Flight monitoring method and system for low-altitude unmanned aerial vehicle

Abstract

The invention relates to the field of unmanned aerial vehicle monitoring, in particular to a low-altitude unmanned aerial vehicle flight monitoring method and system. A low-altitude unmanned aerial vehicle flight monitoring system comprises an obstacle sensing module, a confidence coefficient calculation module, a risk assessment module, an obstacle avoidance distance adjustment module and a flight mode switching module. According to the invention, by quantifying the rainfall interference index, the visibility index, the illumination intensity index, the obstacle complex index and the electromagnetic interference index, the confidence coefficient weight of the millimeter wave radar, the visual sensor and other multi-element sensing equipment is calculated in real time, so that the limitation of traditional fixed priority fusion is broken through; the optimal sensor data can be automatically selected as an obstacle avoidance decision basis according to actual environmental conditions, misjudgment or delayed response caused by sensor conflicts is avoided, the obstacle avoidance decision precision is improved, and the method is particularly suitable for high-reliability flight in complex environments such as urban canyons.

Description

Technical Field

[0001] The present invention relates to the field of unmanned aerial vehicle (UAV) monitoring, and in particular to a low-altitude UAV flight monitoring method and system. Background Art

[0002] With the rapid development of drone technology, low-altitude flight missions are increasingly used in logistics and transportation, agricultural plant protection, urban inspections and other fields; however, the low-altitude environment is complex and changeable, with factors such as dense building structures, dynamically moving obstacles, sudden weather changes and complex electromagnetic interference, posing severe challenges to the autonomous obstacle avoidance and flight safety of drones; traditional drone obstacle avoidance systems mostly rely on a single sensor (such as vision or ultrasound) for obstacle detection, but the reliability of such systems is significantly reduced in harsh environments; for example, rainy and foggy weather reduces the recognition ability of visual sensors, strong electromagnetic interference may cause distortion of ultrasonic signals, and there is a lack of an effective dynamic arbitration mechanism for data conflicts between different sensors, which can easily lead to misjudgments or obstacle avoidance delays.

[0003] In existing technologies, multi-sensor fusion solutions are widely used to improve the robustness of obstacle perception, but their data processing strategies usually adopt fixed weights or simple priority sorting, and do not fully consider the direct impact of dynamic environmental changes on sensor performance. For example, a sudden drop in light intensity may cause the stereo matching algorithm of binocular vision to fail, and the signal penetration capability of millimeter-wave radar in heavy rainfall environments will also be weakened. This type of static fusion method cannot adjust the decision weights of each sensor in real time, resulting in insufficient obstacle avoidance confidence in high-risk scenarios. In addition, obstacle risk assessment models are mostly constructed based on the linear relationship between obstacle distance and speed, which makes it difficult to quantify the combined effects of complex factors such as dynamic obstacle density and sudden wind speed disturbances, thereby limiting the system's ability to make refined judgments on potential collision risks. In addition, how to dynamically switch between multi-level flight modes according to the real-time risk level and coordinately adjust the sensor sampling frequency and flight speed remains a technical difficulty that needs to be solved urgently.

[0004] In order to solve the above problems, the present invention proposes a low-altitude UAV flight monitoring method and system, which can integrate multi-dimensional environmental characteristics, evaluate sensor confidence in real time, dynamically adjust obstacle avoidance strategies and support multi-level risk response, becoming a key technical direction for improving the autonomy and safety of low-altitude UAVs. Summary of the Invention

[0005] The present invention quantifies the rainfall interference index, visibility index, light intensity index, obstacle complexity index and electromagnetic interference index, and calculates the confidence weights of multiple sensing devices such as millimeter-wave radar and visual sensors in real time. It breaks through the limitations of traditional fixed-priority fusion and can automatically select the optimal sensor data as the basis for obstacle avoidance decision-making according to actual environmental conditions, avoiding misjudgment or delayed response due to sensor conflicts, and improving the accuracy of obstacle avoidance decisions. It is particularly suitable for high-reliability flight in complex environments such as urban canyons.

[0006] A low-altitude UAV flight monitoring method, comprising: During low-altitude autonomous flight, the drone uses millimeter-wave radar, binocular vision sensors, lidar, and ultrasonic sensors to perceive obstacle information in real time. The obstacle information includes the number of obstacles within the drone's dynamic obstacle avoidance range, the obstacle status, and the distance and approach speed of each obstacle to the drone. At the current environmental monitoring time point, the system collects environmental information about the drone. Using this information, along with the number and status of obstacles, it builds a five-dimensional environmental feature vector, including the rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index. The five-dimensional environmental feature vector is used to calculate the confidence level of each sensor. Then, before the next environmental monitoring time point, the system uses the conflicting data between sensors and selects the data with the highest confidence level as the final decision value for the drone's automatic obstacle avoidance. The current wind speed is obtained at the current environmental monitoring time point, and the risk assessment of the drone is performed based on obstacle information. The confidence level of each sensor, the current wind speed, and the visibility index are used to adjust the dynamic obstacle avoidance distance of the drone. Set up multi-level flight modes for drones and switch flight modes based on risk assessment results.

[0007] Preferably, a five-dimensional environmental feature vector is established using environmental information as well as the number and state of obstacles. The specific operations are as follows: Environmental information includes rainfall, visibility, light intensity, and electromagnetic interference intensity. The maximum rainfall, maximum visibility, maximum light intensity, and maximum electromagnetic interference intensity are set respectively. Then, the rainfall, visibility, light intensity, and electromagnetic interference intensity collected at the current environmental monitoring time point are normalized. The normalized values ​​of each environmental information are used as the rainfall interference index, visibility index, light intensity index, and electromagnetic interference index respectively. Set the maximum number of obstacles based on the current number of obstacles and the states of each obstacle, where the obstacle states include static and dynamic. Calculate the ratio of the number of obstacles to the maximum number of obstacles as the first ratio. Also calculate the ratio of the number of dynamic obstacles to the total number of obstacles among all current obstacles as the second ratio. Then, calculate the average of the first and second ratios as the obstacle complexity index.

[0008] Preferably, the confidence level of each sensor is calculated using the five-dimensional environmental feature vector, and the specific operation is as follows: Get the environmental sensitivity of each sensor to each environmental feature, where the environmental sensitivity is between 0 and 1; For any one of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor, the environmental sensitivity of the sensor to various environmental features is used as a weight, and a weighted average is performed on the rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index. The result of the weighted average is the interference degree value of the sensor. The confidence level of the sensor is calculated by subtracting the interference degree value from the full confidence score of 1.

[0009] Preferably, a risk assessment is performed on the drone, and the specific operations are as follows: Set the maximum approach speed. For any obstacle around the current drone, calculate the ratio of the obstacle's approach speed to the drone's maximum approach speed. Also calculate the difference between the obstacle's distance and the dynamic obstacle avoidance distance. Then calculate the ratio of this difference to the dynamic obstacle avoidance distance. Multiply the two ratios as the individual risk index for the obstacle. The largest single risk index among all obstacles is taken as the representative single risk index; Set the maximum wind speed, apply the maximum wind speed to normalize the current wind speed, and obtain the normalized wind speed value; The average value of the single risk index, normalized wind speed value and obstacle complexity index is taken as the current risk value of the drone.

[0010] Preferably, the dynamic obstacle avoidance distance of the drone is adjusted, and the specific operations are as follows: Calculate the average confidence of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor; then calculate the difference between the average confidence and 1 as the first reference value for adjusting the dynamic obstacle avoidance distance; use the normalized wind speed value as the second reference value for adjusting the dynamic obstacle avoidance distance; and use the difference between the visibility index and 1 as the third reference value for adjusting the dynamic obstacle avoidance distance; Calculate the average of the first reference value, the second reference value, and the third reference value to obtain the adjustment coefficient of the dynamic obstacle avoidance distance, where the adjustment coefficient is within a range of 0-1. Set the maximum obstacle avoidance distance and the minimum obstacle avoidance distance of the dynamic obstacle avoidance distance, and linearly map the adjustment coefficient between the minimum obstacle avoidance distance and the maximum obstacle avoidance distance. When the adjustment coefficient is 0, the minimum obstacle avoidance distance is taken, and when the adjustment coefficient is 1, the maximum obstacle avoidance distance is taken. The mapping result is the adjusted dynamic obstacle avoidance distance.

[0011] Preferably, a multi-level flight mode is set for the drone, and the flight mode of the drone is switched based on the risk assessment results. The specific operations are as follows: The drone’s multi-level flight modes include efficiency mode, safety mode, and emergency mode; Set a low-risk threshold and a high-risk threshold for the risk value of the drone. If the risk value calculated by the drone for m consecutive times is lower than the low-risk threshold, the drone enters the efficiency mode. If the drone's current risk value is between the low risk threshold and the high risk threshold, the drone switches to safe mode; If the drone's current risk value is higher than the high risk threshold, the drone switches to emergency mode.

[0012] A low-altitude UAV flight monitoring system, comprising: Obstacle perception module, which uses millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor to perceive obstacle information in real time; The confidence calculation module includes an environmental information acquisition unit and a confidence calculation unit. The environmental information acquisition unit is used to collect environmental information of the UAV at the current environmental monitoring time point and use the environmental information as well as the number and status of obstacles to establish a five-dimensional environmental feature vector. The confidence calculation unit is used to calculate the confidence of each sensor using the five-dimensional environmental feature vector. Then, before the next environmental monitoring time point, the data with the highest confidence among the conflicting data between the sensors is selected as the final decision value for the UAV's automatic obstacle avoidance. The risk assessment module is used to obtain the current wind speed at the current environmental monitoring time point and perform risk assessment on the drone based on obstacle information; The obstacle avoidance distance adjustment module is used to adjust the dynamic obstacle avoidance distance of the drone through the confidence level of each sensor, current wind speed and visibility index; The flight mode switching module is used to switch the UAV flight mode based on the risk assessment results.

[0013] The present invention has the following advantages: 1. This invention quantifies the rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index, and calculates the confidence weights of multiple sensing devices such as millimeter-wave radar and visual sensors in real time. This overcomes the limitations of traditional fixed-priority fusion and can automatically select the optimal sensor data as the basis for obstacle avoidance decisions based on actual environmental conditions. This avoids misjudgments or delayed responses caused by sensor conflicts, improves the accuracy of obstacle avoidance decisions, and is particularly suitable for high-reliability flight in complex environments such as urban canyons.

[0014] 2. The present invention constructs a data-driven dynamic obstacle avoidance distance adjustment algorithm and a multi-level flight mode switching strategy through the linkage analysis of sensor confidence, wind speed influencing factors and visibility parameters. The system can intelligently adjust the obstacle avoidance distance in real time to achieve simultaneous optimization of flight efficiency and obstacle avoidance safety. At the same time, based on the three-level response architecture of "efficiency mode-safety mode-emergency mode", the UAV can complete operational tasks at high speed in low-risk scenarios and automatically switch to speed limit or forced landing mode in high-risk scenarios, avoiding the energy waste or insufficient response problems caused by the single control logic of traditional solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a schematic diagram of the structure of the low-altitude UAV flight monitoring system used in an embodiment of the present invention. DETAILED DESCRIPTION

[0016] In order to enable those skilled in the art to better understand the technical solutions in the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0017] Example 1, a low-altitude UAV flight monitoring method, comprising: During autonomous low-altitude flight, the drone uses millimeter-wave radar, binocular vision sensors, lidar, and ultrasonic sensors to perceive obstacle information in real time. This information includes the number of obstacles within the drone's dynamic obstacle avoidance range, the obstacle status, and the distance and approach speed of each obstacle to the drone. By acquiring complementary multi-source environmental data, it provides accurate dynamic obstacle distribution characteristics for subsequent real-time obstacle avoidance decisions, thereby supporting dynamic optimization of flight paths, probability prediction of collision risks, and generation of emergency avoidance trajectories, ensuring the drone's high reliability and adaptive obstacle avoidance capabilities in complex low-altitude environments. At the current environmental monitoring time point, the environmental information of the drone is collected, and the five-dimensional environmental feature vector is established using the environmental information as well as the number and status of obstacles, including rainfall interference index, visibility index, light intensity index, obstacle complexity index and electromagnetic interference index, to quantify the impact of the current environment on the performance of multiple sensors in real time; the confidence of each sensor is calculated using the five-dimensional environmental feature vector; then, before the next environmental monitoring time point, for the conflict data between each sensor, the data with the highest confidence is taken as the final decision value for the drone's automatic obstacle avoidance; during the continuous monitoring cycle, the system uses priority dynamic calibration technology to calibrate the sensor conflict data. For example, if there is a conflict in distance measurement between binocular vision and ultrasonic sensors, the output of the sensor with the highest confidence is always selected as the final obstacle avoidance decision value. This mechanism: 1. Eliminates the bias of traditional fixed priority strategies in complex environments, such as avoiding over-reliance on failed visual data under strong light interference; 2. Dynamically weights the interference of low-reliability sensors and prioritizes the optimal perception results in the current environment, thereby improving the real-time and accuracy of obstacle avoidance decisions; 3. Confidence arbitration replaces simple data fusion, reducing algorithm complexity and avoiding the risk of multi-sensor error superposition. This is particularly suitable for low-altitude flight scenarios with intersecting dynamic obstacles. The current wind speed is obtained at the current environmental monitoring time point, and the drone's risk assessment is performed based on obstacle information. The drone's dynamic obstacle avoidance distance is adjusted using the confidence level of each sensor, the current wind speed, and the visibility index. The adaptive adjustment of the dynamic obstacle avoidance distance essentially balances environmental adaptability with safety redundancy. When sensor confidence is high (low environmental interference), wind speed is low (stable flight attitude), and visibility is good, the obstacle avoidance monitoring range is narrowed (reducing redundant computing overhead), leveraging high-precision sensing capabilities for short-range rapid response. Conversely, in adverse conditions (such as rain and fog that degrade sensor performance, strong wind disturbances, or electromagnetic noise), the monitoring range is expanded to avoid collision risks caused by sudden obstacles (such as birds or drones flying across the sky) or perception delays. This strategy, through dual-mode switching between "precision and robustness," optimizes flight path planning efficiency and enhances obstacle avoidance tolerance in extreme scenarios. Set up multi-level flight modes for drones and switch flight modes based on risk assessment results.

[0018] Use environmental information, the number of obstacles, and the obstacle status to establish a five-dimensional environmental feature vector. The specific operations are as follows: Environmental information includes rainfall, visibility, light intensity and electromagnetic interference intensity; set the maximum rainfall, maximum visibility, maximum light intensity and maximum electromagnetic interference intensity respectively, and then normalize the rainfall, visibility, light intensity and electromagnetic interference intensity collected at the current environmental monitoring time point, and use the normalized values ​​of each environmental information as the rainfall interference index, visibility index, light intensity index and electromagnetic interference index respectively; the setting of maximum rainfall, maximum visibility, maximum light intensity and maximum electromagnetic interference intensity requires comprehensive physical extreme values, sensor performance boundaries and real scene extreme conditions, and its rules follow the third-order logic: first, design the global meteorological and electromagnetic spectrum range based on the physical characteristics of the environment For example, the maximum rainfall can be based on the meteorological definition of extreme rainstorm standard of 50mm / h, and the maximum visibility can refer to the sensor's effective ranging limit of 1km. Secondly, it adapts to the upper tolerance limit of the drone sensor hardware, such as the 50,000 lux light intensity corresponding to the camera sensor overexposure threshold, and the electromagnetic interference intensity is limited to 100V / m allowed in the equipment electromagnetic compatibility test. Finally, dynamic calibration is performed according to actual flight mission requirements. For example, the electromagnetic interference requirement may be relaxed to 120V / m for urban inspection requirements, while the maximum visibility in polar scientific research scenarios can be increased to 5km to cover snow and fog conditions. This achieves a standardized environmental parameter system that meets both theoretical constraints and is compatible with application flexibility. Set the maximum number of obstacles based on the current number of obstacles and the states of each obstacle, where the obstacle states include static and dynamic. Calculate the ratio of the number of obstacles to the maximum number of obstacles as the first ratio. Also calculate the ratio of the number of dynamic obstacles to the total number of obstacles among all current obstacles as the second ratio. Then, calculate the average of the first and second ratios as the obstacle complexity index.

[0019] The confidence of each sensor is calculated using the five-dimensional environmental feature vector. The specific operations are as follows: Obtain the environmental sensitivity of each sensor to various environmental features, which is between 0 and 1. The calibration data of environmental sensitivity comes from experts combining laboratory tests with historical scenario data, and establishing a sensor performance degradation model through multi-condition physical simulation. For example, the rainfall sensitivity of millimeter-wave radar is tested in a water mist simulation chamber to measure the signal attenuation rate under different rainfall amounts, and quantify the negative correlation between the reflected wave signal-to-noise ratio and rainfall intensity. The illumination sensitivity of binocular vision uses an adjustable light box to measure the stereo matching accuracy under different illumination levels (such as 100 lux to 10,000 lux) to calibrate low illumination. The exponential decline in visual confidence during the observation period was investigated; the electromagnetic interference sensitivity of lidar was tested by injecting electromagnetic noise of varying intensities into an electromagnetic shielding room and analyzing the linear threshold relationship between point cloud distortion rate and noise intensity; the visibility sensitivity of ultrasonic waves was tested based on sound wave penetration experiments in haze environments, and the nonlinear effect of large particle concentration on ultrasonic ranging error was statistically analyzed; experts used regression analysis or numerical fitting methods based on physical constraints (such as nonlinear least squares) to ultimately map the functional curves of environmental characteristic parameters and sensor performance, and extracted the sensitivity coefficient matrix as the underlying parameter library for multi-sensor dynamic arbitration; For any one of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor, the environmental sensitivity of the sensor to each environmental feature is used as a weight. The rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index are weighted averaged. The weighted average result is the interference level value of the sensor. The confidence level of the sensor is calculated by subtracting the interference level value from the full confidence score of 1. The confidence calculation formula is as follows: Among them, i=1, 2, 3, 4; respectively represent millimeter wave radar, binocular vision sensor, lidar and ultrasonic sensor, is the confidence of the i-th sensor; j = 1, 2, 3, 4, 5; respectively represent the rainfall interference index, visibility index, light intensity index, obstacle complexity index and electromagnetic interference index, is the environmental sensitivity of the i-th sensor to the j-th environmental feature, arrive They are rainfall interference index, visibility index, light intensity index, obstacle complexity index and electromagnetic interference index.

[0020] Conduct a risk assessment on drones. The specific steps are as follows: Set the maximum closing speed, which represents the maximum relative closing speed allowed by the drone in the flight control system, for example, 15m / s. For any obstacle currently surrounding the drone, calculate the ratio of the obstacle's closing speed relative to the drone's maximum closing speed. Also calculate the difference between the obstacle's distance and the dynamic obstacle avoidance distance. Then calculate the ratio of this difference to the dynamic obstacle avoidance distance. The product of these two ratios is used as the individual risk index for the obstacle. The largest single obstacle risk index among all obstacles is used as the representative single obstacle risk index. Following the "water bucket effect" and the principle of minimum safety margin, only the high risk of one key obstacle is needed to trigger the system's obstacle avoidance response, thus avoiding missed detections due to risk averaging. Set the maximum wind speed, apply the maximum wind speed to normalize the current wind speed, and obtain the normalized wind speed value; The average of the individual risk index, normalized wind speed value, and obstacle complexity index is used as the current risk value of the drone. This model has clear parameter interpretation: the product relationship between speed and distance strengthens the collaborative risk in "high-speed, close-range" scenarios, the maximum value screening mechanism reflects the system's focused defense against extreme working conditions, and the multi-indicator average aggregation balances the weights between flight dynamics (wind speed), safety situation (obstacle risk), and environmental complexity.

[0021] Adjust the dynamic obstacle avoidance distance of the drone. The specific operations are as follows: Calculate the average confidence of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor; then calculate the difference between the average confidence and 1 as the first reference value for adjusting the dynamic obstacle avoidance distance; use the normalized wind speed value as the second reference value for adjusting the dynamic obstacle avoidance distance; and use the difference between the visibility index and 1 as the third reference value for adjusting the dynamic obstacle avoidance distance; Calculate the average of the first reference value, the second reference value, and the third reference value to obtain the adjustment coefficient of the dynamic obstacle avoidance distance, where the adjustment coefficient is within a range of 0-1. Set the maximum obstacle avoidance distance and the minimum obstacle avoidance distance of the dynamic obstacle avoidance distance, and linearly map the adjustment coefficient between the minimum obstacle avoidance distance and the maximum obstacle avoidance distance. When the adjustment coefficient is 0, the minimum obstacle avoidance distance is taken, and when the adjustment coefficient is 1, the maximum obstacle avoidance distance is taken. The mapping result is the adjusted dynamic obstacle avoidance distance.

[0022] Set up multiple flight modes for the drone and switch the flight mode based on the risk assessment results. The specific operations are as follows: The drone’s multi-level flight modes include efficiency mode, safety mode, and emergency mode; Set low-risk thresholds and high-risk thresholds for the drone's risk value. If the drone's risk value is lower than the low-risk threshold for m consecutive times (which can be set to 3-5 times), the drone enters efficiency mode, flying at high speed according to the preset high-speed cruise parameters, and reduces the sampling frequency of each sensor to improve endurance by reducing data redundancy. If the drone's current risk value is between the low-risk threshold and the high-risk threshold, the drone switches to safe mode, flies at a limited speed, and restores the baseline sampling frequency of each sensor to ensure perception continuity; If the drone's current risk value is higher than the high-risk threshold, the drone switches to emergency mode and hovers immediately. If the risk value calculated for m consecutive times after hovering is higher than the high-risk threshold (to prevent false triggering due to brief interference), the drone automatically lands in the nearest safe area.

[0023] A low-altitude UAV flight monitoring system, such as Figure 1 Shown, including: The obstacle perception module uses millimeter-wave radar, binocular vision sensors, lidar, and ultrasonic sensors to sense obstacles in real time. The obstacle information includes the number of obstacles within the UAV's dynamic obstacle avoidance range, the obstacle status, and the distance and approach speed of each obstacle to the UAV. The confidence calculation module includes an environmental information acquisition unit and a confidence calculation unit. The environmental information acquisition unit is used to collect environmental information of the UAV at the current environmental monitoring time point, and use the environmental information as well as the number and status of obstacles to establish a five-dimensional environmental feature vector, including a rainfall interference index, a visibility index, a light intensity index, an obstacle complexity index, and an electromagnetic interference index. The confidence calculation unit is used to calculate the confidence of each sensor using the five-dimensional environmental feature vector. Subsequently, before the next environmental monitoring time point, the data with the highest confidence among the conflicting data between the sensors is taken as the final decision value for the UAV's automatic obstacle avoidance. The risk assessment module is used to obtain the current wind speed at the current environmental monitoring time point and perform risk assessment on the drone based on obstacle information; The obstacle avoidance distance adjustment module is used to adjust the dynamic obstacle avoidance distance of the drone through the confidence level of each sensor, current wind speed and visibility index; The flight mode switching module is used to switch the UAV flight mode based on the risk assessment results.

[0024] It should be understood that those skilled in the art may make improvements or modifications based on the above description, and all such improvements and modifications shall fall within the scope of protection of the appended claims. Any portion of this specification not described in detail is prior art known to those skilled in the art.

Claims

1. A low-altitude UAV flight monitoring method, characterized in that: include: During low-altitude autonomous flight, the drone uses millimeter-wave radar, binocular vision sensors, lidar, and ultrasonic sensors to perceive obstacle information in real time. The obstacle information includes the number of obstacles within the drone's dynamic obstacle avoidance range, the obstacle status, and the distance and approach speed of each obstacle to the drone. At the current environmental monitoring time point, the system collects environmental information about the drone. Using this information, along with the number and status of obstacles, it builds a five-dimensional environmental feature vector, including the rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index. The five-dimensional environmental feature vector is used to calculate the confidence level of each sensor. Then, before the next environmental monitoring time point, the system uses the conflicting data between sensors and selects the data with the highest confidence level as the final decision value for the drone's automatic obstacle avoidance. The current wind speed is obtained at the current environmental monitoring time point, and the risk assessment of the drone is performed based on obstacle information. The confidence level of each sensor, the current wind speed, and the visibility index are used to adjust the dynamic obstacle avoidance distance of the drone. Set up multi-level flight modes for drones and switch flight modes based on risk assessment results.

2. A low-altitude UAV flight monitoring method according to claim 1, characterized in that: Use environmental information, the number of obstacles, and the obstacle status to establish a five-dimensional environmental feature vector. The specific operations are as follows: Environmental information includes rainfall, visibility, light intensity, and electromagnetic interference intensity. The maximum rainfall, maximum visibility, maximum light intensity, and maximum electromagnetic interference intensity are set respectively. Then, the rainfall, visibility, light intensity, and electromagnetic interference intensity collected at the current environmental monitoring time point are normalized. The normalized values ​​of each environmental information are used as the rainfall interference index, visibility index, light intensity index, and electromagnetic interference index respectively. Set the maximum number of obstacles based on the current number of obstacles and the status of each obstacle, where the obstacle status includes static and dynamic; The ratio of the number of obstacles to the maximum number of obstacles is calculated as the first ratio. The ratio of the number of dynamic obstacles to all obstacles among all current obstacles is calculated as the second ratio. The average of the first and second ratios is then calculated as the obstacle complexity index.

3. A low-altitude UAV flight monitoring method according to claim 2, characterized in that: The confidence of each sensor is calculated using the five-dimensional environmental feature vector. The specific operations are as follows: Get the environmental sensitivity of each sensor to each environmental feature, where the environmental sensitivity is between 0 and 1; For any one of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor, the environmental sensitivity of the sensor to various environmental features is used as a weight, and a weighted average is performed on the rainfall interference index, visibility index, light intensity index, obstacle complexity index, and electromagnetic interference index. The result of the weighted average is the interference degree value of the sensor. The confidence level of the sensor is calculated by subtracting the interference degree value from the full confidence score of 1.

4. A low-altitude UAV flight monitoring method according to claim 3, characterized in that: Conduct a risk assessment on drones. The specific steps are as follows: Set the maximum approach speed. For any obstacle around the current drone, calculate the ratio of the obstacle's approach speed to the drone's maximum approach speed. Also calculate the difference between the obstacle's distance and the dynamic obstacle avoidance distance. Then calculate the ratio of this difference to the dynamic obstacle avoidance distance. Multiply the two ratios as the individual risk index for the obstacle. The largest single risk index among all obstacles is taken as the representative single risk index; Set the maximum wind speed, apply the maximum wind speed to normalize the current wind speed, and obtain the normalized wind speed value; The average value of the single risk index, normalized wind speed value and obstacle complexity index is taken as the current risk value of the drone.

5. A low-altitude UAV flight monitoring method according to claim 4, characterized in that: Adjust the dynamic obstacle avoidance distance of the drone. The specific operations are as follows: Calculate the average confidence of the millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor; then calculate the difference between the average confidence and 1 as the first reference value for adjusting the dynamic obstacle avoidance distance; use the normalized wind speed value as the second reference value for adjusting the dynamic obstacle avoidance distance; and use the difference between the visibility index and 1 as the third reference value for adjusting the dynamic obstacle avoidance distance; Calculate the average of the first reference value, the second reference value, and the third reference value to obtain the adjustment coefficient of the dynamic obstacle avoidance distance, where the adjustment coefficient is within a range of 0-1. Set the maximum obstacle avoidance distance and the minimum obstacle avoidance distance of the dynamic obstacle avoidance distance, and linearly map the adjustment coefficient between the minimum obstacle avoidance distance and the maximum obstacle avoidance distance. When the adjustment coefficient is 0, the minimum obstacle avoidance distance is taken, and when the adjustment coefficient is 1, the maximum obstacle avoidance distance is taken. The mapping result is the adjusted dynamic obstacle avoidance distance.

6. A low-altitude UAV flight monitoring method according to claim 5, characterized in that: Set up multiple flight modes for the drone and switch the flight mode based on the risk assessment results. The specific operations are as follows: The drone’s multi-level flight modes include efficiency mode, safety mode, and emergency mode; Set a low-risk threshold and a high-risk threshold for the risk value of the drone. If the risk value calculated by the drone for m consecutive times is lower than the low-risk threshold, the drone enters the efficiency mode. If the drone's current risk value is between the low risk threshold and the high risk threshold, the drone switches to safe mode; If the drone's current risk value is higher than the high risk threshold, the drone switches to emergency mode.

7. A low-altitude UAV flight monitoring system, characterized in that: The system is applied to the low-altitude UAV flight monitoring method according to any one of claims 1 to 6, comprising: Obstacle perception module, which uses millimeter-wave radar, binocular vision sensor, lidar, and ultrasonic sensor to perceive obstacle information in real time; The confidence calculation module includes an environmental information acquisition unit and a confidence calculation unit. The environmental information acquisition unit is used to collect environmental information of the UAV at the current environmental monitoring time point and use the environmental information as well as the number and status of obstacles to establish a five-dimensional environmental feature vector. The confidence calculation unit is used to calculate the confidence of each sensor using the five-dimensional environmental feature vector. Then, before the next environmental monitoring time point, the data with the highest confidence among the conflicting data between the sensors is selected as the final decision value for the UAV's automatic obstacle avoidance. The risk assessment module is used to obtain the current wind speed at the current environmental monitoring time point and perform risk assessment on the drone based on obstacle information; The obstacle avoidance distance adjustment module is used to adjust the dynamic obstacle avoidance distance of the drone through the confidence level of each sensor, current wind speed and visibility index; The flight mode switching module is used to switch the UAV flight mode based on the risk assessment results.

Images