Forest fire monitoring method, device and equipment and readable storage medium

Abstract

The invention relates to a forest fire monitoring method, device and equipment and a readable storage medium. The method comprises the following steps: acquiring fire monitoring data of a target area collected by a monitoring device; the fire monitoring data comprises environment image data and corresponding device pose data; determining a first fire confidence coefficient of the target area based on the environment image data, and determining a fire state based on the first fire confidence coefficient; under the condition that the fire behavior state is that the fire behavior exists, determining the fire point position where the fire behavior exists based on the device pose data; and the monitoring visual angle of the monitoring device is adjusted, so that the fire behavior at the fire point position can be monitored in all directions. By adopting the method, forest fire monitoring can be accurately carried out.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of forest fire prevention, and in particular to a forest fire monitoring method and device, computer equipment, a computer readable storage medium and a computer program product. BACKGROUND

[0002] The forest environment state has the characteristics of wide spatial distribution, complex terrain, interwoven ecological factors and multiple disaster outbreaks. Forests, land, vegetation and habitats are long-term affected by multiple dynamic environmental factors such as climate change, geological activity, human disturbance and fire risk. Forest fires are the most dangerous enemy of forest environments and the most terrible disaster of forestry. They can bring the most harmful and most destructive consequences to forests. Therefore, it is necessary to effectively monitor forest fires.

[0003] In the traditional technology, the forest fire is mainly monitored by artificial patrol or fixed position sensor survey, which has the problem of inaccurate fire monitoring due to long monitoring data update cycle or limited monitoring range. SUMMARY

[0004] Therefore, it is necessary to provide a forest fire monitoring method, device, computer equipment, computer readable storage medium and computer program product capable of accurately monitoring forest fires to solve the above technical problems.

[0005] In a first aspect, the present application provides a forest fire monitoring method, comprising:

[0006] acquiring fire monitoring data of a target area collected by a monitoring device; the fire monitoring data comprising environment image data and corresponding device pose data;

[0007] determining a first fire confidence of the target area based on the environment image data, and determining a fire state based on the first fire confidence;

[0008] in the case that the fire state is that there is a fire, determining a fire point position where the fire exists based on the device pose data;

[0009] adjusting the monitoring angle of the monitoring device to realize all-around monitoring of the fire at the fire point position.

[0010] In a second aspect, the present application further provides a forest fire monitoring device, comprising:

[0011] a monitoring data acquisition module configured to acquire fire monitoring data of a target area collected by a monitoring device; the fire monitoring data comprising environment image data and corresponding device pose data;

[0012] The fire status determination module is used to determine a first fire confidence level of the target area based on the environmental image data, and to determine the fire status based on the first fire confidence level.

[0013] The fire location determination module is used to determine the location of the fire point based on the device pose data when the fire situation is that a fire exists.

[0014] The monitoring angle adjustment module is used to adjust the monitoring angle of the monitoring device to achieve all-round monitoring of the fire situation at the fire point location.

[0015] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the forest fire monitoring method provided in the first aspect.

[0016] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the forest fire monitoring method provided in the first aspect.

[0017] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the forest fire monitoring method provided in the first aspect.

[0018] The aforementioned forest fire monitoring methods, devices, computer equipment, computer-readable storage media, and computer program products acquire fire monitoring data of the target area collected by the monitoring device, determine the first fire confidence level of the target area based on environmental image data in the fire monitoring data, determine the fire status based on the first fire confidence level, and determine the location of the fire point based on the device pose data in the fire monitoring data when the fire status is fire. They then adjust the monitoring angle of the monitoring device to achieve comprehensive monitoring of the fire point location. This enables accurate determination of the first fire confidence level of the target area based on environmental image data, determination of the fire status based on the first fire confidence level, determination of the fire point location in the target area when a fire exists, and then focused monitoring of the fire point location by adjusting the monitoring angle of the monitoring device. This allows for automatic determination of fire point locations from a large target area, and concentrated monitoring of the fire point location, enabling accurate understanding of the fire development and achieving accurate monitoring of forest fires. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is an application environment diagram of a forest fire monitoring method in one embodiment;

[0021] Figure 2 This is a flowchart illustrating a forest fire monitoring method in one embodiment;

[0022] Figure 3 This is a left-side view of a holographic multimodal intelligent sensing device in one embodiment;

[0023] Figure 4 This is a front view schematic diagram of a holographic multimodal intelligent sensing device in one embodiment;

[0024] Figure 5 This is a top view schematic diagram of a holographic multimodal intelligent sensing device in one embodiment;

[0025] Figure 6 This is a structural block diagram of a forest fire monitoring device in one embodiment;

[0026] Figure 7 This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0028] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0029] The forest fire monitoring method provided in this application can be applied to, for example... Figure 1In the application environment shown, terminal 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104 or placed on the cloud or other network servers. Server 104 can acquire fire monitoring data of the target area collected by the monitoring device sent by the terminal. This fire monitoring data includes environmental image data and corresponding device pose data. Server 104 determines the first fire confidence level of the target area based on the environmental image data, and determines the fire status based on the first fire confidence level. If the fire status indicates a fire, it determines the location of the fire point based on the device pose data and adjusts the monitoring angle of the monitoring device to achieve comprehensive monitoring of the fire point. Terminal 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, drones, low-altitude aircraft, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, smart vehicle devices, projection devices, etc. Portable wearable devices can include smartwatches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. Server 104 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. It should be noted that the forest fire monitoring method provided in this application embodiment is applicable not only to application scenarios involving server-terminal interaction, but also to application scenarios involving single terminals, single servers, terminal-to-terminal interaction, or server-to-server interaction.

[0030] In one exemplary embodiment, such as Figure 2 As shown, a forest fire monitoring method is provided, which can be applied to... Figure 1 Taking the server in the example, the explanation includes the following steps 202 to 208. Wherein:

[0031] Step 202: Obtain fire monitoring data of the target area collected by the monitoring device; wherein, the fire monitoring data includes environmental image data and corresponding device pose data.

[0032] The monitoring device refers to a device capable of collecting fire monitoring data. A monitoring device may include one or more devices. For example, a monitoring device may include a visible light camera (e.g., an RGB camera), a multispectral camera, a lidar, an infrared focal plane array, an accelerometer, a gyroscope, a magnetometer, a GNSS (Global Navigation Satellite System) receiver, etc. The target area refers to the forest area where fire monitoring is conducted. The size of the target area can be set according to the actual application scenario. Environmental image data refers to image data collected from the environment of the target area. Environmental image data may include, for example, visible light image data collected by a visible light camera, multispectral image data collected by a multispectral camera, infrared thermal imaging data collected by an infrared focal plane array, and radar reflectivity image data collected by a lidar. Device pose data refers to data characterizing the pose of the monitoring device. Device pose data includes the position data and attitude data of the monitoring device. Device pose data can be acquired through pose sensors installed on the monitoring device.

[0033] It should be noted that fire monitoring data includes environmental image data and corresponding device pose data. Device pose data corresponding to environmental image data refers to device pose data whose acquisition timestamp matches that of the environmental image data. In practical applications, a clock synchronizer can be used to unify the acquisition timestamps of environmental image data and device pose data, achieving spatiotemporal registration of the two data. Furthermore, fire monitoring data can be uniformly mapped to a geographic coordinate system, providing a solid foundation for subsequent fire status identification and 3D situation reconstruction.

[0034] For example, the server can acquire initial image data of the target area collected by the monitoring device, and sequentially perform denoising and light correction processing on the initial image data to obtain environmental image data. This improves the clarity of the environmental image data and enhances the accuracy of fire status identification under day / night or severe weather conditions. For instance, the initial image data can be weighted and filtered to obtain denoised image data, suppressing sensor noise from the monitoring device while preserving edge features of the image data. Then, based on the average ambient brightness feedback from the light sensor, linear light correction is performed on the denoised image to obtain the environmental image data. Optionally, after the monitoring device collects fire monitoring data of the target area, the fire monitoring data can be encrypted using a quantum key, and the encrypted fire monitoring data can be sent to the server. The quantum key can be distributed via a microsatellite link.

[0035] Step 204: Determine the first fire confidence level of the target area based on the environmental image data, and determine the fire status based on the first fire confidence level.

[0036] The first fire confidence level can be used to characterize the fire status of the target area. The fire status can include whether a fire exists or not.

[0037] In an exemplary embodiment, features can be extracted from environmental image data to obtain environmental image features, and fire prediction can be performed based on the environmental image features to obtain a first fire confidence level.

[0038] In one exemplary embodiment, the environmental image data includes multiple types, and features of each type of environmental image data can be extracted separately. The features of the various environmental image data are then fused to obtain fused image features. Fire prediction is then performed based on the fused image features to obtain a first fire confidence level.

[0039] In an exemplary embodiment, features can be extracted from environmental image data to obtain environmental image features, active semantic understanding can be performed on the environmental image features to obtain semantic information of the environmental image data, and fire prediction can be performed based on the environmental image features and semantic information to obtain a first fire confidence level.

[0040] For example, if the confidence level of the first fire is higher than the confidence threshold, the fire status is determined to be present; if the confidence level of the first fire is not higher than the confidence threshold, the fire status is determined to be absent.

[0041] In an exemplary embodiment, the environmental image data includes visible light image data and infrared thermal imaging data. If the first fire confidence level is higher than a confidence threshold, feature extraction is performed on the visible light image data to obtain smoke texture features, and fire prediction is performed on the smoke texture features to obtain a second fire confidence level. Feature extraction is performed on the infrared thermal imaging data to obtain temperature features of the target area, and fire prediction is performed on the temperature features to obtain a third fire confidence level. If the second fire confidence level is greater than the first confidence threshold and the third fire confidence level is greater than the second confidence threshold, the fire status is determined to be present. If the first fire confidence level is not higher than the confidence threshold, feature extraction is performed on the visible light image data to obtain smoke texture features, and fire prediction is performed on the smoke texture features to obtain the second fire confidence level; feature extraction is performed on the infrared thermal imaging data to obtain the temperature features of the target area, and fire prediction is performed on the temperature features to obtain the third fire confidence level; if the second fire confidence level is not greater than the first confidence threshold or the third fire confidence level is not greater than the second confidence threshold, the fire status is determined to be no fire.

[0042] Step 206: If the fire situation is "fire exists", determine the location of the fire point based on the device pose data.

[0043] In practical applications, once a fire is confirmed, it is necessary to determine the location of the fire and issue a fire handling command to promptly address the fire at that location.

[0044] For example, the location of a fire can be determined by combining device pose data and GNSS positioning coordinates. It is easy to understand that GNSS positioning coordinates are typically in a spatial rectangular coordinate system. Based on the device pose data and GNSS positioning coordinates, the spatial rectangular coordinates of the fire can be determined, and then these coordinates can be converted into geographic coordinates. It should be noted that the location of a fire can also be determined using other methods.

[0045] Step 208: Adjust the monitoring angle of the monitoring device to achieve all-round monitoring of the fire situation at the fire point location.

[0046] In practical applications, after determining the location of a fire, the monitoring angles of various monitoring devices can be adjusted. This allows each device to simultaneously scan the fire from different angles, creating complementary multi-view data and effectively eliminating blind spots. Simply put, before determining the fire's location, each monitoring device primarily focuses on monitoring the target area. Adjusting the monitoring angles to achieve comprehensive monitoring of the fire's location enables more accurate monitoring of the fire's condition.

[0047] In the aforementioned forest fire monitoring method, fire monitoring data of the target area collected by the monitoring device is acquired. Based on the environmental image data in the fire monitoring data, the first fire confidence level of the target area is determined, and the fire status is determined based on the first fire confidence level. If the fire status is that a fire exists, the location of the fire point is determined based on the device pose data in the fire monitoring data. The monitoring angle of the monitoring device is adjusted to achieve comprehensive monitoring of the fire at the fire point location. This method can accurately determine the first fire confidence level of the target area based on environmental image data, determine the fire status based on the first fire confidence level, determine the location of the fire point in the target area if a fire exists, and then adjust the monitoring angle of the monitoring device to focus on monitoring the fire at the fire point location. This method can automatically determine the fire point location from a large target area, and then focus on monitoring the fire at the fire point location, accurately understand the development of the fire at the fire point, and achieve accurate monitoring of forest fires.

[0048] In some embodiments, the environmental image data includes visible light image data, infrared thermal imaging data, and radar reflectivity image data; step 204, determining the first fire confidence level of the target area based on the environmental image data, includes:

[0049] Feature extraction is performed on visible light image data to obtain smoke texture features; feature extraction is performed on infrared thermal imaging data to obtain temperature features of the target area; feature extraction is performed on radar reflectivity image data to obtain smoke reflection features; smoke texture features, temperature features, and smoke reflection features are fused to obtain multimodal fusion features; fire prediction is performed on the multimodal fusion features to obtain the first fire confidence level of the target area.

[0050] Visible light image data refers to image data acquired by a visible light camera. Infrared thermal imaging data refers to image data acquired by an infrared imaging device. Radar reflectivity image data is used to characterize the difference between the actual signal reflected back after a radar signal is emitted and the ideal reflected signal. It is easy to understand that the reflectivity of each point cloud corresponding to the radar emitted signal can be characterized in the form of image pixels, thus obtaining the radar reflectivity image data for each point cloud. Typically, the reflectivity of point clouds differs for different objects; for example, the reflectivity of hot smoke and solid obstacles are completely different.

[0051] In practical applications, convolutional neural networks can be used to extract features from visible light image data to obtain smoke texture features, which characterize the texture and brightness fluctuations of smoke in the target area. Feature extraction from infrared thermal imaging data yields temperature features, which characterize the temperature distribution and anomalies in the target area. Feature extraction from radar reflectivity image data provides smoke reflectivity features, which characterize the smoke distribution in the target area. These smoke texture features, temperature features, and smoke reflectivity features can be weighted and fused to obtain a multimodal fusion feature. Then, a prediction network is used to predict the fire situation based on this multimodal fusion feature, obtaining the first fire confidence level for the target area. It should be noted that the weighting weights for the weighted fusion can be set according to the specific application scenario.

[0052] In an exemplary embodiment, features are extracted from visible light image data to obtain a smoke texture feature image. A semantic correlation matrix is ​​calculated between any pixels in the smoke texture feature image. The smoke texture feature image is then weighted based on this semantic correlation matrix to obtain semantically enhanced smoke texture features. Similarly, semantically enhanced temperature features and smoke reflection features can be obtained sequentially. These semantically enhanced smoke texture features, temperature features, and smoke reflection features are then fused to obtain multimodal fusion features. This improves the accuracy of the multimodal fusion features.

[0053] In this embodiment, by fusing smoke texture features, temperature features, and smoke reflection features, a multimodal fusion feature is obtained. Fire prediction is then performed on the multimodal fusion feature to obtain the first fire confidence level of the target area. This enables fire prediction based on multimodal data and improves the accuracy of the first fire confidence level.

[0054] In some embodiments, determining the fire status based on the first fire confidence level in step 204 includes:

[0055] Calculate the confidence difference between the confidence levels of the first fire situation corresponding to two adjacent monitoring periods; if the degree of change of a preset number of consecutive confidence difference values ​​is greater than the degree of change threshold, determine that the fire situation exists.

[0056] The monitoring cycle refers to the period during which forest fires are monitored. The duration of the monitoring cycle can be set according to the actual application scenario. For example, the monitoring cycle for forest fires can be 5 minutes, 30 minutes, 1 hour, 1 day, or real-time. Each monitoring cycle involves one forest fire monitoring session, meaning that there is a monitoring cycle between every two forest fire monitoring sessions.

[0057] For example, the server can calculate the confidence difference between the first fire confidence levels corresponding to two adjacent monitoring periods, determine the degree of change of a preset number of consecutive confidence difference values, and if the degree of change is greater than a change threshold, determine that the fire status is present. If the degree of change is not greater than the change threshold, determine that the fire status is absent. The preset number can be set according to the actual application scenario, for example, the preset number is 3, 4 or 5.

[0058] For example, the degree of change of a preset number of consecutive confidence difference values ​​can be characterized by the rate of change or variance of the preset number of consecutive confidence difference values.

[0059] In one example, the corresponding time stability can be calculated based on the confidence difference value. If the degree of change in a preset number of consecutive time stability values ​​exceeds a change threshold, the fire status is determined to be present. In other words, the degree of change in the confidence difference value can be characterized by the degree of change in the time stability value. For example, if the first monitoring period and the second monitoring period are two adjacent monitoring periods, the difference between the confidence value of the first fire in the first monitoring period and the confidence value of the second fire in the second monitoring period is used as the confidence difference value between the two monitoring periods. The ratio of the confidence difference value to the monitoring period is used as the time stability value. If the degree of change in three consecutive time stability values ​​exceeds the change threshold, the fire status is determined to be present.

[0060] In this embodiment, by calculating the confidence difference between the first fire confidence levels corresponding to two adjacent monitoring periods, if the degree of change of a preset number of consecutive confidence difference values ​​is greater than the degree of change threshold, the fire status is determined to be present. This allows for accurate determination of the fire status based on the difference and fluctuation of the first fire confidence levels corresponding to different monitoring periods.

[0061] In some embodiments, before determining the fire status based on the first fire confidence level, the above method further includes:

[0062] Feature extraction is performed on visible light image data in the environmental image data to obtain smoke texture features, and fire prediction is performed on the smoke texture features to obtain a second fire confidence level; feature extraction is performed on infrared thermal imaging data in the environmental image data to obtain temperature features of the target area, and fire prediction is performed on the temperature features to obtain a third fire confidence level; if the second fire confidence level is greater than the first confidence threshold and the third fire confidence level is greater than the second confidence threshold, the step of determining the fire status based on the first fire confidence level is executed.

[0063] The second fire confidence score is obtained by predicting the fire situation based on smoke texture features. The third fire confidence score is obtained by predicting the fire situation based on temperature features. The first confidence threshold is a confidence threshold set for smoke texture features. The second confidence threshold is a confidence threshold set for temperature features. The first or second confidence threshold can be set according to the actual application scenario.

[0064] In practical applications, after determining the first fire confidence level of the target area based on environmental image data, the server extracts features from the visible light image data in the environmental image data to obtain smoke texture features, and then predicts the fire based on these smoke texture features to obtain the second fire confidence level. Next, it extracts features from the infrared thermal imaging data in the environmental image data to obtain the temperature features of the target area, and then predicts the fire based on these temperature features to obtain the third fire confidence level. If both the second and third fire confidence levels are greater than the first and second confidence thresholds, the fire status is determined based on the first fire confidence level. This approach enables the comprehensive determination of whether a fire status has been triggered based on multimodal and single-modal environmental image data, improving the accuracy of the timing of fire status determination.

[0065] In some embodiments, step 206, determining the location of the fire point based on the device pose data, includes:

[0066] The system acquires satellite positioning coordinates and distances obtained through laser ranging; based on the satellite positioning coordinates, device pose data, and distances, it uses a spatial forward intersection algorithm to determine the ignition location where a fire exists.

[0067] Among them, satellite positioning coordinates refer to the coordinates of the fire point obtained through satellite positioning. The satellite positioning coordinates are essentially the initial coordinates corresponding to the location of the fire point where there is a fire. Then, the initial coordinates are optimized based on the distance obtained from laser ranging and the device pose data to obtain the target coordinates corresponding to the location of the fire point where there is a fire.

[0068] In one example, assuming the satellite positioning coordinates obtained via GNSS are (L0, B0, H0), and the corresponding spatial rectangular coordinates are (X0, Y0, Z0), the azimuth angle α and elevation angle β of the monitoring device can be determined based on the device's pose data. The distance obtained from lidar ranging is D. L The three-dimensional coordinates of the fire points where fires are present are determined using the spatial forward intersection algorithm:

[0069] Formula (1)

[0070] Among them, (X) f ,Y f Z f ( ) represents spatial rectangular coordinates, which can be converted into the geographic coordinates of the fire point (L) f B f ), L f Indicates longitude, B f This indicates the dimension; since the height H is 0 at this point, it can be omitted.

[0071] In this embodiment, the ignition location of the fire is determined by using a spatial forward intersection algorithm based on satellite positioning coordinates, device pose data, and distance. This method can accurately pinpoint the location of the fire.

[0072] In some embodiments, fire monitoring data may further include ambient sound spectrum data, which can be collected via a microphone array. The ambient sound spectrum data is matched against abnormal spectra; when a match is found, a suspicious event is identified. Abnormal spectra include, for example, the sound spectrum corresponding to chainsaw sounds, deflagration sounds, etc. Assuming a fire is confirmed and a suspicious event is identified, the fire event has a higher priority than the suspicious event, meaning the monitoring device is controlled to centrally monitor the fire.

[0073] In some embodiments, the above method further includes:

[0074] After the fire at the identified location is extinguished, the residual heat information, smoke diffusion range, and vegetation spectral recovery information at the fire location are monitored using the post-disaster monitoring mode. Based on the residual heat information, smoke diffusion range, and vegetation spectral recovery information, the current vegetation index is determined. Based on the pre-disaster vegetation index and the current vegetation index at the fire location before the fire occurred, the vegetation recovery index at the fire location is determined.

[0075] In practical applications, after the server detects that a fire at a specific location has been extinguished, it uses a post-disaster monitoring mode to dispatch monitoring devices to monitor residual heat information, smoke diffusion range, and vegetation spectral restoration information at the fire location. Residual heat information refers to the residual thermal radiation at the fire site after it has been extinguished, which can be acquired using an infrared thermal imager. The smoke diffusion range can be characterized by the smoke diffusion area. The smoke diffusion range can also be determined by the intensity of the reflected signal from the laser point cloud. Vegetation spectral restoration information can be characterized by the proportion of pure vegetation spectra and mixed spectra.

[0076] The current vegetation index is used to characterize the biomass of vegetation under current conditions. The server can determine the current vegetation index of the fire zone based on residual heat information, smoke and dust diffusion range, and vegetation spectral recovery information. For example, scores can be assigned based on residual heat information, smoke and dust diffusion range, and vegetation spectral recovery information respectively, and a weighted sum of the scores can be obtained to obtain a total score. Based on the correspondence between the scores and vegetation indices, the current vegetation index corresponding to the total score can be determined.

[0077] The pre-fire vegetation index refers to the vegetation index at the fire location before the fire occurred. The vegetation recovery index is used to characterize the ease of vegetation recovery. The higher the vegetation recovery index, the easier the vegetation recovery; conversely, the lower the vegetation recovery index, the more difficult the vegetation recovery. For example, the vegetation recovery index at the fire location can be determined based on the minimum value of the pre-fire vegetation index, the maximum value of the reference vegetation index under healthy conditions, and the current vegetation index. For example, the vegetation recovery index R at the fire location... t It can be calculated using the following formula (2).

[0078] Formula (2)

[0079] Among them, V t V represents the current vegetation index; min V represents the minimum value of the pre-disaster vegetation index; max This represents the maximum value of the reference vegetation index under healthy conditions.

[0080] In practical applications, the vegetation restoration index R at the fire location is statistically analyzed. t By analyzing the temporal trends of fire locations, we can determine the ecological regeneration potential of fire sites and generate fire trace distribution maps and vegetation restoration assessment reports.

[0081] In this embodiment, after the fire at the identified fire location has been extinguished, the current vegetation index is determined by monitoring the residual heat information, smoke diffusion range, and vegetation spectral recovery information at the fire location through the post-disaster monitoring mode. Based on the pre-disaster vegetation index and the current vegetation index at the fire location before the fire occurred, the vegetation recovery index at the fire location is determined. This enables continued monitoring of the residual impact information of the fire point after the fire has been extinguished, thereby determining the vegetation recovery index of the corresponding area and achieving full-process monitoring of forest fires and forest protection.

[0082] In one example, a holographic multimodal intelligent sensing device for forest disaster prevention and control is provided, which can be used to implement the forest fire monitoring method in the above embodiments. The holographic multimodal intelligent sensing device for forest disaster prevention and control can be considered a monitoring device. This device includes a comprehensive sensing module, an active detection module, a positioning and attitude determination module, an edge computing module, a quantum multimodal communication module, a power management module, an environmental control module, and a structural protection module. These modules work collaboratively through standardized interfaces and a data bus. Specifically, the comprehensive sensing module includes a multispectral intelligent vision sensor, an uncooled infrared focal plane array, a global shutter visible light camera, a high-sensitivity microphone array, a digital temperature and humidity sensor, an atmospheric pressure sensor, and an ultraviolet sensor; the active detection module includes a lidar scanning unit, a two-dimensional microelectromechanical galvanometer system, an electrically adjustable zoom optical lens group, a laser emitter, an echo signal amplifier, and a photon counting detector; the positioning and attitude determination module includes a multi-band global navigation satellite system receiver, a nine-axis inertial measurement unit, a dual-antenna orientation module, and a satellite-based augmentation system demodulator; the edge computing module includes an artificial intelligence processor, a field-programmable gate array chip, and a synchronous dynamic tracking... The device includes a physical memory, embedded multimedia memory, and a neural network accelerator; a quantum multimode communication module comprising an industrial Ethernet switch, a cellular mobile communication module, a low-power wide-area network communication unit, a satellite communication transceiver, a quantum microsatellite link system, and a precision clock synchronizer; a power management module comprising a solar panel, a battery pack, a wide-voltage input DC-DC converter, an intelligent charge / discharge controller, and a supercapacitor buffer; an environmental control module comprising a semiconductor cooling chip, a temperature sensor, a humidity sensor, an anti-condensation heater, and a thermally conductive silicone pad; and a structural protection module comprising an aluminum alloy sealed housing, a lightning protection device, a vibration-damping bracket, an anti-corrosion coating, and a quick-mount base. Exemplarily, a left view of the device is shown below. Figure 3 As shown, the front view is as follows Figure 4 As shown, the top view is as follows Figure 5 As shown. The meanings of the numbers in the figure are as follows:

[0083] 1-Solar panel; 2-Satellite-based augmentation system demodulator; 3-Multi-band global navigation satellite system receiver; 4-Lightning protection device; 5-Photon counting detector; 6-Nine-axis inertial measurement unit; 7-Semiconductor thermoelectric cooler; 8-Wide voltage input DC-DC converter; 9-Quick-install anti-vibration gimbal base; 10-Intelligent charge / discharge controller; 11-Battery pack; 12-Humidity sensor; 13-Laser emitter; 14-Two-dimensional MEMS galvanometer system; 15-LiDAR scanning unit; 16-Quantum microsatellite link system; 17-Digital temperature and humidity sensor; 18-Multispectral intelligent vision sensor; 19-High-sensitivity microphone array; 20-Atmospheric pressure sensor; 21- 21-Uncooled infrared focal plane array; 22-Ultraviolet sensor; 23-Cellular mobile communication module; 24-Low power wide area network communication unit; 25-Global shutter visible light camera; 26-Satellite communication transceiver; 27-Precision clock synchronizer; 28-Synchronous dynamic random access memory; 29-Artificial intelligence processor; 30-Dual antenna directional module; 31-Echo signal amplifier; 32-Aluminum alloy sealed housing; 33-Embedded multimedia memory; 34-Industrial Ethernet switch; 35-Neural network accelerator; 36-Supercapacitor buffer; 37-Field programmable gate array chip; 38-Temperature sensor; 39-Electric zoom optical lens assembly; 40-Anti-condensation heater.

[0084] The device can be implemented through the following six steps in the process of monitoring forest fires.

[0085] Step 1:

[0086] After power-on, the device first executes the system initialization process. The power management module starts, with the battery pack and solar panels working together to provide power, and a wide-voltage input DC-DC converter ensuring system voltage regulation. The artificial intelligence processor and field-programmable gate array chip in the edge computing module complete the boot process, and the synchronous dynamic random access memory and embedded multimedia memory perform integrity self-checks. Subsequently, the integrated sensing module and active detection module start up in sequence, and the multispectral imager, infrared array, temperature, humidity and pressure sensors complete baseline calibration and dark field correction.

[0087] The positioning and attitude determination module uses a multi-frequency GNSS receiver and a dual-antenna orientation module to search for and locate satellites, acquiring precise latitude, longitude, elevation, and attitude angles. The satellite-based augmentation system provides differential correction, improving spatial positioning accuracy. The nine-axis inertial measurement unit completes zero-bias calibration under static conditions. The system's high-precision clock synchronizer extracts a time reference from satellite signals, providing a unified timestamp for multimodal data.

[0088] After initialization, the cellular communication unit, low-power wide-area network unit, and satellite communication transceiver in the multi-mode communication module establish connections sequentially. During satellite communication initialization, the system prioritizes establishing a handshake connection with microsatellites possessing quantum key distribution capabilities. An initial shared key is generated through quantum entangled photon pairs, achieving physical-level encryption of the satellite-to-ground communication link. This quantum key can be generated and verified during system startup, ensuring a quantum-secure foundation for subsequent communication links between monitoring devices and between devices and the cloud platform, providing anti-eavesdropping and anti-tampering protection for network-wide transmission. The system forms a local communication network among multiple monitoring devices within the same area through a self-organizing networking protocol, enabling multi-node spatial collaboration and data sharing. When a fire or abnormal event occurs in the same monitoring area, adjacent devices can automatically network and collaborate, conducting multi-directional synchronous observations of the same target area from different perspectives to improve the spatial coverage and accuracy of fire detection. Finally, the device sends an online heartbeat packet containing the device ID, status code, location information, and timestamp to the cloud platform, registering in the monitoring network and entering standby mode.

[0089] Step Two:

[0090] After the system enters a stable operation phase, multimodal acquisition is initiated according to the task scheduling cycle (i.e., the monitoring cycle). The integrated sensing module simultaneously acquires visible light images, multispectral images, and infrared images; a high-sensitivity microphone array continuously records the ambient sound spectrum; and temperature, humidity, air pressure, and ultraviolet sensors provide environmental baseline information. The active detection module initiates a 1550nm laser scan, with a two-dimensional microelectromechanical galvanometer system controlling the scanning path. The lidar scanning unit and photon counting detector receive echoes and form high-precision point cloud data. All data is uniformly timestamped via a precision clock synchronizer, and spatial coordinates and attitude information are provided by the positioning and attitude determination module to achieve spatiotemporal registration. Data acquired by each sensor is assigned a uniform timestamp by the precision clock synchronizer. To achieve millisecond-level time consistency, among which The synchronized global time. The synchronization correction amount can be measured from the deviation in the formula. Attitude matrix The calculation is based on acceleration, gyroscope, and magnetometer data output from the nine-axis inertial measurement unit, and is obtained through a quaternion attitude calculation algorithm. The update equation for the quaternion attitude is: Where q is the current attitude quaternion, The angular velocity vector, symbol This represents quaternion multiplication. After normalization, the quaternions are converted into rotation matrices. This enables attitude mapping from device coordinates to the geographic coordinate system. The system can also utilize the GNSS antenna baseline vector to constrain and correct the attitude matrix, thereby eliminating inertial drift errors.

[0091] The positioning and attitude determination module outputs the attitude matrix of the device in real time. With translation vector Make any sensor sampling point Coordinate transformation can be used Mapping to a geographic coordinate system enables unified spatiotemporal registration of multimodal observation data, providing high-precision foundational data for fire identification and 3D situation reconstruction. To reduce spatial registration errors between different sensors, the system performs a multi-sensor coordinate calibration process during deployment. An extrinsic parameter matrix is ​​established between the lidar coordinate system and the visible / infrared camera coordinate system using a marker point array. The calibration results are optimized by minimizing reprojection error. During operation, the system uses feature matching and joint optimization algorithms to fine-tune the extrinsic parameters between sensors online, ensuring the stability and consistency of coordinate transformation.

[0092] The edge computing module first performs noise suppression and illumination normalization preprocessing on the multispectral and thermal infrared images. For the original image I(x,y), a weighted filtering algorithm is applied: This is done to suppress sensor noise and preserve edge features. Subsequently, the average ambient brightness feedback from the light sensor is used... Perform linear illumination correction on the image: The coefficients α and β are adaptively adjusted by the edge computing module to achieve real-time compensation for day / night cycles and weather changes. The preprocessed data is then processed based on the attitude matrix. With translation vector Spatial registration is performed to unify the data into a geographic coordinate system, providing high-quality input for subsequent fire identification and target location.

[0093] The system upgrades the traditional camera imaging module to an intelligent visual perception unit with active semantic understanding capabilities in the visual perception stage. This unit, while acquiring images, constructs an active perception pipeline through a deep convolutional neural network and a semantic segmentation network, achieving a leap from "passive imaging" to "active understanding." First, the input image is denoted as... Where H and W are the height and width of the image, respectively, and 3 represents the number of channels. The image is processed by a feature extraction network. Mapped to a high-dimensional semantic feature tensor Where F is the encoded feature map, C is the number of channels, and H' and W' are the feature space resolutions. Subsequently, the semantic segmentation network... Semantic decoding of the feature tensor generates a primary semantic mask: , Where M(i,j) is the semantic category of pixel (i,j), and K is the set of target categories. The pixel belongs to category c k The probability of.

[0094] To achieve proactive semantic reasoning, the system further introduces a context attention module. Calculate the semantic correlation matrix between any pixels in the feature space. A ij Q represents the attention weights between feature positions i and j; i K j These are the query and key vectors extracted from the feature tensor F, respectively; d k The feature dimension normalization factor is used. The attention matrix is ​​used to enhance regional features related to the fire situation, and a weighted mapping F'=A∙V is calculated, where V is the set of value vectors obtained by the F mapping, and F' is the semantically enhanced feature mapping, containing higher fire significance and contextual relevance. Finally, the system generates a structured semantic feature stream based on F'. Where S is the set of structured features output; c k For the detected target category, p k =(x k ,y k ) represents the position of the target in the image coordinate system, s k =P(c k │p k The semantic confidence level is represented by a feature stream. This feature stream replaces the traditional pixel matrix output and is passed to the multimodal fusion and fire identification module through a unified interface for subsequent spatiotemporal correlation and fire level determination. Traditional visual methods mainly rely on color thresholds and brightness change detection, which cannot effectively distinguish between highly similar backgrounds such as smoke and clouds, firelight and sunlight reflection, with an average false detection rate of approximately 18% to 22%. However, by introducing an active semantic attention mechanism through intelligent visual perception, contextual dependencies can be established in the feature space, enabling the system to actively focus on key areas and understand the semantic connotation of the scene, keeping the target false detection rate below 5%. In dynamic forest environments, the stable recognition frame rate of traditional methods averages 12 FPS. Through edge computing optimization and a lightweight deep convolutional network architecture, combined with an active perception strategy, a real-time processing capability of 30 FPS can be achieved. Through the active intelligent perception mechanism, the system can identify the evolution trend in the early stages of disaster spread, achieving an early warning effect.

[0095] Step 3:

[0096] After completing the synchronous acquisition of multimodal data, the edge computing module calls the neural network accelerator to prioritize the execution of the fire identification sub-model. This model is based on a lightweight convolutional neural network structure and uses visible light images... v (x,y), Infrared thermal imaging I t (x,y) and lidar reflectivity map R l(x,y) are the inputs. Features such as smoke texture, temperature anomaly and reflectivity change are extracted respectively. The fire response map S(x,y) is generated by a multimodal feature fusion network.

[0097] The core of the model is a three-channel feature fusion structure: First, the visible light branch extracts the local gradient direction histogram and color deviation matrix to identify smoke texture and brightness fluctuation features; second, the infrared branch calculates the local radiance increment. Third, laser branch detection of reflectivity change rate; This is used to distinguish the differences between thermal smoke and solid barrier reflections. The fusion module performs weighted concatenation of the three-channel feature tensors: F(x,y)=ω v f v (I v )+ω t f t (I t )+ω l f l (R l ), where ω v ω t ω l The channel weights are determined by model training. After normalization by the Softmax layer, the fire confidence score is output: P. f (x,y)=σ(F(x,y)), the system calculates the maximum response value P over the entire monitoring field of view. f =max x,y (P f (x,y)) represents the fire situation confidence level at the current moment.

[0098] The edge computing module monitors the recognition results of each modality in real time. Specifically, the visible light recognition sub-model outputs the fire confidence score P. fvis , representing the probability of the presence of flame or smoke features in a visible light image; the infrared recognition sub-model outputs the fire confidence score P. fthe represents the probability response of high-temperature anomaly regions in thermal imaging data. When P... fvis Greater than the visible light threshold θ1, and P fthe If the fire intensity exceeds the thermal imaging threshold θ2, the system determines the status as "fire confirmed" (i.e., fire status confirmation is initiated), triggering the fire location process. The two threshold parameters θ1 and θ2 are determined through experimental calibration and are used to balance the false negative and false positive rates under different forest lighting and climate conditions.

[0099] During the confirmation process, the device automatically adjusts its components for refined observation. The gimbal drive module aligns the camera and infrared sensor with the center of the high-confidence area, the lidar activates fine-scan mode to acquire local high-density point clouds, and the precision clock synchronizer locks the timestamp to ensure spatiotemporal consistency of the full-modal data. The satellite-based augmentation system automatically connects to the GNSS differential correction signal to improve the device's positioning accuracy. The edge computing module calculates the time stability ∆P of the fire confidence level across multiple consecutive frames. f If the fluctuation of / ∆t is less than the threshold ε for three consecutive sampling periods, the fire event is considered confirmed (i.e., a fire is confirmed to exist). After the fire is confirmed, the device starts the spatial positioning algorithm to determine the location of the fire point.

[0100] Meanwhile, the sound module can use voiceprint matching to trigger a "suspicious event" when it detects abnormal frequencies such as chainsaw sounds or explosions. The device automatically rotates the visible light camera to zoom in and confirm the direction of the sound source, and dynamically adjusts the monitoring priority based on the identification results. In the event of a fire, fire monitoring has a higher priority than suspicious event monitoring.

[0101] Step Four:

[0102] Upon event triggering, the edge computing module executes a tiered transmission strategy based on task priority. For high-priority events (such as fires and deflagrations), the system activates an emergency communication mode with an integrated collaborative observation mechanism. In this mode, the data packets generated by the device automatically encapsulate a "collaborative observation request" instruction frame. A secure channel is established primarily using a microsatellite link with quantum key distribution capabilities. This link implements the handshake process based on the BB84 quantum key distribution protocol. The satellite and ground parties first synchronize their ground and polarization states via a classical channel, then use entangled photon pairs for quantum state transmission and measurement comparison to generate a consistent and uncatchable random key. The quantum key update cycle is set to every 30 seconds, updated in real-time by photon measurement results. This allows for dynamic refreshing of the shared key between the communicating parties, ensuring that transmitted data such as fire location coordinates, confidence levels, visible light, and infrared images cannot be intercepted, copied, or replayed throughout the entire link, significantly improving the security, real-time performance, and integrity of fire data transmission. Upon receiving the command frame, the cloud platform immediately calculates the fire location and the topology of the surrounding monitoring network, and issues "coordinated response commands" to multiple nearby devices within the incident area via the command channel. The devices receiving the commands, based on a unified spatiotemporal reference, autonomously adjust their pan-tilt attitude and sensor modes, simultaneously scanning the fire site from different angles to form multi-view data complementarity, effectively eliminating monitoring blind spots. Addressing the complex terrain and multi-path obstruction environment of the forest area, the quantum microsatellite link employs an adaptive antenna control mechanism. Through real-time ranging and attitude feedback, it adjusts the elevation and azimuth angles of the satellite and ground antennas to compensate for link attenuation caused by terrain undulations, canopy reflection, and aerosol scattering, ensuring a link stability maintenance rate of over 95%.

[0103] Compared to traditional satellite communication links, the quantum microsatellite communication introduced in this embodiment has significant performance improvements and security advantages in the complex environment of forest areas:

[0104] (1) In terms of data security, quantum links are based on the physical non-cloning property of photon quantum state transmission to achieve physical layer anti-eavesdropping, and their data leakage risk is close to zero, while conventional satellite links still rely on software encryption, which has the potential for algorithm-level cracking.

[0105] (2) In terms of real-time performance, quantum links can achieve low-latency transmission of high-priority data (such as fire point coordinates, video frames, etc.) with a latency of ≤2 seconds, while conventional satellites have an average latency of 5~8 seconds;

[0106] (3) In terms of anti-interference capability, the quantum link can maintain a link rate of over 95% under conditions of heavy rain, fog, or terrain obstruction, while the traditional satellite link usually drops to around 70%. The performance comparison between the two fully demonstrates the reliability of quantum satellite link technology in emergency data transmission in forests.

[0107] In terms of transmission strategy, when the main channel is stable, the system will simultaneously utilize cellular communication or industrial Ethernet channels to upload complete data packets, enabling rapid feedback to the disaster site. When the ground signal is unstable, the system automatically switches to a low-power wide-area network relay transmission mechanism, indirectly uploading data through collaborative self-organizing network forwarding between nearby devices. Under extreme conditions, the system directly calls upon the quantum microsatellite secure channel for emergency communication, sending critical fire alarm information and geographical coordinates to the cloud command center. For low-priority events (such as ecological monitoring, temperature and humidity sampling, etc.), the collected data is stored in an embedded multimedia storage unit and uploaded in batches through a second channel during non-emergency periods to achieve a dynamic balance between energy consumption and bandwidth.

[0108] Step 5:

[0109] After receiving the fire information uploaded by the device, the cloud platform activates the holographic data fusion module. This module's fusion engine employs a microservice architecture to process multi-source heterogeneous data in parallel. It overlays multi-modal data collected by ground monitoring devices with data from satellite remote sensing, meteorology, wind fields, and topography at multiple scales in a spatiotemporal manner, constructing a three-dimensional dynamic change field of the monitored area. Specifically, it fuses cloud motion vector field data from the MISR (Multi-angle Imaging Spectro-Radiometer) satellite to extract upper-level wind direction and speed distribution characteristics, using these as dynamic input parameters for the fire prediction model. This optimizes the simulation accuracy of flame propagation direction and spread rate, enhancing the realism of the dynamic fire simulation. Simultaneously, the platform's "cooperative scheduler" generates specific cooperative observation commands based on the optimal coverage algorithm and sends them down to relevant devices. For example, commands may instruct leading devices to broadcast high-definition video streams, commands may instruct flanking devices to monitor the fire line spread trend, and commands may instruct remote devices to assess the overall plume extent. All commands are synchronized through a unified spatiotemporal reference, ensuring seamless fusion of multi-source data and constructing a panoramic dynamic view of the fire scene with no blind spots and a high refresh rate.

[0110] Data fusion employs a weighted Bayesian update model: S t Let D be the environmental state field at the current moment. t The data is from sensor observations. This model achieves dynamic fusion and confidence level correction of multi-source information. Based on the fusion results, the system generates a dynamic holographic monitoring map, displaying in real time the spatial distribution of fire points, the direction of fire spread, wind speed and direction vectors, and changes in combustible material density. Simultaneously, the cloud utilizes terrain analysis and path optimization algorithms based on a digital elevation model (DEM) to automatically calculate the optimal evacuation and rescue routes.

[0111] The path cost function is defined as: , where d k Let s be the distance of the k-th segment of the path. k For the cost of slope, v k ω is the vegetation resistance coefficient. k λ1, λ2, and λ3 are wind direction interference factors, and weighting coefficients. The system solves for the minimum cost path on the spatial grid using the A* or Dijkstra algorithm and updates the route planning results in real time according to changes in fire intensity.

[0112] After the fire subsided, the device automatically switched to post-disaster monitoring mode. Cloud-based dispatch of surrounding monitoring devices enabled continuous observation of key areas, collecting residual thermal signals, smoke and dust diffusion range, and vegetation spectral recovery information. The recovery index of the damaged area was calculated based on a time-series analysis model: R. t =(V t -Vmin ) / (V max -V min ), where V t V represents the current vegetation index. min With V max These represent the extreme values ​​of vegetation indices before the disaster and in a healthy state, respectively. Through analysis of R... t Through temporal trend analysis, the system can calculate ecological regeneration potential and generate fire-affected area distribution maps and restoration assessment reports. Finally, the post-disaster monitoring results and fire process data are archived together and fed back to forestry management and emergency response departments, completing comprehensive forest fire control throughout its entire lifecycle.

[0113] In the above embodiments, a multispectral visual agent with high-precision clock synchronization (delay ≤ 100 milliseconds) and proactive semantic understanding is used to achieve collaborative identification and adaptive judgment of fire characteristics, reducing the false negative rate to below 5%. Simultaneously, a quantum key distribution microsatellite communication link is employed to achieve low-latency transmission of fire data while ensuring physical-level anti-eavesdropping capabilities. Through intelligent semantic analysis and holographic spatiotemporal fusion of multi-source data, a holographic three-dimensional digital spatiotemporal change model of the monitoring area is constructed, enabling early fire identification, precise location, and dynamic ecological collaborative monitoring. This establishes a multi-scale holographic perception system from macroscopic situation to local details, improving the accuracy of forest fire monitoring.

[0114] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0115] Based on the same inventive concept, this application also provides a forest fire monitoring device for implementing the forest fire monitoring method described above. The solution provided by this device is similar to the solution described in the above method; therefore, the specific limitations in one or more embodiments of the forest fire monitoring device provided below can be found in the limitations of the forest fire monitoring method described above, and will not be repeated here.

[0116] In one exemplary embodiment, such as Figure 6 As shown, a forest fire monitoring device 600 is provided, including: a monitoring data acquisition module 602, a fire status determination module 604, a fire location determination module 606, and a monitoring angle adjustment module 608, wherein:

[0117] The monitoring data acquisition module 602 is used to acquire fire monitoring data of the target area collected by the monitoring device; the fire monitoring data includes environmental image data and corresponding device position data.

[0118] The fire status determination module 604 is used to determine the first fire confidence level of the target area based on environmental image data, and to determine the fire status based on the first fire confidence level.

[0119] The fire location determination module 606 is used to determine the location of the fire point based on the device pose data when the fire status is that a fire exists.

[0120] The monitoring angle adjustment module 608 is used to adjust the monitoring angle of the monitoring device to achieve all-round monitoring of the fire situation at the fire point location.

[0121] In some embodiments, the environmental image data includes visible light image data, infrared thermal imaging data, and radar reflectivity image data; the fire status determination module 604 is further configured to extract features from the visible light image data to obtain smoke texture features; extract features from the infrared thermal imaging data to obtain temperature features of the target area; extract features from the radar reflectivity image data to obtain smoke reflection features; fuse the smoke texture features, temperature features, and smoke reflection features to obtain multimodal fusion features; and perform fire prediction using the multimodal fusion features to obtain a first fire confidence level for the target area.

[0122] In some embodiments, the fire status determination module 604 is further configured to calculate the confidence difference between the first fire confidence levels corresponding to two adjacent monitoring cycles; if the degree of change of a preset number of consecutive confidence difference values ​​is greater than the degree of change threshold, the fire status is determined to be that a fire exists.

[0123] In some embodiments, the fire status determination module 604 is further configured to: extract features from visible light image data in environmental image data to obtain smoke texture features, and predict the fire status based on the smoke texture features to obtain a second fire confidence level before determining the fire status based on the first fire confidence level; extract features from infrared thermal imaging data in environmental image data to obtain temperature features of the target area, and predict the fire status based on the temperature features to obtain a third fire confidence level; and determine the fire status based on the first fire confidence level if the second fire confidence level is greater than the first confidence threshold and the third fire confidence level is greater than the second confidence threshold.

[0124] In some embodiments, the fire location determination module 606 is further configured to acquire satellite positioning coordinates and distance obtained by laser ranging; and determine the ignition location where a fire exists based on the satellite positioning coordinates, device pose data and distance using a spatial forward intersection algorithm.

[0125] In some embodiments, the above-mentioned device further includes a post-disaster monitoring module, which is used to monitor the residual heat information, smoke diffusion range and vegetation spectral recovery information of the fire point location through a post-disaster monitoring mode after the fire at the identified fire point location has been extinguished; determine the current vegetation index based on the residual heat information, smoke diffusion range and vegetation spectral recovery information; and determine the vegetation recovery index of the fire point location based on the pre-disaster vegetation index and the current vegetation index at the fire point location before the fire occurred.

[0126] The various modules in the aforementioned forest fire monitoring device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in the processor of a computer device in hardware form or independent of it, or stored in the memory of the computer device in software form, so that the processor can call and execute the corresponding operations of each module.

[0127] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 7 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The database stores data related to forest fire monitoring methods. The I / O interfaces are used for information exchange between the processor and external devices. The communication interface is used for communication with external terminals via a network connection. When the computer program is executed by the processor, it implements a forest fire monitoring method.

[0128] Those skilled in the art will understand that Figure 7 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0129] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above-described method embodiments.

[0130] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps in the above method embodiments.

[0131] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0132] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0133] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0134] The technical features of the above embodiments can be combined in any way. For the sake of brevity, 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 application.

[0135] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for monitoring forest fires, characterized in that, The method includes: Acquire fire monitoring data of the target area collected by the monitoring device; the fire monitoring data includes environmental image data and corresponding device position data; Based on the environmental image data, a first fire confidence level is determined for the target area, and the fire status is determined based on the first fire confidence level. When the fire situation is described as a fire, the location of the fire point is determined based on the device's pose data. Adjust the monitoring angle of the monitoring device to achieve all-round monitoring of the fire situation at the fire point location.

2. The method according to claim 1, characterized in that, The environmental image data includes visible light image data, infrared thermal imaging data, and radar reflectivity image data; determining the first fire confidence level of the target area based on the environmental image data includes: Feature extraction is performed on the visible light image data to obtain smoke texture features; feature extraction is performed on the infrared thermal imaging data to obtain temperature features of the target area; feature extraction is performed on the radar reflectivity image data to obtain smoke reflection features; The smoke texture features, temperature features, and smoke reflection features are fused to obtain multimodal fused features; Fire prediction is performed on the multimodal fusion features to obtain the first fire confidence level of the target area.

3. The method according to claim 1, characterized in that, Determining the fire status based on the first fire confidence level includes: Calculate the confidence difference between the confidence levels of the first fire situation for two adjacent monitoring periods; If the degree of change of a preset number of confidence difference values ​​is greater than the degree of change threshold, the fire status is determined to be present.

4. The method according to claim 1, characterized in that, Before determining the fire status based on the first fire confidence level, the method further includes: Feature extraction is performed on the visible light image data in the environmental image data to obtain smoke texture features, and fire prediction is performed on the smoke texture features to obtain a second fire confidence level; Feature extraction is performed on the infrared thermal imaging data in the environmental image data to obtain the temperature characteristics of the target area, and fire prediction is performed on the temperature characteristics to obtain a third fire confidence level. If the second fire confidence level is greater than the first confidence threshold and the third fire confidence level is greater than the second confidence threshold, then the step of determining the fire status based on the first fire confidence level is executed.

5. The method according to claim 1, characterized in that, Determining the location of a fire point based on the device pose data includes: Obtain satellite positioning coordinates and distances obtained through laser ranging; Based on the satellite positioning coordinates, the device pose data, and the distance, the ignition location where a fire exists is determined by a spatial forward intersection algorithm.

6. The method according to any one of claims 1 to 5, characterized in that, The method further includes: After the fire at the identified fire location is extinguished, the remaining heat information, smoke diffusion range, and vegetation spectral recovery information at the fire location are monitored through the post-disaster monitoring mode. The current vegetation index is determined based on the residual heat information, the smoke and dust diffusion range, and the vegetation spectral recovery information. Based on the pre-fire vegetation index and the current vegetation index at the fire location before the fire occurred, the vegetation recovery index at the fire location is determined.

7. A forest fire monitoring device, characterized in that, The device includes: The monitoring data acquisition module is used to acquire fire monitoring data of the target area collected by the monitoring device; the fire monitoring data includes environmental image data and corresponding device position data; The fire status determination module is used to determine a first fire confidence level of the target area based on the environmental image data, and to determine the fire status based on the first fire confidence level. The fire location determination module is used to determine the location of the fire point based on the device pose data when the fire situation is that a fire exists. The monitoring angle adjustment module is used to adjust the monitoring angle of the monitoring device to achieve all-round monitoring of the fire situation at the fire point location.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images