A Remote Sensing Inversion Method and Device for Moisture Content of Live Fuel Based on an Integrated Model

Abstract

An embodiment of the present invention discloses a method and device for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model. The method includes: S1: Obtaining the moisture content of live combustibles and remote sensing data to establish a remote sensing inversion data set; S2: Using a pre-constructed integrated model to extract remote sensing features from the remote sensing inversion data set; S3: Predicting the moisture content of live combustibles based on the remote sensing features and obtaining a prediction result. The present invention improves the inversion efficiency and performance, is applicable to research areas with diverse environments and complex combustible types, and has broad application prospects.

Description

Technical Field

[0001] The present invention relates to the technical field of predicting the moisture content of live combustibles. More specifically, it relates to a method and device for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model. Background Art

[0002] Wildfires are common natural disasters worldwide, which can pollute the atmosphere, damage the ecological environment, and even cause catastrophic damage to personal property and lives. The moisture content of live combustibles has been proven to be the main driving factor for plant flammability and fire behavior, and is an important indicator affecting the occurrence and spread of wildfires.

[0003] Currently, there are various methods for remotely sensing and retrieving the moisture content of live combustibles based on remote sensing data, mainly including empirical statistics-based, radiative transfer model-based, and machine learning-based retrieval methods. Empirical methods for remotely sensing and retrieving the moisture content of live combustibles usually rely on statistical fitting between the moisture content of live combustibles measured in the field and spectral signals based on reflectance data. An alternative to the empirical method is the method based on the radiative transfer model, which estimates the moisture content of live combustibles using a physical model based on a set of simulated scenarios. This method usually involves comparing the observed reflectance of each pixel with the reflectance simulated in a lookup table and assigning the parameters of the most similar simulated spectrum to each pixel. Machine learning methods include random forest, gradient boosting regression, artificial neural network, and deep learning methods. Empirical statistical methods have simple steps and are widely applied but have poor generalization ability and are only applicable to specific locations. Compared with empirical models, radiative transfer models and machine learning models have higher accuracy, but require manual feature selection and model calibration, and the methods are relatively complex. Summary of the Invention

[0004] In view of this, the first embodiment of the present invention provides a method for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model, including:

[0005] S1: Obtain the moisture content of live combustibles and remote sensing data to establish a remote sensing retrieval dataset;

[0006] S2: Use a pre-constructed integrated model to perform remote sensing feature extraction on the remote sensing retrieval dataset;

[0007] S3: Predict the moisture content of live combustibles based on the remote sensing features and obtain the prediction result.

[0008] In a specific embodiment, pre-constructing the integrated model includes:

[0009] Construct the first layer of the integrated model using multiple base learners;

[0010] Construct the second layer of the integrated model using a meta-learner.

[0011] In a specific embodiment, the remote sensing feature extraction from the remote sensing inversion data set includes:

[0012] Stratified random sampling is performed according to land cover types, and the remote sensing inversion data set is divided into a training set and a test set;

[0013] The first layer of the ensemble model receives the training set and the test set and performs remote sensing feature extraction;

[0014] Based on the results of the remote sensing features output by the first layer, the second layer of the ensemble model performs feature fusion to complete the stacked fusion of the model.

[0015] In a specific embodiment, the base learners include: an LSTM model, a TCN model, and a TCN-LSTM model, where

[0016] The LSTM model extracts remote sensing data features in terms of time;

[0017] The TCN model extracts remote sensing data features in terms of space;

[0018] The TCN-LSTM model includes a residual block structure of TCN, an LSTM layer, a Flatten layer, and a fully connected layer, where

[0019] The residual block structure of TCN includes a causal convolutional layer, which extracts features based on the moisture content of live fuel and multi-source remote sensing variables included in the remote sensing data, obtains the correlation features between the remote sensing variables and the moisture content of live fuel, and compresses the number of parameters through dilated convolution;

[0020] The LSTM layer performs feature extraction using the Dropout mechanism;

[0021] The Flatten layer compresses the output matrix into one dimension;

[0022] The ensemble model extracts the non-linear features between the remote sensing variables and the moisture content of live fuel and realizes moisture content prediction.

[0023] In a specific embodiment, pre-constructing the ensemble model includes: obtaining predicted values through the ensemble model using training set samples, calculating the loss function value by comparing with the true moisture content values of the training set samples;

[0024] Calculating the gradient through backpropagation, updating the parameters according to the gradient, substituting the parameters into the model, and repeating the above process. When the loss function value is less than the expected value, stop the training process to obtain the parameters of the ensemble model.

[0025] In a specific embodiment, in actual application, the live fuel moisture content is predicted based on the fusion features output by the second layer of the integration model, and a prediction result is obtained.

[0026] In a specific embodiment, the method further includes:

[0027] S4: Evaluate the prediction result.

[0028] In a specific embodiment, S4 includes:

[0029] Calculate the deviation, coefficient of determination, root mean square error, and unbiased root mean square error between the prediction result calculated from the test set samples and the true moisture content to evaluate the prediction result of the live fuel moisture content, where

[0030] The formula for the deviation is:

[0031] The formula for the coefficient of determination is:

[0032] The formula for the root mean square error is:

[0033] The formula for the unbiased root mean square error is:

[0034] Where and y i respectively represent the live fuel moisture content value and the measured value, and are respectively the average values of the observed and estimated live fuel moisture content.

[0035] The second embodiment of the present invention provides a computer device, including a processor and a memory storing a computer program. When the processor executes the program, the method described in the first embodiment is implemented.

[0036] The third embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, the method described in the second embodiment is implemented.

[0037] The beneficial effects of the present invention are as follows:

[0038] The present invention provides a method and device for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model. By using the pre-constructed integrated model to extract remote sensing features from the established remote sensing inversion data set, predicting the moisture content of live combustibles according to the remote sensing features, and obtaining the prediction results, the inversion of the moisture content of live combustibles is realized, the diversity and generalization of the live combustible moisture content inversion model are enhanced, the inversion efficiency and performance are improved, and it is applicable to research areas with diverse environments and complex combustible types, having broad application prospects. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0040] Figure 1 The flowchart showing the method for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model according to an embodiment of the present invention;

[0041] Figure 2 The schematic diagram showing the inversion result of the remote sensing inversion data set according to an embodiment of the present invention;

[0042] Figure 3 The schematic diagram showing the structure of a computer device according to another embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0043] In order to make the technical solutions and advantages of the present invention clearer, the following will further describe the embodiments of the present invention in detail with reference to the drawings.

[0044] As Figure 1 shown, an embodiment of the present invention provides a method for remotely sensing and retrieving the moisture content of live combustibles based on an integrated model, including:

[0045] S1: Obtain the moisture content of live combustibles and remote sensing data to establish a remote sensing inversion data set;

[0046] S2: Use the pre-constructed integrated model to extract remote sensing features from the remote sensing inversion data set;

[0047] S3: Predict the moisture content of live combustibles according to the remote sensing features and obtain the prediction results.

[0048] In this embodiment, by using a pre-constructed integrated model to extract remote sensing features from the established remote sensing inversion dataset, predicting the moisture content of live fuels based on the remote sensing features, and obtaining the prediction result, the inversion of the moisture content of live fuels is realized, the diversity and generalization of the live fuel moisture content inversion model are enhanced, the inversion efficiency and performance are improved, and it is applicable to research areas with diverse environments and complex combustible species, having broad application prospects.

[0049] In a specific embodiment, the remote sensing inversion method for the moisture content of live fuels based on the integrated model specifically includes:

[0050] S1: Obtain the moisture content of live fuels and remote sensing data to establish a remote sensing inversion dataset;

[0051] In this embodiment, the remote sensing data includes MODIS data, Landsat-8 data, Sentinel-1A data, and auxiliary data including terrain, soil, and vegetation structure. Among them, the live fuel moisture content data can be sourced from the US National Fire Fuel Moisture Database; the MODIS data can be from the MCD43A4 product jointly observed by MODIS Terra and Aqua; the Landsat data can be from the 16-day surface reflectance data of Landsat-8; the Sentinel-1A data can be from the Sentinel-1A ground range monitoring data; the first major category of auxiliary data includes 3 types of soil data such as silt, sand, and clay content, the second major category is the canopy height and land cover information measuring the vegetation structure, and the third major category is terrain data. The specific acquisition methods are well-known to those skilled in the art and are not overly restricted in this application, as long as they can meet the requirements for obtaining the moisture content of live fuels and remote sensing data.

[0052] S2: Use the pre-constructed integrated model to extract remote sensing features from the remote sensing inversion dataset;

[0053] In a specific embodiment, pre-constructing the integrated model includes:

[0054] Use multiple base learners to construct the first layer of the integrated model;

[0055] Use a meta-learner to construct the second layer of the integrated model.

[0056] In this embodiment, the integrated model includes a base learner and a meta-learner. Among them, the base learner has high prediction accuracy, and at the same time, there should be a variety of base learners. By combining multiple base learners, feature information helpful for the inversion of live fuel moisture content is extracted from the input variables of numerous remote sensing data. The predicted values of the base learner and the true value labels of the original data set are used as the data set of the second-layer meta-learner to train the meta-learner, and the prediction result of the meta-learner is the final output.

[0057] Further, the base learner includes: an LSTM model, a TCN model, and a TCN-LSTM model, where

[0058] The LSTM model extracts remote sensing data features in terms of time;

[0059] The TCN model extracts remote sensing data features in terms of space;

[0060] The TCN-LSTM model includes a residual block structure of TCN, an LSTM layer, a Flatten layer, and a fully connected layer, where

[0061] The residual block structure of TCN includes a causal convolutional layer, which extracts features according to the live fuel moisture content and the multi-source remote sensing variables included in the remote sensing data, obtains the relevant features between the remote sensing variables and the live fuel moisture content, and compresses the number of parameters through dilated convolution;

[0062] The LSTM layer uses the Dropout mechanism for feature extraction;

[0063] The Flatten layer compresses the output matrix into one dimension;

[0064] The integrated model extracts the non-linear features between the remote sensing variables and the live fuel moisture content and realizes moisture content prediction.

[0065] Using the training set samples, the predicted values are obtained through the above steps, and the loss function value is calculated by comparing with the true moisture content values of the training set samples;

[0066] The gradient is calculated through backpropagation, the parameters are updated according to the gradient, and the parameters are substituted into the model. Repeat the above process. When the loss function value is less than the expected value, stop the training process and obtain the parameters of the integrated model.

[0067] In this embodiment, first, the live fuel moisture content data and multi-source remote sensing variables are used as inputs. Feature extraction is performed through the causal convolutional layers included in the residual structure of the TCN to obtain the relevant features of the remote sensing variables and the live fuel moisture content, and the number of parameters can be compressed through dilated convolution to achieve the effect of data dimensionality reduction. Then, multiple LSTM layers combined with the Dropout mechanism are used for prediction to prevent overfitting. Next, through the Flatten layer, the output matrix is compressed into one dimension to facilitate the connection of the subsequent fully connected layer. Finally, the fully connected layer extracts the non-linear features between the data, so as to effectively extract the remote sensing data features from the time and space aspects through the LSTM model and the TCN model, with the ability of long time series memory. The TCN-LSTM model extracts the remote sensing data features from a global perspective, realizes the high-precision inversion of the live fuel moisture content, improves the efficiency and performance of the live fuel moisture content inversion, is applicable to research areas with diverse environments and complex combustible types, and has broad application prospects.

[0068] After the integrated model is pre-constructed, the remote sensing inversion of the live fuel moisture content is started, and the integrated model is used to perform feature learning on the remote sensing inversion data set.

[0069] In a specific embodiment, the remote sensing feature extraction of the remote sensing inversion data set includes:

[0070] Stratified random sampling is performed according to the land cover type, and the remote sensing inversion data set is divided into a training set and a test set;

[0071] The first layer of the integrated model receives the training set and the test set and performs remote sensing feature extraction;

[0072] The second layer of the integrated model proposes results based on the remote sensing features output by the first layer, performs feature fusion, and completes the stacked fusion of the model.

[0073] In this embodiment, stratified random sampling is performed according to the land cover type, such as according to typical types including broad-leaved deciduous forests, coniferous evergreen forests, shrubs, grasslands, and sparse vegetation areas, etc., and the remote sensing inversion data set is divided into a training set and a test set to ensure that the land cover type distributions of the training set and the test set are the same, so that the integrated model has better generalization performance when predicting the moisture content in different cover types.

[0074] S3: Predict the live fuel moisture content according to the remote sensing features and obtain the prediction result.

[0075] In actual application, the live fuel moisture content is predicted according to the fusion features output by the second layer of the integrated model, and the prediction result is obtained.

[0076] S4: Evaluate the prediction result.

[0077] In a specific embodiment, S4 includes:

[0078] Evaluating the predicted result of the live fuel moisture content by calculating the deviation, coefficient of determination, root mean square error, and unbiased root mean square error between the predicted result calculated from the test set samples and the true moisture content, where

[0079] The formula for calculating the deviation is:

[0080] The formula for calculating the coefficient of determination is:

[0081] The formula for calculating the root mean square error is:

[0082] The formula for calculating the unbiased root mean square error is:

[0083] In the formula, and y i represent the live fuel moisture content value and the measured value respectively, and are the average values of the observed and estimated live fuel moisture content respectively.

[0084] Those skilled in the art can understand that the smaller the value of the deviation Bias, the smaller the deviation between the inversion result and the observed value, and the more significant the effect; the larger the value of the coefficient of determination R 2 the higher the correlation between the inversion result and the observed value, and the more significant the effect; the smaller the value of the root mean square error RMSE, the smaller the overall error between the inversion result and the observed value, and the more significant the effect; the smaller the value of the unbiased root mean square error ubRMSE, the smaller the error after removing the bias, and the more significant the effect.

[0085] In a specific embodiment, Table 1 shows the comparison between the existing method for inverting the live fuel moisture content and the method for remotely sensing the inversion of the live fuel moisture content based on the integrated model in this embodiment.

[0086] From the content of Table 1, it can be obtained that when calculating by the method for remotely sensing the inversion of the live fuel moisture content based on the integrated model proposed in this embodiment, the coefficient of determination R 2The values are 0.85 for Bias, 18.88 for Root Mean Square Error (RMSE), and 17.99 for Unbiased Root Mean Square Error (ubRMSE). Compared with the method of using Partial Least Squares Regression for live fuel moisture content inversion in the prior art, it has increased by 47%, 18.19%, and 17.87% respectively. Compared with the inversion method using the TCN-LSTM model in the prior art, it has increased by 4%, 2.1%, and 2.56% respectively. It can be seen from this that the method proposed in this embodiment enhances the diversity and generalization of the live fuel moisture content inversion model, improves the inversion efficiency and performance, is applicable to research areas with diverse environments and complex combustible species, and has broad application prospects.

[0087] Table 1

[0088]

[0089] To verify the effectiveness of the live fuel moisture content inversion method based on the integrated model for live fuel moisture content inversion when there are multiple different types of remote sensing data, in a specific example, predictions are made on a multi-source remote sensing data set, and the prediction results are as Figure 2 shown. Among them, the performance of the integrated model is better than that of the deep learning model and the empirical statistical model. The inversion results are closer to the measured results, with smaller Bias, RMSE, and ubRMSE, and at the same time, the R 2 is higher. The integrated model combines the advantages of TCN causal convolutional temporal feature extraction and LSTM's ability to remember long-term information. The integrated model can effectively improve the LFMC inversion accuracy, and the inversion results show that the integrated model is more suitable for live fuel moisture content inversion. Based on the above results, the live fuel moisture content inversion method based on the integrated model has high effectiveness and stability when performing live fuel moisture content inversion in the case of complex and diverse types of remote sensing data.

[0090] Another embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, in practical applications, the computer-readable storage medium may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this embodiment, the computer-readable storage medium may be any tangible medium that contains or stores a program, and the program may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0091] As Figure 3 shown, a schematic structural diagram of a computer device provided by another embodiment of the present invention. Figure 3 The computer device 12 shown is only an example and should not impose any limitations on the functions and usage scope of the embodiments of the present invention.

[0092] As Figure 3 shown, the computer device 12 is presented in the form of a general-purpose computing device. The components of the computer device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0093] The bus 18 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the various bus structures. For example, these architectures include, but are not limited to, an Industry Standard Architecture (ISA) bus, a Micro Channel Architecture (MAC) bus, an Enhanced ISA bus, a Video Electronics Standards Association (VESA) local bus, and a Peripheral Component Interconnect (PCI) bus.

[0094] The computer device 12 typically includes a variety of computer system-readable media. These media may be any available media that can be accessed by the computer device 12, including volatile and non-volatile media, removable and non-removable media.

[0095] System memory 28 may include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The computer device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 can be used for reading and writing on non-removable, non-volatile magnetic media ( Figure 3 not shown, commonly referred to as a "hard disk drive"). Although Figure 3 not shown in the figure, a disk drive for reading and writing on removable non-volatile disks (such as a "floppy disk") and an optical disk drive for reading and writing on removable non-volatile optical disks (such as CD-ROM, DVD-ROM or other optical media) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data media interfaces. The memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0096] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in the memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include an implementation of a network environment. The program modules 42 generally perform the functions and / or methods in the embodiments described in the present invention.

[0097] The computer device 12 can also communicate with one or more external devices 14 (such as a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the computer device 12, and / or communicate with any device that enables the computer device 12 to communicate with one or more other computing devices (such as a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Also, the computer device 12 can communicate with one or more networks (such as a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As Figure 3 shown, the network adapter 20 communicates with other modules of the computer device 12 through the bus 18. It should be understood that although Figure 3 not shown in the figure, other hardware and / or software modules can be used in conjunction with the computer device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0098] The processor unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, for example, implementing a method for remotely sensing and retrieving the moisture content of live fuel based on an integrated model provided by the embodiments of the present invention.

[0099] Obviously, the above-mentioned embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or variations can be made based on the above description. It is impossible to enumerate all the implementation manners here. Any obvious changes or variations derived from the technical solutions of the present invention still fall within the protection scope of the present invention.

Claims

1. A remote sensing inversion method for the moisture content of living combustibles based on an integrated model, characterized in that, Including: S1: Obtain the moisture content of live combustibles and remote sensing data to establish a remote sensing inversion data set; S2: Use a pre-constructed integrated model to extract remote sensing features from the remote sensing inversion data set; S3: Predict the moisture content of live combustibles based on the remote sensing features and obtain the prediction result; Pre-constructing the integrated model includes: Construct the first layer of the integrated model using multiple base learners; Construct the second layer of the integrated model using a meta-learner; The base learners include: LSTM model, TCN model, and TCN-LSTM model, where The LSTM model extracts remote sensing data features in terms of time; The TCN model extracts remote sensing data features in terms of space; The TCN-LSTM model includes a residual block structure of TCN, an LSTM layer, a Flatten layer, and a fully connected layer, where The residual block structure of TCN includes a causal convolutional layer, which extracts features based on the moisture content of live combustibles and multi-source remote sensing variables included in the remote sensing data, obtains the relevant features between the remote sensing variables and the moisture content of live combustibles, and compresses the parameter quantity through dilated convolution; The LSTM layer combines the Dropout mechanism for feature extraction; The Flatten layer compresses the output matrix into one dimension; The integrated model extracts the non-linear features between the remote sensing variables and the moisture content of live combustibles and realizes the moisture content prediction.

2. The method according to claim 1, wherein Extracting remote sensing features from the remote sensing inversion data set includes: Stratified random sampling according to land cover types, and dividing the remote sensing inversion data set into a training set and a test set; The first layer of the integrated model receives the training set and the test set and performs remote sensing feature extraction; The second layer of the integrated model proposes results based on the remote sensing features output by the first layer, performs feature fusion, and completes the stacked fusion of the model.

3. The method according to claim 1, wherein Pre-constructing the integrated model includes: obtaining predicted values through the integrated model using training set samples, calculating the loss function value by comparing with the true moisture content values of the training set samples; Calculating the gradient through backpropagation, updating the parameters according to the gradient, substituting the parameters into the model, repeating the above process, and stopping the training process when the loss function value is less than the expected value to obtain the parameters of the integrated model.

4. The method according to claim 2, wherein In actual application, predicting the moisture content of live combustibles based on the fusion features output by the second layer of the integrated model and obtaining the prediction result.

5. The method according to any one of claims 1 to 3, characterized in that, The method further includes: S4: Evaluate the prediction result.

6. The method according to claim 5, characterized in that The S4 includes: Calculating the deviation, coefficient of determination, root mean square error, and unbiased root mean square error between the prediction result calculated from the test set samples and the true moisture content to evaluate the prediction result of the moisture content of live combustibles, where The calculation formula for the deviation is as follows: The calculation formula for the coefficient of determination is as follows: The calculation formula for the root mean square error is as follows: The calculation formula for the unbiased root mean square error is as follows: In the formula, and y i represent the moisture content value of live fuel and the measured value respectively, and are the average moisture content of live fuel observed and estimated respectively.

7. A computer device, comprising a processor and a memory storing a computer program, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1-5.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method according to any one of claims 1-5.

Images