Sky-ground collaborative based natural reserve adaptive monitoring system and method

Abstract

The application discloses a kind of natural reserve adaptive monitoring system and method based on sky-ground cooperation, the system includes: service end, for obtaining current satellite remote sensing data that satellite is collected to be monitored natural reserve;According to current satellite remote sensing data, determine whether there is abnormal area in the natural reserve to be monitored;If yes, determine the target unmanned aerial vehicle closest to the abnormal area, target ground mobile device;Generate coordination instruction and send to target unmanned aerial vehicle, target ground mobile device;Target unmanned aerial vehicle, target ground mobile device are used to go to abnormal area based on coordination instruction, to obtain the detailed information of close range of abnormal area, send to service end;Service end is also used to determine whether to give early warning to abnormal area according to the detailed information of close range;If yes, generate the early warning information of abnormal area and send to client end.Therefore, by using the application, the data processing pressure of service end can be reduced, the computing resources are saved, and the monitoring efficiency is improved.

Description

Technical Field

[0001] This application relates to the field of ecosystem protection and management technology, and in particular to an adaptive monitoring system and method for nature reserves based on sky-ground coordination. Background Technology

[0002] In the monitoring work of nature reserves, protecting biodiversity, monitoring species activity, and preventing illegal activities (such as illegal land reclamation) are crucial tasks. Nature reserves typically cover vast geographical areas, including various ecosystems such as forests, wetlands, and grasslands. Monitoring these areas requires the comprehensive use of multiple technical means to achieve comprehensive monitoring and protection of the protected objects.

[0003] In related technologies, the monitoring of nature reserves mainly relies on three monitoring methods: space (satellites), air (drones), and ground (ground-based mobile devices). Currently, the monitoring of nature reserves is carried out by comprehensively analyzing the relevant data from these three monitoring methods.

[0004] However, existing technologies require satellites, drones, and ground mobile devices to operate around the clock. Data centers receive large amounts of data, and the large volume of multi-source data makes data analysis difficult and wastes computing resources. Secondly, there is a lack of effective collaborative mechanisms between space, air, and ground devices. For example, although satellite remote sensing can detect anomalies over a wide area, it cannot determine detailed information. Although drones and ground devices can determine detailed information, they cannot detect anomalies in a timely manner, thus reducing monitoring efficiency. Summary of the Invention

[0005] This application provides an adaptive monitoring system for nature reserves based on a sky-ground collaborative approach. To provide a basic understanding of some aspects of the disclosed embodiments, a brief summary is given below. This summary is not intended as a general commentary, nor is it intended to identify key / important components or describe the scope of protection of these embodiments. Its sole purpose is to present some concepts in a simple form as a prelude to the detailed description that follows.

[0006] In a first aspect, embodiments of this application provide an adaptive monitoring system for nature reserves based on sky-ground coordination, the system comprising:

[0007] The system consists of a server, air-to-ground collaborative devices, and a client; the server communicates with both the air-to-ground collaborative devices and the client; the air-to-ground collaborative devices include satellites, drones, and ground-based mobile devices.

[0008] The server is used to acquire current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset period; based on the current satellite remote sensing data, it determines whether there are abnormal areas in the nature reserve to be monitored; if so, it identifies the target UAV and target ground mobile equipment closest to the abnormal area; and generates collaborative commands to send to the target UAV and target ground mobile equipment.

[0009] The target drone and target ground mobile device are used to travel to the abnormal area based on coordinated commands, obtain close-range detailed information about the abnormal area, and send it to the server.

[0010] The server-side system is also used to determine whether to issue an alert for abnormal areas based on detailed proximity information; if so, it generates an alert for the abnormal area and sends it to the client.

[0011] Optionally, the server can also be used to continue executing the step of acquiring the current satellite remote sensing data collected by the satellite for the nature reserve to be monitored according to a preset cycle, in the case that there are no abnormal areas in the nature reserve to be monitored or no warning is issued for abnormal areas.

[0012] Optionally, identify the target drone or target ground mobile device closest to the anomalous area, including:

[0013] Generate a location request command and send it to all drones and ground mobile devices deployed within the area of ​​the nature reserve to be monitored;

[0014] Receive the latitude and longitude coordinates obtained by each drone and each ground mobile device through its own positioning system, and obtain the location information of each drone and each ground mobile device;

[0015] Determine the latitude and longitude of the center of the anomaly area;

[0016] Based on location information and the latitude and longitude of the center location, determine the target UAV and target ground mobile equipment that are closest to the anomaly area.

[0017] Optionally, determine the latitude and longitude of the center location of the anomalous area, including:

[0018] From the satellite images contained in the current satellite remote sensing data, crop out the current area image of the abnormal region;

[0019] Iterate through all pixels in the current region of the image, with each pixel including its row and column coordinates;

[0020] Based on the affine transformation coefficients of the pre-labeled pixel coordinates and geographic coordinates, each pixel point, including its row coordinates and column coordinates, is subjected to an affine transformation to obtain the abnormal geographic coordinates corresponding to each pixel point.

[0021] Based on the abnormal geographic coordinates corresponding to each pixel, the centroid of the polygon is solved, and the latitude and longitude of the center location of the abnormal region are calculated; where...

[0022] The affine transformation coefficients are obtained by solving using known pixel coordinates and geographic coordinates combined with the least squares method. The expression for the affine transformation is:

[0023]

[0024] in, These are abnormal geographic coordinates. These are the pixel coordinates of each pixel, including row and column coordinates. These are the affine transformation coefficients.

[0025] Optionally, the location information is the target's latitude and longitude;

[0026] Based on the location information and the latitude and longitude of the center location, determine the target UAV and target ground mobile equipment closest to the anomaly area, including:

[0027] Based on the target latitude and longitude and center location latitude and longitude of each drone, and combined with the preset distance calculation formula, the distance between each drone and the abnormal area is calculated.

[0028] Based on the target latitude and longitude and the center location latitude and longitude of each ground mobile device, and combined with the preset distance calculation formula, the distance between each ground mobile device and the abnormal area is calculated.

[0029] The drone with the smallest distance is selected as the closest target drone to the anomaly area, and the ground mobile device with the smallest distance is selected as the closest target ground mobile device to the anomaly area; the preset distance calculation formula is:

[0030]

[0031] in, It is the distance between two points. It is the Earth's radius. It is an arcsine function. It is the difference in latitude between two points. It is the difference in longitude between the two points. and These are the latitudes of the two points. It is a cosine function. It is a sine function.

[0032] Optionally, based on current satellite remote sensing data, determine whether there are any abnormal areas in the nature reserve to be monitored, including:

[0033] The nature reserves to be monitored are divided into a predetermined number of initial areas;

[0034] Radiometric correction, geometric correction and denoising processing are performed on the current satellite remote sensing data to obtain the satellite remote sensing data to be analyzed.

[0035] Obtain target satellite remote sensing data for each initial area from the satellite remote sensing data to be analyzed;

[0036] From the target satellite remote sensing data of each initial area, determine the land cover classification characteristics and spatiotemporal characteristics of each preset land cover type that needs to be focused on, as the basic characteristics of each initial area; each preset land cover type includes at least forest, grassland, water, and wildlife habitat;

[0037] Based on the basic characteristics of each initial area, it is determined whether there are any abnormal areas in the nature reserves to be monitored.

[0038] Optionally, the classification features include at least a vegetation index reflecting the growth status and coverage of vegetation, a water index reflecting the presence and extent of water bodies, and a texture feature index used to describe the surface structure of grassland; the spatiotemporal features include at least a first result of whether it is in a normal growth cycle, and a second result of whether the spatial distance and spatial layout with other land cover types meet the requirements.

[0039] Based on the basic characteristics of each initial area, determine whether there are any abnormal areas in the nature reserves to be monitored, including:

[0040] The vegetation index, water index, and texture feature index of each initial region are compared with the preset normal state threshold range to obtain the comparison results of each initial region.

[0041] If the comparison results for each initial region do not exceed the preset normal state threshold range and both the first and second results are yes, then it is determined that there are no abnormal areas in the nature reserve to be monitored; or...

[0042] If the comparison result of each initial region indicates that it exceeds the preset normal state threshold range, or if both the first and second results are negative, it is determined that there is an abnormal region in the nature reserve to be monitored.

[0043] Optional close-up details include high-definition video, high-definition images, and sensor data;

[0044] Based on detailed close-range information, determine whether to issue an alert for abnormal areas, including:

[0045] High-definition video and high-definition images are input into a pre-trained hazard identification model, which outputs hazard identification results corresponding to close-range detailed information. The pre-trained hazard identification model is obtained by machine learning through convolutional neural networks and model training samples.

[0046] The temperature, humidity, and gas concentration data in the sensor are compared with preset safety thresholds to determine whether there are any abnormal situations such as fire or environmental pollution in the abnormal area.

[0047] If the hazard identification results indicate that a hazardous condition exists in an abnormal area, then a warning should be issued for the abnormal area; or...

[0048] If the hazard identification result indicates that there is no hazard in the abnormal area, it is determined not to issue a warning for the abnormal area.

[0049] Optionally, a pre-trained hazardous state recognition model is generated by following these steps:

[0050] Collect a preset number of historical high-definition videos and images; the historical high-definition videos and images contain both normal and dangerous states of nature reserves.

[0051] For each historical high-definition video and each historical high-definition image, sample labels for normal and dangerous states are marked to obtain model training samples; normal states include normal natural landscapes and wildlife activities; dangerous states include illegal intrusion, logging, fire, and environmental pollution.

[0052] A hazardous state recognition model is created using a convolutional neural network.

[0053] Input the training samples of the model into the dangerous state recognition model to perform machine learning, and output the model loss value;

[0054] Generate a pre-trained dangerous state recognition model when the model loss value reaches its minimum.

[0055] Alternatively, if the model loss value has not reached its minimum, the model loss value is backpropagated to update the model parameters of the hazard identification model, and the step of inputting the model training samples into the hazard identification model for machine learning continues until the model loss value reaches its minimum.

[0056] Secondly, an adaptive monitoring method for nature reserves based on sky-ground coordination is proposed, the method comprising:

[0057] The server acquires current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset period; based on the current satellite remote sensing data, it determines whether there are abnormal areas in the nature reserve to be monitored; if so, it identifies the target UAV and target ground mobile equipment closest to the abnormal area; and generates collaborative commands to send to the target UAV and target ground mobile equipment.

[0058] The target drone and target ground mobile device travel to the abnormal area based on coordinated commands to obtain close-range detailed information about the abnormal area and send it to the server.

[0059] The server determines whether to issue an alert for abnormal areas based on nearby detailed information; if so, it generates an alert for the abnormal area and sends it to the client.

[0060] In this embodiment, on the one hand, the server acquires satellite remote sensing data at a preset cycle to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0061] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0062] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0063] Figure 1 This is a schematic diagram of the system structure of an adaptive monitoring system for nature reserves based on sky-ground coordination, provided in an embodiment of this application.

[0064] Figure 2 This is a satellite image of a nature reserve to be monitored, provided in an embodiment of this application;

[0065] Figure 3This is a schematic diagram of a nature reserve to be monitored, divided into grid areas of equal size, according to an embodiment of this application.

[0066] Figure 4 This is a schematic flowchart illustrating a server-side abnormal region determination process provided in an embodiment of this application;

[0067] Figure 5 This is a schematic diagram of relevant information of a sky-ground collaborative device displayed in the background, provided in an embodiment of this application;

[0068] Figure 6 This is a schematic diagram illustrating the process of locating a target drone or ground mobile device closest to an abnormal area, as provided in an embodiment of this application.

[0069] Figure 7 This is a data schematic diagram of environmental sensor data provided in an embodiment of this application;

[0070] Figure 8 This is a schematic diagram of the method flow of an adaptive monitoring method for nature reserves based on sky-ground coordination, provided in an embodiment of this application.

[0071] Figure 9 This is a schematic diagram of an early warning center displayed on an administrator's client according to an embodiment of this application;

[0072] Figure 10 This is a flowchart illustrating a model training method for a hazardous state identification model provided in an embodiment of this application.

[0073] Figure 11 This is a schematic diagram of the structure of an adaptive monitoring device for nature reserves based on sky-ground coordination, provided in an embodiment of this application. Detailed Implementation

[0074] The following description and accompanying drawings fully illustrate specific embodiments of this application to enable those skilled in the art to practice them.

[0075] It should be understood that the described embodiments are merely some, not all, of the embodiments in this application. All other embodiments obtained by those skilled in the art based on the embodiments in this application without inventive effort are within the scope of protection of this application.

[0076] In the following description, when referring to the accompanying drawings, the same numbers in different drawings denote the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of systems and methods consistent with some aspects of this application as detailed in the appended claims.

[0077] In the description of this application, it should be understood that the terms "first," "second," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances. Furthermore, in the description of this application, unless otherwise stated, "multiple" refers to two or more. "And / or" describes the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, or B existing alone. The character " / " generally indicates that the preceding and following related objects are in an "or" relationship.

[0078] Currently, the monitoring of nature reserves mainly relies on three monitoring methods: satellite, drone, and ground-based mobile equipment. The monitoring of nature reserves is achieved by comprehensively analyzing the data obtained from these three methods.

[0079] The inventors realized that existing technologies require satellites, drones, and ground mobile devices to operate around the clock. Data centers receive large amounts of data, and the large volume of multi-source data leads to significant difficulties in data analysis and serious waste of computing resources. Secondly, there is a lack of effective collaborative mechanisms between space, air, and ground devices. For example, although satellite remote sensing can detect anomalies over a wide area, it cannot determine detailed information. Although drones and ground devices can determine detailed information, they cannot detect anomalies in a timely manner, thus reducing monitoring efficiency.

[0080] In this embodiment, on the one hand, the server acquires satellite remote sensing data according to a preset period to determine whether there are abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed to obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative linkage mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency. A detailed description using exemplary embodiments follows.

[0081] Please see Figure 1 , Figure 1This is a schematic diagram of the system structure of an adaptive monitoring system for nature reserves based on sky-ground coordination, provided in an embodiment of this application. The system includes: a server, sky-ground coordination equipment, and a client; the server is communicatively connected to the sky-ground coordination equipment and the client; the sky-ground coordination equipment includes satellites, drones, and ground mobile devices.

[0082] The server is the core control unit of the system, responsible for data processing, command generation, and communication coordination with various devices. The space-air-ground collaborative equipment is a multi-layered monitoring system comprised of satellites, drones, and ground-based mobile devices. Through collaborative work, it achieves comprehensive monitoring of nature reserves. Satellites, located in space, provide large-scale, macroscopic remote sensing data for initial detection of anomalies. Drones, also in the air, can quickly travel to the site after anomalies are detected to acquire high-resolution images and videos, as well as other detailed information. Ground-based mobile devices, equipped with sensors and data acquisition capabilities, can penetrate deep into anomaly areas to collect precise data such as temperature, humidity, and gas concentrations. Clients are end-users or management organizations that receive early warning information from the server, allowing them to promptly understand any anomalies in nature reserves.

[0083] In some embodiments of this application, the server is configured to acquire current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset period; determine whether there is an abnormal area in the nature reserve to be monitored based on the current satellite remote sensing data; if so, determine the target UAV and target ground mobile device closest to the abnormal area; generate a collaborative command and send it to the target UAV and target ground mobile device; the target UAV and target ground mobile device are configured to travel to the abnormal area based on the collaborative command to obtain close-range detailed information about the abnormal area and send it to the server; the server is also configured to determine whether to issue an early warning for the abnormal area based on the close-range detailed information; if so, generate an early warning information for the abnormal area and send it to the client.

[0084] Satellite remote sensing data comprises images and data collected by satellites regarding the nature reserve to be monitored, including spectral information, topographic information, and satellite imagery. Anomaly areas are regions within the nature reserve identified through analysis of satellite remote sensing data as potentially exhibiting anomalies, such as fires, illegal intrusions, or environmental pollution. Coordination commands are instructions generated by the server and sent to the target drones and ground-based mobile devices, directing them to monitor the anomaly areas. Close-range detailed information consists of detailed data acquired by the target drones and ground-based mobile devices within the anomaly areas, including high-definition video, images, and sensor data. Early warning information is alert information generated and sent to the client when the server determines that an anomaly area presents a dangerous condition.

[0085] In one possible implementation, the server receives remote sensing data of the nature reserve to be monitored from a satellite via a communication link at a preset period (e.g., once a day). The remote sensing data includes multispectral images, thermal infrared images, etc. The server preprocesses the acquired satellite remote sensing data, such as radiometric correction, geometric correction, and denoising. The nature reserve is divided into multiple initial areas, and features such as vegetation indices and water quality indices are extracted for each area. Based on a preset normal state threshold range, each area is judged to be anomaly-free. If the feature value of a certain area exceeds the threshold range, the area is considered to be anomaly-free.

[0086] The server generates a location request command and sends it to all drones and ground mobile devices deployed within the nature reserve. Each device obtains its current latitude and longitude coordinates using its own positioning system (such as GPS) and sends them back to the server. The server calculates the distance of each device to the center of the anomaly area and selects the closest drone or ground mobile device as the target device.

[0087] The server generates coordination instructions, including path planning and task requirements for the target device to reach the abnormal area. These instructions are then sent to the target drone and target ground mobile device via a communication link. Upon receiving the instructions, the drone and ground mobile device proceed to the abnormal area. The drone captures high-definition video and images, while the ground mobile device collects sensor data (such as temperature, humidity, and gas concentration). The acquired close-range detailed information is then sent back to the server.

[0088] The server inputs high-definition video and images into a pre-trained hazard identification model to determine whether a hazardous condition exists in the abnormal area. It compares sensor data with preset safety thresholds to determine if there are any anomalies such as fires or environmental pollution. If the hazard identification result indicates a hazard, or if the sensor data exceeds the safety threshold, an early warning message is generated. The server sends the early warning message to the client via a communication link, notifying relevant personnel or organizations to take appropriate measures.

[0089] In this embodiment, the server acquires satellite remote sensing data at a preset period to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This approach eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission. Consequently, it significantly reduces the amount of data received by the data center. This reduction in data volume lowers the difficulty of data analysis, allows for efficient use of computing resources, and avoids wasting computing resources.

[0090] In this embodiment, the satellite first conducts wide-area monitoring, identifies anomaly areas, and then determines the closest drones and ground-based mobile devices. It then generates coordinated commands instructs these devices to proceed to the anomaly area to obtain detailed close-range information. This coordinated mechanism fully leverages the strengths of each device, combining the satellite's macroscopic monitoring capabilities with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that anomalies are detected promptly but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0091] In some embodiments of this application, the server is further configured to continue executing the step of acquiring current satellite remote sensing data of the nature reserve to be monitored at a preset period, even when there are no abnormal areas in the nature reserve to be monitored or no warning is issued for abnormal areas. The nature reserve to be monitored is, for example... Figure 2 As shown.

[0092] This process enables continuous adaptive monitoring of nature reserves, thereby improving the intelligence level of monitoring.

[0093] In some embodiments of this application, the specific process of determining whether there are abnormal areas in a nature reserve to be monitored based on current satellite remote sensing data includes: dividing the nature reserve to be monitored into a predetermined number of initial areas; performing radiometric correction, geometric correction, and denoising processing on the current satellite remote sensing data to obtain satellite remote sensing data to be analyzed; obtaining target satellite remote sensing data for each initial area from the satellite remote sensing data to be analyzed; determining the land cover classification characteristics and spatiotemporal characteristics of each predetermined land cover type to be of interest from the target satellite remote sensing data of each initial area, as the basic characteristics of each initial area; each predetermined land cover type includes at least forest, grassland, water area, and wildlife habitat; and determining whether there are abnormal areas in the nature reserve to be monitored based on the basic characteristics of each initial area.

[0094] Radiometric correction processes satellite remote sensing data to eliminate radiometric errors caused by sensor characteristics, atmospheric conditions, and other factors, making the image's radiometric values ​​closer to the true surface reflectance. Geometric correction processes satellite remote sensing images to eliminate geometric distortions caused by sensor attitude, terrain undulations, and other factors, ensuring the image's geometric position matches its actual geographic coordinates. Denoising processing processes remote sensing data to remove or reduce noise (such as random errors and stripe noise) to improve data clarity and usability.

[0095] In one possible implementation, Figure 2 The nature reserves to be monitored are shown to be divided into grid areas of equal size, as shown in the following example. Figure 3As shown. Satellite remote sensing data, including multispectral and thermal infrared images, is acquired. Radiometric correction is performed on the images to eliminate the influence of sensor readings and atmospheric conditions. Geometric correction is performed to ensure accurate geometric positioning of the images. Noise reduction processing is performed to remove random noise from the images. Data from each grid region is analyzed to extract the following features: a vegetation index to assess vegetation growth and cover; a water index to identify and assess the presence and extent of water bodies; and texture features to describe the surface structure of the grassland. The system assesses whether vegetation is in a normal growth cycle and whether the spatial distribution of ground features conforms to normal spatiotemporal characteristics. If the vegetation index of a grid region is below the normal range and the vegetation growth cycle does not conform to normal conditions, that region is identified as an anomalous region. If the water index of a grid region is abnormally high and the spatial distribution of water bodies does not conform to normal conditions, that region is identified as an anomalous region.

[0096] Among them, the classification features include at least the vegetation index reflecting the growth status and coverage of vegetation, the water index reflecting the presence and extent of water bodies, and the texture feature index used to describe the surface structure of grassland; the spatiotemporal features include at least the first result of whether it is in a normal growth cycle, and the second result of whether the spatial distance and spatial layout with other land cover types meet the requirements.

[0097] In some embodiments of this application, the specific process of determining whether there are abnormal areas in the nature reserve to be monitored based on the basic characteristics of each initial area includes: comparing the vegetation index, water index, and texture feature index of each initial area with a preset normal state threshold range to obtain the comparison result of each initial area; when the comparison result of each initial area indicates that it does not exceed the preset normal state threshold range and both the first result and the second result are yes, it is determined that there are no abnormal areas in the nature reserve to be monitored; or, when the comparison result of each initial area indicates that it exceeds the preset normal state threshold range or both the first result and the second result are no, it is determined that there are abnormal areas in the nature reserve to be monitored.

[0098] The preset normal state threshold range is a normal range set based on historical data and experience, used to determine whether the vegetation index, water body index and texture feature index are in a normal state.

[0099] In one possible implementation, the preset normal state threshold ranges are: vegetation index 0.5 to 0.8, water index 0.3 to 0.6, and texture feature index 0.2 to 0.4. Region A has a vegetation index of 0.7 (within the normal range), a water index of 0.4 (within the normal range), and a texture feature index of 0.3 (within the normal range). The first result is "Yes" (vegetation is in a normal growth cycle), and the second result is "Yes" (spatial layout meets requirements). Region B has a vegetation index of 0.4 (below the normal range), a water index of 0.7 (above the normal range), and a texture feature index of 0.5 (above the normal range). The first result is "No" (vegetation is not in a normal growth cycle), and the second result is "No" (spatial layout does not meet requirements). In this case, all feature values ​​in Region A are within the normal range, and both the first and second results are "Yes," so the region is determined to be normal. Region B has a vegetation index below the normal range, but a water index and a texture feature index above the normal range, and both the first and second results are "No," so the region is determined to be abnormal.

[0100] For example Figure 4 As shown, Figure 4 This application provides a schematic flowchart of an anomaly area determination process for a server-side application. First, the nature reserve to be monitored is divided into multiple initial areas. Then, target satellite remote sensing data for each initial area is acquired. Next, land cover classification features and spatiotemporal features are extracted. Land cover classification features include vegetation index, water body index, and texture feature index. Spatiotemporal features include a first result characterizing whether the area is in a normal growth cycle and a second result characterizing whether spatial distance and layout meet requirements. Finally, the vegetation index, water body index, and texture feature index are compared with preset threshold intervals to determine if they exceed the threshold intervals. The first and second results are also checked for errors. If the first result is positive, an anomaly area is determined to exist; otherwise, no anomaly area is determined to exist.

[0101] In some embodiments of this application, the specific process of determining the target UAV and target ground mobile device closest to the abnormal area includes: generating a location request command and sending it to all UAVs and all ground mobile devices deployed within the area of ​​the nature reserve to be monitored; receiving the latitude and longitude coordinates obtained by each UAV and each ground mobile device through its own positioning system to obtain the location information of each UAV and each ground mobile device; determining the latitude and longitude of the center location of the abnormal area; and calculating the target UAV and target ground mobile device closest to the abnormal area based on the location information and the latitude and longitude of the center location. The relevant information of the sky-ground collaborative device displayed in the background includes, for example... Figure 5 As shown.

[0102] The location request instruction is used to request all drones and ground mobile devices deployed within nature reserves to report their current location.

[0103] In some embodiments of this application, the specific process of determining the latitude and longitude of the center location of the anomalous region includes: cropping the current region image of the anomalous region from the satellite images contained in the current satellite remote sensing data; traversing all pixels in the current region image, each pixel including row coordinates and column coordinates; performing an affine transformation on each pixel including row coordinates and column coordinates according to the affine transformation coefficients of the pre-labeled pixel coordinates and geographic coordinates to obtain the anomalous geographic coordinates corresponding to each pixel; and calculating the latitude and longitude of the center location of the anomalous region based on the anomalous geographic coordinates corresponding to each pixel; wherein, the affine transformation coefficients are obtained by using known pixel coordinates and geographic coordinates combined with the least squares method, and the expression for the affine transformation is:

[0104]

[0105] in, These are abnormal geographic coordinates. These are the pixel coordinates of each pixel, including row and column coordinates. These are the affine transformation coefficients.

[0106] Here, row and column coordinates represent the horizontal (column) and vertical (row) positions of a pixel within the image matrix. Affine transformations can convert image coordinates into actual geographic coordinates.

[0107] In some embodiments of this application, the specific process of determining the target UAV and target ground mobile device closest to the abnormal area based on location information and center location latitude and longitude includes: calculating the distance between each UAV and the abnormal area based on the target latitude and longitude and center location latitude and longitude of each UAV, combined with a preset distance calculation formula; calculating the distance between each ground mobile device and the abnormal area based on the target latitude and longitude and center location latitude and longitude of each ground mobile device, combined with a preset distance calculation formula; selecting the UAV with the smallest distance as the target UAV closest to the abnormal area, and selecting the ground mobile device with the smallest distance as the target ground mobile device closest to the abnormal area; wherein, the preset distance calculation formula is:

[0108]

[0109] in, It is the distance between two points. It is the Earth's radius. It is an arcsine function. It is the difference in latitude between two points. It is the difference in longitude between the two points. and These are the latitudes of the two points. It is a cosine function. It is a sine function.

[0110] In one possible implementation, the server generates a location request command: "Please report your current location." This command is sent wirelessly to all deployed drones and ground mobile devices within the nature reserve to be monitored. Upon receiving the command, drones A, B, and C, and ground mobile devices 1, 2, and 3 obtain their own locations via GPS. Their locations are shown in Table 1. The server extracts images of the anomalous area from satellite remote sensing data and calculates the center location of the anomalous area using image processing algorithms: (30.1235°N, 120.5679°E). The distances of each device to the center of the anomalous area are calculated: drone A is approximately 100 meters away, drone B is approximately 50 meters away, drone C is approximately 150 meters away, ground mobile device 1 is approximately 200 meters away, ground mobile device 2 is approximately 250 meters away, and ground mobile device 3 is approximately 300 meters away. At this point, the server selects the device with the shortest distance: the target drone is drone B (50 meters away), and the target ground mobile device is ground mobile device 1 (200 meters away).

[0111] Table 1

[0112]

[0113] For example Figure 6 As shown, Figure 6 This application provides a schematic diagram of the process for locating the target UAV and target ground mobile device closest to an anomaly area. Image data of the anomaly area is extracted from satellite remote sensing images, and all pixels within that area are traversed. The row and column coordinates of each pixel in the image are recorded. Using an affine transformation formula, the image coordinates of each pixel are converted to actual geographic coordinates. The affine transformation coefficients are obtained by combining the known pixel coordinates and geographic coordinates with the least squares method. Using the geographic coordinates of all pixels, the center position of the anomaly area is calculated using a polygon centroid algorithm. The latitude and longitude coordinates of the center position of the anomaly area are obtained. The distances of all UAVs and ground mobile devices from the center position of the anomaly area are calculated. The distance between each UAV and the anomaly area is calculated. The distance between each ground mobile device and the anomaly area is calculated. The UAV closest to the center of the anomaly area is selected from all UAVs. The device closest to the center of the anomaly area is selected from all ground mobile devices. The final target UAV is determined for monitoring the anomaly area. The final target ground mobile device is determined for monitoring the anomaly area.

[0114] The detailed close-up information includes high-definition video, high-definition images, and sensor data.

[0115] In some embodiments of this application, the specific process of determining whether to issue a warning for an abnormal area based on close-range detailed information includes: inputting high-definition video and high-definition images into a pre-trained hazard identification model, and outputting hazard identification results corresponding to the close-range detailed information; the pre-trained hazard identification model is obtained through machine learning using a convolutional neural network and model training samples; comparing the temperature, humidity, and gas concentration present in the sensor data with preset safety thresholds to determine whether there is an abnormal situation such as fire or environmental pollution in the abnormal area; if the hazard identification result indicates that there is a dangerous situation in the abnormal area, determining to issue a warning for the abnormal area; or, if the hazard identification result indicates that there is no dangerous situation in the abnormal area, determining not to issue a warning for the abnormal area. The sensor data includes, for example, Figure 7 As shown.

[0116] The hazard identification model is a pre-trained model used to identify and determine whether a hazardous state exists in an abnormal area. A Convolutional Neural Network (CNN) is a deep learning model suitable for processing and analyzing image data. It automatically extracts image features and performs classification by learning from a large number of training samples. The training samples used to train the hazard identification model include samples of both normal and hazardous states. This helps the model learn how to distinguish between normal and hazardous states.

[0117] In one possible implementation, a drone flies to the anomalous area and acquires high-resolution video, high-resolution images, and sensor data (such as temperature, humidity, and gas concentration). The acquired high-resolution video and images are then input into a pre-trained hazard identification model. This model is obtained through machine learning using a convolutional neural network and a large number of training samples. The model analyzes the input data and outputs a hazard identification result. For example, the model identifies signs of fire. The sensor data (such as temperature, humidity, and gas concentration) is compared with preset safety thresholds. Assuming that both temperature and gas concentration exceed the preset safety thresholds, the comparison results of the sensor data indicate a fire risk in the anomalous area. Since the hazard identification result indicates signs of fire, and the sensor data shows abnormal temperature and gas concentration, an early warning is issued for the anomalous area.

[0118] In some embodiments of this application, the specific process of generating a pre-trained hazard identification model includes: collecting a preset number of historical high-definition videos and images; the historical high-definition videos and images contain normal and hazardous states of nature reserves; for each historical high-definition video and image, labeling samples of normal and hazardous states to obtain model training samples; normal states include normal natural landscapes and wildlife activities; hazardous states include illegal intrusion, logging, fire, and environmental pollution; using a convolutional neural network to create a hazard identification model; inputting the model training samples into the hazard identification model for machine learning, and outputting the model loss value; generating a pre-trained hazard identification model when the model loss value reaches its minimum; or, if the model loss value does not reach its minimum, backpropagating the model loss value to update the model parameters of the hazard identification model, and continuing to execute the step of inputting the model training samples into the hazard identification model for machine learning until the model loss value reaches its minimum.

[0119] In this embodiment, on the one hand, the server acquires satellite remote sensing data at a preset cycle to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0120] Please see Figure 8 This document provides a flowchart illustrating an adaptive monitoring method for nature reserves based on sky-ground coordination, as described in an embodiment of this application. Figure 8 As shown, the detection method in this application embodiment may include the following steps:

[0121] S101, the server acquires the current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset cycle; based on the current satellite remote sensing data, it determines whether there are abnormal areas in the nature reserve to be monitored; if so, it identifies the target UAV and target ground mobile equipment closest to the abnormal area; and generates a collaborative command to send to the target UAV and target ground mobile equipment.

[0122] S102: The target drone and target ground mobile device travel to the abnormal area based on the coordinated command to obtain close-range detailed information about the abnormal area and send it to the server.

[0123] S103, the server determines whether to issue a warning for the abnormal area based on the detailed information of the nearby area; if so, it generates a warning information for the abnormal area and sends it to the client.

[0124] For example, the client generates warning information for abnormal areas, such as... Figure 9 As shown.

[0125] In this embodiment, on the one hand, the server acquires satellite remote sensing data at a preset cycle to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0126] Please see Figure 10 This is a flowchart illustrating a training method for a hazardous state identification model, as provided in an embodiment of this application. Figure 10 As shown, the process includes the following steps:

[0127] S201, collect a preset number of historical high-definition videos and images; the historical high-definition videos and images contain both normal and dangerous states of nature reserves.

[0128] S202: For each historical high-definition video and each historical high-definition image, label the samples for normal and dangerous states to obtain model training samples; normal states include normal natural landscapes and wildlife activities; dangerous states include illegal intrusion, logging, fire, and environmental pollution.

[0129] S203 uses a convolutional neural network to create a hazardous state recognition model;

[0130] S204, Input the model training samples into the dangerous state recognition model to perform machine learning, and output the model loss value;

[0131] S205, if the model loss value reaches the minimum, generate a pre-trained hazardous state recognition model; or, if the model loss value does not reach the minimum, backpropagate the model loss value to update the model parameters of the hazardous state recognition model, and continue to execute the step of inputting the model training samples into the hazardous state recognition model for machine learning until the model loss value reaches the minimum.

[0132] In this embodiment, on the one hand, the server acquires satellite remote sensing data at a preset cycle to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0133] The following are embodiments of the apparatus described in this application, which can be used to execute the embodiments of the method described in this application. For details not disclosed in the apparatus embodiments of this application, please refer to the embodiments of the method described in this application.

[0134] Please see Figure 11 This illustration shows a schematic diagram of the server provided in an exemplary embodiment of this application. The server can be implemented as all or part of an electronic device through software, hardware, or a combination of both. The server 1 includes a data acquisition module 10, an abnormal area judgment module 20, an instruction sending module 30, and an early warning module 40.

[0135] Data acquisition module 10 is used to acquire current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset cycle;

[0136] The anomaly area judgment module 20 is used to determine whether there are anomalies in the nature reserve to be monitored based on the current satellite remote sensing data; if so, it identifies the target UAV and target ground mobile equipment closest to the anomaly area.

[0137] Command sending module 30 is used to generate collaborative commands and send them to the target UAV and the target ground mobile device;

[0138] The early warning module 40 is used to determine whether to issue an early warning for an abnormal area based on the close-range detailed information fed back by the target UAV and the target ground mobile device; if so, it generates an early warning information for the abnormal area and sends it to the client.

[0139] It should be noted that the above-described adaptive monitoring device for nature reserves based on sky-ground coordination, when executing the adaptive monitoring method for nature reserves based on sky-ground coordination, is only illustrated by the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. Furthermore, the adaptive monitoring device for nature reserves based on sky-ground coordination and the adaptive monitoring method for nature reserves based on sky-ground coordination provided in the above embodiments belong to the same concept, and the implementation process is detailed in the method embodiments, which will not be repeated here.

[0140] The sequence numbers of the embodiments in this application are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0141] In this embodiment, on the one hand, the server acquires satellite remote sensing data at a preset cycle to determine if there are any abnormal areas. Only when an abnormal area is detected is the server instructing drones and ground mobile devices to proceed and obtain detailed close-range information. This method eliminates the need for drones and ground mobile devices to operate around the clock, reducing unnecessary data collection and transmission, thereby significantly reducing the amount of data received by the data center. The reduction in data volume lowers the difficulty of data analysis, allowing for efficient use of computing resources and avoiding waste. On the other hand, the satellite first conducts large-scale monitoring, identifies abnormal areas, determines the drones and ground mobile devices closest to the abnormal areas, and generates collaborative instructions to direct them to proceed to the abnormal areas to obtain detailed close-range information. This collaborative mechanism fully leverages the advantages of each device, combining the macroscopic monitoring capabilities of the satellite with the microscopic monitoring capabilities of drones and ground equipment. This not only ensures that abnormal situations are detected in a timely manner but also enables rapid acquisition of detailed information for early warning, thereby improving monitoring efficiency.

[0142] This application also provides a computer-readable medium having program instructions stored thereon, which, when executed by a processor, implement the adaptive monitoring method for nature reserves based on sky-ground coordination provided in the above-described method embodiments.

[0143] This application also provides a computer program product containing instructions that, when run on a computer, causes the computer to execute the adaptive monitoring method for nature reserves based on sky-ground coordination described in the various method embodiments above.

[0144] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program for adaptive monitoring of nature reserves based on sky-ground coordination can be stored in a computer-readable storage medium. When executed, the program can include the processes of the embodiments of the above methods. The storage medium for the program for adaptive monitoring of nature reserves based on sky-ground coordination can be a magnetic disk, optical disk, read-only memory, or random access memory, etc.

[0145] The above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application shall still fall within the scope of this application.

Claims

1. An adaptive monitoring system for nature reserves based on sky-ground coordination, characterized in that, The system includes: The system comprises a server, a sky-ground collaborative device, and a client; the server is communicatively connected to both the sky-ground collaborative device and the client; the sky-ground collaborative device includes satellites, drones, and ground-based mobile devices. The server is used to acquire current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset period; based on the current satellite remote sensing data, determine whether there are abnormal areas in the nature reserve to be monitored, including: dividing the nature reserve to be monitored into a preset number of initial areas; performing radiometric correction, geometric correction, and noise reduction on the current satellite remote sensing data to obtain satellite remote sensing data to be analyzed; acquiring target satellite remote sensing data for each initial area from the satellite remote sensing data to be analyzed; determining the land cover classification features and spatiotemporal features of each preset land cover type to be focused on from the target satellite remote sensing data of each initial area, as the basic features of each initial area; each preset land cover type includes at least forest, grassland, water, and wildlife habitat; the classification features include at least a vegetation index reflecting the growth status and coverage of vegetation, a water body index reflecting the presence and extent of water bodies, and a texture feature index describing the surface structure of grassland; the spatiotemporal features include at least whether it is in a positive state. The process involves several steps: First, determining whether the natural protected area under monitoring exhibits a normal growth cycle; second, determining whether the spatial distance and layout with other land cover types meet requirements; third, determining whether an abnormal area exists based on the basic characteristics of each initial area, including: comparing the vegetation index, water index, and texture feature index of each initial area with a preset normal state threshold range to obtain a comparison result for each initial area; fourth, determining that the natural protected area under monitoring does not have an abnormal area if the comparison result of each initial area does not exceed the preset normal state threshold range and both the first and second results are yes; or fifth, determining that the natural protected area under monitoring has an abnormal area if the comparison result of each initial area exceeds the preset normal state threshold range or both the first and second results are no; if yes, identifying the target UAV and target ground mobile device closest to the abnormal area; and generating a collaborative command to send to the target UAV and target ground mobile device. The target drone and target ground mobile device are used to travel to the abnormal area based on the cooperative command, to obtain close-range detailed information about the abnormal area, and send it to the server. The server is also used to determine whether to issue a warning for the abnormal area based on the nearby detailed information; if so, it generates a warning message for the abnormal area and sends it to the client.

2. The system according to claim 1, characterized in that, The server is also used to continue executing the step of acquiring the current satellite remote sensing data collected by the satellite for the nature reserve to be monitored according to a preset period, provided that there are no abnormal areas in the nature reserve to be monitored or no warning is issued for the abnormal areas.

3. The system according to claim 1, characterized in that, The determination of the target drone and target ground mobile device closest to the abnormal area includes: A location request command is generated and sent to all drones and ground mobile devices deployed within the area of ​​the nature reserve to be monitored; Receive the latitude and longitude coordinates obtained by each UAV and each ground mobile device through their own positioning system, and obtain the location information of each UAV and each ground mobile device; Determine the latitude and longitude of the center location of the abnormal region; Based on the location information and the latitude and longitude of the center location, the target UAV and target ground mobile device closest to the abnormal area are determined.

4. The system according to claim 3, characterized in that, Determining the latitude and longitude of the center location of the abnormal region includes: From the satellite images contained in the current satellite remote sensing data, crop out the current area image of the abnormal region; Traverse all pixels within the current region of the image, where each pixel includes row and column coordinates; Based on the affine transformation coefficients of the pre-labeled pixel coordinates and geographic coordinates, each pixel point, including its row coordinates and column coordinates, is subjected to an affine transformation to obtain the abnormal geographic coordinates corresponding to each pixel point. Based on the abnormal geographic coordinates corresponding to each pixel, the centroid of the polygon is solved to calculate the latitude and longitude of the center position of the abnormal region; wherein... The affine transformation coefficients are obtained by solving using known pixel coordinates and geographic coordinates combined with the least squares method. The expression for the affine transformation is: in, These are abnormal geographic coordinates. These are the pixel coordinates of each pixel, including row and column coordinates. These are the affine transformation coefficients.

5. The system according to claim 3, characterized in that, The location information is the target's latitude and longitude; The step of determining the target UAV and target ground mobile device closest to the abnormal area based on the location information and the latitude and longitude of the center location includes: Based on the target latitude and longitude and the center location latitude and longitude of each UAV, and combined with a preset distance calculation formula, the distance between each UAV and the abnormal area is calculated. Based on the target latitude and longitude and the center location latitude and longitude of each ground mobile device, and combined with a preset distance calculation formula, the distance between each ground mobile device and the abnormal area is calculated. The drone with the smallest distance is selected as the target drone closest to the abnormal area, and the ground mobile device with the smallest distance is selected as the target ground mobile device closest to the abnormal area; wherein, the preset distance calculation formula is: in, It is the distance between two points. It is the Earth's radius. It is an arcsine function. It is the difference in latitude between two points. It is the difference in longitude between the two points. and These are the latitudes of the two points. It is a cosine function. It is a sine function.

6. The system according to claim 1, characterized in that, The detailed close-up information includes high-definition video, high-definition images, and sensor data; The step of determining whether to issue an early warning for the abnormal area based on the detailed proximity information includes: The high-definition video and high-definition images are input into a pre-trained hazard identification model, which outputs the hazard identification result corresponding to the close-range detailed information. The pre-trained hazard identification model is obtained by machine learning through a convolutional neural network and model training samples. The temperature, humidity, and gas concentration data from the sensors are compared with preset safety thresholds to determine whether there is an abnormal situation such as fire or environmental pollution in the abnormal area. If the hazard identification result indicates that a hazard exists in the abnormal area, a warning will be issued for the abnormal area; or... If the danger identification result indicates that there is no danger in the abnormal area, it is determined not to issue a warning for the abnormal area.

7. The system according to claim 6, characterized in that, Generate a pre-trained hazardous state recognition model by following these steps: Collect a predetermined number of historical high-definition videos and images; the historical high-definition videos and images include both normal and dangerous states of nature reserves. For each historical high-definition video and each historical high-definition image, sample labels for normal and dangerous states are marked to obtain model training samples; the normal state includes normal natural landscapes and wildlife activities; the dangerous state includes illegal intrusion, logging, fire, and environmental pollution; A hazardous state recognition model is created using a convolutional neural network. The training samples of the model are input into the dangerous state identification model to perform machine learning, and the model loss value is output. When the model loss value reaches its minimum, a pre-trained dangerous state recognition model is generated. Alternatively, if the model loss value has not reached its minimum, the model loss value is backpropagated to update the model parameters of the hazard identification model, and the step of inputting the model training samples into the hazard identification model for machine learning continues until the model loss value reaches its minimum.

8. A method for adaptive monitoring of nature reserves based on sky-ground coordination, implemented using the system described in any one of claims 1-7, characterized in that, The method includes: The server acquires current satellite remote sensing data collected by the satellite on the nature reserve to be monitored according to a preset period; based on the current satellite remote sensing data, it determines whether there is an abnormal area in the nature reserve to be monitored; if so, it identifies the target UAV and target ground mobile device closest to the abnormal area; and generates a collaborative command and sends it to the target UAV and target ground mobile device. Based on the coordinated command, the target drone and the target ground mobile device travel to the abnormal area to obtain close-range detailed information about the abnormal area and send it to the server. The server determines whether to issue a warning for the abnormal area based on the nearby detailed information; if so, it generates a warning message for the abnormal area and sends it to the client.

Images