Unmanned aerial vehicle multi-target rescue decision-making method

Abstract

The invention relates to the technical field of unmanned aerial vehicle rescue, in particular to an unmanned aerial vehicle multi-target rescue decision-making method, which comprises the steps of transmitting real-time image data to a remote command terminal according to a high-definition real-time transmission technology, and receiving target confirmation information fed back by the remote command terminal based on the real-time image data; generating a multi-target rescue priority list according to the target confirmation information, the real-time position data and the initial distribution coordinate data; and generating a personalized rescue execution instruction for each to-be-rescued target according to the multi-target rescue priority list, the real-time image data and the remaining power data of the intelligent unmanned aerial vehicle. According to the invention, the personalized rescue execution instruction is generated according to the survival state scores and the environmental obstacle levels of different targets, so that the rescue operation is more targeted and effective; and in a complex environment, such as a night low-illumination environment, the unmanned aerial vehicle can identify a heat source and adjust the illumination direction and brightness by using infrared thermal imaging data, a night rescue mode is realized, and the rescue success rate is improved.

Description

Technical Field

[0001] This invention relates to the field of drone rescue technology, specifically a drone multi-target rescue decision-making method. Background Technology

[0002] Drones, also known as unmanned aerial vehicles, are aircraft that can be remotely controlled or fly autonomously, and usually do not require human crew. Drone rescue refers to an emerging method of using drone technology for emergency rescue, disaster recovery, or humanitarian assistance. This rescue method utilizes the flight capabilities and image processing technology of drones to respond to various emergencies quickly and efficiently.

[0003] Currently, traditional methods often use fixed monitoring or reporting systems, which are not updated in a timely manner, making it impossible for the command center to grasp the situation on the ground in real time, affecting the timeliness and accuracy of decision-making. Moreover, in the face of complex and ever-changing rescue environments, the rescue plans of traditional methods are often relatively rigid and lack the ability to dynamically adjust strategies according to real-time situations, which can easily lead to an inability to effectively respond to emergencies. Furthermore, among multiple rescue targets, traditional methods may not be able to effectively assess and prioritize rescue efforts, resulting in the most needy targets not receiving timely assistance.

[0004] Furthermore, in complex environments such as at night or in severe weather, traditional rescue methods may be ineffective, making it difficult to accurately identify targets and assess environmental obstacles. In addition, traditional communication methods often rely on means such as radio or telephone, which may be subject to signal interference or blockage, affecting information exchange between on-site command and rescue personnel. Moreover, when faced with a large amount of real-time data, traditional methods often lack sufficient processing capabilities, resulting in the inability to quickly extract important information, thereby affecting the overall rescue efficiency. Summary of the Invention

[0005] To achieve the above objectives, the present invention provides the following technical solution: a multi-target rescue decision-making method for unmanned aerial vehicles (UAVs), comprising: Receive rescue mission instructions from a remote command terminal, which include the number of targets to be rescued and the scope of the rescue area; The real-time location data of the intelligent drone and the initial distribution coordinate data of the target to be rescued are obtained through a single Beidou positioning and navigation system. Based on real-time location data and initial distribution coordinate data, the intelligent drone is controlled to fly to the target area and collect real-time image data of the rescue area through dual-light gimbal. The real-time image data is transmitted to the remote command terminal using high-definition real-time transmission technology, and the target confirmation information is received from the remote command terminal based on the real-time image data. A multi-target rescue priority list is generated based on target confirmation information, real-time location data, and initial distribution coordinate data. Personalized rescue execution instructions are generated for each target to be rescued based on a multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone.

[0006] Preferably, after transmitting real-time image data to the remote command terminal using high-definition real-time transmission technology and receiving target confirmation information from the remote command terminal based on the real-time image data, the method further includes: If packet loss or interruption is detected in the high-definition real-time transmission link, immediately switch to the short message channel of the single Beidou positioning and navigation system to transmit critical status data. An emergency rescue strategy is generated based on key status data and the last image frame recorded by the dual-light PTZ camera. The system uses an onboard remote voice module to send out high-frequency distress signals and location broadcasts, including latitude and longitude coordinates generated by a single BeiDou positioning and navigation system.

[0007] Preferably, after generating a multi-target rescue priority list based on target confirmation information, real-time location data, and initial distribution coordinate data, the method further includes: Multiple targets awaiting rescue were detected at the same coordinate point in the initial distribution coordinate data. The environmental carrying capacity of the coordinate point was determined based on real-time image data. If the environmental carrying capacity is greater than the preset carrying capacity, the coordinate point will be marked as a rendezvous point according to the multi-target rescue priority list. If the environmental carrying capacity is less than the preset carrying capacity, the coordinate point is divided into several sub-coordinates according to the positioning accuracy of the single Beidou positioning and navigation system, and the order of the multi-target rescue priority list is assigned to the sub-coordinates to generate a dispersed rescue path.

[0008] Preferably, personalized rescue execution instructions are generated for each target to be rescued based on a multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone, including: Based on real-time image data, identify the survival status score of the target to be rescued and the level of obstacles in the surrounding environment; When the survival status score is detected to be lower than the preset score threshold and the surrounding environmental obstacle level is high, the precise positioning data of the single Beidou positioning and navigation system is retrieved to generate hovering coordinates. When the survival status score is detected to be higher than the preset score threshold and the surrounding environmental obstacle level is low, the landing coordinates are generated according to the multi-target rescue priority list. The first rescue sub-command is generated by combining the remaining power data, hovering coordinates and landing coordinates. The first rescue sub-command includes the lighting equipment activation parameters and the dual-light gimbal zoom parameters. The onboard lighting equipment is activated to provide supplemental lighting based on the first rescue sub-instruction. Pre-set guidance voice is played to the target to be rescued via the remote voice module, generating personalized rescue execution instructions for each target to be rescued.

[0009] Preferably, the first rescue sub-command is generated by combining the remaining battery power data, hovering coordinates, and landing coordinates, including: Acquire the current attitude angle and focal length data of the dual-light gimbal; calculate the gimbal rotation angle increment based on the hovering or landing coordinates; The maximum power of lighting equipment is limited based on the remaining power data, and the bandwidth usage of high-definition real-time transmission technology is also limited. The device control parameters are generated based on the gimbal rotation angle increment, maximum power, and bandwidth usage. The first rescue sub-command includes the device control parameters and remote voice triggering conditions.

[0010] Preferably, after identifying the survival status score of the target to be rescued and the level of obstacles in the surrounding environment based on real-time image data, the method further includes: The presence of a low-light environment at night was detected in the real-time image data, and the distribution of heat sources was identified based on the infrared thermal imaging data of the dual-light gimbal. Adjust the optical axis direction of the airborne lighting equipment according to the distribution of heat sources to ensure that the light spot covers the main heat source areas; The lighting brightness is adjusted based on the temperature change trend of the main heat source area and the clarity of the high-definition real-time transmitted images to generate nighttime rescue mode instructions; among them, personalized rescue execution instructions include nighttime rescue mode instructions and reassurance messages based on remote voice.

[0011] Preferably, the method further includes: Establish a remote voice link between the rescue drone and the field command center to acquire voice interaction data and generate environmental perception information; Image feature extraction processing is performed on real-time video data to obtain first rescue feature information, and semantic recognition processing is performed on voice interaction data to obtain second rescue feature information; Based on the first rescue feature information and the second rescue feature information, the multi-target rescue decision information of the rescue execution drone is determined and sent to the rescue execution drone.

[0012] Preferably, based on the first rescue feature information and the second rescue feature information, the multi-target rescue decision information of the rescue execution drone is determined, including: Acquire single BeiDou positioning and navigation data of the rescue drone, where single BeiDou positioning and navigation data is the trajectory prediction information generated by the rescue drone based on its own position; Based on the first and second rescue feature information, a target priority sequence is determined; based on single BeiDou positioning and navigation data and the target priority sequence, multi-target rescue decision information for the rescue execution drone is determined.

[0013] Preferably, after sending multi-target rescue decision information to the rescue drone, the process also includes: Acquire rescue execution feedback information generated by the rescue drone based on multi-target rescue decision information; generate rescue progress and status information based on the rescue execution feedback information; wherein, the rescue execution feedback information includes the current rescue target number, the dual-optical gimbal imaging quality level, and the signal strength of the single Beidou positioning and navigation system; Once the trapped personnel or on-site commander corresponding to the target to be rescued are identified, rescue progress and status information are sent to the trapped personnel or on-site commander via a remote voice link.

[0014] Compared with the prior art, the beneficial effects of the present invention are: This invention enables drones to quickly and accurately fly to target areas by receiving rescue mission instructions from a remote command terminal and combining real-time location and target distribution data, thereby improving rescue efficiency. Furthermore, through high-definition real-time transmission technology, the drone can transmit image data to the command terminal in real time, ensuring the command center receives the latest information and can make timely decisions. Moreover, by generating a multi-target rescue priority list, the drone can dynamically adjust the priority of rescue targets based on real-time conditions, ensuring that the targets most in need of rescue receive priority assistance. This invention generates personalized rescue execution instructions based on the survival status scores of different targets and the level of environmental obstacles, making rescue operations more targeted and effective. In complex environments, such as low-light nighttime environments, the drone can use infrared thermal imaging data to identify heat sources and adjust the direction and brightness of illumination to achieve a nighttime rescue mode, improving the success rate of rescue. Furthermore, when there is packet loss or interruption in the high-definition real-time transmission link, the drone can quickly switch to the short message channel to transmit critical status data, ensuring that the transmission of critical data is not affected and enhancing the reliability of the system. This invention establishes a remote voice link with the field command center to enable voice interaction, enhances information communication between on-site command and drones, and improves decision-making efficiency; furthermore, by extracting and analyzing features from real-time image data and voice interaction data, it comprehensively judges the environmental and target situation and optimizes the rescue decision-making process. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of the overall method in one embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] Example 1, please refer to Figure 1 This invention provides a technical solution: a multi-target rescue decision-making method for unmanned aerial vehicles (UAVs), comprising: S1. Receive rescue mission instructions from the remote command terminal, wherein the rescue mission instructions include the number of targets to be rescued and the scope of the rescue area; S2. Obtain real-time location data of the intelligent drone and initial distribution coordinate data of the target to be rescued through a single Beidou positioning and navigation system; S3. Based on real-time location data and initial distribution coordinate data, control the intelligent drone to fly to the target area and collect real-time image data of the rescue area through dual-light gimbal. S4. Transmit real-time video data to the remote command terminal using high-definition real-time transmission technology, and receive target confirmation information from the remote command terminal based on the real-time video data. S5. Generate a multi-target rescue priority list based on target confirmation information, real-time location data, and initial distribution coordinate data; S6. Generate personalized rescue execution instructions for each target to be rescued based on the multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone.

[0018] It should be noted that the remote command center will transmit a rescue mission instruction, including the number of targets to be rescued and the area to be rescued; for example, the instruction might be: "Rescue 5 trapped people in area X;" The drone obtains its real-time location data through the BeiDou Navigation Satellite System; this is crucial for the drone to determine its current flight position; for example, the drone's current position is "latitude: 30.1234, longitude: 120.5678"; after receiving the mission, the drone also needs to obtain the initial coordinates of the targets to be rescued. Data; this data may originate from prior reconnaissance or on-site feedback; for example, the initial coordinates of the targets to be rescued are: Target 1: Latitude: 30.1300, Longitude: 120.5700; Target 2: Latitude: 30.1310, Longitude: 120.5720; Target 3: Latitude: 30.1320, Longitude: 120.5730; Target 4: Latitude: 30.1330, Longitude: 120.5740; Target 5: Latitude: 30.1340, Longitude: 120.5750; Based on real-time location data and the coordinates of the target to be rescued, the drone automatically plans its flight route to the target area for reconnaissance. At this time, the drone uses a dual-light gimbal (infrared and visible light cameras) to collect real-time image data of the area, helping to confirm the specific situation of the target. The real-time image data collected by the drone is sent back to the remote command terminal via high-definition real-time transmission technology. The command center can view the situation in the area in real time to make further decisions; for example, the command center may see a trapped person in danger. After receiving the real-time image data, the command center will confirm the location and status of the target to be rescued based on the image data and feed the confirmation information back to the drone; for example, it may be confirmed that targets 1 and 3 require immediate rescue, while targets 2, 4, and 5 are in better condition and can be dealt with later. Based on the target confirmation information and combined with real-time location and initial coordinate data, the system generates a priority list of targets to be rescued; for example: Priority 1: Target 1 (critical); Priority 2: Target 3 (requires rescue); Priority 3: Target 2 (relatively safe, can be rescued later); Finally, based on the priority list, real-time image data, and the drone's remaining battery power, personalized rescue execution instructions are generated for each target to be rescued; for example: for Target 1, the instruction is "land quickly and provide emergency medical support"; for Target 3, the instruction is "send a rope to help it escape danger"; for Target 2, the instruction is "monitor its condition and carry out rescue after sufficient battery power".

[0019] In an optional embodiment, after transmitting real-time image data to a remote command terminal using high-definition real-time transmission technology, and receiving target confirmation information from the remote command terminal based on the real-time image data, the method further includes: If packet loss or interruption is detected in the high-definition real-time transmission link, immediately switch to the short message channel of the single Beidou positioning and navigation system to transmit critical status data. An emergency rescue strategy is generated based on key status data and the last image frame recorded by the dual-light PTZ camera. The system uses an onboard remote voice module to send out high-frequency distress signals and location broadcasts, including latitude and longitude coordinates generated by a single BeiDou positioning and navigation system.

[0020] It should be noted that during real-time image data transmission by the drone, network instability may occur, such as data packet loss or transmission interruption. In this case, the drone will activate its internal monitoring system to detect the problem. For example, if the system records a 50% packet loss rate in the past few seconds, the drone will immediately switch to the short message channel of the BeiDou positioning and navigation system. This channel typically has higher reliability and can send critical status data, such as the drone's location information, battery level, and flight status, even in harsh environments. For example, the drone might send the following data via short message: Location information: Latitude 30.1300, Longitude 120.5700; Remaining battery: 75%; Flight status: Normal. Based on critical status data and the last image frame recorded by the dual-light gimbal, the UAV system generates an emergency rescue strategy in conjunction with the current situation. Assuming the last image shows that target 1 is in mortal danger and requires immediate rescue, and the UAV's critical status data indicates that the remaining battery power is sufficient to support emergency operations, the generated emergency rescue strategy might be: prioritize the emergency landing and medical assistance of target 1; postpone the rescue of target 3 until the battery power is sufficient; to ensure that on-site personnel can hear the presence of the UAV and the ongoing rescue operation, the UAV can use its onboard remote voice module to emit a high-frequency distress signal; such a signal could be a continuous alarm or a specific distress audio to attract the attention of the trapped personnel. Simultaneously, the drone will also broadcast its current location via a voice module; the location broadcast includes latitude and longitude coordinates generated by the BeiDou positioning and navigation system; for example, the drone may issue the following voice broadcast: "Rescue drone has arrived, current location: latitude 30.1300, longitude 120.5700, please confirm your location;" For example: Suppose a landslide occurs in a certain area, and several climbers are trapped; the drone is dispatched to perform a rescue mission; during transmission, a packet loss suddenly occurs, causing the video stream to be interrupted; the drone promptly switches to the short message channel to send critical status data; the last image frame shows a climber trying to find a safe shelter; based on this information, the drone generates an emergency strategy and decides to prioritize rescuing this climber.

[0021] In an optional embodiment, after generating a multi-target rescue priority list based on target confirmation information, real-time location data, and initial distribution coordinate data, the method further includes: Multiple targets awaiting rescue were detected at the same coordinate point in the initial distribution coordinate data. The environmental carrying capacity of the coordinate point was determined based on real-time image data. If the environmental carrying capacity is greater than the preset carrying capacity, the coordinate point will be marked as a rendezvous point according to the multi-target rescue priority list. If the environmental carrying capacity is less than the preset carrying capacity, the coordinate point is divided into several sub-coordinates according to the positioning accuracy of the single Beidou positioning and navigation system, and the order of the multi-target rescue priority list is assigned to the sub-coordinates to generate a dispersed rescue path.

[0022] It should be noted that when a drone detects multiple targets to be rescued at the same coordinate point based on initial distribution coordinate data—for example, in a collapsed building where several people may be trapped in the same location—the drone needs to process this information to ensure the effectiveness of the rescue. The drone analyzes the environmental carrying capacity of the coordinate point using real-time image data. Environmental carrying capacity typically refers to factors such as the number of people the location can safely accommodate and structural stability. For example, images captured by the drone's dual-light gimbal may show that the location is a partially collapsed building, potentially supporting a maximum of three people. If, after assessment, the environmental carrying capacity of the coordinate point is greater than the preset carrying capacity (e.g., 4 people), the drone will mark this coordinate point as a rendezvous and rescue point. In this case, the drone can simultaneously rescue all the targets to be rescued. For example: rendezvous and rescue point coordinates: latitude 30.1300, longitude 120.5700; number of targets to be rescued: 5 people. If the environmental carrying capacity is less than the preset carrying capacity (e.g., a maximum of 3 people), the drone needs to perform further processing. In this case, it will break down the coordinate point into several sub-coordinates based on the positioning accuracy of the single Beidou positioning and navigation system to facilitate the gradual implementation of the rescue. Assuming the drone assesses that the maximum carrying capacity of the location is 2 people, and there are 4 people trapped, the system might split the original coordinates into two sub-coordinates: Sub-coordinate 1: latitude 30.1301, longitude 120.5700 (corresponding to target 1 and target 2); Sub-coordinate 2: latitude 30.1302, longitude 120.5700 (corresponding to target 3 and target 4). After splitting the coordinates, the drone will assign the rescue order of each target to the sub-coordinates according to the multi-target rescue priority list; for example: target 1 (priority 1) and target 2 (priority 2) are assigned to sub-coordinate 1; target 3 (priority 3) and target 4 (priority 4) are assigned to sub-coordinate 2. Then, the drone will generate a distributed rescue path based on these sub-coordinates; for example, the drone can perform the rescue according to the following path: first fly to sub-coordinate 1 to rescue target 1 and target 2; after completing the rescue, then go to sub-coordinate 2 to rescue target 3 and target 4. For example, in a rescue mission after an earthquake, a drone discovered that five people were trapped in a certain area with the same initial coordinates. The drone conducted an environmental assessment and found that the building's environmental carrying capacity was only for two people. Therefore, the drone split the point into two sub-coordinates and determined the rescue order according to a priority list. Target 1 and Target 2 were assigned to the first sub-coordinate, and Target 3 and Target 4 were assigned to the second sub-coordinate, and the corresponding flight path was determined.

[0023] In an optional embodiment, personalized rescue execution instructions are generated for each target to be rescued based on a multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone, including: Based on real-time image data, identify the survival status score of the target to be rescued and the level of obstacles in the surrounding environment; When the survival status score is detected to be lower than the preset score threshold and the surrounding environmental obstacle level is high, the precise positioning data of the single Beidou positioning and navigation system is retrieved to generate hovering coordinates. When the survival status score is detected to be higher than the preset score threshold and the surrounding environmental obstacle level is low, the landing coordinates are generated according to the multi-target rescue priority list. The first rescue sub-command is generated by combining the remaining power data, hovering coordinates and landing coordinates. The first rescue sub-command includes the lighting equipment activation parameters and the dual-light gimbal zoom parameters. The onboard lighting equipment is activated to provide supplemental lighting based on the first rescue sub-instruction. Pre-set guidance voice is played to the target to be rescued via the remote voice module, generating personalized rescue execution instructions for each target to be rescued.

[0024] It should be noted that the drone uses real-time imagery data to assess the survival status of the target to be rescued (e.g., assessing the consciousness and breathing of the trapped person) and the level of obstacles in the surrounding environment (e.g., the extent of building collapse and debris accumulation). The survival status score is usually based on a series of indicators, such as from 0 to 10, where 0 indicates no signs of life and 10 indicates good health. For example, the survival status score of target A to be rescued is 3 (low), and the level of obstacles in the surrounding environment is high (e.g., a large number of collapsed buildings and debris); the survival status score of target B to be rescued is 8 (high), and the level of obstacles in the surrounding environment is low (e.g., open area with no obvious obstacles). Based on the survival status score and the level of environmental obstacles, the drone will take different actions: If the survival status score is lower than a preset threshold (e.g., 5 points) and the surrounding environmental obstacle level is high, the drone will retrieve precise positioning data from the BeiDou Navigation Satellite System to generate hovering coordinates. This is to ensure that the drone can hover in a safe location to better assess the situation and prepare for rescue. For example, for target A to be rescued, with a survival status score of 3 and a high obstacle level, the drone generates hovering coordinates: latitude 30.1301, longitude 120.5701. If the survival status score is higher than the preset threshold (e.g., 5 points) and the surrounding environmental obstacle level is low, landing coordinates are generated to prepare for direct rescue. For example, for target B to be rescued, with a survival status score of 8 and a low obstacle level, the drone generates landing coordinates: latitude 30.1302, longitude 120.5702. The drone combines remaining battery power data, hovering coordinates, and landing coordinates to generate the first rescue sub-command. This command includes parameters for activating lighting equipment (such as turning on a high-intensity light) and zoom parameters for the dual-light gimbal (such as adjusting the camera's focus to clearly observe the target). For example, for target A to be rescued, the generated first rescue sub-command might be: Lighting equipment: Turn on high brightness mode (high-intensity illumination); Dual-light gimbal zoom parameters: Zoom to 4x, focusing on monitoring the status of target A. For target B to be rescued, the generated command might be slightly different, depending on the environmental conditions to determine whether to activate the lighting and zoom settings. Based on the first rescue sub-command, the drone activates its onboard lighting equipment to provide supplemental illumination, ensuring clear observation of the target's status in dim environments and improving the success rate of subsequent rescues. Through a remote voice module, the drone plays pre-set guidance messages to each target awaiting rescue, helping them remain calm and instructing them on self-rescue or waiting for rescue. For example, for target A, the message might be: "Please remain calm. Our rescue drone has arrived. Please do not move and stay where you are and wait for rescue." For target B, the message might be: "You can safely leave the current area. The drone will be here to support you."

[0025] In an optional embodiment, a first rescue sub-command is generated by combining remaining battery power data, hovering coordinates, and landing coordinates, including: Acquire the current attitude angle and focal length data of the dual-light gimbal; calculate the gimbal rotation angle increment based on the hovering or landing coordinates; The maximum power of lighting equipment is limited based on the remaining power data, and the bandwidth usage of high-definition real-time transmission technology is also limited. The device control parameters are generated based on the gimbal rotation angle increment, maximum power, and bandwidth usage. The first rescue sub-command includes the device control parameters and remote voice triggering conditions.

[0026] It should be noted that the dual-light gimbals on drones are typically equipped with high-precision sensors. These sensors can provide the gimbal's current attitude angles (such as pitch, roll, and yaw) and the focal length data of the gimbal lens. This information is crucial for adjusting the drone's camera angle and sharpness. For example, the current gimbal attitude angle data is: pitch angle 15 degrees, roll angle 0 degrees, and yaw angle 30 degrees; the current focal length data is: focal length 20mm. Based on the previously generated hovering or landing coordinates, the drone needs to adjust the gimbal's angle of view to better observe the target to be rescued. This process involves calculating the required rotation angle increment of the gimbal, which is usually determined based on the geometric relationship between the target position and the gimbal's current attitude. For example, if the target to be rescued is directly in front of the hovering coordinates, but the current gimbal view is not aligned with the target, the required rotation angle needs to be calculated to align the gimbal with the target. For example, if the target's position is southeast of the drone's hovering position, and the current gimbal yaw angle is 30 degrees, the calculated target angle is 45 degrees, then the gimbal needs to rotate an increment of 15 degrees (45 degrees - 30 degrees). The drone's remaining battery power is a crucial factor during rescue operations. To ensure the drone maintains sufficient power during the mission, the maximum power of the lighting equipment must be limited. Furthermore, high-definition real-time transmission technologies (such as video streaming) consume bandwidth, so their bandwidth usage also needs to be limited based on the remaining battery power to prevent the drone from losing control or signal. For example, if the remaining battery is 30%, the drone's lighting equipment is set to a maximum power of 50W (limited according to the battery percentage). To ensure power supply, the high-definition real-time transmission bandwidth is limited to 2Mbps. Based on the gimbal rotation angle increment, maximum power, and bandwidth usage, device control parameters are generated. These parameters include specific instructions for gimbal rotation, power settings for lighting equipment, and bandwidth limits for video transmission. For example, device control parameters might include: gimbal rotation angle increment: 15 degrees; lighting equipment power: 50W; high-definition real-time transmission bandwidth: 2Mbps. Finally, combining all the above information, a first rescue sub-instruction is generated, containing device control parameters and remote voice trigger conditions. This instruction will guide the drone on how to operate to optimize the rescue effect. For example, the first rescue sub-instruction might be generated as follows: "Adjust gimbal yaw angle by 15 degrees to align with the target;" "Set lighting equipment power to 50W to ensure illumination;" "Limit video bandwidth to 2Mbps to ensure stable transmission;" "Trigger voice prompt: 'Please note that we are adjusting the viewing angle to better observe you, please remain calm;'"

[0027] In an optional embodiment, after identifying the survival status score of the target to be rescued and the level of obstacles in the surrounding environment based on real-time image data, the method further includes: The presence of a low-light environment at night was detected in the real-time image data, and the distribution of heat sources was identified based on the infrared thermal imaging data of the dual-light gimbal. Adjust the optical axis direction of the airborne lighting equipment according to the distribution of heat sources to ensure that the light spot covers the main heat source areas; The lighting brightness is adjusted based on the temperature change trend of the main heat source area and the clarity of the high-definition real-time transmitted images to generate nighttime rescue mode instructions; among them, personalized rescue execution instructions include nighttime rescue mode instructions and reassurance messages based on remote voice.

[0028] It should be noted that in low-light conditions or at night, the drone's camera will detect insufficient ambient light. In such cases, the infrared thermal imaging system needs to be activated to better identify and locate potential heat sources (such as trapped people or other living beings). For example, when the drone is flying at night, the real-time image shows a very dark surrounding environment, making it impossible to clearly capture the target. Through the infrared thermal imaging sensor of the dual-light gimbal, the drone can obtain the distribution of heat sources in the surrounding environment in real time. These heat sources may include people, animals, or other heat-generating objects. For example, in the thermal imaging image, the drone detected three main heat sources, one of which was located in the southeast direction with a temperature of 38°C, and the other two heat sources were located in the northwest and due north directions, with temperatures of 36°C and 37°C, respectively. Based on the identified heat source distribution, the drone will automatically adjust the optical axis of its onboard lighting equipment to ensure that the light spot covers the main heat source area. This process is to provide sufficient illumination to help search and rescue personnel better confirm the target location. For example, the drone may adjust its optical axis to the southeast to illuminate a heat source with a temperature of 38°C, while keeping other heat sources within its field of view. After identifying the main heat source areas, it is necessary to monitor the temperature change trends in these areas. If the heat source temperature remains stable or increases, it may mean that the target still exists; if the temperature drops, it may indicate that the target's status has changed (e.g., someone has left the area). Based on this information, the brightness of the lighting equipment will be adjusted to ensure effective observation and rescue under different environmental conditions. For example, if the temperature of a heat source in the southeast direction rises to 39°C in a short period of time, it indicates that the trapped person may be active, and the lighting brightness will be increased to 100%; if the temperature drops to 35°C, the brightness will be reduced to 50%. By combining heat source distribution, optical axis direction, and brightness settings, nighttime rescue mode instructions are generated. These instructions guide the drone to perform relevant operations at night to ensure the smooth progress of rescue work. For example, nighttime rescue mode instructions include: "Adjust the optical axis direction to the southeast to illuminate the main heat source area" and "Set the brightness of the lighting equipment to 100% to ensure clear observation." In addition to technical instructions, personalized rescue execution instructions also need to include reassuring dialogue based on remote voice to calm trapped personnel and reduce their anxiety and fear. For example, remote voice trigger content includes: "Hello, we are here to help you; please remain calm, we are illuminating your location, you will be found soon."

[0029] In an optional embodiment, the method further includes: Establish a remote voice link between the rescue drone and the field command center to acquire voice interaction data and generate environmental perception information; Image feature extraction processing is performed on real-time video data to obtain first rescue feature information, and semantic recognition processing is performed on voice interaction data to obtain second rescue feature information; Based on the first rescue feature information and the second rescue feature information, the multi-target rescue decision information of the rescue execution drone is determined and sent to the rescue execution drone.

[0030] It should be noted that this step involves establishing a stable communication link, enabling the drone to interact with the on-site command center in real time via voice. This communication method allows on-site commanders to obtain real-time status information of the drone and its surrounding environment, and also to operate the drone via voice commands. For example, if the rescue command center establishes a voice link with the drone via a wireless network, the commander can directly issue commands to the drone, such as "Please move southeast and search for heat sources." After establishing the voice link, the drone will receive voice data from the command center in real time. Through voice recognition technology, the system can analyze this voice content and extract useful information, such as the rescue location, specific areas requiring attention, or other important instructions, thereby generating environmental awareness information. For example, if the commander says, "Pay attention near the building in the southeast direction; there may be trapped individuals," the drone's system will identify the key information of "southeast direction" and "near the building" and record it as environmental awareness data. During flight, the drone continuously captures real-time images. Image processing algorithms extract features from these images to identify key rescue characteristics, such as suspicious heat sources, human silhouettes, or other relevant environmental features. For example, the drone uses a thermal imaging camera to capture a high-temperature area and extracts its image features (such as location, size, and shape) to form the first rescue feature information. In addition to image data, voice interaction data also needs to be processed to identify specific instructions and intentions. This can be achieved through natural language processing (NLP) technology to obtain the second rescue feature information. For example, the system analyzes the commander's voice instructions and identifies the semantic meaning of "there may be trapped persons," forming the second rescue feature information and marking it as the "potential trapped person location." By combining primary rescue feature information (feature extraction from real-time image data) and secondary rescue feature information (semantic recognition from voice interaction data), the UAV system can make more accurate rescue decisions. These decisions consider factors such as the priority and location of multiple targets to formulate the most effective rescue strategy. For example, based on extracted image features (heat source near a building in the southeast direction) and voice features (location of potential trapped individuals), the UAV decides to prioritize a detailed search in the southeast direction and formulate a corresponding rescue plan. Finally, the determined multi-target rescue decision information is sent back to the rescue execution UAV to ensure it acts according to instructions. For example, the rescue decision information includes: "Head southeast, focus on searching near the building, pay attention to changes in the heat source, and prepare for rescue."

[0031] In an optional embodiment, determining multi-target rescue decision information for the rescue-executing UAV based on first rescue feature information and second rescue feature information includes: Acquire single BeiDou positioning and navigation data of the rescue drone, where single BeiDou positioning and navigation data is the trajectory prediction information generated by the rescue drone based on its own position; Based on the first and second rescue feature information, a target priority sequence is determined; based on single BeiDou positioning and navigation data and the target priority sequence, multi-target rescue decision information for the rescue execution drone is determined.

[0032] It should be noted that single BeiDou positioning and navigation data refers to the UAV's own position information obtained through the BeiDou satellite navigation system. This data not only includes the UAV's current position information (such as latitude, longitude, and altitude), but also calculates its trajectory prediction information, that is, the UAV's possible flight path in the future. For example, assuming the UAV is currently located at longitude 120.5 and latitude 30.5, after analysis, it is predicted that in the next 5 minutes, if it continues to fly along the current heading, it is expected to reach a position at longitude 120.6 and latitude 30.55. At this stage, combined with the first and second rescue feature information obtained from image feature extraction and voice interaction data, the system will analyze the importance of different rescue targets and assign them priorities. For example, the identified heat source, the location of the trapped person, and the degree of their trapped status will all affect the priority ranking of the targets. For example: Suppose a drone detects the locations of three potential trapped individuals through image analysis: Location A (high temperature, suspected trapped individuals, priority 1); Location B (partial building collapse, possible trapped individuals, priority 2); Location C (no one around, normal temperature, priority 3). Based on this information, the system generates a target priority sequence: Location A > Location B > Location C. Combining the drone's positioning and navigation data with the target priority sequence, the system will formulate specific multi-target rescue decision information. This decision will clarify which target the drone should prioritize during the rescue process and plan the corresponding flight path. For example: Based on the above priority sequence and single BeiDou positioning and navigation data, the system decides to first proceed to Location A for rescue. Suppose the drone is currently at location (120.5, 30.5), while location A is at (120.55, 30.52). The drone will calculate the optimal flight path and formulate the following decision information: "Proceed to Location A (longitude 120.55, latitude 30.52), confirm and implement rescue measures."

[0033] In an optional embodiment, after sending the multi-target rescue decision information to the rescue execution drone, the method further includes: Acquire rescue execution feedback information generated by the rescue drone based on multi-target rescue decision information; generate rescue progress and status information based on the rescue execution feedback information; wherein, the rescue execution feedback information includes the current rescue target number, the dual-optical gimbal imaging quality level, and the signal strength of the single Beidou positioning and navigation system; Once the trapped personnel or on-site commander corresponding to the target to be rescued are identified, rescue progress and status information are sent to the trapped personnel or on-site commander via a remote voice link.

[0034] It should be noted that rescue execution feedback information refers to real-time data generated by the UAV after performing a specific rescue mission. This information includes: Current rescue target number: identifying the current location or object the UAV is rescuing; Dual-light gimbal imaging quality level: assessing the image quality captured by the UAV's camera, usually expressed in levels (e.g., high, medium, low); Single BeiDou positioning and navigation system signal strength: reflecting the strength of the BeiDou satellite signal received by the UAV, which directly affects the UAV's positioning accuracy. For example, if the UAV is heading to location A while performing a rescue mission, the generated rescue execution feedback information might be: Current rescue target number: 001; Dual-light gimbal imaging quality level: High; Single BeiDou positioning and navigation system signal strength: Strong. Based on the feedback information obtained from the rescue operation, the system will generate comprehensive status information on the rescue progress, including whether the drone has reached the target location, image quality, and the effectiveness of the navigation system. For example, based on the feedback information, the system might generate the following rescue progress and status information: "Rescue target number 001 has been reached; image quality is excellent and the image can be used for analysis; navigation system signal is stable." After confirming the rescue target, the drone system needs to establish contact with the trapped personnel or on-site commander via a voice link. This step ensures timely information transmission and communication, allowing all participants to understand the current rescue status. For example, the drone might contact the on-site commander via a remote voice link to confirm: "We have reached target location 001, image quality is good, and we are conducting further reconnaissance and rescue preparations." Finally, the system will send the generated rescue progress and status information to the trapped personnel or on-site commander via a voice link to maintain transparent communication and guide subsequent rescue operations.

[0035] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made within the scope of knowledge possessed by those skilled in the art without departing from the spirit of the present invention.

Claims

1. A multi-target rescue decision-making method for unmanned aerial vehicles (UAVs), characterized in that, include: Receive rescue mission instructions from a remote command terminal, which include the number of targets to be rescued and the scope of the rescue area; The real-time location data of the intelligent drone and the initial distribution coordinate data of the target to be rescued are obtained through a single Beidou positioning and navigation system. Based on real-time location data and initial distribution coordinate data, the intelligent drone is controlled to fly to the target area and collect real-time image data of the rescue area through dual-light gimbal. The real-time image data is transmitted to the remote command terminal using high-definition real-time transmission technology, and the target confirmation information is received from the remote command terminal based on the real-time image data. A multi-target rescue priority list is generated based on target confirmation information, real-time location data, and initial distribution coordinate data. Personalized rescue execution instructions are generated for each target to be rescued based on a multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone.

2. The UAV multi-target rescue decision-making method according to claim 1, characterized in that, After transmitting real-time video data to a remote command terminal using high-definition real-time transmission technology, and receiving target confirmation information from the remote command terminal based on the real-time video data, the process also includes: If packet loss or interruption is detected in the high-definition real-time transmission link, immediately switch to the short message channel of the single Beidou positioning and navigation system to transmit critical status data. An emergency rescue strategy is generated based on key status data and the last image frame recorded by the dual-light PTZ camera. The system uses an onboard remote voice module to send out high-frequency distress signals and location broadcasts, including latitude and longitude coordinates generated by a single BeiDou positioning and navigation system.

3. The UAV multi-target rescue decision-making method according to claim 2, characterized in that, After generating a multi-target rescue priority list based on target confirmation information, real-time location data, and initial distribution coordinate data, it also includes: Multiple targets awaiting rescue were detected at the same coordinate point in the initial distribution coordinate data. The environmental carrying capacity of the coordinate point was determined based on real-time image data. If the environmental carrying capacity is greater than the preset carrying capacity, the coordinate point will be marked as a rendezvous point according to the multi-target rescue priority list. If the environmental carrying capacity is less than the preset carrying capacity, the coordinate point is divided into several sub-coordinates according to the positioning accuracy of the single Beidou positioning and navigation system, and the order of the multi-target rescue priority list is assigned to the sub-coordinates to generate a decentralized rescue path.

4. The UAV multi-target rescue decision-making method according to claim 3, characterized in that, Based on a multi-target rescue priority list, real-time image data, and the remaining battery power data of the intelligent drone, personalized rescue execution instructions are generated for each target to be rescued, including: Based on real-time image data, identify the survival status score of the target to be rescued and the level of obstacles in the surrounding environment; When the survival status score is detected to be lower than the preset score threshold and the surrounding environmental obstacle level is high, the precise positioning data of the single Beidou positioning and navigation system is retrieved to generate hovering coordinates. When the survival status score is detected to be higher than the preset score threshold and the surrounding environmental obstacle level is low, the landing coordinates are generated according to the multi-target rescue priority list. The first rescue sub-command is generated by combining the remaining power data, hovering coordinates and landing coordinates. The first rescue sub-command includes the lighting equipment activation parameters and the dual-light gimbal zoom parameters. The onboard lighting equipment is activated to provide supplemental lighting based on the first rescue sub-instruction. Pre-set guidance voice is played to the target to be rescued via the remote voice module, generating personalized rescue execution instructions for each target to be rescued.

5. The UAV multi-target rescue decision-making method according to claim 4, characterized in that, The first rescue sub-command is generated by combining remaining battery power data, hovering coordinates, and landing coordinates, including: Acquire the current attitude angle and focal length data of the dual-light gimbal; calculate the gimbal rotation angle increment based on the hovering or landing coordinates; The maximum power of lighting equipment is limited based on the remaining power data, and the bandwidth usage of high-definition real-time transmission technology is also limited. The device control parameters are generated based on the gimbal rotation angle increment, maximum power, and bandwidth usage. The first rescue sub-command includes the device control parameters and remote voice triggering conditions.

6. The UAV multi-target rescue decision-making method according to claim 5, characterized in that, After identifying the survival status score of the target to be rescued and the level of obstacles in the surrounding environment based on real-time image data, the following are also included: The presence of a low-light environment at night was detected in the real-time image data, and the distribution of heat sources was identified based on the infrared thermal imaging data of the dual-light gimbal. Adjust the optical axis direction of the airborne lighting equipment according to the distribution of heat sources to ensure that the light spot covers the main heat source areas; The lighting brightness is adjusted based on the temperature change trend of the main heat source area and the clarity of the high-definition real-time transmitted images to generate nighttime rescue mode instructions; among them, personalized rescue execution instructions include nighttime rescue mode instructions and reassurance messages based on remote voice.

7. The UAV multi-target rescue decision-making method according to claim 6, characterized in that, The method further includes: Establish a remote voice link between the rescue drone and the field command center to acquire voice interaction data and generate environmental perception information; Image feature extraction processing is performed on real-time video data to obtain first rescue feature information, and semantic recognition processing is performed on voice interaction data to obtain second rescue feature information; Based on the first rescue feature information and the second rescue feature information, the multi-target rescue decision information of the rescue execution drone is determined and sent to the rescue execution drone.

8. The UAV multi-target rescue decision-making method according to claim 7, characterized in that, Based on the first rescue characteristic information and the second rescue characteristic information, multi-target rescue decision information for the rescue execution drone is determined, including: Acquire single BeiDou positioning and navigation data of the rescue drone, where single BeiDou positioning and navigation data is the trajectory prediction information generated by the rescue drone based on its own position; Based on the first and second rescue feature information, a target priority sequence is determined; based on single BeiDou positioning and navigation data and the target priority sequence, multi-target rescue decision information for the rescue execution drone is determined.

9. A multi-target rescue decision-making method for unmanned aerial vehicles according to claim 8, characterized in that, After sending multi-target rescue decision information to the rescue drone, it also includes: Acquire rescue execution feedback information generated by the rescue execution drone based on multi-target rescue decision information; generate rescue progress and status information based on the rescue execution feedback information; wherein, the rescue execution feedback information includes the current rescue target number, the dual-optical gimbal imaging quality level, and the signal strength of the single Beidou positioning and navigation system; Once the trapped personnel or on-site commander corresponding to the target to be rescued are identified, rescue progress and status information are sent to the trapped personnel or on-site commander via a remote voice link.

Images