature set based on the temperature weight allocation parameters to generate evapotranspiration prediction values, and evaluating the prediction accuracy based on preset root mean square error (RMSE), preset mean absolute error (MAE), and preset coefficient of determination (RSQ) to obtain accuracy evaluation results; S35, performing dynamic feedback calibration on the temperature weight allocation parameters based on the accuracy evaluation results to generate optimized weight parameters, and performing prediction processing on the standard three-temperature feature set based on the optimized weight parameters to generate evapotranspiration prediction values.

[0008] The further technical solution is as follows: S31, based on remote sensing data of the target area collected by the sensor, extracts initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features according to radiation characteristics, including: S41, based on remote sensing data of the target area collected by the sensor, performing radiation anomaly detection processing on the remote sensing data to generate an effective radiation dataset; S42, calculating surface radiation features, canopy radiation features, and atmospheric radiation features based on the effective radiation dataset; S43, extracting initial surface temperature features based on surface radiation features, extracting initial canopy temperature features based on canopy radiation features, and extracting initial atmospheric temperature features based on atmospheric radiation features.

[0009] The further technical solution is as follows: S32, performing data standardization processing on the initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics to generate a standard three-temperature feature set, including: S51, performing data standardization processing on the initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics; S52, identifying spatial anomalies in the initial surface temperature characteristics, correcting the anomaly data based on a geographic weighted regression algorithm, and generating optimized surface temperature characteristics; S53, detecting vegetation shading interference in the initial canopy temperature characteristics, correcting the canopy temperature through a radiative transfer model, and generating effective canopy temperature characteristics; S54, performing time series smoothing processing on the initial atmospheric temperature characteristics to eliminate transient fluctuation interference and generate stable atmospheric temperature characteristics; S55, fusing the optimized surface temperature characteristics, effective canopy temperature characteristics, and stable atmospheric temperature characteristics to generate a standard three-temperature feature set.

[0010] A further technical solution involves dynamically filtering and weighting the surface temperature, canopy temperature, and atmospheric temperature characteristics to generate temperature weighting parameters, including: S61, dynamically filtering and weighting the surface temperature, canopy temperature, and atmospheric temperature characteristics, and calculating and parsing the surface temperature vector, canopy temperature vector, and atmospheric temperature vector through sequence attention; S62, based on the surface temperature vector, canopy temperature vector, and atmospheric temperature vector, calculating the surface temperature dominance coefficient, canopy temperature regulation coefficient, and atmospheric temperature control coefficient through a multi-layer attention network; and S63, integrating the surface temperature dominance coefficient, canopy temperature regulation coefficient, and atmospheric temperature control coefficient to generate temperature weighting parameters.

[0011] The further technical solution is as follows: The step of evaluating prediction accuracy and obtaining accuracy evaluation results based on preset root mean square error (RMSE), preset mean absolute error (MAE), and preset coefficient of determination (RSQ) includes: S71, analyzing the fluctuation distribution characteristics of evapotranspiration prediction values ​​and extracting time series deviation features based on preset RMSE, MAE, and RSQ; S72, generating an overall deviation index corresponding to RMSE, a local anomaly index corresponding to MAE, and a model fit index corresponding to RSQ based on the time series deviation features; S73, converting the overall deviation index into an overall deviation level, the local anomaly index into a local anomaly level, and the model fit index into a model fit level according to preset index range mapping rules; S74, integrating the overall deviation level, local anomaly level, and model fit level to generate accuracy evaluation results.

[0012] A further technical solution is as follows: the step of performing predictive processing on the standard three-temperature feature set based on optimized weight parameters to generate evapotranspiration prediction values ​​includes: S81, performing predictive processing on the standard three-temperature feature set based on optimized weight parameters, and performing dynamic weighted fusion of surface temperature, canopy temperature, and atmospheric temperature features through a sequence attention mechanism; S82, generating a surface feature enhancement vector, a canopy environment coupling vector, and an atmospheric temperature response vector based on the dynamic weighted fusion result; and S83, integrating the surface feature enhancement vector, the canopy environment coupling vector, and the atmospheric temperature response vector to generate evapotranspiration prediction values.

[0013] This invention employs a three-layer system architecture, comprising a data acquisition layer, a model processing layer, and an application service layer. The overall workflow includes four main stages: data preprocessing, model training and optimization, prediction and evaluation, and real-time monitoring and early warning. Specifically, at the data acquisition end, this invention innovatively adopts a three-temperature model theory, simultaneously acquiring three core physical quantities—surface temperature, canopy temperature, and atmospheric temperature—as the model's primary input features, greatly simplifying data dependencies. In terms of prediction, this invention employs and specifically fine-tunes the TabNet neural network architecture. This architecture utilizes a sequence attention mechanism, enabling multi-step reasoning similar to a decision tree, automatically selecting the temperature features that have the greatest impact on evapotranspiration under different conditions, and dynamically weighting them. To efficiently train this model and address the scarcity of labeled evapotranspiration data, this invention designs a two-stage training strategy combining self-supervised and supervised learning. First, self-supervised pre-training is performed using unlabeled temperature data to allow the model to learn the inherent patterns in the data. Then, supervised learning fine-tuning is performed using labeled evapotranspiration data to construct an accurate predictive mapping relationship. (During supervised learning fine-tuning, the model loss rapidly decreases in the first few cycles and eventually converges to an extremely low level (approximately 0.005). Furthermore, the errors on both the training and test sets decrease rapidly and converge, with a very small difference between them, indicating that the model does not suffer from severe overfitting and has good generalization ability.) Finally, this invention utilizes the inherent properties of TabNet to quantify the contribution of different temperature features, achieving high interpretability of the model's decisions. It also establishes a graphical interface that can process evapotranspiration predictions based on the trained model individually or in batches, thus forming a closed-loop intelligent system from data acquisition to risk warning.

[0014] This invention innovatively combines a physics-driven three-temperature model with a data-driven TabNet fine-tuning architecture. On the one hand, it uses the three-temperature model to capture the physical mechanism of evapotranspiration, and on the other hand, it uses TabNet's attention mechanism to achieve dynamic weighting and interpretability analysis of influencing factors. At the same time, this application supports deployment on edge computing devices or monitoring stations to achieve real-time online monitoring and prediction of key forest areas, improving the system's timeliness and applicability.

[0015] In summary, existing monitoring methods have significant limitations in forest fire monitoring scenarios. Therefore, the solution described in this application can predict regional evapotranspiration in real time and accurately in complex and ever-changing forest ecosystems, so as to provide a reliable basis for fire risk assessment and early warning. Attached Figure Description

[0016] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 This is a flowchart of the forest fire monitoring method provided by the present invention.

[0019] Figure 2 Another flowchart of the forest fire monitoring method provided by the present invention.

[0020] Figure 3 The overall system workflow diagram of the forest fire monitoring method provided by the present invention is shown.

[0021] Figure 4 A comparison chart showing the model training and evaluation of the forest fire monitoring method provided by this invention.

[0022] Figure 5 The image shows the model training results of the forest fire monitoring method provided by this invention.

[0023] Figure 6 A feature importance and interpretability analysis diagram for the forest fire monitoring method provided by this invention.

[0024] Figure 7 A simplified diagram of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0025] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] It should be understood that, when used in this specification and the appended claims, the terms "comprising" and "including" indicate the presence of the described features, integrals, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0027] It should also be understood that the terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the invention. As used in this specification and the appended claims, the singular forms “a,” “an,” and “the” are intended to include the plural forms unless the context clearly indicates otherwise.

[0028] It should also be further understood that the term "and / or" as used in this specification and the appended claims refers to one or any combination of the associated listed items and all possible combinations, and includes such combinations.

[0029] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrases "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0030] In this specification and the appended claims, there may be multiple ways of expressing the same technical feature or technical term, such as using a superordinate generalization, a subordinate limitation, or a synonym substitution. Those skilled in the art can clearly understand the substantially the same technical meaning referred to by different expressions based on their professional knowledge and in conjunction with the overall content of the specification and the drawings. The differences in different expressions are only reflected in the diversity of words and do not constitute a substantial modification or limitation to the technical solution, nor will they affect the certainty of the scope of protection of this patent claim or the full disclosure of the technical content of the specification.

[0031] Example 1

[0032] Please see Figures 1 to 6 As shown, among them Figure 1This invention proposes a forest fire monitoring method based on a three-temperature model. The method includes: S1, extracting initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on radiation characteristics to obtain a standard three-temperature feature set; S2, generating temperature weight allocation parameters based on the standard three-temperature feature set, and obtaining evapotranspiration prediction values ​​based on the temperature weight allocation parameters; S3, comparing the evapotranspiration prediction values ​​with a preset fire risk threshold range, and outputting a visualized fire risk distribution map as the forest fire monitoring result.

[0033] The main technical problem addressed by this invention is how to accurately and in real-time predict regional evapotranspiration in complex and ever-changing forest ecosystems, providing a reliable basis for fire risk assessment and early warning. As is known to those skilled in the art, evapotranspiration is a physical quantity that can be predicted. Evapotranspiration, abbreviated as ET, is a core component of the water cycle, referring to the combined process by which liquid water is converted into gaseous water through surface evaporation and plant transpiration (i.e., soil evaporation, canopy intercepted evaporation, and plant transpiration). Predicting evapotranspiration refers to a physical model based on a three-temperature model, i.e., the difference in surface temperature, to estimate evapotranspiration. Its core principle is to quantify the latent heat flux consumed by the phase change of water by comparing the temperature difference between the actual surface and an ideal reference surface without evaporation / transpiration (actual surface temperature, reference surface temperature, and near-surface air temperature), thereby inverting evapotranspiration.

[0034] In one embodiment, see Figure 2 S1, which extracts initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on radiation characteristics to obtain a standard three-temperature feature set, includes: S31, extracting initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on remote sensing data of the target area collected by sensors; S32, performing data standardization processing on the initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features to generate a standard three-temperature feature set.

[0035] In one embodiment, see further. Figure 2S2, which generates temperature weight allocation parameters based on a standard three-temperature feature set and obtains evapotranspiration prediction values ​​based on the temperature weight allocation parameters, includes: S33, inputting the standard three-temperature feature set into an evapotranspiration prediction model fine-tuned based on the TabNet architecture, dynamically filtering and weighting the surface temperature, canopy temperature, and atmospheric temperature features to generate temperature weight allocation parameters; S34, performing weighted fusion processing on the standard three-temperature feature set based on the temperature weight allocation parameters to generate evapotranspiration prediction values, and evaluating the prediction accuracy based on preset root mean square error (RMSE), preset mean absolute error (MAE), and preset coefficient of determination (RSQ) to obtain an accuracy evaluation result; S35, performing dynamic feedback calibration on the temperature weight allocation parameters based on the accuracy evaluation result to generate optimized weight parameters, and performing prediction processing on the standard three-temperature feature set based on the optimized weight parameters to generate evapotranspiration prediction values.

[0036] In one embodiment, step S31, which involves collecting remote sensing data of the target area using sensors and extracting initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on radiation characteristics, includes: S41, collecting remote sensing data of the target area using sensors, performing radiation anomaly detection processing on the remote sensing data, and generating an effective radiation dataset; S42, calculating surface radiation features, canopy radiation features, and atmospheric radiation features based on the effective radiation dataset; and S43, extracting initial surface temperature features based on surface radiation features, initial canopy temperature features based on canopy radiation features, and initial atmospheric temperature features based on atmospheric radiation features.

[0037] In one embodiment, step S32, which performs data standardization processing on the initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features to generate a standard three-temperature feature set, includes: S51, performing data standardization processing on the initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features; S52, identifying spatial outliers in the initial surface temperature features, correcting the outlier data based on a geographic weighted regression algorithm, and generating optimized surface temperature features; S53, detecting vegetation shading interference in the initial canopy temperature features, correcting the canopy temperature through a radiative transfer model, and generating effective canopy temperature features; S54, performing time series smoothing processing on the initial atmospheric temperature features to eliminate transient fluctuation interference and generate stable atmospheric temperature features; and S55, fusing the optimized surface temperature features, effective canopy temperature features, and stable atmospheric temperature features to generate a standard three-temperature feature set.

[0038] In one embodiment, the dynamic screening and weight allocation of surface temperature, canopy temperature, and atmospheric temperature features to generate temperature weight allocation parameters includes: S61, dynamically screening and weighting surface temperature, canopy temperature, and atmospheric temperature features by calculating and parsing the surface temperature vector, canopy temperature vector, and atmospheric temperature vector through sequence attention; S62, calculating the surface temperature dominance coefficient, canopy temperature regulation coefficient, and atmospheric temperature control coefficient based on the surface temperature vector, canopy temperature vector, and atmospheric temperature control coefficient through a multi-layer attention network; S63, integrating the surface temperature dominance coefficient, canopy temperature regulation coefficient, and atmospheric temperature control coefficient to generate temperature weight allocation parameters.

[0039] In one embodiment, the step of evaluating prediction accuracy and obtaining accuracy evaluation results based on preset root mean square error (RMSE), preset mean absolute error (MAE), and preset coefficient of determination (RSQ) includes: S71, analyzing the fluctuation distribution characteristics of evapotranspiration prediction values ​​and extracting time series deviation features based on preset RMSE, MAE, and RSQ; S72, generating an overall deviation index corresponding to RMSE, a local anomaly index corresponding to MAE, and a model fit index corresponding to RSQ based on the time series deviation features; S73, converting the overall deviation index into an overall deviation level, the local anomaly index into a local anomaly level, and the model fit index into a model fit level according to preset index range mapping rules; and S74, integrating the overall deviation level, local anomaly level, and model fit level to generate accuracy evaluation results.

[0040] In one embodiment, the step of performing predictive processing on the standard three-temperature feature set based on optimized weight parameters to generate evapotranspiration prediction values ​​includes: S81, performing predictive processing on the standard three-temperature feature set based on optimized weight parameters, and performing dynamic weighted fusion of surface temperature, canopy temperature, and atmospheric temperature features through a sequence attention mechanism; S82, generating a surface feature enhancement vector, a canopy environment coupling vector, and an atmospheric temperature response vector based on the dynamic weighted fusion result; and S83, integrating the surface feature enhancement vector, the canopy environment coupling vector, and the atmospheric temperature response vector to generate evapotranspiration prediction values.

[0041] This application also proposes a monitoring system for implementing the forest fire monitoring method based on a three-temperature model as described in any of the above embodiments. By executing the forest fire monitoring method, the monitoring system can accurately predict regional evapotranspiration in real time within a complex and ever-changing forest ecosystem, thus providing a reliable basis for fire risk assessment and early warning.

[0042] The overall workflow of this invention is as follows: Figure 3The "System Overall Workflow Diagram" is shown. To evaluate the superiority of the present invention, a comparative experiment was also designed. Figure 4 The "Model Training and Evaluation Comparison Chart" visually demonstrates the performance comparison of the TabNet model used in this invention with other benchmark models on three key evaluation metrics: Root Mean Square Error (RMSE), Mean Absolute Error (MAE), and Coefficient of Determination (RSQ), also known as R². These other benchmark models include Linear Regression, Decision Tree, and Support Vector Regression (SVR). All comparison models used the same training and test sets and employed the same evaluation metrics. As can be seen from the chart, the TabNet model exhibits significant superiority across all evaluation metrics, corresponding to RMSE (Root Mean Square Error), MAE (Mean Absolute Error), and R² (Ratio of Determination).

[0043] In terms of root mean square error (RMSE), TabNet's RMSE value is approximately 0.0624, which is about 31.73% lower than linear regression (approximately 0.0914), about 20.61% lower than decision trees (approximately 0.0786), and about 17.57% lower than SVR (approximately 0.0757). This indicates that TabNet performs better in terms of overall accuracy of predictions.

[0044] In terms of mean absolute error (MAE), the TabNet model has an MAE value of approximately 0.0392, which is also significantly lower than other comparative models. Specifically, TabNet's MAE is about 36.26% lower than linear regression (about 0.0615), about 9.05% lower than decision trees (about 0.0431), and a significant 38.85% lower than SVR (about 0.0641), further demonstrating TabNet's advantage in controlling the magnitude of prediction error.

[0045] Regarding the coefficient of determination (R²), the TabNet model's R² value is approximately 0.6762, significantly higher than other models, indicating its stronger explanatory power for data variation. Compared to linear regression (R² approximately 0.3041), TabNet's R² is improved by approximately 122.36%; compared to decision trees (R² approximately 0.4861), it is improved by approximately 39.11%; and compared to SVR (R² approximately 0.5227), it is also improved by approximately 29.37%.

[0046] These comparative results clearly demonstrate the superiority of the method of this invention over some existing methods in the task of evapotranspiration prediction. Furthermore, to verify the practicality of the model, a graphical interface is provided that can process evapotranspiration predictions based on the trained model individually or in batches, thus forming a closed-loop system from data acquisition to model prediction.

[0047] The distribution of prediction errors can be determined by... Figure 5 We will analyze the "Prediction Error Distribution" histogram in the "Model Training Results" chart. The chart shows that the prediction error (predicted value - true value) is mainly concentrated around 0, exhibiting an approximately normal distribution. This indicates that the model's prediction results do not have obvious systematic biases, and the absolute value of the error for most prediction samples is relatively small.

[0048] The results of the feature importance and interpretability analysis are as follows: Figure 6 The “Feature Importance and Interpretability Analysis Chart” shows the global feature importance analysis results. The chart, presented as a bar chart, visually illustrates the relative contribution of each input feature (labeled F0, F1, F2) to evapotranspiration prediction across the entire dataset. According to the chart, feature F2 (corresponding to temp3 in the code, atmospheric temperature) has the highest importance, followed by F1 (temp2, canopy temperature), and finally F0 (temp1, surface temperature). This indicates that all features have a positive correlation with interpretability in evapotranspiration prediction.

[0049] In summary, existing monitoring methods have significant limitations in forest fire monitoring scenarios. Therefore, the solution described in this application can predict regional evapotranspiration in real time and accurately in complex and ever-changing forest ecosystems, so as to provide a reliable basis for fire risk assessment and early warning.

[0050] Example 2

[0051] Please see Figure 7 , Figure 7 This invention provides a block diagram of an electronic device. The electronic device can be a terminal or a server. The terminal can be a smartphone, tablet computer, laptop computer, desktop computer, personal digital assistant, wearable device, or other electronic device with communication capabilities. It includes a processor 111, a communication interface 112, a memory 113, and a communication bus 114. The processor 111, communication interface 112, and memory 113 communicate with each other via the communication bus 114.

[0052] Memory 113 is used to store computer programs.

[0053] In one embodiment of the present invention, the processor 111, when executing the program stored in the memory 113, implements the method provided in any of the foregoing method embodiments.

[0054] It should be understood that in the embodiments of this application, processor 111 may be a central processing unit (CPU), and processor 502 may also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0055] It will be understood by those skilled in the art that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program may be stored in a storage medium, which is a computer-readable storage medium. The computer program is executed by at least one processor in the computer system to implement the process steps of the embodiments of the above methods.

[0056] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions, but such implementations should not be considered beyond the scope of this invention.

[0057] In the several embodiments provided by this invention, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For example, the division of each unit is only a logical functional division, and there may be other division methods in actual implementation. For example, units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0058] The steps in the method of this invention can be adjusted, merged, or reduced in order according to actual needs. The units in the device of this invention can be merged, divided, or reduced according to actual needs. Furthermore, the functional units in the various embodiments of this invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0059] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, a terminal, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention.

[0060] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0061] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Since these modifications and variations fall within the scope of the claims and their equivalents, this invention also intends to include these modifications and variations.

[0062] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A forest fire monitoring method based on a three-temperature model, characterized in that, The method includes: S1, based on radiation characteristics, extract initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics to obtain a standard set of three temperature characteristics; S2, generates temperature weight allocation parameters based on the standard three-temperature feature set, and obtains the evapotranspiration prediction value based on the temperature weight allocation parameters; S3 compares the predicted evapotranspiration values ​​with the preset fire risk threshold range and outputs a visualized fire risk distribution map as the forest fire monitoring result.

2. The method according to claim 1, characterized in that, S1 extracts initial surface temperature features, initial canopy temperature features, and initial atmospheric temperature features based on radiation characteristics to obtain a standard three-temperature feature set, including: S31, based on remote sensing data of the target area collected by the sensor, extracts the initial surface temperature characteristics, initial canopy temperature characteristics and initial atmospheric temperature characteristics according to the radiation characteristics; S32 performs data standardization processing on the initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics to generate a standard set of three temperature characteristics.

3. The method according to claim 2, characterized in that, S2 generates temperature weighting parameters based on a standard three-temperature feature set, and obtains evapotranspiration prediction values ​​based on the temperature weighting parameters, including: S33 inputs the standard three-temperature feature set into the evapotranspiration prediction model based on the TabNet architecture for fine-tuning, and dynamically filters and assigns weights to the surface temperature, canopy temperature and atmospheric temperature features to generate temperature weight assignment parameters. S34. Based on the temperature weight allocation parameters, the standard three-temperature feature set is subjected to weighted fusion processing to generate evapotranspiration prediction values. Based on the preset root mean square error RMSE, the preset mean absolute error MAE, and the preset coefficient of determination RSQ, the prediction accuracy is evaluated to obtain the accuracy evaluation results. S35 performs dynamic feedback calibration on the temperature weight allocation parameters based on the accuracy assessment results, generates optimized weight parameters, and performs prediction processing on the standard three-temperature feature set based on the optimized weight parameters to generate evapotranspiration prediction values.

4. The method according to claim 3, characterized in that, S31, based on remote sensing data of the target area collected by sensors, extracts initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics according to radiation characteristics, including: S41, Based on the remote sensing data of the target area collected by the sensor, perform radiation anomaly detection processing on the remote sensing data to generate an effective radiation dataset; S42, calculate the surface radiation characteristics, canopy radiation characteristics and atmospheric radiation characteristics based on the effective radiation dataset; S43, extracts initial surface temperature features based on surface radiation features, extracts initial canopy temperature features based on canopy radiation features, and extracts initial atmospheric temperature features based on atmospheric radiation features.

5. The method according to claim 4, characterized in that, S32 performs data standardization processing on the initial surface temperature characteristics, initial canopy temperature characteristics, and initial atmospheric temperature characteristics to generate a standard set of three temperature characteristics, including: S51, perform data standardization processing on the initial surface temperature characteristics, initial canopy temperature characteristics and initial atmospheric temperature characteristics; S52, identify spatial anomalies in the initial surface temperature characteristics, correct the abnormal data based on the geographic weighted regression algorithm, and generate optimized surface temperature characteristics; S53 detects vegetation shading interference in the initial canopy temperature characteristics, corrects the canopy temperature through a radiative transfer model, and generates effective canopy temperature characteristics; S54 performs time-series smoothing on the initial atmospheric temperature characteristics to eliminate transient fluctuations and generate stable atmospheric temperature characteristics. S55 integrates and optimizes surface temperature characteristics, effective canopy temperature characteristics, and stable atmospheric temperature characteristics to generate a standard three-temperature feature set.

6. The method according to claim 5, characterized in that, The dynamic screening and weighting of surface temperature, canopy temperature, and atmospheric temperature characteristics to generate temperature weighting parameters includes: S61 dynamically filters and assigns weights to the characteristics of surface temperature, canopy temperature, and atmospheric temperature, and calculates and analyzes the surface temperature vector, canopy temperature vector, and atmospheric temperature vector through sequence attention. S62 calculates the dominant coefficient of surface temperature, the regulation coefficient of canopy temperature, and the dominant coefficient of atmospheric temperature based on the surface temperature vector, the canopy temperature vector, and the atmospheric temperature vector through a multi-layer attention network. S63 integrates the dominant surface temperature coefficient, the canopy temperature regulation coefficient, and the dominant atmospheric temperature coefficient to generate temperature weighting parameters.

7. The method according to claim 6, characterized in that, The prediction accuracy is evaluated based on preset root mean square error (RMSE), preset mean absolute error (MAE), and preset coefficient of determination (RSQ), and the accuracy evaluation results are obtained, including: S71, based on the preset root mean square error RMSE, preset mean absolute error MAE, and preset coefficient of determination RSQ, analyze the fluctuation distribution characteristics of evapotranspiration forecast values ​​and extract time series deviation characteristics. S72, based on the time series deviation characteristics, generates the overall deviation index corresponding to the root mean square error RMSE, the local anomaly index corresponding to the mean absolute error MAE, and the model fit index corresponding to the coefficient of determination RSQ. S73, according to the preset index range mapping rules, convert the overall deviation index into the overall deviation level, the local anomaly index into the local anomaly level, and the model fit index into the model fit level. S74, integrate the overall deviation level, local anomaly level and model fitting level to generate accuracy evaluation results.

8. The method according to claim 7, characterized in that, The step of performing prediction processing on the standard three-temperature feature set based on optimized weight parameters to generate evapotranspiration prediction values ​​includes: S81 performs prediction processing on the standard three-temperature feature set based on optimized weight parameters, and performs dynamic weighted fusion of surface temperature, canopy temperature and atmospheric temperature features through a sequence attention mechanism. S82 generates a surface feature enhancement vector, a canopy environment coupling vector, and an atmospheric temperature response vector based on the dynamic weighted fusion results. S83 integrates surface feature enhancement vectors, canopy environment coupling vectors, and atmospheric temperature response vectors to generate evapotranspiration prediction values.

9. A monitoring system, characterized in that, The monitoring system is used to implement the forest fire monitoring method based on the three-temperature model as described in any one of claims 1-8.

Images