Fire monitoring and alarm system and method based on intelligent assessment of fire risk level

Abstract

The present application relates to the field of intelligent fire risk alarm, and specifically discloses a fire monitoring and alarm system and method based on intelligent fire risk level assessment, which uses artificial intelligence technology based on the field of deep learning to extract and analyze features of forest remote sensing images and forest meteorological data of a predetermined time period to obtain classification results representing forest fire risk levels, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and discovered, and alarm information can be generated, which can avoid false alarms and missed alarms caused by human subjective factors.

Description

Technical Field

[0001] The present application relates to the field of intelligent fire risk alarm, and more specifically to a fire monitoring and alarm system and method based on intelligent assessment of fire risk level. Background Art

[0002] Forest fire is one of the major disasters facing forest resources. It is characterized by strong suddenness, difficulty in control, wide impact and great destructiveness. They not only pose a serious threat to people's lives and property safety, but also have a significant impact on the carbon cycle process of terrestrial ecosystems. Once it occurs, forest fires will spread rapidly, causing huge economic losses and ecological damage, and its consequences are usually difficult to estimate. Therefore, the prevention and effective management of forest fires are crucial to protecting the environment and ensuring human well-being. Traditional means of forest fire prediction are often obtained through data acquisition that can be carried out through multiple channels and means of forest fire monitoring, such as satellite remote sensing monitoring, high-definition cameras, drone cruises, manual inspections, etc. Fire prediction often relies on expert experience and human judgment on whether there is a possibility of fire risk. This may lead to strong subjectivity in the prediction results and increase the risk of false alarms and missed reports.

[0003] Therefore, an optimized forest fire monitoring and alarm solution is expected. Summary of the invention

[0004] In order to solve the above technical problems, the present application is proposed. The embodiment of the present application provides a fire monitoring and alarm system and method based on intelligent assessment of fire risk level, which uses artificial intelligence technology based on the field of deep learning to extract and analyze features of forest remote sensing images and forest meteorological data of a predetermined time period to obtain a classification result representing the forest fire risk level, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and discovered, and alarm information can be generated, which can avoid false alarms and missed alarms caused by human subjective factors.

[0005] According to one aspect of the present application, a fire monitoring and alarm method based on intelligent assessment of fire risk level is provided, which includes:

[0006] Obtain forest remote sensing images and forest meteorological data for a predetermined period of time;

[0007] Extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector;

[0008] Extracting the time series characteristics of the forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector;

[0009] The spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector are comprehensively analyzed to obtain the forest fire risk level.

[0010] According to another aspect of the present application, a fire monitoring and alarm system based on intelligent assessment of fire risk level is also provided, which includes:

[0011] A forest data acquisition module is used to acquire forest remote sensing images and forest meteorological data for a predetermined period of time;

[0012] A forest remote sensing feature extraction module, used for extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector;

[0013] A forest meteorological data extraction module, used to extract the time series characteristics of the forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector;

[0014] The forest fire risk level analysis module is used to comprehensively analyze the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector to obtain the forest fire risk level.

[0015] Compared with the prior art, the fire monitoring and alarm system and method based on intelligent fire risk level assessment provided by the present application uses artificial intelligence technology based on the field of deep learning to extract and analyze features of forest remote sensing images and forest meteorological data of a predetermined time period to obtain classification results representing forest fire risk levels, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and discovered, and alarm information can be generated, which can avoid false alarms and missed alarms caused by human subjective factors. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] By describing the embodiments of the present application in more detail in conjunction with the accompanying drawings, the above and other purposes, features and advantages of the present application will become more apparent. The accompanying drawings are used to provide a further understanding of the embodiments of the present application and constitute a part of the specification. Together with the embodiments of the present application, they are used to explain the present application and do not constitute a limitation of the present application. In the accompanying drawings, the same reference numerals generally represent the same components or steps.

[0017] Figure 1 The present invention is a flowchart of a fire monitoring and alarm method based on intelligent fire risk level assessment according to an embodiment of the present application.

[0018] Figure 2 This is a flowchart for extracting the image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector in the fire monitoring and alarm method based on intelligent fire risk level assessment according to an embodiment of the present application.

[0019] Figure 3This is a flow chart for extracting the time series characteristics of forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector in the fire monitoring and alarm method based on intelligent fire risk level assessment according to an embodiment of the present application.

[0020] Figure 4 It is a block diagram of a fire monitoring and alarm system based on intelligent assessment of fire risk level according to an embodiment of the present application.

[0021] Figure 5 is a block diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0022] Below, the exemplary embodiments according to the present application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application, and it should be understood that the present application is not limited to the exemplary embodiments described here.

[0023] Figure 1 FIG. 1 is a flow chart of a fire monitoring and alarm method based on intelligent fire risk level assessment according to an embodiment of the present application. Figure 1 As shown, according to the fire monitoring and alarm method based on intelligent assessment of fire risk level according to the embodiment of the present application, the method includes: S110, acquiring a forest remote sensing image and forest meteorological data of a predetermined time period; S120, extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector; S130, extracting time series features of the forest meteorological data of the predetermined time period to obtain a forest meteorological feature vector; and S140, comprehensively analyzing the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector to obtain a forest fire risk level.

[0024] In step S110, forest remote sensing images and forest meteorological data for a predetermined time period are obtained. As mentioned in the above background technology, traditional forest fire prediction methods often use multiple channels and methods to obtain data for forest fire monitoring, such as satellite remote sensing monitoring, high-definition cameras, drone cruises, manual inspections, etc. Fire prediction often relies on expert experience and human judgment on whether there is a possibility of fire risk. This may lead to strong subjectivity in the prediction results and increase the risk of false alarms and missed alarms. Therefore, an optimized forest fire monitoring and alarm solution is expected.

[0025] In response to the above technical problems, a fire monitoring and alarm solution based on intelligent assessment of fire risk level is proposed. It uses artificial intelligence technology based on deep learning to extract and analyze features of forest remote sensing images and forest meteorological data in a predetermined time period to obtain classification results representing forest fire risk levels, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and generated, and false alarms and missed alarms caused by human subjective factors can be avoided.

[0026] Specifically, first, forest remote sensing images and forest meteorological data for a predetermined time period are obtained. The forest meteorological data refers to real-time data such as air pressure, wind speed, temperature and humidity, soil moisture, and historical data such as lightning strikes and extreme climate. Through forest remote sensing images, real-time data such as moisture content and density of combustibles can be obtained, as well as elements such as local vegetation distribution and terrain characteristics. Obtaining forest remote sensing images and forest meteorological data for a predetermined time period is crucial for assessing and predicting forest fire risk levels. Forest remote sensing images can provide spatial information such as forest coverage, vegetation type, terrain characteristics, and information such as moisture content and density of combustibles, while meteorological data provides key meteorological conditions such as temperature, humidity, and wind speed. Through real-time or recent remote sensing images and meteorological data, changes in fire risk conditions can be monitored and fire risk levels can be updated in a timely manner.

[0027] More specifically, the methods for obtaining the above data include: 1. Satellite remote sensing: Obtain remote sensing images of forest areas through sensors carried on satellites, such as Landsat, MODIS, Sentinel, etc. 2. Meteorological station observations: Use real-time and historical meteorological observation data collected by ground meteorological stations. 3. Public data platforms: Access platforms such as CHELSA (High Resolution Land Surface Climatology Data), which provide global downscaled climate data and can download the required data through the website. 4. Drone aerial photography: Use remote sensing equipment carried by drones to obtain remote sensing images of specific areas. When obtaining data, attention should be paid to: 1. Data quality: Ensure that the data is accurate and the resolution and time resolution meet the model requirements. 2. Data consistency: Maintain consistent standards for data collection to avoid mixing data from different sources or types. 3. Data storage and processing: Consider the storage format, size and processing requirements of the data to ensure that there are sufficient computing resources for analysis. 4. Data update: Update the data regularly to reflect the latest meteorological conditions and vegetation changes.

[0028] In step S120, image features of the forest remote sensing image are extracted to obtain a spatially enhanced forest remote sensing feature vector.

[0029] Figure 2The flowchart is a method for extracting the image features of the forest remote sensing image to obtain the spatially enhanced forest remote sensing feature vector in the fire monitoring and alarm method based on the intelligent assessment of the fire risk level according to the embodiment of the present application. Figure 2 As shown, extracting the image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector includes: S121, preprocessing the forest remote sensing image to obtain a preprocessed forest remote sensing image; S122, performing spatial feature extraction on the preprocessed forest remote sensing image to obtain a spatially enhanced forest remote sensing feature map; and S123, performing global mean pooling processing on the feature matrix along the channel dimension on the spatially enhanced forest remote sensing feature map to obtain a spatially enhanced forest remote sensing feature vector.

[0030] More specifically, the forest remote sensing image is preprocessed to obtain a preprocessed forest remote sensing image. Preprocessing can unify the size, resolution and pixel range of the image to make it suitable for the input layer of the model. By eliminating irrelevant information in the image, the noise of model training can be reduced, and the generalization ability and prediction performance of the model can be improved. At the same time, preprocessing can reduce the complexity of image data, thereby reducing the time and computing resources required for model training. Moreover, by enhancing certain features in the image (such as contrast, brightness), the model can be made easier to recognize and learn these features. Data enhancement techniques in the preprocessing step (such as rotation, scaling, cropping, adding noise, etc.) can increase the diversity of training data and improve the robustness of the model. Preprocessing can help remove interfering factors in the image, such as atmospheric noise, cloud occlusion, sensor saturation, etc. Moreover, preprocessing can compress image data and reduce the demand for storage space. The preprocessing step may include standardization or normalization processing to convert the image data to a uniform distribution, which helps the convergence of the model. Preprocessing can adjust the spatial resolution of the image to match the spatial resolution of meteorological data to facilitate feature fusion. Through the above preprocessing steps, it can be ensured that the remote sensing image data is input into the deep learning model in the optimal format and state, thereby improving the accuracy and efficiency of forest fire risk level assessment.

[0031] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, spatial feature extraction is performed on the preprocessed forest remote sensing image to obtain a spatially enhanced forest remote sensing feature map, including: the preprocessed forest remote sensing image is passed through a forest feature extraction module using a spatial attention mechanism to obtain a spatially enhanced forest remote sensing feature map.

[0032] More specifically, the preprocessed forest remote sensing image is passed through a forest feature extraction module using a spatial attention mechanism to obtain a spatially enhanced forest remote sensing feature map, and the forest feature extraction module is a convolutional neural network model using a spatial attention mechanism. The attention mechanism enables the model to focus on the most important areas in the image for fire risk level assessment, thereby improving the expressiveness of the features and the prediction accuracy of the model. Through the attention mechanism, the model can filter out the most useful information for the current task and ignore those less relevant backgrounds or noises. The attention mechanism allows the model to take into account a wider range of contextual information when making predictions, which is especially important when processing remote sensing images with complex backgrounds. By using the spatial attention mechanism, the convolutional neural network model can process remote sensing images more intelligently and extract the most valuable features for fire risk level assessment, thereby improving the effectiveness of the entire fire monitoring and alarm system.

[0033] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest feature extraction module is a convolutional neural network model using a spatial attention mechanism.

[0034] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, spatial feature extraction is performed on the preprocessed forest remote sensing image to obtain a spatially enhanced forest remote sensing feature map, including: using each layer of the forest feature extraction module to perform the following on the input data in the forward transfer of the layer: convolution processing is performed on the input data based on the convolution kernel to obtain a convolution feature map; the convolution feature map is passed through the spatial attention module to obtain the spatial attention score matrix; the spatial attention score matrix and each feature matrix of the preprocessed forest remote sensing image along the channel dimension are multiplied by the spatial attention feature map at each position point; the spatial attention feature map is pooled based on the local feature matrix to obtain a pooled feature map; and the pooled feature map is nonlinearly activated to obtain an activated feature map; wherein the input of the first layer of the forest feature extraction module is the preprocessed forest remote sensing image, and the output of the last layer of the forest feature extraction module is the spatially enhanced forest remote sensing feature map.

[0035] More specifically, the spatially enhanced forest remote sensing feature map is subjected to global mean pooling of the feature matrix along the channel dimension to obtain a spatially enhanced forest remote sensing feature vector. Global mean pooling can reduce the spatial dimension of the data, that is, the width and height of the image, thereby reducing the dimension of the feature and reducing the computational burden of the model. Through global pooling, the model can obtain global statistical features of the image, such as global average color or texture, which are very important for understanding the overall content of the image. Through global mean pooling, the two-dimensional spatially enhanced forest remote sensing feature map can be effectively converted into a one-dimensional spatially enhanced forest remote sensing feature vector, providing a concise and information-rich feature representation for subsequent fire risk level classification or other analysis tasks.

[0036] In step S130, the time series features of the forest meteorological data in the predetermined time period are extracted to obtain a forest meteorological feature vector.

[0037] Figure 3 The flowchart is a method for extracting the time series characteristics of the forest meteorological data in the predetermined time period to obtain the forest meteorological feature vector in the fire monitoring and alarm method based on the intelligent assessment of the fire risk level according to the embodiment of the present application. Figure 3 As shown, extracting the time series features of the forest meteorological data of the predetermined time period to obtain a forest meteorological feature vector includes: S131, passing the forest meteorological data through a transformer-based forest meteorological feature semantic encoder to obtain a plurality of forest meteorological semantic feature vectors; S132, arranging the plurality of forest meteorological semantic feature vectors in two dimensions to obtain a forest meteorological feature matrix; and S133, passing the forest meteorological feature matrix through a forest meteorological feature extractor based on a convolutional neural network model to obtain a forest meteorological feature vector.

[0038] More specifically, the forest meteorological data is passed through a transformer-based forest meteorological feature semantic encoder to obtain multiple forest meteorological semantic feature vectors, and the forest meteorological feature semantic encoder is a transformer-based context encoder model including an embedding layer. The transformer (Transformer) architecture, especially its embedding layer, can capture the deep-level features and patterns in the meteorological data. This is crucial for understanding the intrinsic connection of the data and making accurate fire risk level assessments. Meteorological data is usually time series data, and the transformer model is good at processing this type of data and can understand the time dependency between data points. The embedding layer can map the original data to a high-dimensional space, improve the model's ability to express data features, and thus extract richer and more detailed feature representations. The transformer model supports parallel processing, which enables it to effectively process large-scale meteorological data sets. The context encoder can integrate meteorological data from different time points, provide more comprehensive context information, and help improve the accuracy of fire risk level assessments.

[0039] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest meteorological feature semantic encoder is a converter-based context encoder model including an embedding layer.

[0040] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest meteorological data is passed through a transformer-based forest meteorological feature semantic encoder to obtain a plurality of forest meteorological semantic feature vectors, including: performing one-hot encoding on each data item in the forest meteorological data to obtain a plurality of forest meteorological vectors; and, inputting the plurality of forest meteorological vectors into the transformer-based forest meteorological feature semantic encoder to obtain the plurality of forest meteorological semantic feature vectors.

[0041] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the multiple forest meteorological vectors are input into the converter-based forest meteorological feature semantic encoder to obtain the multiple forest meteorological semantic feature vectors, including: arranging the multiple forest meteorological vectors as input vectors; converting the input vectors into query vectors and key vectors respectively through a learnable embedding matrix; calculating the product between the query vector and the transposed vector of the key vector to obtain a self-attention association matrix; normalizing the self-attention association matrix to obtain a standardized self-attention association matrix; inputting the standardized self-attention association matrix into a Softmax activation function for activation to obtain a self-attention feature matrix; and multiplying the self-attention feature matrix with each forest meteorological vector in the multiple forest meteorological vectors to obtain the multiple forest meteorological semantic feature vectors.

[0042] More specifically, the multiple forest meteorological semantic feature vectors are arranged in two dimensions to obtain a forest meteorological feature matrix. A two-dimensional matrix is ​​a structured data format that facilitates subsequent data processing and analysis. It provides a clear row and column structure for the data, making data storage and access more efficient. Through the two-dimensional arrangement, semantic feature vectors of different time points or different features can be combined to form a richer feature representation, which helps the model capture the spatiotemporal changes in meteorological conditions. When performing this conversion, care should be taken to ensure that each row or column of the matrix has a clear meaning, for example, each row represents a meteorological feature at a time point, or each column represents a feature vector of a specific meteorological parameter. In addition, the consistency and integrity of the data should be guaranteed to avoid information loss or errors.

[0043] More specifically, the forest meteorological feature matrix is ​​passed through a forest meteorological feature extractor based on a convolutional neural network (CNN) model to obtain a forest meteorological feature vector. CNN can capture the spatial hierarchical structure of data through multi-layer convolution operations, which is very important for understanding the complex relationships and patterns in meteorological data. Through the convolution and pooling operations of CNN, key information can be extracted from the original high-dimensional data to achieve dimensionality reduction while retaining the most important features. By inputting the forest meteorological feature matrix into the CNN model, a forest meteorological feature vector that has been intelligently compressed and feature-enhanced can be obtained. This vector contains the most useful information for fire risk level assessment and can be used as input for subsequent models to predict forest fire risk levels.

[0044] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest meteorological feature matrix is ​​passed through a forest meteorological feature extractor based on a convolutional neural network model to obtain a forest meteorological feature vector, including: using each layer of the forest meteorological feature extractor based on the convolutional neural network model to perform convolution processing and nonlinear activation processing on the input data in the forward pass of the layer so that the forest meteorological feature vector is output by the last layer of the forest meteorological feature extractor based on the convolutional neural network model, wherein the input of the first layer of the forest meteorological feature extractor based on the convolutional neural network model is the forest meteorological feature matrix.

[0045] In step S140, the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector are comprehensively analyzed to obtain a forest fire risk level.

[0046] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector are comprehensively analyzed to obtain the forest fire risk level, including: fusing the forest meteorological feature vector and the spatially enhanced forest remote sensing feature vector to obtain a forest fire risk factor analysis feature matrix; and classifying the forest fire risk factor analysis feature matrix to obtain a result representing the forest fire risk level.

[0047] More specifically, the forest meteorological feature vector and the spatially enhanced forest remote sensing feature vector are fused to obtain a forest fire risk factor analysis feature matrix. Fusion of two different types of feature vectors can integrate complementary information from meteorological conditions and remote sensing images to provide a more comprehensive forest fire risk assessment. By combining meteorological and remote sensing data, the accuracy of fire risk level prediction can be improved because the model can consider more influencing factors. At the same time, in some cases, meteorological data and remote sensing data may contain redundant information. Fusion of these data can reduce redundancy and extract the most useful features. Through this fusion, a powerful forest fire risk factor analysis feature matrix can be constructed, which contains rich, multi-dimensional information, providing a solid data foundation for the intelligent assessment of fire risk levels.

[0048] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest meteorological feature vector and the spatially enhanced forest remote sensing feature vector are fused to obtain a forest fire risk factor analysis feature matrix, including: fusing the forest meteorological feature vector and the spatially enhanced forest remote sensing feature vector with the following fusion formula to obtain a forest fire risk factor analysis feature matrix; wherein the fusion formula is:

[0049]

[0050] in represents vector multiplication, M represents the characteristic matrix of forest fire risk factor analysis, F 1 represents the forest meteorological feature vector, F 2 represents the spatially enhanced forest remote sensing feature vector, Represents the transpose of the spatially enhanced forest remote sensing feature vector.

[0051] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest fire risk factor analysis feature matrix is ​​classified to obtain a result representing the forest fire risk level, including: screening the forest fire risk factor analysis feature matrix based on feature distribution to obtain an optimized forest fire risk factor analysis feature matrix; and, passing the optimized forest fire risk factor analysis feature matrix through a classifier to obtain a classification result, and the classification result is used to represent the forest fire risk level.

[0052] In particular, in the technical solution of the present application, the feature extraction process includes processing forest meteorological data and forest remote sensing images, extracting features and fusing them for forest fire risk factor analysis. However, due to downsampling operations such as pooling, the following problems may occur. Specifically, the pooling operation may lead to information loss because only certain key information is retained and other information is discarded during the pooling process. This may cause the eigenvalues ​​at different positions to be compressed into the same pooled value, reducing the distribution correlation between the eigenvalues. The pooling operation reduces the spatial resolution of the feature map and loses position information and detail information. In the forest fire risk factor analysis feature matrix, important spatial information may be discarded, resulting in a reduction in the distribution correlation between eigenvalues ​​at different positions. The pooling operation will cause ambiguity in position information, making it difficult to accurately match features with their positions in the original feature map. This may lead to a reduction in the distribution correlation between eigenvalues, affecting the expressive power of the final forest fire risk factor analysis feature matrix. The pooling operation may lead to the sparsity of features, that is, after downsampling, the feature information at some positions may become sparse and the amount of information is insufficient. This will reduce the distribution correlation between the eigenvalues, affecting the accurate classification of the forest fire risk factor analysis feature matrix by the classifier. Based on this, the forest fire risk factor analysis feature matrix is ​​screened for classification influence features based on feature distribution.

[0053] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest fire risk factor analysis feature matrix is ​​subjected to classification influence feature screening based on feature distribution to obtain an optimized forest fire risk factor analysis feature matrix, including: determining the mean and variance of the feature set composed of all eigenvalues ​​of the forest fire risk factor analysis feature matrix; calculating the natural exponential function value with the mean difference of the feature set as the exponent to obtain the mean difference natural exponential function value; calculating the natural exponential function value with the variance of the feature set as the exponent to obtain the variance natural exponential function value; calculating the difference of the variance natural exponential function value minus one to obtain to a first difference; dividing the eigenvalue at a predetermined position of the forest fire risk factor analysis feature matrix by the first difference and then subtracting the mean difference natural exponential function value to obtain a second difference; multiplying the second difference by the eigenvalue at a predetermined position of the forest fire risk factor analysis feature matrix and then taking the absolute value to obtain a first absolute value; calculating the weighted sum of the first absolute value and the absolute value of the eigenvalue at a predetermined position of the forest fire risk factor analysis feature matrix to obtain a feature adjustment weighted sum; and activating the feature adjustment weighted sum through a softmax function to obtain the eigenvalue at a predetermined position of the classification influence feature screening matrix.

[0054] Specifically, in the fire monitoring and alarm method based on intelligent assessment of fire risk level, the forest fire risk factor analysis feature matrix is ​​subjected to classification influence feature screening based on feature distribution to obtain an optimized forest fire risk factor analysis feature matrix, including: using the following optimization formula to perform classification influence feature screening based on feature distribution on the forest fire risk factor analysis feature matrix to obtain a classification influence feature screening matrix; wherein the optimization formula is:

[0055]

[0056] Among them, m i,j is the eigenvalue of the (i, j) position of the forest fire risk factor analysis feature matrix, M represents the forest fire risk factor analysis feature matrix, μ and σ are the feature set m i,j ∈M, α and β represent weight hyperparameters, softmax represents the normalized exponential function, and m i,j ′ represents the eigenvalue of the (i, j) position of the classification impact feature screening matrix; and, the positional point multiplication between the classification impact feature screening matrix and the forest fire risk factor analysis feature matrix is ​​calculated to obtain the optimized forest fire risk factor analysis feature matrix.

[0057] In the feature extraction process of deep learning, although downsampling operations such as pooling can help reduce the amount of calculation and improve the efficiency of operation, they may also cause the loss of spatial information in the feature matrix, thereby affecting the classification accuracy. More specifically, in the feature extraction process, downsampling operations such as pooling will reduce the distribution correlation between the eigenvalues ​​at each position of the final forest fire risk factor analysis feature matrix, thereby affecting the accuracy of the classification judgment of the forest fire risk factor analysis feature matrix through the classifier.

[0058] In view of this, in the technical solution of the present application, the forest fire risk factor analysis feature matrix is ​​subjected to classification influence feature screening based on feature distribution, which first analyzes the distribution of each eigenvalue in the original feature matrix to determine which features contribute to the classification result; then, according to the result of the feature distribution analysis, the label probability response corresponding to each eigenvalue is calculated, that is, the contribution degree of the eigenvalue to each possible label, and finally, the calculated label probability response is used as a new feature layer and stacked with the original feature matrix to form an enhanced feature representation. In this way, even if downsampling is performed during the feature extraction process, the final forest fire risk factor analysis feature matrix can better retain the spatial structure information of the original data, thereby effectively improving the distribution correlation between the eigenvalues ​​at each position of the forest fire risk factor analysis feature matrix, so as to improve the judgment accuracy of the classifier.

[0059] More specifically, the optimized forest fire risk factor analysis feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to represent the forest fire risk level. The classifier can automatically output the fire risk level according to the input optimized forest fire risk factor analysis feature matrix, reducing the need for manual judgment and improving efficiency. By inputting the optimized forest fire risk factor analysis feature matrix into the classifier, complex data can be effectively converted into intuitive fire risk level classification results, providing strong technical support for the prevention and early detection of forest fires.

[0060] Specifically, in the above-mentioned fire monitoring and alarm method based on intelligent assessment of fire risk level, the optimized forest fire risk factor analysis feature matrix is ​​passed through a classifier to obtain a classification result, and the classification result is used to represent the forest fire risk level, including: using the classifier to process the optimized forest fire risk factor analysis feature matrix with the following classification formula to obtain the classification result; wherein the classification formula is:

[0061] O=softmax{(W c ,B c )|Project(F)}

[0062] Among them, Project(F) represents the projection of the optimized forest fire risk factor analysis feature matrix into a vector, W c is the weight matrix, B c represents a bias vector, softmax represents a normalized exponential function, and O represents the classification result.

[0063] In summary, the fire monitoring and alarm method based on intelligent fire risk level assessment according to the embodiment of the present application has been explained, which uses artificial intelligence technology based on the field of deep learning to extract and analyze features of forest remote sensing images and forest meteorological data of a predetermined time period to obtain a classification result representing the forest fire risk level, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and discovered, and alarm information can be generated, which can avoid false alarms and missed alarms caused by human subjective factors.

[0064] Based on the same inventive concept, the embodiments of the present application also provide a fire monitoring and alarm system based on intelligent assessment of fire risk level, which can be used to implement the method described in the above embodiments, as described in the following embodiments. Since the principle of solving the problem by the fire monitoring and alarm system based on intelligent assessment of fire risk level is similar to the above method, the implementation of the fire monitoring and alarm system based on intelligent assessment of fire risk level can refer to the implementation of the above method, and the repeated parts will not be repeated. As used below, the term "unit" or "module" can be a combination of software and / or hardware that implements predetermined functions. Although the device described in the following embodiments is preferably implemented in software, the implementation of hardware, or a combination of software and hardware, is also possible and conceivable.

[0065] Figure 4 FIG. 1 is a block diagram of a fire monitoring and alarm system based on intelligent fire risk level assessment according to an embodiment of the present application. Figure 4 As shown, according to the embodiment of the present application, the fire monitoring and alarm system 100 based on intelligent assessment of fire risk level includes: a forest data acquisition module 110, used to acquire forest remote sensing images and forest meteorological data of a predetermined time period; a forest remote sensing feature extraction module 120, used to extract image features of the forest remote sensing images to obtain spatially enhanced forest remote sensing feature vectors; a forest meteorological data extraction module 130, used to extract time series features of the forest meteorological data of the predetermined time period to obtain forest meteorological feature vectors; and a forest fire risk level analysis module 140, used to comprehensively analyze the spatially enhanced forest remote sensing feature vectors and the forest meteorological feature vectors to obtain the forest fire risk level.

[0066] Here, those skilled in the art will appreciate that the specific functions and operations of the various units and modules in the fire monitoring and alarm system 100 based on intelligent fire risk level assessment have been described in detail above. Figures 1 to 3 The description of the fire monitoring and alarm method based on intelligent assessment of fire risk level has been introduced in detail, and therefore, its repeated description will be omitted.

[0067] In summary, the fire monitoring and alarm system based on intelligent fire risk level assessment according to the embodiment of the present application has been explained, which uses artificial intelligence technology based on the field of deep learning to extract and analyze features of forest remote sensing images and forest meteorological data of a predetermined time period to obtain a classification result representing the forest fire risk level, and judge whether to issue an alarm with high-risk fire risk information based on the fire risk level shown. In this way, based on the assessment and forecast of forest and grassland fire risk levels, fire risks in high-risk areas can be intelligently identified and generated, and false alarms and missed alarms caused by human subjective factors can be avoided.

[0068] The devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is an electronic device, and the electronic device may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0069] In a typical example, the electronic device specifically includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements the above data processing steps when executing the program.

[0070] Figure 5 FIG. 1 is a block diagram of an electronic device according to an embodiment of the present application. Figure 5 As shown, based on the above data processing method, this specification also provides an electronic device, at least one processor; a memory connected to the at least one processor in communication; wherein the memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute any of the above-mentioned fire monitoring and alarm methods based on intelligent assessment of fire risk level. The processor can take the form of, for example, a microprocessor or a processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro) processor, a logic gate, a switch, an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), a programmable logic controller, and an embedded microcontroller, etc. The memory includes but is not limited to a random access memory (Random Access Memory, RAM), a read-only memory (Read-Only Memory, ROM), a cache, a hard disk (Hard Disk Drive, HDD) or a memory card. The electronic device can be a server of the method described in the above embodiment, etc. Of course, the electronic device can also include module structures such as an input device and a transmission module. In this embodiment, the functions and effects specifically implemented by the electronic device can be explained in comparison with other embodiments, and will not be repeated here.

[0071] In addition, the embodiments of this specification also provide a computer-readable storage medium on which computer instructions are stored, and when the instructions are executed, the steps of the method described in any one or more of the above embodiments are implemented. The storage medium may include a physical device for storing information, which is usually to digitize the information and then store it in a medium using electrical, magnetic or optical methods. The storage medium may include: devices that store information using electrical energy, such as various types of memories, such as RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, magnetic tapes, magnetic core memories, bubble memories, and U disks; devices that store information optically, such as CDs or DVDs. Of course, there are other types of readable storage media, such as quantum memories, graphene memories, and so on.

[0072] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0073] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, devices or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0074] The present application may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0075] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0076] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A fire monitoring and alarm method based on intelligent assessment of fire risk level, characterized in that: include: Obtain forest remote sensing images and forest meteorological data for a predetermined period of time; Extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector; Extracting the time series characteristics of the forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector; Comprehensively analyzing the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector to obtain a forest fire risk level; The forest fire risk level is obtained by comprehensively analyzing the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector, including: The forest meteorological characteristic vector and the spatially enhanced forest remote sensing characteristic vector are merged to obtain a forest fire risk factor analysis characteristic matrix; Classifying the forest fire risk factor analysis characteristic matrix to obtain a result representing the forest fire risk level; The forest fire risk factor analysis characteristic matrix is ​​classified to obtain a result representing the forest fire risk level, including: Performing classification influence feature screening based on feature distribution on the forest fire risk factor analysis feature matrix to obtain an optimized forest fire risk factor analysis feature matrix; Passing the optimized forest fire risk factor analysis feature matrix through a classifier to obtain a classification result, wherein the classification result is used to represent the forest fire risk level; The forest fire risk factor analysis feature matrix is ​​subjected to classification influence feature screening based on feature distribution to obtain an optimized forest fire risk factor analysis feature matrix, including: Determine the mean and variance of a feature set consisting of all eigenvalues ​​of the forest fire risk factor analysis feature matrix; Calculating a natural exponential function value using the mean difference of the feature set as an exponent to obtain a mean difference natural exponential function value; Calculating a natural exponential function value using the variance of the feature set as an exponent to obtain a variance natural exponential function value; Calculate the difference of the variance natural exponential function value minus one to obtain a first difference; The eigenvalue of the predetermined position of the forest fire risk factor analysis characteristic matrix is ​​divided by the first difference value and then subtracted from the mean difference natural exponential function value to obtain a second difference value; The second difference is multiplied by the eigenvalue of the predetermined position of the forest fire risk factor analysis characteristic matrix and then the absolute value is taken to obtain a first absolute value; Calculating a weighted sum of the first absolute value and the absolute value of the eigenvalue at a predetermined position of the forest fire risk factor analysis characteristic matrix to obtain a feature-adjusted weighted sum; The feature adjustment weights are activated through a softmax function to obtain the feature values ​​of the predetermined positions of the classification-influencing feature screening matrix.

2. The fire monitoring and alarm method based on intelligent fire risk level assessment according to claim 1 is characterized in that: Extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector includes: Preprocessing the forest remote sensing image to obtain a preprocessed forest remote sensing image; Performing spatial feature extraction on the preprocessed forest remote sensing image to obtain a spatially enhanced forest remote sensing feature map; The spatially enhanced forest remote sensing feature map is subjected to global mean pooling processing of the feature matrix along the channel dimension to obtain a spatially enhanced forest remote sensing feature vector.

3. The fire monitoring and alarm method based on intelligent fire risk level assessment according to claim 2 is characterized in that: Extracting spatial features from the preprocessed forest remote sensing image to obtain a spatially enhanced forest remote sensing feature map includes: The preprocessed forest remote sensing image is passed through a forest feature extraction module using a spatial attention mechanism to obtain a spatially enhanced forest remote sensing feature map.

4. The fire monitoring and alarm method based on intelligent fire risk level assessment according to claim 3 is characterized in that: The forest feature extraction module is a convolutional neural network model using a spatial attention mechanism.

5. The fire monitoring and alarm method based on intelligent fire risk level assessment according to claim 4 is characterized in that: Extracting the time series features of the forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector includes: Passing the forest meteorological data through a transformer-based forest meteorological feature semantic encoder to obtain a plurality of forest meteorological semantic feature vectors; Arranging the plurality of forest meteorological semantic feature vectors in two dimensions to obtain a forest meteorological feature matrix; The forest meteorological feature matrix is ​​passed through a forest meteorological feature extractor based on a convolutional neural network model to obtain a forest meteorological feature vector.

6. The fire monitoring and alarm method based on intelligent fire risk level assessment according to claim 5 is characterized in that: The forest meteorological feature semantic encoder is a transformer-based context encoder model including an embedding layer.

7. A fire monitoring and alarm system based on intelligent assessment of fire risk level, characterized in that: include: A forest data acquisition module is used to acquire forest remote sensing images and forest meteorological data for a predetermined period of time; A forest remote sensing feature extraction module, used for extracting image features of the forest remote sensing image to obtain a spatially enhanced forest remote sensing feature vector; A forest meteorological data extraction module, used to extract the time series characteristics of the forest meteorological data in the predetermined time period to obtain a forest meteorological feature vector; A forest fire risk level analysis module, used for comprehensively analyzing the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector to obtain a forest fire risk level; The forest fire risk level is obtained by comprehensively analyzing the spatially enhanced forest remote sensing feature vector and the forest meteorological feature vector, including: The forest meteorological characteristic vector and the spatially enhanced forest remote sensing characteristic vector are merged to obtain a forest fire risk factor analysis characteristic matrix; Classifying the forest fire risk factor analysis characteristic matrix to obtain a result representing the forest fire risk level; The forest fire risk factor analysis characteristic matrix is ​​classified to obtain a result representing the forest fire risk level, including: Performing classification influence feature screening based on feature distribution on the forest fire risk factor analysis feature matrix to obtain an optimized forest fire risk factor analysis feature matrix; Passing the optimized forest fire risk factor analysis feature matrix through a classifier to obtain a classification result, wherein the classification result is used to represent the forest fire risk level; The forest fire risk factor analysis feature matrix is ​​subjected to classification influence feature screening based on feature distribution to obtain an optimized forest fire risk factor analysis feature matrix, including: Determine the mean and variance of a feature set consisting of all eigenvalues ​​of the forest fire risk factor analysis feature matrix; Calculating a natural exponential function value using the mean difference of the feature set as an exponent to obtain a mean difference natural exponential function value; Calculating a natural exponential function value using the variance of the feature set as an exponent to obtain a variance natural exponential function value; Calculate the difference of the variance natural exponential function value minus one to obtain a first difference; The eigenvalue of the predetermined position of the forest fire risk factor analysis characteristic matrix is ​​divided by the first difference value and then subtracted from the mean difference natural exponential function value to obtain a second difference value; The second difference is multiplied by the eigenvalue of the predetermined position of the forest fire risk factor analysis characteristic matrix and then the absolute value is taken to obtain a first absolute value; Calculating a weighted sum of the first absolute value and the absolute value of the eigenvalue at a predetermined position of the forest fire risk factor analysis characteristic matrix to obtain a feature-adjusted weighted sum; The feature adjustment weights are activated through a softmax function to obtain the feature values ​​of the predetermined positions of the classification-influencing feature screening matrix.

Images