Low-altitude security patrol unmanned aerial vehicle system based on millimeter wave radar

Abstract

The application discloses a kind of low altitude security patrol unmanned aerial vehicle systems based on millimeter wave radar, it is related to low altitude security patrol unmanned aerial vehicle technology, including multi-rotor flight platform, airborne millimeter wave radar perception, airborne edge computing processing, adaptive track regulation and control, air-ground two-way communication, airborne security response unit and ground command control unit, radar carries stabilizing gimbal with high-precision detection and online self-calibration ability, edge computing unit can be self-adaptively to cancel clutter, quickly identify low, slow and small target and security hidden danger, system can dynamically regulate track, autonomously avoid obstacle, support local autonomous operation when communication link breaks, ground unit can realize multi-machine intelligent scheduling, multi-source data fusion and no-gap patrol coverage.The application solves the problems of traditional optical detection being easily disturbed by environment, ordinary radar having high false alarm rate and patrol having blind area and operation interruption, can stably operate in bad weather and night, realizes target accurate identification, hierarchical disposal and global situation monitoring.

Description

Technical Field

[0001] This invention relates to the field of low-altitude security patrol drone technology, and in particular to a low-altitude security patrol drone system based on millimeter-wave radar. Background Technology

[0002] In recent years, with the gradual liberalization of low-altitude airspace control, the demand for low-altitude security in scenarios such as airport airspace, urban core areas, major event sites, and border control zones has continued to grow. Problems such as illegal take-off of low-speed and small drones, intrusion of floating objects, and failure to detect ground security risks have occurred frequently, posing a serious threat to public safety and production and operation order. The industry urgently needs a low-altitude inspection solution that can achieve all-time and all-scenario coverage, with the core requirements of accurate identification of small targets, timely dynamic response, and stable and reliable operation.

[0003] Current mainstream low-altitude inspection solutions mostly use drone systems equipped with visible light and infrared cameras to collect images of the inspection area through optical imaging. Security risks are then identified by manual or AI recognition algorithms. This type of solution has the advantages of low deployment cost and intuitive recognition results, and has been applied in some security scenarios. However, optical detection is easily interfered with by environmental factors such as rain, fog, strong light, and darkness at night. It has a very short detection range for low-altitude weak targets with extremely small radar cross-sections, and it cannot directly obtain the target's motion parameters. The recognition response delay is high, making it difficult to meet the real-time handling needs of dynamic risks.

[0004] Some solutions employ inspection drones equipped with ordinary millimeter-wave radar, relying on radar ranging and velocity measurement capabilities to achieve all-weather, 24 / 7 detection, thus compensating to some extent for the shortcomings of optical detection. However, these solutions lack clutter suppression algorithms optimized for airborne scenarios, resulting in severe interference from rotor micro-motion clutter and clutter from tall buildings and ground objects, leading to a persistently high false alarm rate. Flight path adjustments are limited to target tracking alone, increasing the risk of collisions. Operational capability is lost upon communication link interruption, and multi-drone relay operations create management gaps. Equipment calibration requires manual disassembly and adjustment, making long-term operational accuracy difficult to guarantee. These solutions are unsuitable for demanding, routine low-altitude security management scenarios. Existing technologies cannot simultaneously achieve detection accuracy, operational reliability, and continuous management capabilities, failing to meet the needs of low-altitude security inspections in complex scenarios. Therefore, a specifically optimized system solution is urgently needed to fill this gap. Summary of the Invention

[0005] This invention proposes a low-altitude security patrol drone system based on millimeter-wave radar to solve the problems mentioned in the prior art.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a low-altitude security patrol drone system based on millimeter-wave radar, comprising a multi-rotor flight platform, an airborne millimeter-wave radar sensing unit, an airborne edge computing processing unit, an adaptive trajectory control unit, an air-to-ground two-way communication unit, a ground command and control unit, and an airborne security response unit. The multi-rotor flight platform adopts a carbon fiber integrated frame, combined with redundant power units and low rotor electromagnetic interference design, to provide an electromagnetically compatible installation environment for each airborne unit and flight power that meets the requirements of long endurance. It supports constant altitude and variable altitude cruise in the designated core low-altitude security control area according to the pre-set inspection route under level 6 wind conditions, and all cruise parameters support millisecond-level real-time dynamic adjustment. The airborne millimeter-wave radar sensing unit adopts a 4-transmit, 8-receive wide-beam patch array antenna architecture. The antenna beam has an elevation coverage of no less than 120 degrees and an azimuth coverage of no less than 100 degrees. It continuously transmits 77GHz frequency-modulated continuous wave signals at a fixed frame frequency of 20Hz, receives target echo signals within the circumferential coverage area, and simultaneously completes echo analog-to-digital conversion, preliminary processing of range and velocity dimension fast Fourier transform, and collects raw point cloud data of target range, radial velocity, azimuth angle, and elevation angle. It also completes initial filtering of strong reflection fixed ground clutter through fixed threshold filtering. The airborne edge computing processing unit is equipped with an embedded neural network acceleration chip. It performs adaptive clutter cancellation, density clustering, multi-frame trajectory association, and feature matching on the pre-processed radar point cloud data. It identifies the type, real-time location, and motion parameters of intruding low, slow, and small flying targets, illegal floating objects in the airspace, and ground security hazards. It calculates the target hazard confidence level by combining the target size, motion speed, and intrusion path, and outputs the graded hazard level judgment result. The adaptive trajectory control unit optimizes target tracking stability, flight safety, and inspection coverage integrity. It dynamically calculates the optimal cruising altitude, flight speed, and observation attitude adjustment of the multi-rotor flight platform, guiding the millimeter-wave radar sensing unit to continuously lock onto and track the target in the potential danger area. The entire adjustment process ensures that the flight platform's attitude remains stable without violent shaking. The air-to-ground two-way communication unit adopts an adaptive frequency-hopping broadband data transmission link, which dynamically adjusts the frequency hopping rate and transmission power according to the link signal-to-noise ratio. It transmits the sensing data, processing results, and equipment status information collected by the airborne terminal back to the ground command and control unit in real time in encrypted frame format. After completing the command verification, it forwards the control commands issued by the ground to the corresponding airborne unit. The airborne security response unit matches the hierarchical preset handling strategy according to the level of hidden danger, and performs close-range target directional driving and on-site multi-dimensional evidence collection operations while maintaining a safe tracking distance. The ground command and control unit completes the overall scheduling of multiple inspection drones, unified display of airspace situation, and automatic archiving of operation data. It also supports the issuance of emergency response instructions and the retrospective of the response process under manual intervention mode.

[0007] Furthermore, the airborne millimeter-wave radar sensing unit is integrated and installed on a two-axis stabilization gimbal below the nose of the multi-rotor flight platform. The gimbal adopts a brushless direct-drive motor and a composite vibration damping structure, which can suppress the vibration angle of the radar mounting surface caused by the vibration of the flight platform's own rotor and airflow disturbance to within 0.05 degrees. The gimbal can complete stepless rotation adjustment within the range of 0 degrees to 360 degrees horizontally and -90 degrees to +30 degrees pitch according to the attitude commands of the adaptive trajectory control unit, with a rotation response delay of no more than 20 milliseconds. The radar module adopts a fully digital intermediate frequency architecture, with a working frequency band covering 76 to 81 GHz, a maximum frequency modulation bandwidth supporting 4 GHz, a range resolution better than 0.2 meters, and a velocity resolution better than 0.05 meters per second. It can stably detect low-altitude weak targets such as UAVs and Kongming lanterns with a radar cross-section of not less than 0.001 square meters, and a maximum effective detection range of not less than 3 kilometers, meeting the long-range early warning requirements for low-altitude security.

[0008] Furthermore, the airborne edge computing processing unit incorporates an adaptive moving target display filtering module. It employs an improved minimum mean square error iteration criterion to update the filtering weights in real time, adaptively canceling ground clutter, rotor micro-motion clutter from the flight platform itself, and micro-Doppler clutter caused by gimbal vibration in the echo signal. The filtering weight update formula is as follows: in, This represents the filter weight vector within the k-th coherent processing interval. The iteration step size factor is dynamically determined based on the current clutter intensity to balance the convergence speed and steady-state error. The residual between the filtered output and the desired target signal. This represents the echo sampling vector within the current coherent processing interval. The corresponding sampling vector is the conjugate transpose; the filtering module constructs a clutter reference signal by collecting real-time data on the rotor speed and gimbal attitude of the flight platform. It can complete clutter cancellation without relying on pre-stored external clutter maps, and the clutter suppression ratio is not less than 60dB, which can reduce the false alarm rate caused by rotor micro-motion clutter by two orders of magnitude.

[0009] Furthermore, the airborne edge computing processing unit employs an improved density-based noisy spatial clustering algorithm to cluster the clutter-filtered point cloud. During the clustering process, the neighborhood radius threshold is adaptively adjusted based on the radial velocity difference between point clouds to avoid misclustering adjacent point clouds with large velocity differences into the same target. Combining the cluster center position changes of 10 consecutive frames, a uniform motion model is used to construct the target motion trajectory. Kalman filtering is used to smooth the trajectory and remove outliers caused by point cloud outliers. The extracted target size, average radar cross section, and motion velocity features are matched with a pre-set illegal target feature library using cosine similarity to complete target type identification. The identification response delay does not exceed 80 milliseconds, meeting the real-time detection requirements of low-altitude high-speed dynamic targets.

[0010] Furthermore, the adaptive trajectory control unit incorporates a dynamic obstacle avoidance submodule and a multi-target trajectory optimization module. It integrates millimeter-wave radar point cloud data with differential satellite positioning, inertial measurement unit, and magnetometer sensor data from the flight platform in real time. The unit calculates the comprehensive cost of trajectory adjustment based on target tracking stability, flight safety, and inspection coverage integrity. The comprehensive cost calculation formula is as follows: in, The lower the overall cost of trajectory adjustment, the better the overall performance of the trajectory scheme. , , The weight coefficients corresponding to the optimization objective are summed to 1, and are dynamically adjusted according to the current hazard level. The higher the hazard level, the greater the weight α corresponding to the target tracking stability. To adjust the angular deviation between the radar beam axis and the target direction, For radar half-power beamwidth, To adjust the minimum distance between the flight platform and surrounding obstacles, To preset a safe flight distance threshold, To adjust the distance between the adjusted route and the preset inspection route, The current inspection segment is the total length; the module iterates through all feasible flight attitude adjustment amounts, selects the adjustment scheme with the lowest overall cost, and outputs flight control commands. In the process of adjusting the flight path to track potential targets, it automatically avoids overhead cables, tall trees, and building protrusions in the airspace, so that the flight platform always maintains a distance of no less than the safety threshold from various obstacles.

[0011] Furthermore, the air-to-ground bidirectional communication unit performs link quality monitoring every 10 milliseconds, comprehensively evaluating link quality by measuring received signal strength, signal-to-noise ratio, and packet loss rate. When the link quality is lower than a preset threshold and reliable data interaction cannot be completed for three consecutive communication cycles, the local autonomous handling mode of the airborne edge computing processing unit is automatically triggered. Hazard identification, tracking, and response operations are completed sequentially according to the priority of hazard level. During the link loss, all operation data is encrypted and stored in the airborne non-volatile memory. After the link is restored, the breakpoint resume transmission mechanism is used to automatically retransmit all operation data stored during the link loss. The retransmission process uses a lossless data compression algorithm to reduce transmission bandwidth usage and avoid inspection operation interruptions and data loss.

[0012] Furthermore, the airborne security response unit integrates a 4K high-definition visible light zoom evidence-gathering camera, a directional acoustic wave decoy module, and a high-frequency high-intensity light warning module. All modules maintain unified installation coordinates with the millimeter-wave radar sensing unit, completing the spatial mapping between the detection coordinate system and the operational coordinate system. It automatically matches the corresponding operational mode for different levels of security risks: for illegal low-speed small flying targets intruding into the core control area, a directional acoustic wave decoy is initiated while maintaining a safe tracking distance. The acoustic wave beam is aligned in real-time with the direction of the target detected by the radar to avoid the decoy signal affecting surrounding unrelated areas; for ownerless floating objects and ground security risks, a high-frequency high-intensity light warning is initiated, simultaneously triggering the visible light camera to acquire target images and record precise location coordinates. All evidence-gathering data is automatically encrypted and stored with timestamps and location information to meet subsequent traceability and evidence-gathering requirements.

[0013] Furthermore, the ground command and control unit has a built-in airspace situation fusion module, which converts radar detection data transmitted by multiple UAVs and target data collected by fixed-point monitoring equipment in the control area into the WGS84 geodetic coordinate system to achieve time synchronization and spatial alignment. It uses a weighted fusion algorithm to fuse the target position and velocity data in the overlapping detection areas to generate a continuous low-altitude situation map covering the entire control airspace. It assigns a globally unique identifier to the same target with overlapping multi-source data and performs deduplication. It dynamically displays the real-time movement trajectory and handling status of the target, eliminating blind spots and single-point detection errors that exist in single UAV detection.

[0014] Furthermore, the ground command and control unit has a built-in intelligent task scheduling module that monitors the remaining battery power, equipment health status, current operating location, and completed inspection coverage of each inspection drone in real time. When the remaining battery power of a single drone reaches the preset return threshold or an abnormal equipment condition occurs that affects the operation, the system automatically schedules the standby drone closest to the corresponding inspection point to plan the optimal flight route to take over the operation, with the goal of minimizing the shortest flight path and the airspace coverage gap. During the task handover, the radar detection areas of the two drones maintain an overlap of no less than 30 seconds to achieve seamless continuous coverage of the inspection area and avoid security control loopholes.

[0015] Furthermore, the airborne millimeter-wave radar sensing unit automatically performs an online calibration process before each cruise mission takes off. By identifying pre-deployed corner reflectors with precisely known position coordinates within the control area as calibration references, and comparing the deviation between the radar-detected reference position and the known actual position, the radar's angle measurement zero point and distance measurement offset parameters are automatically adjusted. The calibration process does not require manual disassembly of the radar equipment or manual parameter adjustment. The entire calibration process takes no more than 1 minute. After calibration, a calibration report containing calibration parameters and calibration error results is automatically generated and transmitted back to the ground command and control unit for storage and archiving via the air-to-ground two-way communication unit, ensuring that the detection accuracy of the equipment remains stable during long-term operation.

[0016] Compared with existing technologies, the beneficial effects of this invention are: This invention, by incorporating a wide-beam 77GHz millimeter-wave radar sensing unit and combining adaptive rotor clutter and ground clutter cancellation algorithms with intelligent recognition logic for low-speed, small targets, achieves high-accuracy and low-latency recognition of weak targets in complex low-altitude scenarios. It effectively solves the shortcomings of existing technologies, such as severe environmental interference with optical detection, high false alarm and missed alarm rates with ordinary radar, and slow response in small target recognition. It can stably complete detection operations in complex environments such as severe weather and nighttime. Furthermore, through multi-target optimized adaptive trajectory control, autonomous operation during communication interruptions, and a seamless multi-aircraft scheduling mechanism, this invention dynamically adjusts trajectory tracking risks while simultaneously ensuring flight safety and the integrity of inspection coverage. It can autonomously complete tiered handling in the event of communication interruptions, and the multi-aircraft handover process has no control gaps. This effectively solves the shortcomings of existing technologies, such as target collisions with obstacles, operational halts due to communication interruptions, and control loopholes in inspection coverage, significantly improving the overall operational stability and reliability of the system.

[0017] This invention achieves precise targeted response, blind-spot-free global situational awareness, and stable long-term equipment operation accuracy through multi-module coordinate unified registration hierarchical security response, multi-source data fusion for global situational awareness generation, and automatic online radar calibration mechanism. It effectively solves the defects of existing technologies such as poor alignment accuracy for expulsion and evidence collection, blind spots in the field of vision of single-machine situational awareness construction, and cumbersome equipment calibration process, and significantly reduces the difficulty of operation and maintenance and labor costs.

[0018] This system is adaptable to various application scenarios such as airport airspace control, security for major events, patrols in urban core areas, and low-altitude border protection. It can be quickly deployed without major modifications to the existing security control system and has extremely high industry promotion value. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the overall operation of the low-altitude security patrol drone system proposed in this invention. Figure 2 This is a flowchart of the radar online calibration and sensing preprocessing proposed in this invention; Figure 3 This is a flowchart of the target identification and adaptive clutter filtering proposed in this invention; Figure 4 This is a flowchart of the adaptive trajectory optimization and obstacle avoidance control proposed in this invention; Figure 5 This is a flowchart of the air-ground coordinated scheduling and hierarchical security response proposed in this invention. Detailed Implementation

[0020] 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.

[0021] In the description of this invention, it should be understood that the terms "center," "longitudinal," "lateral," "length," "width," "thickness," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "clockwise," and "counterclockwise," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0022] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, features defined with "first" and "second" may explicitly or implicitly include one or more of the stated features. In the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified. Furthermore, the terms "installed," "connected," and "linked" should be interpreted broadly; for example, they may refer to a fixed connection, a detachable connection, or an integral connection; they may refer to a mechanical connection or an electrical connection; they may refer to a direct connection or an indirect connection through an intermediate medium; and they may refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances. The invention will now be described in further detail with reference to the accompanying drawings.

[0023] Reference Figures 1 to 5 A low-altitude security patrol drone system based on millimeter-wave radar includes a multi-rotor flight platform, an airborne millimeter-wave radar sensing unit, an airborne edge computing processing unit, an adaptive trajectory control unit, an air-to-ground two-way communication unit, a ground command and control unit, and an airborne security response unit. The multi-rotor flight platform uses a T800 carbon fiber integrated frame, equipped with a dual-redundant brushless power unit and carbon fiber propellers. All metal connectors on the frame are covered with radar-absorbing material to prevent distortion interference from the metal structure to the millimeter-wave antenna pattern. The power system can support a maximum flight time of 45 minutes and maintain attitude stability in 6-level gusts, providing an electromagnetically compatible and vibration-isolated installation environment for each airborne unit, as well as flight power that meets long-endurance requirements. The flight platform incorporates a combined navigation module consisting of differential GNSS, a high-precision IMU, and a magnetometer, achieving centimeter-level positioning accuracy. It supports constant-altitude and variable-altitude cruises along pre-set inspection routes in complex low-altitude scenarios such as urban core areas and airport airspace. All cruise parameters support millisecond-level real-time dynamic adjustment. The various airborne units communicate via a hybrid airborne network composed of gigabit Ethernet and CAN bus. High-bandwidth radar point cloud and high-definition video data are transmitted via Ethernet, while low-latency flight control commands and gimbal control commands are transmitted via CAN bus. All data transmission employs a cyclic redundancy check mechanism to ensure integrity and prevent equipment malfunctions caused by data errors.

[0024] The airborne millimeter-wave radar sensing unit adopts a 4-transmit, 8-receive wide-beam patch array antenna architecture. The antenna beam has an elevation coverage of no less than 120 degrees and an azimuth coverage of no less than 100 degrees. It continuously transmits 77GHz frequency-modulated continuous wave signals at a fixed frame frequency of 20Hz, receives target echo signals within the circumferential coverage area, and simultaneously completes echo analog-to-digital conversion, and preliminary processing of range and velocity dimension fast Fourier transform. It collects raw point cloud data of target range, radial velocity, azimuth angle, and elevation angle, and completes initial filtering of strong reflection fixed ground clutter through fixed threshold filtering.

[0025] The airborne edge computing processing unit is equipped with an embedded neural network acceleration chip with 6 TOPS INT8 computing power. It performs adaptive clutter cancellation, density clustering, multi-frame trajectory association, and feature matching on the pre-processed radar point cloud data. It identifies the type, real-time location, and motion parameters of intruding low, slow, and small flying targets, illegal floating objects in the airspace, and ground security hazards. It calculates the target hazard confidence level by combining the target size, motion speed, and intrusion path, and outputs the red, orange, and yellow three-level hazard level judgment results.

[0026] The adaptive trajectory control unit optimizes target tracking stability, flight safety, and inspection coverage integrity. It dynamically calculates the optimal cruising altitude, flight speed, and observation attitude adjustment of the multi-rotor flight platform, guiding the millimeter-wave radar sensing unit to continuously lock onto and track the target in the potential danger area. The entire adjustment process uses uniform acceleration control logic to ensure the flight platform's attitude remains stable without violent shaking.

[0027] The air-to-ground two-way communication unit adopts an adaptive frequency-hopping broadband data transmission link, operating in the 1.4GHz dedicated government wireless band. It dynamically adjusts the frequency hopping rate and transmission power based on the link's signal-to-noise ratio, transmitting the sensor data, processing results, and equipment status information collected by the airborne unit back to the ground command and control unit in real time using the national cryptographic SM2 encrypted frame format. After command verification, it forwards the control commands issued from the ground to the corresponding airborne unit. The airborne security response unit matches a pre-set, graded response strategy according to the level of hazard, performing close-range target directional removal and multi-dimensional evidence collection operations while maintaining a safe tracking distance.

[0028] The ground command and control unit is deployed in the security command center and adopts a server cluster architecture to complete the full-domain scheduling of multiple inspection drones, unified display of airspace situation, automatic encryption and archiving of operation data, and support the issuance of emergency response instructions and full-link traceability of the response process under manual intervention mode.

[0029] This invention also discloses that the airborne millimeter-wave radar sensing unit is integrated and installed on a two-axis stabilization gimbal below the nose of a multi-rotor flight platform. The gimbal adopts a brushless direct-drive motor and a multi-layer silicone rubber composite vibration damping structure, and has a built-in 2000Hz sampling rate IMU attitude feedback module to form a closed-loop image stabilization control, which can suppress the vibration angle of the radar mounting surface caused by the flight platform's own rotor vibration and airflow disturbance to within 0.05 degrees. The gimbal can perform stepless rotation adjustment in the range of 0 degrees to 360 degrees horizontally and -90 degrees to +30 degrees pitch according to the attitude command of the adaptive trajectory control unit, with a rotation response delay of no more than 20 milliseconds. The radar module adopts a fully digital intermediate frequency architecture, with an operating frequency band covering 76 to 81 GHz and a maximum FM bandwidth supporting 4 GHz. It has a range resolution better than 0.2 meters and a velocity resolution better than 0.05 meters per second. It can stably detect low-altitude weak targets such as micro drones and Kongming lanterns with a radar cross-section of not less than 0.001 square meters, and has a maximum effective detection range of not less than 3 kilometers, providing sufficient reaction time for low-altitude security operations. The radar module has a built-in temperature compensation circuit, which can stably maintain the consistency of radio frequency parameters within an ambient temperature range of -20℃ to 55℃, avoiding the decrease in detection accuracy caused by temperature drift.

[0030] This invention also discloses an adaptive moving target display filtering module built into the airborne edge computing processing unit. It uses an improved minimum mean square error iteration criterion to update the filtering weights in real time, adaptively canceling ground clutter, rotor micro-motion clutter from the flight platform itself, and micro-Doppler clutter caused by gimbal vibration in the echo signal. The filtering weight update formula is as follows: in, This is the complex filtering weight vector within the kth coherent processing interval. Its dimension is the same as the number of echo sampling points. Its function is to perform amplitude and phase weighting on the echo signal of each sampling point to achieve clutter cancellation. The step size factor is the iteration step size factor, which ranges from 0 to 1. It is dynamically determined based on the current clutter intensity to balance the convergence speed and steady-state error. When the clutter power is higher than -40dBm, the step size is 0.005 to ensure steady-state filtering accuracy. When the clutter power is lower than -60dBm, the step size is 0.02 to accelerate the convergence speed. The residual between the filtered output signal and the constructed clutter reference signal is retained only in the residual, which contains the target echo component with motion characteristics. This represents the complex echo sampling vector within the current coherent processing interval, where each element corresponds to a range gate sampling value. The conjugate transpose of the sampling vector ensures phase matching for weight updates. The filtering module constructs a dynamic clutter reference signal using real-time acquired flight platform motor speed feedback values ​​and gimbal vibration acceleration data. It can cancel clutter without relying on pre-stored external clutter maps, achieving a clutter suppression ratio of no less than 60dB. This reduces the false alarm rate caused by rotor micro-motion clutter by two orders of magnitude, preventing slow-moving small targets from being masked by clutter and resulting in missed detections.

[0031] This invention also discloses that the airborne edge computing processing unit uses an improved density-based noisy spatial clustering algorithm to cluster the clutter-filtered point cloud. During the clustering process, the neighborhood radius threshold is adaptively adjusted based on the radial velocity difference between point clouds. When the radial velocity difference between two point clouds is greater than 2 m / s, they are not classified into the same cluster even if the spatial distance is less than the clustering neighborhood threshold, thus avoiding mis-clustering adjacent point clouds with large velocity differences into the same target. After clustering, the target motion trajectory is constructed using a uniform motion model based on the cluster center position changes of 10 consecutive frames. The trajectory jumps caused by outliers in the point cloud are eliminated through Kalman filtering. The state vector of the Kalman filter contains the target's planar coordinates and x and y direction velocity components, and the observation vector is the target coordinate value measured by the radar in real time. After the trajectory is constructed, the extracted target size, average radar cross section, motion speed, and maneuvering angular velocity features are matched with a pre-set illegal target feature library using cosine similarity. When the similarity is greater than 0.8, the corresponding target type is determined. The identification response delay does not exceed 80 milliseconds, meeting the real-time detection requirements of low-altitude high-speed dynamic targets. The pre-set database of illegal target features covers typical low-altitude security targets such as multi-rotor drones, fixed-wing model aircraft, Kongming lanterns, tethered balloons, and illegally launched flocks of pigeons. The feature boundaries of different targets have been calibrated through a large number of field tests, which can effectively distinguish between birds and artificially launched targets and reduce invalid false alarms.

[0032] This invention also discloses that the adaptive trajectory control unit incorporates a dynamic obstacle avoidance submodule and a multi-target trajectory optimization module, which fuse millimeter-wave radar point cloud data with position and attitude data output by the flight platform's integrated navigation module in real time. The comprehensive cost of trajectory adjustment is calculated based on target tracking stability, flight safety, and inspection coverage integrity as optimization targets. The comprehensive cost calculation formula is as follows: in, The overall cost of adjusting a dimensionless trajectory is as follows: the smaller the cost, the better the overall performance of the trajectory scheme. , , To correspond to the weight coefficients of the optimization objectives, the sum of the three is always 1. The weights are dynamically adjusted according to the current level of the hidden danger. Under the red level 1 hidden danger, the target tracking weight α is 0.7, the safety weight β is 0.2, and the coverage weight γ is 0.1. Under the orange level 2 hidden danger, all three weights are 0.33 to balance the various objectives. Under the yellow level 3 hidden danger, the coverage weight γ is 0.7, the safety weight β is 0.2, and the tracking weight α is 0.1, so as to minimize the deviation from the original inspection route. To adjust the angular deviation between the radar beam axis and the target direction, The radar half-power beamwidth is fixed at 12 degrees to ensure that the target is always within the coverage area of ​​the main lobe of the beam and is not lost when Δθ is less than 6 degrees. To adjust the minimum distance between the flight platform and surrounding obstacles, The preset safe flight distance threshold is fixed at 10 meters to ensure that the flight platform always maintains a safe distance from obstacles. To adjust the distance between the adjusted route and the preset inspection route, Given the total length of the current inspection segment, the inspection coverage loss is kept within an acceptable range. The module iterates through all feasible combinations of flight attitude and speed adjustments, selects the adjustment scheme with the lowest overall cost, and outputs flight control commands. During the process of adjusting the trajectory to track potential targets, it automatically avoids overhead cables with a diameter of more than 2 cm, tall trees, and building protrusions in the airspace, preventing the flight platform from blindly maneuvering to track targets and causing collisions.

[0033] This invention also discloses that the air-to-ground bidirectional communication unit performs link quality monitoring every 10 milliseconds. The link quality is comprehensively evaluated by measuring the received signal strength, signal-to-noise ratio, and packet loss rate. When the received signal strength is below -90dBm, the signal-to-noise ratio is below 5dB, the packet loss rate is greater than 30%, and reliable data interaction cannot be completed for three consecutive communication cycles, the local autonomous handling mode of the airborne edge computing processing unit is automatically triggered. The hazard identification, tracking, and response operations are completed in the order of priority: red hazards first, orange hazards second, and yellow hazards last. During the link loss, all operation data is encrypted and stored in an airborne non-volatile memory with a capacity of 128GB, which can store no less than 72 hours of continuous operation data. After the link is restored, the breakpoint resume transmission mechanism is used to automatically retransmit all operation data stored during the link loss. During the retransmission process, the LZ4 lossless data compression algorithm is used to compress the data to one-third of the original size, reducing the transmission bandwidth occupation and avoiding the occurrence of inspection operation interruptions and data loss.

[0034] This invention also discloses that the airborne security response unit integrates a 4K high-definition visible light zoom evidence-gathering camera, a directional acoustic wave deterrent module, and a high-frequency strong light warning module. All modules are rigidly coaxially mounted with the millimeter-wave radar sensing unit. During the deployment phase, the spatial mapping between the detection coordinate system and the operation coordinate system is completed through field calibration, with a registration error of less than 0.5 degrees, ensuring that the centers of action of the camera, acoustic wave, and strong light modules are synchronously aligned with the target when the radar is aligned with the target. The module automatically matches the corresponding operating mode for different levels of security risks: For red-level illegal low-speed small flying targets that have entered the core control area, a directional sound wave is activated to drive them away while maintaining a safe tracking distance of 100 meters. The beamwidth of the sound wave module is 15 degrees, and the sound pressure level in the direction of the target at a distance of 100 meters can reach 110dB, which can interfere with the operation of the drone's barometer or serve as a warning to the illegal operators. The sound pressure level outside the 30-degree angle can be reduced to below 60dB to avoid noise interference to the surrounding unrelated areas; for orange-level ownerless floating objects, a high-frequency strong light warning is activated with a strobe frequency set to 10Hz to remind ground personnel to take precautions; for yellow-level ground security risks, the visible light camera is simultaneously triggered to zoom to the maximum magnification to collect clear images of the target and record the precise location coordinates. All evidence data, along with high-precision timestamps and positioning information, is automatically stored in an encrypted partition. The data cannot be tampered with after being written, meeting the requirements for subsequent traceability and evidence collection. This invention also discloses a built-in airspace situation fusion module in the ground command and control unit. This module converts radar detection data transmitted from multiple UAVs and target data collected by fixed-point monitoring equipment within the control area into the WGS84 geodetic coordinate system. By using the NTP network for time synchronization, the time synchronization error of all data sources is controlled to within 1 millisecond. After completing time synchronization and spatial alignment, a weighted fusion algorithm is used to fuse the target position and velocity data in the overlapping detection area. The closer the detection data source is to the target, the higher its weight, generating a continuous low-altitude situation map covering the entire control airspace. During the fusion process, when the position difference between two targets is less than 5 meters, the velocity difference is less than 1 m / s, and the difference in movement direction is less than 10 degrees, they are determined to be the same target and assigned a globally unique identifier for deduplication. The real-time movement trajectory and handling status of the target are dynamically displayed, eliminating blind spots and single-point detection errors that exist in single UAV detection.

[0035] This invention also discloses a built-in intelligent task scheduling module in the ground command and control unit. This module monitors the remaining battery power, equipment health status, current operating location, and completed inspection coverage of each inspection drone in real time. When the remaining battery power of a single drone reaches the threshold for a 10-minute return trip, or when an abnormal equipment condition occurs that affects the operation, the system automatically schedules the standby drone closest to the corresponding inspection point to plan the optimal obstacle avoidance flight route to take over the operation, with the shortest flight path and the smallest airspace coverage gap as the optimization objectives. During the task handover, the returning drone maintains its original cruise state and tracking task. When the replacement drone flies to a position 50 meters away, it synchronizes all currently tracked target parameters and inspection progress to the replacement drone through point-to-point communication. After the replacement drone adjusts its radar beam direction to complete the target follow-up tracking, it maintains overlapping observation for no less than 30 seconds to confirm that all targets are stably tracked and that no inspection area is missed. Then, the original task drone executes the return trip procedure, achieving seamless continuous coverage of the inspection area and avoiding security control loopholes.

[0036] This invention also discloses that the airborne millimeter-wave radar sensing unit automatically performs an online calibration process before each cruise mission takeoff. At each routinely deployed UAV takeoff and landing point, a trihedral corner reflector is fixed at a position 50 meters directly in front of the takeoff apron and 2 meters above it. The radar cross-section of the corner reflector for millimeter waves is consistently maintained at 10 square meters. Its WGS84 coordinates are precisely measured to the centimeter level using RTK equipment during deployment and stored in the parameter library of the ground command and control unit. After the UAV is powered on at the takeoff point, it automatically acquires the precise coordinates of the corner reflector, controls the gimbal to rotate to complete a full-circle scan, and identifies the intensity of the radar echo. The highest fixed point target corresponds to the corner reflector. By comparing the deviations of the azimuth, elevation, and range values ​​of the corner reflector detected by the radar with the known actual values, the radar's angle measurement zero point and range measurement offset parameters are automatically adjusted. The calibration process does not require manual disassembly of the radar equipment or manual parameter adjustment. The entire calibration process takes no more than 1 minute. After calibration, a calibration report containing calibration parameters and calibration error results is automatically generated and transmitted back to the ground command and control unit for storage and archiving via the air-to-ground two-way communication unit. When the calibration error exceeds the threshold, a secondary calibration process is automatically triggered to ensure that the detection accuracy of the equipment remains stable during long-term operation.

[0037] Reference Figure 1This diagram illustrates the global collaborative working mechanism of a low-altitude security patrol UAV system based on millimeter-wave radar. The multi-rotor flight platform patrols the controlled area according to a pre-set route, providing payload platforms and flight propulsion for each airborne unit. During patrol, the airborne millimeter-wave radar sensing unit continuously scans and detects the surrounding airspace and ground, outputting point cloud data. The airborne edge computing processing unit performs deep processing on the point cloud to identify various security risk targets and classify them. The adaptive trajectory control unit dynamically calculates the cost based on the target's location and optimizes the flight attitude to achieve stable target tracking. The airborne security response unit carries out target removal and evidence collection operations. Simultaneously, the air-to-ground two-way communication unit transmits situational data in real time and receives dispatch instructions from the ground command and control unit, forming an integrated air-to-ground closed-loop security control system.

[0038] Reference Figure 2 This diagram details the pre-flight calibration and in-flight preprocessing steps of the radar sensing unit. Before takeoff for an inspection mission, the system automatically compares the detection position deviation using a corner reflector reference with known precise coordinates. This calibration of the radar's angle zero point and range offset is performed without manual intervention, generating a calibration report and sending it back for archiving, thus ensuring long-term high-precision operation of the equipment. During flight operations, the two-axis stabilization gimbal responds at high frequency to attitude commands, significantly suppressing vibrations caused by the rotor and airflow. The radar array antenna continuously transmits frequency-modulated continuous wave signals, and upon receiving the echoes, simultaneously performs analog-to-digital conversion and fast Fourier transform. A fixed threshold filter is then used to initially remove strong ground clutter, outputting high-precision raw point cloud data for backend processing.

[0039] Reference Figure 3 This figure illustrates the internal processing logic of the edge computing unit in extracting and identifying low-altitude potential hazards. First, the system constructs a reference signal by combining real-time attitude data from the flight platform and gimbal. An adaptive filter with dynamically updated weights accurately cancels rotor micro-motion clutter and gimbal vibration clutter. After cancellation, an adaptive radius density clustering algorithm is used to cluster the clean point cloud, preventing misclassification of neighboring point clouds with significantly different speeds. Next, a motion model is established based on position changes across multiple frames, and Kalman filtering is introduced to smooth out outliers. Finally, feature parameters such as the target's size, radar cross-section, and velocity are extracted and matched with a pre-set database of prohibited target features using cosine similarity, quickly outputting the target type and hazard level assessment results.

[0040] Reference Figure 4This figure illustrates the computational process of a UAV dynamically adjusting its trajectory and intelligently avoiding obstacles in a complex low-altitude environment. The adaptive trajectory control unit integrates radar point cloud data with multi-source sensor data such as satellite positioning and inertial navigation in real time to comprehensively perceive the three-dimensional spatial environment surrounding the UAV. When tracking potential targets, the system uses three core elements—target tracking stability, flight safety, and inspection coverage integrity—as optimization objectives, and dynamically assigns weight coefficients to each objective based on different hazard levels. The system iterates through and calculates the comprehensive cost of all feasible attitude adjustment schemes, selecting the scheme with the lowest cost and converting it into flight control commands. This mechanism ensures that the UAV can automatically avoid obstacles such as tall trees and overhead cables while stably monitoring and driving away potential targets.

[0041] Reference Figure 5 This diagram illustrates the collaborative workflow of air-to-ground data link assurance, on-site airborne response, and global ground control. The communication unit monitors link quality at high frequency. In the event of severe interference and link failure, it automatically triggers the onboard autonomous response mode, performing hazard tracking and on-site encrypted data storage. Once the link is restored, the data is automatically compressed and retransmitted, ensuring complete and unloss-free inspection data. At the on-site response end, based on the target category, it initiates acoustic directional deterrence or high-frequency high-intensity light warnings, and simultaneously controls cameras to collect location and evidence data. At the ground control end, the system integrates multi-source radar data to generate a continuous low-altitude situation map of the entire domain in a unified coordinate system. The scheduling module monitors the battery level and health of each UAV in real time, automatically planning the optimal path to dispatch standby UAVs when necessary, maintaining overlapping coverage time and completely eliminating blind spots in airspace security.

[0042] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A low-altitude security patrol drone system based on millimeter-wave radar, characterized in that, include: The multi-rotor flight platform provides installation support and flight power for each airborne unit, and performs constant-altitude cruise within the designated low-altitude security control area according to the pre-set inspection route, supporting real-time dynamic adjustment of cruise parameters; The airborne millimeter-wave radar sensing unit adopts a wide-beam multi-transmit and multi-receive array antenna architecture to continuously transmit millimeter-wave detection signals at a fixed frame frequency, receive target echo signals in front of the inspection route and within the circumferential coverage area, and simultaneously collect raw point cloud data of target distance, radial velocity, azimuth angle, and elevation angle to complete the initial filtering of fixed ground clutter. The airborne edge computing processing unit performs cluster analysis and multi-frame trajectory association processing on the pre-processed radar point cloud data to identify the types, real-time locations and motion parameters of intruding low, slow and small flying targets, illegal floating objects in the airspace and ground security hazards, and generates hazard level judgment results. The adaptive trajectory control unit dynamically adjusts the cruising altitude, flight speed and observation attitude of the multi-rotor flight platform according to the hazard level judgment result, guiding the millimeter-wave radar sensing unit to continuously and stably track the hazard area. The air-to-ground two-way communication unit adopts an anti-interference frequency hopping communication link to transmit the sensing data, processing results and equipment operation status information collected by the airborne terminal back to the ground command and control unit in real time. At the same time, it receives control instructions issued by the ground terminal and forwards them to the corresponding airborne unit. The airborne security response unit matches preset handling strategies according to the level of the hidden danger, and performs close-range target driving away and on-site evidence collection operations; the ground command and control unit completes the unified scheduling of multiple inspection drones, displays the low-altitude security situation in the whole area, and supports manual intervention to issue handling instructions.

2. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The airborne millimeter-wave radar sensing unit is integrated and installed on a two-axis stabilization gimbal below the nose of the multi-rotor flight platform. The gimbal can rotate and adjust within the range of 0 to 360 degrees horizontally and -90 to +30 degrees pitch according to the attitude commands of the adaptive trajectory control unit. The radar operates in the frequency band of 76 to 81 GHz, with a range resolution better than 0.5 meters and a speed resolution better than 0.1 meters per second.

3. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 2, characterized in that, The airborne edge computing processing unit has a built-in adaptive moving target display filtering module. It performs adaptive cancellation processing on ground clutter and micro-motion clutter of the flight platform's own rotor in the echo signal by iteratively updating the clutter covariance matrix, retaining the target point cloud with effective motion characteristics, and the clutter suppression ratio is not less than 50dB.

4. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 3, characterized in that, The airborne edge computing processing unit uses a density-based noisy spatial clustering algorithm to divide the clutter-filtered point cloud into clusters. It constructs the target motion trajectory by combining the changes in the cluster center position of multiple consecutive frames. The target type is identified by feature matching with a pre-set illegal target feature library. The identification response delay does not exceed 100 milliseconds.

5. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The adaptive trajectory control unit has a built-in dynamic obstacle avoidance submodule that integrates millimeter-wave radar point cloud data with satellite positioning and inertial measurement unit sensor data on the flight platform in real time. In the process of adjusting the trajectory to track potential targets, it automatically avoids overhead cables, tall trees and building protrusions in the airspace, so that the flight platform always maintains a distance of no less than the safety threshold from various obstacles.

6. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The air-to-ground two-way communication unit monitors the quality of the communication link in real time. When the link quality is lower than a preset threshold, the local autonomous handling mode of the airborne edge computing processing unit is automatically triggered. The unit completes the hazard identification and response work according to the preset priority. After the link is restored, all the operation data stored during the link loss period is automatically retransmitted.

7. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The airborne security response unit integrates a high-definition visible light evidence-gathering camera, a directional acoustic wave deterrence module, and a strong light warning module. It automatically matches the corresponding operating mode for different levels of security risks, initiates directional acoustic wave deterrence for low, slow, and small targets flying illegally, and initiates strong light warning for ownerless floating objects and ground security risks while simultaneously acquiring high-definition evidence-gathering images.

8. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The ground command and control unit has a built-in airspace situation fusion module, which performs unified coordinate transformation and spatiotemporal alignment on radar detection data transmitted back by multiple UAVs and data from fixed-point monitoring equipment within the control area, generating a continuous situation map covering the entire control airspace, and deduplicating targets with overlapping multi-source data.

9. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 8, characterized in that, The ground command and control unit has a built-in intelligent task scheduling module that monitors the remaining battery power and equipment health status of each inspection drone in real time. When the remaining battery power of a single drone reaches the return threshold, it automatically schedules the standby drone closest to the corresponding inspection point to take over the operation, achieving seamless and continuous coverage of the inspection area.

10. The low-altitude security patrol drone system based on millimeter-wave radar according to claim 1, characterized in that, The airborne millimeter-wave radar sensing unit automatically performs online calibration before each cruise mission. It adjusts the radar's angle measurement error and distance measurement error by identifying fixed calibration reference objects at known locations within the control area. The calibration process does not require manual disassembly of the equipment. After calibration, a calibration report is automatically generated and sent back to the ground command and control unit.

Images