Forest fire monitoring method and device based on deep learning algorithm, equipment and medium
By installing infrared cameras or thermal imaging cameras in the power grid and combining deep learning algorithms, the problem of insufficient accuracy and real-time performance of traditional wildfire monitoring methods is solved, accurate distance measurement and positioning of wildfires is achieved, and the grid mountain is improved The scientificity and efficiency of fire prevention and control.
Patent Information
- Application Number
- CN202411629043.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-15
- Publication Date
- 2025-02-07
AI Technical Summary
Traditional wildfire monitoring methods have problems such as high labor costs, limited monitoring range and low accuracy, which are difficult to provide effective support for emergency response.
By installing infrared cameras or thermal imaging cameras in the power grid, we capture image data of the heat and spectral information of the surrounding environment of wildfires, use deep learning algorithms to perform image processing and feature extraction, identify wildfire targets, and perform distance measurement, positioning and spread prediction. .
Accurate ranging and positioning of wildfires has been achieved, the accuracy and real-time monitoring have been improved, and scientific basis for the power grid wildfire prevention and control work, improving emergency response efficiency, and reducing losses caused by fires.
Smart Images
Figure CN119399700A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of power transmission line monitoring, and in particular to a forest fire monitoring method, device, equipment and medium based on a deep learning algorithm. Background Art
[0002] Wildfires are a common natural disaster that poses a serious threat to people's lives and property, especially in mountainous and forested areas. They not only cause large-scale vegetation damage, but also threaten nearby residents and power grid facilities. Therefore, the ability to detect wildfires early and accurately measure and locate them is crucial for taking timely and effective firefighting measures.
[0003] Traditional wildfire monitoring methods mainly rely on manual patrols, remote sensing technology, and sensor networks. However, these methods have some limitations, such as high labor costs, limited monitoring range, and low accuracy. In addition, traditional methods can often only determine the approximate location of wildfires, but lack the ability to accurately measure and locate wildfires, making it difficult to provide effective support for emergency response.
[0004] In recent years, with the development and application of deep learning algorithms, wildfire monitoring technology based on image processing and pattern recognition has made significant progress. Deep learning algorithms can extract features from data collected by sensors and realize automatic detection, classification and positioning of wildfires. Compared with traditional methods, wildfire monitoring technology based on deep learning has higher accuracy and real-time performance, and can realize wildfire monitoring and early warning in a wider range. Therefore, a new method is proposed that can effectively utilize deep learning algorithms and combine power grid facilities to achieve accurate ranging and positioning of wildfires, so as to improve the efficiency and effectiveness of wildfire monitoring and emergency response. Summary of the invention
[0005] Based on this, it is necessary to propose a wildfire monitoring method, device, equipment and medium based on deep learning algorithm to address the above problems.
[0006] To achieve the above objectives, the first aspect of the present application provides a forest fire monitoring method based on a deep learning algorithm, the method comprising:
[0007] Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid;
[0008] Processing and feature extracting the image data to obtain a feature vector;
[0009] Analyzing the feature vectors through a deep learning model to identify wildfire targets;
[0010] Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target;
[0011] The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
[0012] In some implementations, the processing and feature extraction of the image data to obtain a feature vector specifically includes:
[0013] Performing denoising processing on the image data;
[0014] Standardize the denoised image data to obtain image data with a uniform scale and range;
[0015] The abstract information in the image data is converted into a quantifiable feature vector.
[0016] In some embodiments, the denoising of the image data specifically includes: Obtain denoised image data; where G(x,y) is the output of the Gaussian filter, σ is the standard deviation of the Gaussian distribution, and (x,y) is the spatial coordinate of the filter.
[0017] In some embodiments, the denoised image data is standardized to obtain image data with a uniform scale and range, specifically comprising: Obtain image data with uniform scale and range; where x is the original data, u is the mean, and σ is the standard deviation.
[0018] In some embodiments, analyzing the feature vectors by a deep learning model to identify wildfire targets specifically includes:
[0019] The feature vector is extracted through a recurrent neural network (RNN) model, and the abstract features in the feature vector are gradually extracted through a series of convolutional layers, pooling layers, and fully connected layers;
[0020] The extracted image features are analyzed through a deep learning model to identify wildfire targets.
[0021] In some embodiments, the method of performing ranging and positioning according to the wildfire target to obtain the wildfire location information of the infrared camera or thermal imaging camera and the wildfire target specifically includes: Determine the distance d between the infrared camera or thermal imaging camera and the wildfire target; wherein the focal length of the camera is f, the pixel size of the wildfire in the image is h pixels, and the actual size of the wildfire is H;
[0022] according to Determine the actual distance D between the infrared camera or thermal imaging camera and the wildfire target.
[0023] In some embodiments, predicting the spread speed and direction of a wildfire based on the wildfire location information and topographic data specifically includes:
[0024] Determine the influence of the distance the wildfire moves per unit time on the spread of the wildfire based on the wildfire location information and topographic data;
[0025] By analyzing the shape of the wildfire in the image data and the surrounding environment, the spread direction of the wildfire can be estimated.
[0026] To achieve the above-mentioned purpose, the second aspect of the present application provides a wildfire monitoring device based on a deep learning algorithm, the device comprising:
[0027] An image data capture module, used to capture image data of thermal and spectral information of the surrounding environment of the wildfire through an infrared camera or a thermal imaging camera installed in the power grid;
[0028] A feature vector acquisition module, used for processing and extracting features from the image data to obtain a feature vector;
[0029] A wildfire target recognition module, used to analyze the feature vector through a deep learning model to identify the wildfire target;
[0030] A position information acquisition module is used to perform ranging and positioning according to the wildfire target, and obtain the wildfire position information of the infrared camera or thermal imaging camera and the wildfire target;
[0031] The wildfire prediction module is used to predict the spread speed and direction of the wildfire based on the wildfire location information and topographic data.
[0032] To achieve the above-mentioned object, the third aspect of the present application provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the following steps:
[0033] Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid;
[0034] Processing and feature extracting the image data to obtain a feature vector;
[0035] Analyzing the feature vectors through a deep learning model to identify wildfire targets;
[0036] Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target;
[0037] The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
[0038] To achieve the above-mentioned purpose, the fourth aspect of the present application provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor executes the following steps:
[0039] Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid;
[0040] Processing and feature extracting the image data to obtain a feature vector;
[0041] Analyzing the feature vectors through a deep learning model to identify wildfire targets;
[0042] Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target;
[0043] The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
[0044] The embodiments of the present invention have the following beneficial effects:
[0045] The present invention captures image data of heat and spectral information of the surrounding environment of a wildfire through an infrared camera or a thermal imaging camera installed in a power grid; processes and extracts features of the image data using a deep learning algorithm to obtain a feature vector; analyzes the feature vector using a deep learning model to identify a wildfire target; performs ranging and positioning based on the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target; accurately measures and positions the wildfire based on the wildfire location information and topographic data, and predicts the spread speed and direction of the wildfire, which can provide a scientific basis for wildfire prevention and control work in power grids, improve emergency response efficiency, and reduce losses caused by fires, and can be widely used in the field of wildfire prevention of power grid transmission lines. BRIEF DESCRIPTION OF THE DRAWINGS
[0046] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0047] in:
[0048] Figure 1 A schematic diagram of a flow chart of a wildfire monitoring method based on a deep learning algorithm in one embodiment;
[0049] Figure 2 is a structural diagram of a wildfire monitoring device based on a deep learning algorithm in one embodiment;
[0050] Figure 3 is a schematic diagram of the structure of a computer device in one embodiment;
[0051] Figure 4 FIG. 4 is a schematic diagram of the structure of a computer-readable storage medium in one embodiment. DETAILED DESCRIPTION
[0052] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0053] In the embodiment of the present application, a wildfire monitoring method based on a deep learning algorithm is provided. Figure 1 , Figure 1 1 is a flow chart of a wildfire monitoring method based on a deep learning algorithm in an embodiment. The wildfire monitoring method based on a deep learning algorithm includes steps S1 to S5.
[0054] Step S1, capturing image data of heat and spectral information of the surrounding environment of the wildfire by means of an infrared camera or a thermal imaging camera installed in the power grid;
[0055] Specifically, real-time wildfire-related data, including heat and spectral information, can be obtained from sensors on power grid equipment. Various sensor technologies, such as infrared cameras and thermal imagers, can be used to capture images and other related data of wildfires. The acquired image data will be used as input for subsequent processing for wildfire detection, positioning and analysis.
[0056] Step S2, processing and feature extraction of the image data to obtain a feature vector; specifically, preprocessing, feature extraction and analysis of the data obtained by the image acquisition module, including algorithms such as image denoising, image enhancement, and feature extraction. In this step, the image quality is improved, the scale is unified, and feature extraction is performed. The purpose of improving the image quality is to remove noise in the image and enhance the characteristics of the image. The method includes denoising and median filtering, and noise is removed by a Gaussian filter.
[0057] In some implementations, the processing and feature extraction of the image data to obtain a feature vector specifically includes:
[0058] Performing denoising processing on the image data;
[0059] Standardize the denoised image data to obtain image data with a uniform scale and range;
[0060] The abstract information in the image data is converted into a quantifiable feature vector.
[0061] In some embodiments, the denoising of the image data specifically includes: Obtain denoised image data; where G(x,y) is the output of the Gaussian filter, σ is the standard deviation of the Gaussian distribution, and (x,y) is the spatial coordinate of the filter.
[0062] In some embodiments, the denoised image data is standardized to obtain image data with a uniform scale and range, specifically comprising: Obtain image data with uniform scale and range; where x is the original data, u is the mean, and σ is the standard deviation.
[0063] Specifically, median filtering removes noise by replacing the value of a pixel with the median of the pixel values in its neighborhood.
[0064] Scaling ensures that the deep learning model can process the input data in a consistent manner. Methods include unit variance and scaling. Unit variance is achieved by subtracting the mean and dividing by the standard deviation to ensure that the variance of the data is 1. The formula is:
[0065]
[0066] Where x is the original data, u is the mean, and σ is the standard deviation. Scaling adjusts the image data to the specified size through bilinear interpolation to meet the input requirements of the deep learning model.
[0067] Feature extraction uses a deep learning model to perform automatic feature learning to convert image data into a high-dimensional feature vector that can represent the image content, and learns discriminative feature representations from the original image data.
[0068] Step S3, analyzing the feature vector through a deep learning model to identify the wildfire target;
[0069] In some embodiments, feature extraction is performed on the feature vector through a recurrent neural network (RNN) model, and abstract features in the feature vector are gradually extracted through a series of convolutional layers, pooling layers, and fully connected layers;
[0070] The extracted image features are analyzed through a deep learning model to identify wildfire targets.
[0071] Step S4, performing ranging and positioning according to the wildfire target, obtaining the wildfire position information of the infrared camera or thermal imaging camera and the wildfire target; specifically, using the output results of the deep learning algorithm to achieve accurate ranging and positioning of the wildfire, using triangulation ranging method, parallax method and other methods, by analyzing the position information of the wildfire in the image and the parameters of the camera, the distance between the wildfire and the camera is calculated.
[0072] In some embodiments, the method of performing ranging and positioning according to the wildfire target to obtain the wildfire location information of the infrared camera or thermal imaging camera and the wildfire target specifically includes: Determine the distance d between the infrared camera or thermal imaging camera and the wildfire target; wherein the focal length of the camera is f, the pixel size of the wildfire in the image is h pixels, and the actual size of the wildfire is H;
[0073] according to Determine the actual distance D between the infrared camera or thermal imaging camera and the wildfire target.
[0074] Specifically, RNN (recurrent neural network) is capable of processing deep learning of sequence data and is suitable for processing image data with time series relationships. In the wildfire location ranging system, RNN is used to extract spatiotemporal features in images to capture the evolution of wildfires. It is divided into convolutional layers and pooling layers for extracting spatial features of images, and RNN layers for processing time series data. Among them, the convolutional layer extracts features in the image through convolution operations, while the pooling layer reduces the size of the feature map through downsampling operations. The convolutional layer and the pooling layer extract local and global features in the image for subsequent wildfire detection and ranging processing.
[0075] The purpose of wildfire detection is to identify whether there is a wildfire in an image and its location. The YOLO algorithm divides the image into grids and predicts the bounding box and category probability of the target in each grid, thereby achieving simultaneous detection of multiple targets. The SSD algorithm detects targets of different scales by performing detection on feature maps of different scales and predicting bounding boxes of different sizes. In deep learning, the wildfire detection algorithms YOLO and SSD are built based on convolutional neural networks (CNNs), sliding windows on images and using deep learning models for classification and positioning to achieve rapid detection of wildfires.
[0076] Finally, the distance between the wildfire and the camera is calculated based on the image data and camera parameters using triangulation and parallax methods. The formula is:
[0077]
[0078] Among them, D represents the actual distance of the object, B represents the parallax between two viewpoints of the same object in the image, f represents the focal length of the camera, and d represents the distance between the object and the camera.
[0079] Step S5, predicting the spread speed and direction of the wildfire based on the wildfire location information and topographic data; specifically, using the output results, combined with historical data and meteorological conditions and other factors, to achieve predictive analysis of the wildfire trend, spread direction and speed, including fire spread models, risk assessment models, etc., for simulating the spread process of wildfires and predicting future situations.
[0080] In some embodiments, predicting the spread speed and direction of a wildfire based on the wildfire location information and topographic data specifically includes:
[0081] Determine the influence of the distance the wildfire moves per unit time on the spread of the wildfire based on the wildfire location information and topographic data;
[0082] By analyzing the shape of the wildfire in the image data and the surrounding environment, the spread direction of the wildfire can be estimated.
[0083] The wildfire monitoring method based on deep learning algorithm provided by the present invention can improve the scientificity and efficiency of wildfire prevention and control in power grids. The wildfire location and ranging system based on deep learning algorithm can realize accurate detection, positioning and predictive analysis of wildfires, improve the accuracy and timeliness of early warning, and timely discover wildfire risks through real-time monitoring and predictive analysis, provide scientific basis and technical support, help relevant departments and personnel better monitor and respond to the spread of wildfires, and ensure the safety and stability of power grids and society.
[0084] In some embodiments, early warning processing of wildfires can also be implemented based on the results of estimating the spread direction of wildfires. When the system detects that the risk of wildfires is high, the early warning module automatically sends out an early warning signal and takes corresponding processing measures, such as sending early warning information to relevant departments and personnel, and initiating emergency response plans.
[0085] Specifically, the warning can first determine the warning trigger conditions based on the wildfire location, spread speed, diffusion direction and other parameters. For example, if the distance between the wildfire and the densely populated area is less than a certain threshold, or the spread speed exceeds a certain limit, the warning may be triggered. Once the warning trigger conditions are determined, the warning module will continuously monitor the wildfire location, spread speed and diffusion direction and other parameters in real time. The system will continuously obtain the latest wildfire information from the data source and conduct real-time monitoring according to the warning trigger conditions to ensure that the system can detect potential wildfire risks in a timely manner and make corresponding warning responses. Finally, the warning module will update the prediction analysis results and warning information based on the monitored wildfire information. If the system detects that the wildfire has reached the warning trigger conditions, it will trigger the warning mechanism and send the wildfire location, spread speed, estimated arrival time and other warning information and recommended response measures to relevant departments and personnel.
[0086] The technical solution of this embodiment can improve monitoring accuracy. Through deep learning algorithm processing and feature extraction, wildfires can be identified more accurately. Real-time capture and analysis of image data can quickly detect wildfires and shorten response time. The location of the wildfire can be accurately located through ranging and positioning, thereby improving firefighting efficiency. At the same time, the spread trend can be predicted. The speed and direction of wildfire spread can be predicted in combination with terrain data, which helps to formulate effective firefighting and evacuation strategies, forming an efficient and accurate wildfire monitoring method that helps to reduce the losses caused by wildfires.
[0087] In the embodiment of the present application, a wildfire monitoring device based on a deep learning algorithm is provided. Figure 2 , Figure 2 This is a diagram of a wildfire monitoring device based on a deep learning algorithm in an embodiment, wherein the wildfire monitoring device based on a deep learning algorithm includes an image data capture module 201, a feature vector acquisition module 202, a wildfire target recognition module 203, a location information acquisition module 204 and a wildfire prediction module 205.
[0088] The image data capturing module 201 is configured to capture image data of heat and spectral information of the surrounding environment of the wildfire through an infrared camera or a thermal imaging camera installed in the power grid;
[0089] Specifically, the image data capture module 201 can monitor the occurrence and development of wildfires in real time. The sensor continuously collects image data of the surrounding environment. When a wildfire occurs, its characteristics will be captured and transmitted to the system. At the same time, the image acquisition module can measure the heat of the wildfire and the spectral information generated. Different combustion states produce different spectral characteristics. By analyzing the spectral information, we can further understand the composition and characteristics of the wildfire and more accurately assess the impact range and risk level of the wildfire. The image data capture module 201 is the basis for deep learning analysis of the power grid wildfire positioning system, providing important data support for wildfire monitoring and emergency response.
[0090] In some embodiments, the image data capture module 201 collects image information to capture image data of thermal and spectral information of the surrounding environment of the wildfire in the following steps:
[0091] Sensor installation: infrared cameras or thermal imaging cameras are installed in the power grid to capture image data such as heat and spectral information of the environment around the wildfire. When the system starts the sensor, the sensor continuously captures images to ensure that comprehensive data is obtained;
[0092] Real-time data transmission: the collected image data is transmitted to the data processing unit of the system in real time through the data transmission system, and is selectively stored on the local device or remote server for subsequent analysis or retrospective use;
[0093] Data quality monitoring: During the data collection process, the collection system monitors the quality of image data to ensure the quality and accuracy of the collected data;
[0094] Through the above steps, image data of the wildfire surrounding environment is collected and transmitted to the processing unit, providing necessary data support for wildfire ranging.
[0095] A feature vector obtaining module 202 is configured to process and extract features from the image data to obtain a feature vector;
[0096] Specifically, the feature vector acquisition module 202 performs preprocessing such as denoising and standardization on the raw data acquired by the image acquisition module. After preprocessing, the feature vector acquisition module 202 performs feature extraction on the data to extract representative features from the raw data, including features of the shape, color, texture, etc. of the wildfire, or other features related to the wildfire. Feature extraction can reduce the dimension of the data and extract information that is important for wildfire detection and positioning.
[0097] A wildfire target identification module 203 is configured to analyze the feature vector through a deep learning model to identify the wildfire target;
[0098] Specifically, the wildfire target recognition module 203 further understands the characteristics of the wildfire data by analyzing the feature data, discovers the correlation and regularity between the data, and finds the relationship between the wildfire location and the image data. Data analysis includes statistical analysis, time series analysis, spatial analysis and other methods to better explore the relationship between the wildfire location and data. Finally, the wildfire target recognition module 203 passes the results to the location information acquisition module 204 and the wildfire prediction module 205.
[0099] In some embodiments, the first step in the wildfire target recognition module 203 is data preparation. The collected raw image data is preprocessed to improve the quality and accuracy of the image data and provide good input for the subsequent deep learning algorithm; then, the image data is feature extracted using a recurrent neural network (RNN) model, and abstract features in the image data are gradually extracted through a series of convolutional layers, pooling layers, and fully connected layers. The convolutional layer can detect edges and textures in the image, while the pooling layer can reduce the dimension and complexity of the features; the wildfire detection and classification uses a deep learning model to analyze the extracted image features to achieve wildfire detection and classification. The deep learning model can accurately identify wildfire targets in images by learning a large amount of image data. The YOLO (You Only Look Once) or SSD (Single Shot MultiBox Detector) model simultaneously detects and classifies targets in one image, thereby quickly and accurately identifying wildfire targets. Finally, the wildfire is located. Based on the output of the deep learning model, the location of the wildfire in the image is determined, and the distance between the wildfire and the sensor is further inferred. Positioning methods include triangulation and parallax methods.
[0100] The position information obtaining module 204 is configured to perform ranging and positioning according to the wildfire target, and obtain the wildfire position information between the infrared camera or the thermal imaging camera and the wildfire target;
[0101] Specifically, the location information acquisition module 204 realizes accurate distance measurement and positioning of the wildfire according to the results output by the deep learning algorithm. This module receives the feature information output by the data processing unit and extracts it using the deep learning algorithm. By analyzing the size, shape and other parameters of the wildfire features in the image, combined with the location information output by the deep learning algorithm, the relative distance between the wildfire and the power grid equipment is inferred. When judging the distance between the wildfire and the power grid equipment, multiple factors need to be considered, such as topography, wind direction and wind speed, etc. After the result is judged, the threat degree of the wildfire to the power grid is evaluated according to the distance.
[0102] The wildfire prediction module 205 is configured to predict the spreading speed and direction of the wildfire according to the wildfire location information and the topographic data.
[0103] Specifically, the wildfire prediction module 205 receives the wildfire location information output by the ranging and positioning module, analyzes the information, and predicts the development trend, spread direction and speed of the wildfire. When predicting the spread trend of the wildfire, the prediction and analysis module needs to consider factors such as the complexity of the terrain, changes in wind speed and direction, changes in air humidity and temperature, and conduct a comprehensive analysis on them, combined with the output results of the ranging and positioning module, to make an accurate prediction.
[0104] In some embodiments, first, the wildfire prediction module 205 uses tools such as geographic information systems (GIS) and combines the wildfire location information and topographic data output by the deep learning algorithm to estimate the spread speed of the wildfire by calculating the distance the wildfire moves per unit time. The speed estimation can take into account many factors, including meteorological conditions such as the topography around the wildfire, wind direction and speed, temperature and humidity.
[0105] Secondly, the wildfire prediction module 205 analyzes the direction of the spread of the wildfire, taking into account the influence of the surrounding terrain and wind direction and other factors. By analyzing the shape of the wildfire on the image and the surrounding environment, the system estimates the direction of the spread of the wildfire, provides directional guidance for emergency response, and helps decision makers formulate effective prevention and control strategies.
[0106] The wildfire prediction module 205 uses existing fire spread models based on historical data and mathematical models, such as fire spread models or risk assessment models, combined with the output results of the deep learning algorithm to further predict the spread of wildfires, the spread range and speed of wildfires. Finally, the system continuously updates the prediction results based on real-time monitoring data and adjusts the early warning and emergency response strategies in a timely manner.
[0107] In some embodiments, the wildfire monitoring device based on the deep learning algorithm also includes an early warning module 206: the early warning module 206 receives the output results of the wildfire location and distance of the ranging and positioning module and the wildfire spread trend, possible impact range and risk of the prediction and analysis module, and evaluates the threat level of the wildfire to the surrounding areas and power grid equipment based on the received information, and generates corresponding early warning information, and generates corresponding warning levels and contents based on the severity.
[0108] Specifically, fire warning is achieved based on the wildfire location and prediction analysis results output by the deep learning algorithm, combined with real-time monitoring data. The triggering conditions for the warning are the wildfire location, spread speed, and diffusion direction. While processing the warning, the system continues to monitor the real-time situation of the wildfire and continuously updates the results of the prediction analysis and warning information. Continuous monitoring and updating are the key to ensuring the accuracy and timeliness of warning processing. Finally, the system evaluates and summarizes the warning processing process, continuously improves and enhances system performance, including analyzing the accuracy and timeliness of the warning and the effectiveness of the response measures, thereby optimizing the system's warning processing process and improving the overall ability of wildfire prevention and control.
[0109] In some implementations, the feature vector obtaining module 202 is further configured to perform denoising on the image data;
[0110] Standardize the denoised image data to obtain image data with a uniform scale and range;
[0111] The abstract information in the image data is converted into a quantifiable feature vector.
[0112] Specifically, first, the original image data is denoised. In the process of image acquisition, it is often affected by environmental noise, such as light changes, equipment interference, etc., which leads to a decrease in image quality. Removing image noise is an essential step. Various image processing techniques, such as median filtering and Gaussian filtering, are used to improve the clarity and quality of the image.
[0113] Secondly, standardization is performed to ensure that the image data has a uniform scale and range. Standardization eliminates the scale differences between different images, making the deep learning algorithm more stable when processing images. Common standardization methods include scaling the image pixel values to between 0 and 1 or performing zero mean and unit variance processing;
[0114] Finally, feature extraction converts the abstract information in the image into quantifiable feature vectors. These feature vectors contain information such as edges, textures, colors, etc. in the image, which can effectively describe the characteristics of the image. The feature extraction method is an automatic feature learning method based on deep learning.
[0115] In some embodiments, the feature vector obtaining module 202 is further configured to: Obtain denoised image data; where G(x,y) is the output of the Gaussian filter, σ is the standard deviation of the Gaussian distribution, and (x,y) is the spatial coordinate of the filter.
[0116] In some embodiments, the feature vector obtaining module 202 is further configured to: Obtain image data with uniform scale and range; where x is the original data, u is the mean, and σ is the standard deviation.
[0117] Specifically, the distance measurement and positioning is realized according to the output results of the deep learning model, and the triangulation distance measurement method is used. This method uses the camera imaging principle and the geometric relationship between the camera and the target to calculate the distance between the wildfire and the camera through the triangle similarity principle. Specifically, assuming that the distance between the camera and the wildfire is d, the focal length of the camera is f, the pixel size of the wildfire in the image is h pixels, and the actual size of the wildfire is H, the distance d is calculated using the following formula:
[0118]
[0119] The triangulation distance measurement method estimates the distance between the wildfire and the camera through the imaging law of the camera. It is suitable for situations where the relative position between the camera and the wildfire is fixed and the distance is relatively close, and can achieve relatively accurate distance measurement. In addition to the triangulation distance measurement method, the parallax method can also be used for distance measurement. The parallax method estimates the distance of the wildfire by comparing the parallax of the image at different positions. Parallax refers to the position difference of the same object at different viewpoints. Its relationship with the distance of the object can be calculated by the following formula:
[0120]
[0121] Among them, D represents the actual distance of the object, B represents the parallax between two viewpoints of the same object in the image, f represents the focal length of the camera, and d represents the distance between the object and the camera.
[0122] The ranging and positioning method is based on the location information output by the deep learning model and the camera imaging principle, which can achieve accurate positioning and ranging of the wildfire location, providing important support for subsequent early warning and emergency response.
[0123] In some embodiments, the wildfire target recognition module 203 is further configured to perform feature extraction on the feature vector through a recurrent neural network (RNN) model, and gradually extract abstract features in the feature vector through a series of convolutional layers, pooling layers, and fully connected layers;
[0124] The extracted image features are analyzed through a deep learning model to identify wildfire targets.
[0125] In some implementations, the location information obtaining module 204 is further configured to Determine the distance d between the infrared camera or thermal imaging camera and the wildfire target; wherein the focal length of the camera is f, the pixel size of the wildfire in the image is h pixels, and the actual size of the wildfire is H;
[0126] according to Determine the actual distance D between the infrared camera or thermal imaging camera and the wildfire target.
[0127] Specifically,
[0128] In some embodiments, the wildfire prediction module 205 is further configured to determine the wildfire spread speed based on the wildfire location information and topographic data;
[0129] By analyzing the shape of the wildfire in the image data and the surrounding environment, the spread direction of the wildfire can be estimated.
[0130] For other details about how the modules in the wildfire monitoring device based on deep learning algorithm implement the above technical solution, please refer to the description of the wildfire monitoring method based on deep learning algorithm provided above, which will not be repeated here.
[0131] In an embodiment of the present application, a computer device is provided. Figure 3 , Figure 3 3 is a schematic diagram of the structure of a computer device in an embodiment. The device includes a memory 301 and a processor 302. The memory 301 stores a computer program. When the computer program is executed by the processor 302, the processor 302 performs the following steps:
[0132] Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid;
[0133] Processing and feature extracting the image data to obtain a feature vector;
[0134] Analyzing the feature vectors through a deep learning model to identify wildfire targets;
[0135] Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target;
[0136] The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
[0137] Among them, the processor 302 can also be called a CPU (Central Processing Unit), and the processor 302 may be an integrated circuit chip with signal processing capabilities; the processor 302 can also be a general-purpose processor, DSP (Digital Signal Process), ASIC (Application Specific Integrated Circuit), FPGA (Field Programmable Gate Array) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, among which the general-purpose processor can be a microprocessor or the processor 302 can also be any conventional processor, etc.
[0138] In an embodiment of the present application, a computer readable storage medium is provided. Figure 4 , Figure 4The structure diagram of a computer-readable storage medium in an embodiment is a schematic diagram, on which a readable computer program 401 is stored; wherein the computer program 401 may be stored in the above storage medium in the form of a software product, and includes a number of instructions for causing a computer device (which may be a personal computer, a server machine, or a network device, etc.) or a processor to perform the following steps:
[0139] Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid;
[0140] Processing and feature extracting the image data to obtain a feature vector;
[0141] Analyzing the feature vectors through a deep learning model to identify wildfire targets;
[0142] Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target;
[0143] The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
[0144] The aforementioned storage media include: U disk, mobile hard disk, magnetic disk or CD, ROM (Read-Only Memory), RAM (Random Access Memory) and other media that can store program code, or terminal devices such as computers, service machines, mobile phones, and tablets.
[0145] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0146] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0147] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.
Claims
1. A wildfire monitoring method based on deep learning algorithm, characterized in that: The method includes: Capture image data of the heat and spectral information of the wildfire surrounding environment through infrared cameras or thermal imaging cameras installed in the power grid; Processing and feature extracting the image data to obtain a feature vector; Analyzing the feature vectors through a deep learning model to identify wildfire targets; Perform ranging and positioning according to the wildfire target to obtain wildfire location information of the infrared camera or thermal imaging camera and the wildfire target; The spread speed and direction of the wildfire are predicted based on the wildfire location information and topographic data.
2. The forest fire monitoring method based on deep learning algorithm according to claim 1, characterized in that: The processing and feature extraction of the image data to obtain a feature vector specifically includes: Performing denoising processing on the image data; Standardize the denoised image data to obtain image data with a uniform scale and range; The abstract information in the image data is converted into a quantifiable feature vector.
3. The forest fire monitoring method based on deep learning algorithm according to claim 2 is characterized in that: The denoising process of the image data specifically includes: Obtain denoised image data; where G(x,y) is the output of the Gaussian filter, σ is the standard deviation of the Gaussian distribution, and (x,y) is the spatial coordinate of the filter.
4. The forest fire monitoring method based on deep learning algorithm according to claim 3 is characterized in that: The standardization of the denoised image data to obtain image data with a uniform scale and range specifically includes: Obtain image data with uniform scale and range; where x is the original data, u is the mean, and σ is the standard deviation.
5. The forest fire monitoring method based on deep learning algorithm according to claim 4 is characterized in that: The deep learning model is used to analyze the feature vector to identify the wildfire target, specifically including: The feature vector is extracted through a recurrent neural network (RNN) model, and the abstract features in the feature vector are gradually extracted through a series of convolutional layers, pooling layers, and fully connected layers; The extracted image features are analyzed through a deep learning model to identify wildfire targets.
6. The forest fire monitoring method based on deep learning algorithm according to claim 5 is characterized in that: The method of performing ranging and positioning according to the wildfire target to obtain the wildfire position information of the infrared camera or thermal imaging camera and the wildfire target specifically includes: Determine the distance d between the infrared camera or thermal imaging camera and the wildfire target; wherein the focal length of the camera is f, the pixel size of the wildfire in the image is h pixels, and the actual size of the wildfire is H; according to Determine the actual distance D between the infrared camera or thermal imaging camera and the wildfire target.
7. The forest fire monitoring method based on deep learning algorithm according to claim 6 is characterized in that: The predicting of the spreading speed and direction of the wildfire according to the wildfire location information and the topographic data specifically includes: Determine the influence of the distance the wildfire moves per unit time on the spread of the wildfire based on the wildfire location information and topographic data; By analyzing the shape of the wildfire in the image data and the surrounding environment, the spread direction of the wildfire can be estimated.
8. A forest fire monitoring device based on a deep learning algorithm, characterized in that: The device comprises: An image data capture module, used to capture image data of thermal and spectral information of the surrounding environment of the wildfire through an infrared camera or a thermal imaging camera installed in the power grid; A feature vector acquisition module, used for processing and extracting features from the image data to obtain a feature vector; A wildfire target recognition module, used to analyze the feature vector through a deep learning model to identify the wildfire target; A position information acquisition module is used to perform ranging and positioning according to the wildfire target, and obtain the wildfire position information of the infrared camera or thermal imaging camera and the wildfire target; The wildfire prediction module is used to predict the spread speed and direction of the wildfire based on the wildfire location information and topographic data.
9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 7.
10. A computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to perform the steps of the method according to any one of claims 1 to 7.