Forest fire point identification method and device, electronic equipment and readable storage medium

Abstract

The invention discloses a forest fire point identification method and device, electronic equipment and a readable storage medium, and the method comprises the steps: receiving a first thermal image which is a thermal image collected by an unmanned plane carrying a thermal imager in a forest patrol process; the first thermal image is input into a first image recognition module, the probability value that the first thermal image output by the first image recognition module contains a fire point is obtained, and the first image recognition module is used for determining the probability value that the input thermal image contains the fire point; and under the condition that the probability value is greater than a preset threshold value, determining that the suspected fire point exists in the first thermal image.

Description

Technical Field

[0001] This application relates to the field of fire safety data processing, and in particular to a method, device, electronic device, and readable storage medium for identifying forest fire spots. Background Technology

[0002] Forest fires are defined as fires that spread uncontrollably within forest areas, causing harm and damage to forests, forest ecosystems, and human beings. Forest fires are a type of natural disaster characterized by their suddenness, destructiveness, and the difficulty of their control and response.

[0003] In related technologies, drones equipped with binocular cameras are typically used for inspections to identify forest fires. However, binocular cameras are easily affected by environmental factors such as fire smoke, resulting in a low accuracy rate in identifying forest fires. Summary of the Invention

[0004] This application discloses a method, device, electronic equipment, and readable storage medium for identifying forest fires, which can improve the accuracy of forest fire identification.

[0005] To solve the above problems, this application adopts the following technical solution: In a first aspect, embodiments of this application disclose a forest fire spot identification method, comprising: receiving a first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during forest patrol; inputting the first thermal image into a first image recognition module to obtain a probability value of the first thermal image containing a fire spot output by the first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire spot; and determining that there is a suspected fire spot in the first thermal image if the probability value is greater than a preset threshold.

[0006] Secondly, this application discloses a forest fire detection device, comprising: a receiving module for receiving a first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during forest patrol; an obtaining module for obtaining a probability value of the first thermal image containing a fire point by inputting the first thermal image into a first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire point; and a determining module for determining that the first thermal image contains a suspected fire point if the probability value is greater than a preset threshold.

[0007] Thirdly, embodiments of this application provide an electronic device including a processor and a memory, the memory storing programs or instructions executable on the processor, the programs or instructions, when executed by the processor, implementing the steps of the method described in the first aspect.

[0008] Fourthly, embodiments of this application provide a readable storage medium on which a program or instructions are stored, which, when executed by a processor, implement the steps of the method described in the first aspect.

[0009] Fifthly, embodiments of this application provide a computer program product, the computer program product including a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, which, when executed by a computer, cause the computer to perform the steps of the method described in the first aspect.

[0010] The technical solution adopted in this application can achieve the following beneficial effects: This application provides a method for identifying forest fire spots. It receives a first thermal image, collected by a drone equipped with a thermal imager during forest patrol, and inputs this image into a first image recognition module. The module outputs a probability value indicating that the thermal image contains a fire spot. If this probability value exceeds a preset threshold, the module determines that the thermal image contains a suspected fire spot. Because this method determines the presence of a suspected fire spot based on a thermal image, it effectively overcomes visual obstacles such as smoke, improving the accuracy of forest fire identification. Furthermore, by inputting the thermal image into the first image recognition module to determine the probability value of a fire spot, the method improves identification efficiency. Attached Figure Description

[0011] Figure 1 This is a flowchart illustrating a forest fire detection method disclosed in an embodiment of this application; Figure 2 This is a flowchart illustrating a method for implementing a forest fire risk inspection and rescue system as disclosed in an embodiment of this application. Figure 3 This is a schematic diagram of the structure of a forest fire detection device disclosed in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device disclosed in an embodiment of this application. Detailed Implementation

[0012] The technical solutions of the embodiments of this application will be clearly described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0013] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such terms can be used interchangeably where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the electrically connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0014] The forest fire identification method, apparatus, electronic device, and readable storage medium disclosed in this application will be described in detail below with reference to the accompanying drawings and through specific embodiments and application scenarios.

[0015] This application discloses a method for identifying forest fire spots. Figure 1 This is a flowchart illustrating a forest fire detection method disclosed in an embodiment of this application. Figure 1 As shown, the method includes the following steps: S120. Receive the first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during a forest patrol.

[0016] In this application, the drone is equipped with a high-precision thermal imager and a high-definition camera. Following the patrol route pre-set by the administrator, it autonomously cruises and scans the target forest area. The thermal imager captures the infrared heat radiated by ground objects and converts it into thermal images. During the patrol, the drone collects thermal image data in real time and transmits this data to the first image recognition module of the forest fire risk inspection and rescue system via wireless network for fire point identification.

[0017] S140. By inputting the first thermal image into the first image recognition module, the probability value of the first thermal image containing a fire point is obtained by the output of the first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire point.

[0018] For example, the first image recognition module can be a pre-trained deep learning model.

[0019] S160. If the probability value is greater than a preset threshold, it is determined that there is a suspected fire point in the first thermal image.

[0020] For example, the preset threshold can be 0.8. If a suspected fire point is determined in the first thermal image, an alarm process is triggered.

[0021] In this application, thermal images are collected by a drone equipped with a thermal imager, and fire identification is performed using a first image recognition module. This effectively overcomes visual obstacles such as smoke and improves the accuracy, efficiency, and reliability of forest fire identification.

[0022] This application provides a method for identifying forest fire spots. It receives a first thermal image, collected by a drone equipped with a thermal imager during forest patrol, and inputs this image into a first image recognition module. The module outputs a probability value indicating that the thermal image contains a fire spot. If this probability value exceeds a preset threshold, the module determines that the thermal image contains a suspected fire spot. Because this method determines the presence of a suspected fire spot based on a thermal image, it effectively overcomes visual obstacles such as smoke, improving the accuracy of forest fire identification. Furthermore, by inputting the thermal image into the first image recognition module to determine the probability value of a fire spot, the method improves identification efficiency.

[0023] In one implementation, the first image recognition module includes a pre-trained CNN (Convolutional Neural Network) model and a pre-trained SVM (Support Vector Machine) classifier. The step of inputting the first thermal image into the first image recognition module to obtain the probability value of the first thermal image containing a fire point, output by the first image recognition module, may include: inputting the first thermal image into the pre-trained CNN model to obtain a feature vector output by the pre-trained CNN model corresponding to the first thermal image, wherein the pre-trained CNN model is used to extract feature information from the input thermal image and output a feature vector corresponding to the feature information; and inputting the feature vector into the pre-trained SVM classifier to obtain the probability value of the first thermal image containing a fire point, output by the pre-trained SVM classifier, wherein the pre-trained SVM classifier is used to determine the probability value of containing a fire point based on the input feature vector.

[0024] In this application, a first thermal image is input into a pre-trained CNN model. The CNN model can automatically learn and extract key features from the first thermal image through multi-layer convolution and pooling operations, such as the texture, shape, and edge information of the high-temperature area. These features can effectively characterize the visual features of the fire point. The last layer of the CNN model outputs a feature vector, which contains all the key feature information extracted from the input image.

[0025] The feature vector output by the CNN model is input into a pre-trained SVM classifier. The SVM classifier determines whether the first heat map contains a fire based on the feature vector. The SVM classifier finds an optimal hyperplane to divide the feature space into two categories: fire and non-fire. The SVM classifier outputs a probability value P(fire), representing the probability that the first heat map contains a fire. ,in: This describes the process by which a CNN model extracts features from an input image (Image) and outputs a feature vector. This describes the process by which an SVM classifier identifies fire points based on feature vectors extracted by a CNN model, and outputs probability values. .

[0026] This application uses a thermal imager to acquire thermal images and a pre-trained CNN+SVM model for fire identification, which can effectively overcome the interference of environmental factors such as smoke and light, and improve the accuracy and reliability of fire identification.

[0027] In one implementation, the first thermal image carries first location information acquired by the drone. After determining that a suspected fire point exists in the first thermal image, the implementation may further include: dispatching a drone swarm to a target area corresponding to the first location information, wherein the drones in the drone swarm are equipped with imaging devices, the imaging devices including cameras and thermal imagers; receiving on-site data corresponding to the target area sent by the drone swarm, wherein the on-site data includes image data and video data, the on-site data being acquired by the drone swarm based on the imaging devices; if it is determined based on the on-site data that a fire actually exists and there are trapped personnel, determining the terrain information, obstacle information, fire spread information, and second location information of the trapped personnel based on the on-site data and wind direction information; and determining the escape route of the trapped personnel based on the terrain information, obstacle information, second location information, fire spread information, wind direction information, and third location information of the safe area.

[0028] After the first image recognition module identifies a suspected fire point and triggers an alarm, the forest fire risk inspection and rescue system enters the auxiliary fire-fighting decision-making phase. Based on the first location information of the suspected fire point, more drones equipped with high-definition cameras and thermal imagers are dispatched to the target area corresponding to the first location information, forming a drone swarm. This swarm works collaboratively to confirm the fire situation and collect information, transmitting fire information back to the control center in real time. Upon reaching the target area, the drone swarm uses high-definition cameras and thermal imagers to observe and photograph the suspected fire point from multiple angles and at close range, acquiring clearer and more comprehensive image and video data, which is transmitted back to the control center in real time. Manual personnel then confirm the existence of the fire based on the returned on-site data (including image and video data). If the fire is confirmed, the exact location of the ignition point is determined. It should be noted that the image and video data transmitted back by the drone swarm both carry location information of the shooting location.

[0029] In this application, after confirming the existence of a fire, the drone swarm will continue to collect information such as the location of the fire, the size of the fire, the spread of the fire, and the surrounding environment. Specifically, it will use a GPS positioning system to accurately locate the latitude and longitude coordinates of the fire and map the fire area. Based on information such as flame height, burning area, and spread speed, it will assess the size and development trend of the fire. Based on factors such as wind direction and terrain, it will predict the direction and speed of fire spread. By taking images of the surrounding environment of the fire, such as terrain, vegetation, buildings, and roads, and combining them with the topography and geomorphology information of the forest area, it will provide a reference for the planning of fire fighting strategies.

[0030] After receiving the on-site data sent by the drone cluster, the second thermal image from the on-site data can be input into the second image recognition module to search for whether there are trapped people near the fire. The second image recognition module is a CNN+SVM image recognition module specifically trained for human recognition in forests. The CNN in the second image recognition module is used to extract the feature information of the input thermal image and output the feature vector corresponding to the feature information. The SVM in the second image recognition module is used to determine the probability value of containing a person based on the feature vector output by the CNN in the second image recognition module. If a person is found, the second location information of the trapped person is recorded.

[0031] If, based on on-site data of the target area, a fire is confirmed to exist and trapped personnel are identified, the following methods are used: First, based on the on-site data and wind direction information, determine the terrain, obstacles, fire spread, and the second location of the trapped personnel. Second, based on the terrain, obstacles, second location, fire spread, wind direction, and the third location of the safe zone, determine the escape route for the trapped personnel and use the drone's equipped audio equipment to guide them to safety. In this application, terrain information may include slope, altitude, etc.; obstacle information may include buildings, trees, etc.; fire spread information may include spread speed and direction; and wind direction information may include wind speed and wind angle. It should be noted that the terrain and obstacle information can be directly determined based on on-site data captured by the drone swarm, and the fire spread information can be determined based on wind direction and terrain information.

[0032] By adopting the scheme of this application, when determining the escape route for trapped personnel, the terrain information of the fire site, obstacle information, the second location information of the trapped personnel, fire spread information, wind direction information, and the third location information of the safe area are considered. This ensures that the planned escape route is safe and reliable, provides scientific decision support for rescue personnel, and effectively improves the efficiency and safety of fire rescue.

[0033] In one implementation, determining the escape route of the trapped personnel based on the terrain information, the obstacle information, the second location information, the fire spread information, the wind direction information, and the third location information of the safe area may include: generating a grid map based on the terrain information and the obstacle information, wherein each grid in the grid map represents a passable or impassable area; and determining the escape route of the trapped personnel based on the grid map, the second location information, the fire spread information, the wind direction information, and the third location information of the safe area.

[0034] In this application, the specific process for determining the escape route for trapped personnel is as follows: S11. Constructing a map model The forest fire risk inspection and rescue system converts the terrain information, fire spread information, obstacle information and wind direction information of the fire site and surrounding area into a grid map. Each grid in the grid map represents a passable or impassable area and includes terrain information (e.g., slope, altitude), fire spread information (e.g., spread speed, spread direction), wind direction information (e.g., wind speed, wind angle) and obstacle information (e.g., buildings, trees).

[0035] S12, Define a heuristic function This application uses Manhattan distance as a heuristic function. Used to estimate the current node The estimated cost (distance) to reach the destination (safe zone).

[0036]

[0037] in, Indicates the current node coordinates This indicates the coordinates of the endpoint (safe zone). It should be noted that the nodes here correspond to the grid cells in the raster map.

[0038] S13. Define the cost function This application employs the following cost function. It is used to calculate the distance from the starting point (i.e., the second location information of the trapped personnel) to the current node. The actual cost.

[0039]

[0040] in, This represents the distance from the starting point to the parent node. The actual cost, Indicates from the parent node To the current node The cost of movement, taking into account factors such as terrain, fire spread rate, obstacles, and wind direction, is calculated as follows: ,in, Indicates the parent node To the current node distance, Indicates the current node The slope value indicates the cost; the steeper the slope, the higher the cost. Indicates the current node The faster the fire spreads, the higher the cost. Wind direction is also taken into account; if the wind direction aligns with the direction of fire spread, the cost increases. The value is reduced if the value is less than the value. The value, Indicates the current node Is it an obstacle? If it is an obstacle, then set the cost to infinity. Indicates the current node The cost of wind direction increases if the wind direction is opposite to the direction of movement. The value is reduced if the value is less than the value. The value, , , , , These are weighting coefficients used to balance the importance of distance, slope, fire spread rate, obstacles, and wind direction, and can be adjusted according to the actual situation.

[0041] S14. Define the evaluation function This application uses the following evaluation function. Used to evaluate nodes Priority.

[0042]

[0043] in, This represents the distance from the starting point to the current node. The actual cost, Indicates the current node Estimated cost to reach the destination (Manhattan distance).

[0044] S15, Search Path The forest fire risk inspection and rescue system employs an A* algorithm search strategy, starting from the starting point and continuously expanding the nodes until the endpoint (safe zone) is found. It should be noted that the open list stores nodes to be expanded, while the closed list stores already expanded nodes.

[0045] The process is as follows: S151. Add the starting point to the open list.

[0046] S152. Select from the open list Expand the node with the smallest value.

[0047] S153. Remove the node from the open list and add it to the closed list.

[0048] S154. Traverse the neighboring nodes of this node.

[0049] If a neighboring node is not traversable or is already on the closed list, then ignore that node.

[0050] If a neighboring node is not in the open list, add it to the open list and calculate its value. , and value.

[0051] If a neighboring node is already in the open list, compare the newer one. Value and the original Value, if new If the value is smaller, then update the parent node and the child node of that node. , value.

[0052] S155. Repeat S52-S54 until the endpoint (safe zone) is found or the open list is empty.

[0053] S16. Generate an escape route Once the destination is found, the escape route from the starting point to the destination can be generated by tracing back from the destination to the starting point, and the determined escape route can be sent to the trapped personnel.

[0054] The proposed solution employs an improved A* algorithm, which considers factors such as distance, slope, fire spread rate, obstacles, and wind direction in the cost function, and uses Manhattan distance as a heuristic function to ensure that the planned path is safe and reliable.

[0055] Since wind direction information at a fire scene changes in real time, the system needs to continuously monitor wind direction information and dynamically adjust escape routes according to changes in wind direction.

[0056] After determining the initial escape route for trapped personnel, the drone swarm monitors the on-site data and wind direction information of the fire scene in real time, and transmits the updated information back to the control center. If there are significant changes in the fire's spread speed, direction, or wind direction, or if new obstacles are discovered, the system re-determines the escape route for the trapped personnel based on the updated information, following steps S11-S16, and sends the updated information to the trapped personnel. The drones then use their equipped audio equipment to guide the trapped personnel in adjusting their escape route. Alternatively, the system determines the escape route for the trapped personnel in real time based on the acquired on-site data and wind direction information, following steps S11-S16. If the change in path direction between the newly determined escape route and the current escape route exceeds a first threshold, or the change in path length between the newly determined escape route and the current escape route exceeds a second threshold, the system updates the escape route for the trapped personnel, adopts the newly determined escape route as the current escape route, and sends it to the trapped personnel. The drones then use their equipped audio equipment to guide the trapped personnel in adjusting their escape route. For example, the first threshold can be 30 degrees, and the second threshold can be 10%. By adopting the path update scheme of this application, frequent path changes can be avoided.

[0057] Based on the traditional A* algorithm, this application considers the fire spread rate, terrain factors, obstacle information, and wind direction information, improves the cost function, and uses the dynamic adjustment function of escape routes and the setting of path change thresholds to make the planned escape routes safer and more reliable, better adaptable to the dynamic changes of the fire scene, reduce the frequency of path changes, and improve escape efficiency.

[0058] In one implementation, after receiving the on-site data corresponding to the target area sent by the drone cluster, the method may further include: if the on-site data confirms the existence of a fire, randomly generating a fire extinguishing plan; generating a new fire extinguishing plan by randomly perturbing the current fire extinguishing plan; if the objective function corresponding to the new fire extinguishing plan is less than the objective function corresponding to the current fire extinguishing plan, determining the new fire extinguishing plan as the current fire extinguishing plan, wherein the objective function corresponding to the fire extinguishing plan is determined based on the target information corresponding to the fire extinguishing plan, the target information including the effective fire extinguishing area, total cost, estimated fire extinguishing time, the relationship between the location of the fire extinguishing action and the direction of fire spread, and the terrain information of the location of the fire extinguishing action, the objective function being used to evaluate the merits of the fire extinguishing plan; returning to the step of generating a new fire extinguishing plan by randomly perturbing the current fire extinguishing plan, until a preset number of iterations is met.

[0059] In this application, the existence of a fire is confirmed manually based on the transmitted on-site data (including image and video data). If the fire is real, the location of the ignition point is precisely determined. It should be noted that the image and video data transmitted by the drone swarm both carry location information of the shooting location.

[0060] In this application, the process for determining the fire extinguishing plan is as follows, assuming the fire is confirmed to exist based on on-site data corresponding to the target area: S21, Scheme Coding Encode the fire extinguishing plan as a solution vector. , where the solution vector Each element in the table represents a firefighting operation, for example: =1 indicates that in the region Ground personnel are conducting firefighting operations. =2 indicates that in the region Drones were used to drop fire extinguishing bombs. =3 indicates that in the region Establish a buffer zone.

[0061] S22. Define the objective function objective function Used to evaluate the merits of fire extinguishing strategies, for example:

[0062] in, Indicates the effective fire suppression area of ​​the fire suppression plan. This indicates the total cost of the firefighting plan, including labor costs, material costs, etc. R(X) represents the estimated firefighting time of the firefighting plan, and R(X) represents the risk assessment of the firefighting plan. For example, if the firefighting operation is located downwind of the direction of fire spread, the risk level is high; if the firefighting operation is located in an area with a steep slope, the risk level is also high. , , , These are weighting coefficients used to balance the importance of different objectives. It should be noted that a correlation is established between the location of the firefighting operation and the direction of fire spread, and between the risk level and the terrain information of the firefighting operation location. This application determines the risk assessment of the firefighting plan based on the relationship between the location of the firefighting operation and the direction of fire spread, as well as the terrain information of the firefighting operation location.

[0063] S23, Initialization Scheme If the fire is confirmed to actually exist based on the on-site data corresponding to the target area, an initial fire extinguishing plan is randomly generated. .

[0064] S24, Iterative Optimization Based on the process of simulated annealing algorithm, iterative optimization is performed.

[0065] S241. Generate a new plan: Update the current firefighting plan. Perform random perturbation to generate a new fire extinguishing plan. .

[0066] S242. Calculate the energy difference: Calculate the difference in objective function value between the new fire extinguishing plan and the current fire extinguishing plan. ,in, This represents the difference in the objective function values. This represents the objective function value corresponding to the new fire extinguishing plan. This represents the objective function value corresponding to the current fire extinguishing plan.

[0067] S243. Accept the new firefighting plan: If (i.e., the objective function value corresponding to the new fire extinguishing plan) Less than the objective function value corresponding to the current fire extinguishing plan If so, the new firefighting plan will be accepted. The new firefighting plan This has been determined as the current firefighting plan.

[0068] S25, Iteration Termination When the number of iterations reaches its maximum value and the preset number of iterations is met, the iteration is terminated, and the final fire extinguishing solution is output.

[0069] This application employs an improved simulated annealing algorithm for fire suppression planning, incorporating fire spread direction and terrain information into the objective function, thereby enabling the planning of safer and more effective fire suppression schemes.

[0070] In one implementation, before receiving the first thermal image, the method may further include: determining a target fire risk level based on the temperature of the forest area during a first preset time period, the relative humidity during the first preset time period, the wind speed during the first preset time period, the rainfall during a second preset time period, and the rainfall during a third preset time period, wherein the second preset time period is a portion of the third preset time period; and determining the drone patrol frequency corresponding to the target fire risk level based on the target fire risk level and the target correspondence relationship, wherein the target correspondence relationship is the correspondence between the fire risk level and the drone patrol frequency.

[0071] For example, the temperature (T) in the first preset time period can be the daily maximum temperature (°C), the relative humidity (RH) in the first preset time period can be the daily minimum relative humidity (%), the wind speed (WS) in the first preset time period can be the daily maximum wind speed (m / s), the rainfall (P1) in the second preset time period can be the rainfall in the past 24 hours (mm), and the rainfall (P7) in the third preset time period can be the rainfall in the past week (mm).

[0072] In this application, the forest fire risk level can be assessed using the temperature and humidity of the forest area reported by the passive IoT terminal, the wind speed collected by the weather station, and the precipitation over the past week and the past 24 hours, so as to ensure the accuracy of the determined forest fire risk level.

[0073] In this application, after obtaining indicators such as the daily maximum temperature, daily minimum relative humidity, daily maximum wind speed, rainfall in the past 24 hours, and rainfall in the past week in the forest area, the obtained indicator values ​​can be converted into membership degrees based on their corresponding membership functions to represent the degree of membership of the indicator to different fire risk levels. Then, based on the converted membership degrees and the weight values ​​corresponding to each indicator, a comprehensive evaluation value is determined. Based on the comprehensive evaluation value, the corresponding target fire risk level is determined, and then the frequency of drone patrols corresponding to the target fire risk level is determined. This can improve the efficiency of fire early warning and effectively reduce the risk of fire.

[0074] In this application, the specific process for determining the frequency of drone patrols is as follows: S31. Determine the membership function Each indicator is assigned a membership function based on its impact on fire risk, which converts the indicator value into a membership degree, representing the degree of membership of the indicator to different fire risk levels.

[0075] The membership function corresponding to temperature (T) is:

[0076] The membership function corresponding to relative humidity (RH) is:

[0077] The membership function corresponding to wind speed (WS) is:

[0078] The membership function corresponding to the rainfall (P1) in the past 24 hours is:

[0079] The membership function corresponding to the rainfall in the past week (P7) is:

[0080] S32. Weight Determination Based on experience and historical data, the weight values ​​for each indicator are redefined, and the default values ​​for the weight vector are as follows: It can be adjusted according to the specific conditions of the forest area.

[0081] S33, Fuzzy Comprehensive Evaluation

[0082] Where A is the membership matrix, and each column represents a fire risk level (e.g., low, medium, high, extremely high).

[0083] S34. Determination of Fire Risk Level The final fire risk level is determined based on the comprehensive evaluation value B. Low risk: The maximum value in B corresponds to the "low" level; Medium risk: The maximum value in B corresponds to the "medium" level; High risk: The maximum value in B corresponds to the "high" level; Extremely high risk: The maximum value in B corresponds to the "extremely high" level.

[0084] S35. Determine the frequency of drone patrols. Based on the assessed fire risk level, the system automatically adjusts the frequency of drone patrols. Low risk: patrol once every two days; Medium risk: maintain normal patrols once a day; High risk: increase patrol frequency to 2-3 times a day; Very high risk: use multiple drones for continuous, uninterrupted, full-coverage patrols, and notify relevant departments to prepare for emergencies.

[0085] It should be noted that regardless of the fire risk level, the drones patrol along a predetermined route.

[0086] In this application, the method and process for implementing the forest fire risk inspection and rescue system are as follows: Figure 2As shown, forest meteorological data (including daily maximum temperature, daily minimum relative humidity, daily maximum wind speed, rainfall in the past 24 hours, and rainfall in the past week) are acquired. The forest fire risk level is assessed according to the procedures in S31-S35 above, and a drone patrol plan is determined. Based on the thermal images collected during the drone patrol, image recognition is performed. If a suspected fire point is identified, an alarm procedure is triggered, and a drone swarm is dispatched to the target area corresponding to the first location information of the suspected fire point for reinforcement. On-site data corresponding to the target area is collected. If the fire is confirmed to exist based on the on-site data, a search and rescue operation is conducted. If trapped personnel are found, an escape route is determined according to the procedures in S11-S16 above, and the trapped personnel are guided to evacuate. After the trapped personnel are evacuated, a fire extinguishing plan is determined according to the procedures in S21-S25 above, and the determined fire extinguishing plan is executed. If no trapped personnel are found, a fire extinguishing plan is determined according to the procedures in S21-S25 above, and the determined fire extinguishing plan is executed.

[0087] This application discloses a forest fire risk inspection and rescue system for the prevention, early warning, rescue, and decision-making support for forest fires. The system comprises a patrol module, an image recognition module, an alarm module, and an emergency planning module. It integrates multiple drones equipped with thermal imagers and high-definition cameras, meteorological data, and passive IoT terminals to achieve comprehensive monitoring and rapid response to forest fires. The patrol module determines drone patrol plans, the image recognition module identifies suspected fire points and trapped personnel based on thermal images collected by drones, the alarm module dispatches a drone swarm, and the emergency planning module determines escape routes for trapped personnel and firefighting strategies.

[0088] Furthermore, the forest fire risk inspection and rescue system proposed in this application can be deployed on a cloud platform, where data processing and decision support can be provided. Moreover, because this application uses a low-cost thermal imager and mature image recognition algorithms, the overall cost is low and it is easy to deploy on a large scale.

[0089] The forest fire identification method provided in this application can be executed by a forest fire identification device. This application uses a forest fire identification device executing the forest fire identification method as an example to illustrate the forest fire identification device provided in this application.

[0090] Figure 3 This is a schematic diagram of the structure of a forest fire detection device disclosed in an embodiment of this application. Figure 3 As shown, the forest fire detection device 300 includes: a receiving module 310, an obtaining module 320, and a determining module 330.

[0091] In this application, the receiving module 310 is used to receive a first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during forest patrol; the obtaining module 320 is used to obtain a probability value of the first thermal image containing a fire point by inputting the first thermal image into a first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire point; and the determining module 330 is used to determine that there is a suspected fire point in the first thermal image if the probability value is greater than a preset threshold.

[0092] In one implementation, the first image recognition module includes a pre-trained CNN model and a pre-trained SVM classifier. The obtaining module 320 obtains the probability value of the first thermal image containing a fire point by inputting the first thermal image into the first image recognition module, including: inputting the first thermal image into the pre-trained CNN model to obtain a feature vector corresponding to the first thermal image output by the pre-trained CNN model, wherein the pre-trained CNN model is used to extract feature information of the input thermal image and output a feature vector corresponding to the feature information; and inputting the feature vector into the pre-trained SVM classifier to obtain the probability value of the first thermal image containing a fire point output by the pre-trained SVM classifier, wherein the pre-trained SVM classifier is used to determine the probability value of containing a fire point based on the input feature vector.

[0093] In one implementation, the first thermal image carries first location information acquired by the drone. The device further includes: a scheduling module, configured to schedule a drone cluster to a target area corresponding to the first location information after determining that a suspected fire point exists in the first thermal image; wherein the drones in the drone cluster are equipped with imaging devices, including cameras and thermal imagers; a receiving module 310, configured to receive on-site data corresponding to the target area sent by the drone cluster, wherein the on-site data includes image data and video data, and the on-site data is acquired by the drone cluster based on the imaging device; a determining module 330, configured to, based on the on-site data and wind direction information, determine the terrain information, obstacle information, fire spread information, and second location information of the trapped personnel, when it is determined based on the on-site data that a fire actually exists and there are trapped personnel; and a determining module 330, configured to, based on the terrain information, obstacle information, second location information, fire spread information, wind direction information, and third location information of the safe area, determine the escape route of the trapped personnel.

[0094] In one implementation, the determining module 330 determines the escape route of the trapped person based on the terrain information, the obstacle information, the second location information, the fire spread information, the wind direction information, and the third location information of the safe area. This includes: generating a grid map based on the terrain information and the obstacle information, wherein each grid in the grid map represents a passable or impassable area; and determining the escape route of the trapped person based on the grid map, the second location information, the fire spread information, the wind direction information, and the third location information of the safe area.

[0095] In one implementation, the device further includes: a generation module, configured to randomly generate a fire extinguishing plan after receiving on-site data corresponding to the target area sent by the UAV cluster, and if the fire is confirmed to actually exist based on the on-site data; the generation module is further configured to generate a new fire extinguishing plan by randomly perturbing the current fire extinguishing plan; the determination module 330 is further configured to determine the new fire extinguishing plan as the current fire extinguishing plan if the objective function corresponding to the new fire extinguishing plan is less than the objective function corresponding to the current fire extinguishing plan, wherein the objective function corresponding to the fire extinguishing plan is determined based on the target information corresponding to the fire extinguishing plan, the target information including the effective fire extinguishing area, total cost, estimated fire extinguishing time, the relationship between the location of the fire extinguishing action and the direction of fire spread, and the terrain information of the location of the fire extinguishing action, the objective function being used to evaluate the merits of the fire extinguishing plan; and an iteration module, configured to return to the step of generating a new fire extinguishing plan by randomly perturbing the current fire extinguishing plan, until a preset number of iterations is met.

[0096] In one implementation, the determining module 330 is further configured to, before receiving the first thermal image, determine the target fire risk level based on the temperature within a first preset time period, the relative humidity within the first preset time period, the wind speed within the first preset time period, the rainfall within a second preset time period, and the rainfall within a third preset time period, wherein the second preset time period is a portion of the third preset time period; the determining module 330 is further configured to, based on the target fire risk level and the target correspondence, determine the drone patrol frequency corresponding to the target fire risk level, wherein the target correspondence is the correspondence between the fire risk level and the drone patrol frequency.

[0097] The forest fire identification device provided in this application embodiment can realize all the processes implemented in the forest fire identification method embodiment, and will not be described again here to avoid repetition.

[0098] Optionally, such as Figure 4As shown, this application embodiment also provides an electronic device 400, including a processor 401 and a memory 402. The memory 402 stores a program or instructions that can run on the processor 401. When the program or instructions are executed by the processor 401, they implement the various steps of the above-described forest fire identification method embodiment and can achieve the same technical effect. To avoid repetition, they will not be described again here.

[0099] It should be noted that the electronic devices in the embodiments of this application include mobile electronic devices and non-mobile electronic devices.

[0100] This application also provides a readable storage medium storing a program or instructions. When the program or instructions are executed by a processor, they implement the various processes of the above-described forest fire identification method embodiments and achieve the same technical effect. To avoid repetition, they will not be described again here.

[0101] The processor is the processor in the electronic device described in the above embodiments. The readable storage medium includes computer-readable storage media, such as computer read-only memory (ROM), random access memory (RAM), magnetic disk, or optical disk.

[0102] This application also provides a computer program product, which includes a computer program stored on a non-transitory computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer, cause the computer to perform the steps of the forest fire identification method described above.

[0103] The above embodiments of this application focus on describing the differences between the various embodiments. As long as the different optimization features between the various embodiments are not contradictory, they can be combined to form a better embodiment. For the sake of brevity, they will not be described in detail here.

[0104] The above description is merely an embodiment of this application and is not intended to limit the scope of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for identifying forest fire spots, characterized in that, include: Receive a first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during a forest patrol. By inputting the first thermal image into the first image recognition module, the probability value of the first thermal image containing a fire point is obtained from the output of the first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire point; If the probability value is greater than a preset threshold, it is determined that there is a suspected fire point in the first thermal image.

2. The method according to claim 1, characterized in that, The first image recognition module includes a pre-trained CNN model and a pre-trained SVM classifier. The step of inputting the first thermal image into the first image recognition module to obtain the probability value of the first thermal image containing a fire point, as output by the first image recognition module, includes: By inputting the first thermal image into the pre-trained CNN model, a feature vector corresponding to the first thermal image is obtained from the output of the pre-trained CNN model. The pre-trained CNN model is used to extract feature information from the input thermal image and output a feature vector corresponding to the feature information. By inputting the feature vector into the pre-trained SVM classifier, the probability value of the first thermal image containing a fire point is obtained from the output of the pre-trained SVM classifier, wherein the pre-trained SVM classifier is used to determine the probability value of containing a fire point based on the input feature vector.

3. The method according to claim 1, characterized in that, The first thermal image carries first location information acquired by the drone. After determining that a suspected fire point exists in the first thermal image, the method further includes: The drone swarm is dispatched to the target area corresponding to the first location information, wherein the drones in the drone swarm are equipped with imaging devices, including cameras and thermal imagers; The system receives on-site data corresponding to the target area sent by the drone cluster, wherein the on-site data includes image data and video data, and the on-site data is acquired by the drone cluster based on the imaging device. If, based on the on-site data, it is determined that a fire actually exists and that there are trapped personnel, then, based on the on-site data and wind direction information, the terrain information, obstacle information, fire spread information, and the second location information of the trapped personnel are determined. Based on the terrain information, obstacle information, second location information, fire spread information, wind direction information, and third location information of the safe area, the escape route of the trapped personnel is determined.

4. The method according to claim 3, characterized in that, The process of determining the escape route for the trapped personnel based on the terrain information, obstacle information, second location information, fire spread information, wind direction information, and third location information of the safe area includes: Based on the terrain information and the obstacle information, a grid map is generated, wherein each grid in the grid map represents a passable or impassable area; Based on the grid map, the second location information, the fire spread information, the wind direction information, and the third location information of the safe area, the escape route of the trapped personnel is determined.

5. The method according to claim 3, characterized in that, After receiving the on-site data corresponding to the target area sent by the drone cluster, the method further includes: If the fire is confirmed to be real based on the on-site data, a fire extinguishing plan is generated randomly. By randomly perturbing the current fire extinguishing plan, a new fire extinguishing plan is generated; If the objective function corresponding to the new fire extinguishing plan is less than the objective function corresponding to the current fire extinguishing plan, the new fire extinguishing plan is determined as the current fire extinguishing plan. The objective function corresponding to the fire extinguishing plan is determined based on the objective information corresponding to the fire extinguishing plan. The objective information includes the effective fire extinguishing area, total cost, estimated fire extinguishing time, the relationship between the location of the fire extinguishing operation and the direction of fire spread, and the terrain information of the location of the fire extinguishing operation. The objective function is used to evaluate the merits of the fire extinguishing plan. Return to the step of generating a new fire extinguishing plan by randomly perturbing the current fire extinguishing plan, until the preset number of iterations is met.

6. The method according to claim 1, characterized in that, Before receiving the first thermal image, the method further includes: The target fire risk level is determined based on the temperature during a first preset time period, the relative humidity during the first preset time period, the wind speed during the first preset time period, the rainfall during a second preset time period, and the rainfall during a third preset time period, wherein the second preset time period is a portion of the third preset time period. Based on the target fire risk level and the target correspondence, the drone patrol frequency corresponding to the target fire risk level is determined, wherein the target correspondence is the correspondence between the fire risk level and the drone patrol frequency.

7. A forest fire detection device, characterized in that, include: A receiving module is used to receive a first thermal image, wherein the first thermal image is a thermal image collected by a drone equipped with a thermal imager during a forest patrol. The module is used to obtain the probability value of the first thermal image containing a fire point by inputting the first thermal image into the first image recognition module, wherein the first image recognition module is used to determine the probability value of the input thermal image containing a fire point; The determination module is used to determine that there is a suspected fire point in the first thermal image when the probability value is greater than a preset threshold.

8. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a program or instructions that can run on the processor, the program or instructions being executed by the processor to implement the steps of the forest fire identification method as described in any one of claims 1-6.

9. A readable storage medium, characterized in that, The readable storage medium stores a program or instructions that, when executed by a processor, implement the steps of the forest fire identification method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions that, when executed by a computer, cause the computer to perform the steps of the forest fire identification method as described in any one of claims 1-6.

Images