A kind of apron safety supervision and aircraft monitoring system based on multi-source fusion

Abstract

This invention belongs to the field of aircraft ground monitoring technology, specifically relating to a multi-source fusion-based apron safety supervision and aircraft monitoring system. It includes a multi-source perception layer, a data transmission layer, a data fusion processing layer, an intelligent analysis and decision-making layer, and a visualization layer. The multi-source perception layer employs various monitoring devices. The data fusion processing layer achieves spatiotemporal alignment and weighted fusion positioning of multi-source data; a low-speed, small target identification module distinguishes between drones and birds; and a non-cooperative object detection module integrates thermal imaging, millimeter-wave radar, and fiber optic vibration data to track untagged intruders. The intelligent analysis and decision-making layer dynamically generates electronic fences, calculates collision risk indices, and includes an automatic monitoring module for post-flight periods or nighttime shutdown periods, constructing enhanced safety zones and generating monitoring reports. This invention achieves centimeter-level positioning, accurate identification of low-speed, small targets, all-weather tracking of non-cooperative objects, and closed-loop monitoring of the entire aircraft process, significantly improving apron safety and operational efficiency.

Description

Technical Field

[0001] This invention belongs to the field of airport apron safety supervision and aircraft ground monitoring technology, specifically relating to an apron safety supervision and aircraft monitoring system based on multi-source fusion. Background Technology

[0002] With the increasing density of apron operations, apron safety supervision and aircraft ground monitoring are facing increasingly severe challenges. As an important part of the airport flight area, the apron is the core area for ground support operations such as aircraft parking, passenger pick-up and drop-off, cargo loading and unloading, refueling, and cleaning. Its operational safety is directly related to the overall safety level of civil aviation transportation.

[0003] Traditional aircraft ground monitoring primarily employs a fixed one-person-one-aircraft monitoring model. This means that each aircraft is monitored by a single monitor while parked, responsible for preventing unauthorized personnel and vehicles from approaching the aircraft and promptly detecting and handling any abnormalities. While this model has played a positive role in ensuring aircraft safety, its drawbacks have become increasingly apparent with the continuous growth of air traffic and the increasing demands for operational efficiency. The traditional monitoring model incurs extremely high labor costs, requiring a large number of monitoring personnel daily at large hub airports, placing immense pressure on staffing levels. Manual monitoring suffers from blind spots, as a single monitor cannot simultaneously observe multiple directions around the aircraft. The effectiveness of personnel monitoring significantly decreases under adverse weather conditions, and monitoring quality is difficult to guarantee at night and in rainy or foggy weather. Furthermore, with the continuous expansion of airport apron areas and the widening of the monitoring scope, the traditional model is increasingly unable to meet the operational needs of modern airports in terms of coverage and management efficiency.

[0004] In recent years, some large airports have begun to explore aircraft area monitoring models, adopting a multi-layered prevention and control approach that combines fixed duty, vehicle patrols, and video surveillance to replace the traditional monitoring model. However, existing area monitoring systems still have the following technical shortcomings: Although existing video surveillance systems can provide real-time images of the apron area, they lack the ability to accurately perceive the three-dimensional spatial position of aircraft, and cannot achieve millimeter-level or decimeter-level positioning of aircraft positions, making it difficult to achieve refined safety monitoring in key stages such as aircraft entry and exit; existing systems lack the ability to effectively identify and distinguish threats from low, slow, and small targets, and there is a significant deficiency in the supervision of non-cooperative targets. Summary of the Invention

[0005] The purpose of this invention is to overcome the above-mentioned shortcomings of the prior art and provide an apron safety supervision and aircraft monitoring system based on multi-source fusion. Through the deep integration of various sensing technologies, it can achieve precise three-dimensional positioning, real-time trajectory tracking, intelligent risk warning and full-process closed-loop management of aircraft and ground vehicles in the apron area. Furthermore, it can solve the problem of insufficient supervision of low, slow and small target threats and non-cooperative objects, and comprehensively improve the intelligence level and operational efficiency of apron safety supervision and aircraft ground monitoring.

[0006] The present invention discloses a multi-source fusion-based apron safety monitoring and aircraft surveillance system, comprising: The multi-source sensing layer includes UWB positioning base stations, lidar, video cameras, millimeter-wave radar, passive radar detection devices, acoustic array detection devices, thermal imaging cameras, and distributed fiber optic sensing devices. The data transmission layer is used to transmit data from the perception layer to the central processing platform in real time. The data fusion processing layer includes a spatiotemporal alignment module, a target recognition and tracking module, a fusion localization module, a low-speed-small target recognition and tracking module, and a non-cooperative object detection and tracking module; The intelligent analysis and decision-making layer includes a dynamic electronic fence management module, a collision risk analysis module, a monitoring status management module, an early warning and handling module, and an automatic monitoring module for post-flight periods or nighttime shutdown periods; The visualization layer includes a 3D digital twin engine and a user terminal; The low-speed, small target identification and tracking module distinguishes between drones and birds by fusing passive radar detection data with acoustic array acoustic signature features. The non-cooperative object detection and tracking module detects and tracks personnel who are not carrying electronic tags by fusing thermal imaging data, millimeter-wave radar data, and fiber optic sensing data. The automatic monitoring module is automatically activated after flights are grounded or during nighttime shutdown periods. It increases the frequency and sensitivity of sensor sampling and provides fully automatic unattended monitoring of parked aircraft.

[0007] Specifically, the multi-source sensing layer includes several types of sensing devices installed in the apron area for real-time acquisition of apron operation data. The multi-source sensing layer includes: an ultra-wideband (UWB) positioning base station, installed at each aircraft stand and its surrounding area, for bidirectional communication with aircraft-borne UWB tags and vehicle UWB tags to obtain the three-dimensional spatial coordinates of aircraft and vehicles; a lidar scanning device, installed at fixed positions at each aircraft stand, for acquiring three-dimensional point cloud data of the stand area and identifying aircraft approach attitude and ground obstacles; a high-definition video camera device, installed at each aircraft stand and key passageway area, for acquiring real-time video images of the apron area; and a millimeter-wave radar detection device, installed at taxiway intersections and aircraft stand access passageways, for detecting the motion parameters of moving targets. The system includes: a passive radar detection device, installed at high points around the apron and in key locations within the airspace, used to passively receive the scattered echoes of electromagnetic signals from drones and birds, enabling long-range detection of low-speed, small targets; an acoustic array detection device, installed in various aircraft positions and at the edge of the airspace, consisting of an array of multiple high-sensitivity microphones, used to collect high-frequency acoustic signatures generated by drone rotor rotation and low-frequency acoustic signatures generated by bird flapping; a thermal imaging camera device, installed in various aircraft positions and key passageways on the apron, used to detect the thermal radiation characteristics of the human body at night or in adverse weather conditions; and a distributed fiber optic sensing device, installed below ground level around the apron perimeter and aircraft positions, using optical fibers as the sensing medium to detect ground vibration and pressure by measuring changes in backscattered light within the fiber. The data transmission layer, including 5G-AeroMACS communication base stations and fiber optic communication networks set up in the apron area, is used to transmit various types of data collected by the multi-source sensing layer to the central processing platform in real time. The multi-source data spatiotemporal alignment module unifies data collected by different sensing devices to the same time reference and spatial coordinate system; the target recognition and tracking module performs target detection and recognition on video images, point cloud data, and radar echo data, extracting the target's type, size, outline, and motion parameters; the fusion positioning module performs fusion calculations on UWB positioning data, lidar point cloud data, video recognition results, and millimeter-wave radar data, outputting the fused position and fused trajectory of each target; the low-speed-small target recognition and tracking module fuses passive radar detection data and acoustic array acoustic signature features to distinguish between UAVs and birds; the non-cooperative object detection and tracking module fuses thermal imaging data, millimeter-wave radar data, and fiber optic sensing data to detect and track personnel without electronic tags. The dynamic electronic fence management module is used to dynamically generate aircraft safety protection zones based on aircraft type, parking position, and operational phase; the collision risk analysis module is used to calculate the minimum distance between targets and the collision risk index based on the fused position and motion parameters of each target; the monitoring status management module is used to record and manage the monitoring status information of the aircraft from entry to exit, including the operational status of each support link, personnel and vehicle entry and exit records, and risk event records; the early warning and handling module is used to generate graded early warning information and push it to relevant terminals when the collision risk index exceeds a preset threshold or the monitoring status is abnormal; the automatic monitoring module for post-flight periods or nighttime shutdown periods is used to automatically activate after flight shutdown, improve the sensor sampling frequency and sensitivity, and provide fully automatic unattended monitoring of parked aircraft. The visualization layer, including a 3D digital twin engine and a user terminal, is used to visualize the apron operation status in the form of a 3D digital twin and push early warning information and monitoring instructions to operation and management personnel.

[0008] Preferably, the UWB positioning base station adopts the Time Difference of Arrival (TDOA) positioning principle. A positioning network is formed by setting up at least four UWB positioning base stations in the apron area. The base stations achieve nanosecond-level time synchronization through a high-precision time synchronization protocol. UWB positioning tags are installed on aircraft and vehicles. These tags transmit UWB pulse signals at a preset frequency. Each base station receives the signals and records their arrival times, uploading the arrival time data to a positioning calculation server. The positioning calculation server calculates the spatial coordinates of the tag relative to the base station network based on the time difference of the same tag's signal arrival at each base station, outputting the tag's three-dimensional coordinates and motion speed information. The UWB positioning system can achieve decimeter-level or even centimeter-level positioning accuracy in unobstructed apron environments, far exceeding the positioning accuracy of traditional surface surveillance radar and multi-point positioning systems. This effectively solves the problems of inconsistent aircraft positioning references and large position errors.

[0009] Preferably, the lidar scanning device is fixed on both sides of each aircraft position, with the scanning direction facing the aircraft position area, to acquire three-dimensional point cloud data of the aircraft position area. The lidar measures the distance from the target surface point to the radar by emitting laser pulses and receiving the echo, and obtains the three-dimensional point cloud of the surrounding environment by rotating the scanning mechanism. The data processing unit filters and denoises the point cloud data and performs ground segmentation processing. Then, it identifies the aircraft fuselage outline and the ground support vehicle outline through a point cloud clustering algorithm, and calculates the coordinates of the aircraft fuselage center point and the nose angle to correct the positioning error of the UWB positioning data and provide accurate orientation information of the aircraft. When the aircraft is at low speed or stationary, the aircraft's nose direction may not be consistent with the direction of motion. Traditional heading estimation methods based on the direction of motion fail, while lidar point cloud data can directly identify the actual fuselage orientation of the aircraft, effectively solving the problem of difficulty in determining the heading in low-speed and stationary states.

[0010] Preferably, the multi-source data spatiotemporal alignment module performs the following alignment steps: First, a global coordinate system for the apron is established, and the coordinates of all fixed reference points in the apron GIS map are entered into the coordinate system, including the corner points of the aircraft stand markers, the fixed points of the jet bridge, the coordinates of the UWB positioning base station, the coordinates of the lidar installation location, and the coordinates of the video camera installation location, forming a unified coordinate reference benchmark; Second, a high-precision time benchmark is obtained through GPS timing devices set at key locations on the apron, and all sensing devices are synchronized with the time benchmark through the NTP time synchronization protocol or the PTP precise time protocol to ensure that the time error of each data source is controlled within a preset range; Then, for video image data, a mapping relationship between image pixel coordinates and the global coordinate system of the apron is established through camera calibration technology, and the target position detected by the video is transformed into the global coordinate system; For lidar point cloud data, the polar coordinate system coordinates of the lidar itself are transformed into the global coordinate system through a coordinate transformation matrix; For UWB positioning data, the positioning solution server directly outputs the three-dimensional coordinates in the global coordinate system; Finally, the multi-source data after spatial coordinate unification and time alignment are associated by timestamp to form a time-space synchronized multi-source dataset.

[0011] Preferably, the fusion positioning module uses a weighted fusion method to calculate the fused position coordinates of the target. Specifically, the steps are as follows: confidence levels are assessed for target position data acquired at the same time t from the UWB positioning system, LiDAR detection system, and video recognition system; the confidence weights of each data source are calculated, and these weights are dynamically adjusted based on the historical positioning accuracy statistics of the data sources and current environmental conditions; the target position coordinates from each data source are weighted and summed according to their confidence weights to obtain the fused position coordinates. When a target is detected by multiple data sources simultaneously, the fusion positioning result can integrate the advantages of each data source, reducing the error and noise impact of a single data source and improving the stability and reliability of the positioning.

[0012] Preferably, the target recognition and tracking module employs different recognition and tracking strategies for different types of targets: For aircraft targets, the coarse position provided by UWB positioning data is used as the initial position prior. Fine target detection and contour extraction are performed using lidar point cloud data in the vicinity of the initial position to identify the precise geometric center position and fuselage orientation of the aircraft. At the same time, millimeter-wave radar data is used to provide the aircraft's velocity vector and heading angle parameters, ultimately forming complete motion state information of the aircraft target, including three-dimensional coordinates, fuselage size, nose heading angle, velocity vector, and acceleration. For ground vehicle targets, UWB positioning data is used as the main positioning source, and video recognition results are used to confirm the target type and correct the position. For vehicles without UWB tags, detection and tracking are performed through video recognition and millimeter-wave radar.

[0013] Preferably, the low-speed, small target identification and tracking module performs the following steps: First, the passive radar detection device continuously monitors the apron clearance area. When an unknown aerial target is detected, it outputs the target's initial position coordinates, speed, and estimated radar cross section. Second, based on the target position output by the passive radar, the system guides the corresponding acoustic array detection device to point towards the target direction for acoustic signature acquisition. Beamforming technology is used to enhance the acoustic signal in the target direction to obtain the target's acoustic signature feature sequence. Third, the acquired acoustic signature features are compared with a pre-trained classification model. The classification model outputs the probability value that the target belongs to a drone or a bird. When the probability of a drone exceeds a preset threshold... The system first identifies a target as a drone threat, and then identifies a target as a bird risk when the probability of a bird exceeds a preset threshold. The fourth step involves activating a tracking mode for drone targets, fusing continuous position data from passive radar and directional data from the acoustic array to form a continuous drone trajectory, and calculating the minimum distance between the drone and the airport airspace boundary and aircraft in real time. A drone intrusion warning is triggered when a drone enters the no-fly zone or approaches an aircraft. The fifth step involves recording the bird activity trajectory and density distribution for birds, generating a bird heatmap. A bird strike risk warning is triggered when a large flock of birds continuously moves near the aircraft's takeoff and landing path.

[0014] Preferably, the non-cooperative object detection and tracking module performs the following steps: First, for passive intruders, the system detects and locates them by fusing human attribute recognition data sources and motion positioning data sources. The human attribute recognition data source includes at least one of a thermal imaging camera and a video camera. The thermal imaging camera detects the thermal radiation characteristics of the human body and outputs the thermal image contour and pixel coordinates, while the video camera is used for human detection when lighting conditions are met. The motion positioning data source includes at least one of a millimeter-wave radar and a distributed fiber optic sensing device. The millimeter-wave radar detects the distance, speed, and azimuth of the moving target and outputs position parameters, while the distributed fiber optic sensing device detects ground vibration signals and outputs the vibration source location coordinates.

[0015] The second step involves spatiotemporal alignment and fusion of the aforementioned multi-source data. The location detected by the distributed fiber optic sensing device or millimeter-wave radar is used as the initial positioning. Human targets detected by thermal imaging cameras or video cameras are searched for in the vicinity of this initial positioning, and the detected moving targets and human targets are associated and matched. After successful association, a temporary tracking ID is assigned to each non-cooperative object, and its movement trajectory is continuously recorded.

[0016] The third step is to determine whether the movement behavior of non-cooperative objects constitutes a security threat. Threat determination rules include: non-cooperative objects entering the aircraft security protection area, staying in the prohibited area for more than a preset threshold, abnormal movement patterns, and occurring during non-operational periods such as post-flight periods or nighttime shutdown periods. When any of the threat determination rules are met, a non-cooperative object intrusion warning is generated.

[0017] The fourth step is to identify personnel who refuse to be located through behavioral analysis. When a person is detected to be continuously approaching the aircraft without carrying a valid UWB tag, or when the corresponding UWB tag signal suddenly disappears, the target is marked as having a location anomaly and the monitoring level is increased.

[0018] Preferably, the automatic monitoring module for the post-flight period or nighttime shutdown period is automatically activated after the end of each day's flight operations (usually from 0:00 AM to 5:00 AM the next day) and executes the following monitoring strategies: the system automatically switches to the post-flight period or nighttime shutdown period monitoring mode, increases the sampling frequency and sensitivity of each sensor device, increases the UWB positioning data sampling frequency from 10 times per second to 20 times per second, switches the thermal imaging camera from standby mode to full-time acquisition mode, and increases the detection sensitivity of the distributed fiber optic sensor to the highest level; The system enters a silent baseline learning mode, recording background environmental data of the apron area during the post-flight period or the first 30 minutes after the start of the nighttime shutdown period. This includes the background temperature distribution of thermal imaging, the environmental vibration and noise level of fiber optic sensing, and the clutter distribution of millimeter-wave radar. After learning is complete, these baseline data are used as a reference, and any abnormal signals that deviate from the baseline are identified as potential threats. The enhanced safety protection zone is generated based on the safety protection zone during the parking support phase. The safety protection zone during the parking support phase is a closed area that extends outward from the outer perimeter of the aircraft fuselage by a preset distance. The linear distance between the outer boundary of the enhanced safety protection zone and the outer perimeter of the aircraft fuselage is 1.5 times the preset distance. The entry of any personnel or vehicle into the enhanced security protection area will immediately trigger a Level 3 warning and automatically generate a monitoring report for the post-flight period or the nighttime shutdown period.

[0019] Preferably, the dynamic electronic fence management module dynamically generates aircraft safety protection zones based on the aircraft type and current operational phase. Aircraft types are categorized into several levels based on fuselage length and wingspan, with different types corresponding to different sized safety protection zones. The aircraft operational phase includes the parking phase, the parking support phase, and the departure / pushback phase, with different shapes and range parameters for the safety protection zones at each phase. The safety protection zone for the parking phase is a conical area extending forward along the taxiway line, used to warn of collision risks between the aircraft and facilities at the front of the parking position during parking. The safety protection zone for the parking support phase is a closed area extending outwards from the outer perimeter of the aircraft fuselage at a certain distance. The aircraft is zoned according to its door location and ground support area, including a door protection zone and an engine danger zone. Different zones have different intrusion thresholds and warning rules. The safety protection zone during the pushback phase is the warning zone on both sides of the pushback passage behind the aircraft. The boundary parameters of the safety protection zone are dynamically updated based on the real-time monitored actual position of the aircraft, moving with the aircraft's movement. The dynamic configuration information of the safety protection zone is pushed to the vehicle-mounted terminal of the ground vehicle through a data interface. When the vehicle approaches the boundary of the safety protection zone, the vehicle-mounted terminal issues an audible and visual warning. When the vehicle intrudes into the safety protection zone, the vehicle-mounted terminal issues a continuous alarm and automatically records the intrusion event.

[0020] Preferably, the collision risk analysis module performs the following risk calculation steps: acquiring fused positioning information and motion parameters of all aircraft targets and ground vehicle targets; for each pair of targets that may collide, calculating the minimum expected distance and expected collision time between the targets based on the target's motion state parameters; comparing the minimum expected distance with a preset safe distance threshold, and comparing the expected collision time with a preset time threshold, and comprehensively evaluating to generate a collision risk index; generating a level one warning when the collision risk index exceeds a first threshold, a level two warning when it exceeds a second threshold, and a level three warning when it exceeds a third threshold; the collision risk analysis module is also used to calculate the collision risk between the aircraft and the fixed facilities at the parking position, with the position coordinates of the fixed facilities pre-entered into the system database, and determining whether the aircraft exceeds the parking position safety envelope based on the aircraft's real-time position and nose orientation.

[0021] Preferably, the monitoring status management module establishes an independent monitoring file for each parked aircraft. The monitoring file includes: basic aircraft information, including flight number, aircraft type, registration number, parking position, estimated arrival time, and estimated departure time; support operation status records, including the start time, end time, and status of support operations such as refueling, water supply, sewage disposal, catering, baggage handling, passenger boarding bridge docking, and jet bridge docking; personnel and vehicle entry and exit records, including the identity information of all personnel and vehicle information approaching the aircraft and their timestamps for entering and exiting the security protection area; risk event records, including the type, level, occurrence time, and handling results of warning events; and time node records for aircraft arrival and departure. All records in the monitoring file are associated with the corresponding time period's video surveillance recording index, supporting time-based review of on-site video.

[0022] Preferably, the early warning information generated by the early warning and handling module includes the early warning level, early warning type, early warning target identifier, early warning location coordinates, and recommended handling measures. The early warning information is pushed to at least one of the following terminals via the 5G-AeroMACS network: the display terminal of the apron operation monitoring center, the mobile terminal of the relevant aircraft crew, the vehicle-mounted terminal of the relevant ground vehicle, and the mobile handheld terminal of the relevant monitoring personnel. After the early warning information is pushed, the system continuously tracks the status changes of the early warning event. When the early warning event is effectively handled, the early warning status is automatically lifted. When the early warning event continues to deteriorate, the early warning level is automatically upgraded and the push range is expanded.

[0023] Preferably, the 3D digital twin engine constructs a 3D virtual scene of apron operation based on a high-precision GIS map of the apron and a 3D digital model library of aircraft and vehicles. The 3D digital model library pre-stores 3D models of different types of aircraft and 3D models of various types of ground support vehicles. Each 3D model is labeled with the coordinates of key parts of the model, including the coordinates of the fuselage center point, nose point, tail point, wingtip point, engine position, and door position of the aircraft model, and the coordinates of the front point, rear point, and left and right side points of the vehicle model. The target fusion position coordinates and nose orientation angle output by the fusion positioning module are matched and bound with the corresponding 3D models, driving the 3D models to be presented in the virtual scene with the correct attitude and position. The visualization display layer also provides the following interactive functions: scene perspective switching, target information query, historical trajectory playback, and video linkage call.

[0024] The beneficial effects of this invention are as follows: 1. This invention integrates multiple sensing technologies, including UWB high-precision positioning, lidar point cloud detection, video recognition, and millimeter-wave radar, to achieve precise three-dimensional spatial positioning and continuous, stable tracking of aircraft and ground vehicles on the tarmac. The UWB positioning system can achieve decimeter-level or even centimeter-level positioning accuracy in unobstructed environments, significantly outperforming traditional surface surveillance radar and multi-point positioning systems, providing a precise position reference for refined aircraft monitoring. Lidar point cloud data can directly identify the aircraft's fuselage outline and orientation, solving the problem of difficulty in determining the heading of aircraft at low speeds and when stationary using traditional methods. Multi-source data fusion effectively compensates for the performance degradation of a single sensor under conditions such as obstruction and adverse weather, improving the system's robustness and all-weather adaptability.

[0025] 2. This invention achieves intelligent management of aircraft safety protection areas through dynamic electronic fence technology. It can automatically generate safety areas of different shapes and ranges according to aircraft type and operational phase, and monitor and warn of intrusion behavior within the area in real time. The dynamic configuration information of the safety protection area is simultaneously pushed to the vehicle-mounted terminal of the ground vehicle, guiding the vehicle to actively avoid the collision and reducing the risk of collision between aircraft and ground vehicles from the source. Attached Figure Description

[0026] Figure 1 This is a schematic diagram of the overall architecture of the multi-source fusion-based apron safety supervision and aircraft monitoring system of the present invention.

[0027] Figure 2 This is a schematic diagram of the workflow of the system of the present invention.

[0028] Figure 3 This is a schematic diagram illustrating the phased setup of the dynamic electronic fence in this invention.

[0029] Figure 4 This is a comparison chart showing the consistency between system monitoring data and manually verified data during the experimental verification of this invention. Detailed Implementation

[0030] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0031] Example 1: As Figures 1-4 As shown in the figure, this embodiment provides a multi-source fusion-based apron safety supervision and aircraft monitoring system, which includes five layers: multi-source perception layer, data transmission layer, data fusion processing layer, intelligent analysis and decision-making layer, and visualization display layer.

[0032] The multi-source sensing layer is deployed in various aircraft stands, taxiway intersections, and key passageways on the apron. Specifically, 4 to 6 UWB positioning base stations are set up around each aircraft stand, with the distance between base stations controlled within 50 meters. The positioning network formed by multiple base stations achieves comprehensive coverage of the aircraft stand area. The UWB positioning base stations adopt the TDOA (Time Difference of Arrival) positioning principle, and the base stations achieve nanosecond-level time synchronization through a high-precision time synchronization protocol. UWB positioning tags are installed on aircraft and apron vehicles. The tags actively transmit UWB pulse signals at a frequency of 10 to 50 times per second. After receiving the signals, each base station records the signal arrival time and uploads the data to the positioning calculation server. The positioning calculation server calculates the three-dimensional spatial coordinates and movement speed of the tags using a hyperbolic positioning algorithm. The lidar scanning device is mounted on fixed poles on both sides of each aircraft stand, with the scanning direction facing the stand area and the scanning angle covering the aircraft's parking position and its surrounding area. The lidar adopts a 360-degree rotating scanning method, acquiring hundreds of thousands of three-dimensional point cloud data points per second. The lidar measures the distance from the target surface point to the radar by emitting laser pulses and receiving the echo. The data processing unit filters and denoises the raw point cloud data, removing outliers and ground points, and then uses a point cloud clustering algorithm to identify the aircraft fuselage outline and the ground vehicle outline, and calculates the coordinates of the aircraft fuselage center point and the nose angle. High-definition video cameras are installed at various aircraft positions and key passageways on the apron. These cameras are high-definition network cameras with a resolution of 4 megapixels or higher, featuring night vision capabilities and maintaining clear image quality even in low-light conditions. The cameras cover key locations such as aircraft positions, taxiway intersections, vehicle passageways, and personnel passageways, achieving comprehensive apron monitoring without blind spots. Some key locations utilize panoramic stitching cameras, employing multi-lens stitching technology to obtain 180-degree or 360-degree panoramic monitoring images. Millimeter-wave radar detection devices are installed on fixed poles at the intersection of the apron taxiway and the access passage to the aircraft position. They are used to detect the motion parameters of moving targets. Millimeter-wave radar measures the target's distance, speed and azimuth by transmitting frequency-modulated continuous wave signals and receiving echoes. It can maintain stable detection performance under adverse weather conditions such as heavy rain and fog, making up for the performance degradation of video and lidar under adverse weather conditions.

[0033] Passive radar detection devices are installed at high points around the apron and in key locations within the airspace to passively receive the scattered echoes of electromagnetic signals from drones and birds. Passive radar does not actively emit electromagnetic waves; it uses existing broadcast television signals, communication base station signals, or satellite navigation signals in the urban environment as external radiation sources. When drones or birds pass through the coverage areas of these signals, they scatter the signals. The passive radar receiving antenna array captures the scattered signals and analyzes the time delay and Doppler frequency shift to retrieve the target's position and velocity information. The acoustic array detection device is set up in the apron area and the edge of the clearance zone. It consists of multiple high-sensitivity microphones arranged in a cross or circle. Each array contains at least 4 microphones. The direction and distance of the sound source are calculated by the time difference of arrival method. Thermal imaging cameras are installed in various aircraft positions and key passages on the apron. They use uncooled infrared detectors and can detect the temperature difference between the human body and the environment, locating the personnel in a completely dark environment. The distributed fiber optic sensing device is installed below the ground (approximately 0.2 meters deep) around the apron perimeter and aircraft position. It uses optical fiber as the sensing medium and detects ground vibration and pressure by measuring changes in backscattered light in the optical fiber. The positioning accuracy can reach several meters. The data transmission layer includes 5G-AeroMACS communication base stations and fiber optic communication networks located in the apron area. 5G-AeroMACS is an aviation airport mobile communication system based on 5G technology, featuring high bandwidth, low latency, and high reliability, and is specifically designed for airport surface communication. High-bandwidth data such as video data is transmitted to the central processing platform via the fiber optic network, while UWB positioning data, lidar point cloud data, millimeter-wave radar data, and some control commands are wirelessly transmitted via the 5G-AeroMACS network. The data fusion processing layer includes a multi-source data spatiotemporal alignment module, a target recognition and tracking module, a fusion positioning module, a low-slow-small target recognition and tracking module, and a non-cooperative object detection and tracking module, which are set on the server cluster of the central processing platform; The non-cooperative object detection and tracking module performs detection and positioning by fusing human attribute recognition data sources and motion positioning data sources. The human attribute recognition data source includes at least one of a thermal imaging camera and a video camera. The thermal imaging camera is used to detect human thermal radiation characteristics, and the video camera is used to detect human bodies. The motion positioning data source includes at least one of a millimeter-wave radar and a distributed fiber optic sensing device. The millimeter-wave radar is used to detect moving target parameters, and the distributed fiber optic sensing device is used to detect ground vibration signals. The non-cooperative object detection and tracking module spatiotemporally correlates the human attribute recognition results with the motion positioning results. After successful fusion, it assigns a temporary tracking ID to the non-cooperative object and continuously records its motion trajectory. When the non-cooperative object enters the aircraft security protection area or stays in the prohibited area for more than a preset threshold, it generates a non-cooperative object intrusion warning.

[0034] The intelligent analysis and decision-making layer includes a dynamic electronic fence management module, a collision risk analysis module, a monitoring status management module, an early warning and handling module, and an automatic monitoring module for post-flight periods or nighttime shutdown periods; The visualization layer includes a 3D digital twin engine and user terminals. The user terminals include a large-screen display terminal in the apron operation monitoring center, workstation terminals for management personnel, and handheld mobile terminals for monitoring personnel.

[0035] Example 2: This example details the specific implementation of the multi-source data spatiotemporal alignment module and the fusion positioning module; The multi-source data spatiotemporal alignment module performs the following alignment steps; The first step is to establish a global coordinate system for the apron. This involves inputting the coordinates of all fixed reference points from the high-precision GIS map of the apron into the system, including the coordinates of the four corner points of the gate markings, the coordinates of the fixed points of the jet bridges, the three-dimensional spatial coordinates of each UWB positioning base station, the three-dimensional spatial coordinates of each lidar installation location, the installation location coordinates of each video camera, and the three-dimensional spatial coordinates of each millimeter-wave radar installation location. All coordinates use a unified coordinate system, such as a local coordinate system or a national geodetic coordinate system with the airport runway center point as the origin. The scale accuracy of the apron GIS map reaches 1:500, accurately reflecting the spatial distribution of geographical elements such as roads, gate markings, and buildings in the apron area. The second step is to establish a unified time reference. A high-precision time reference is obtained through GPS or BeiDou timing devices located at key positions on the apron, achieving microsecond-level accuracy. All sensing devices maintain synchronization with the time reference via Network Time Protocol (NTP) or Precision Time Protocol (PTP). UWB positioning base stations achieve nanosecond-level synchronization accuracy through wired time synchronization cables, ensuring the accuracy of time difference measurements. Video cameras use the IEEE 1588 PTP protocol for time synchronization, ensuring that the time error between multiple video images is controlled within 1 millisecond. LiDAR and millimeter-wave radar employ hardware-triggered synchronization, with the central processing platform sending synchronization pulse signals to trigger data acquisition, ensuring microsecond-level time alignment accuracy between point cloud data and radar data. The third step is a unified transformation of spatial coordinates. For video image data, the mapping relationship between image pixel coordinates and the global coordinates of the apron is first established through camera calibration technology. The camera calibration adopts the Zhang Zhengyou calibration method, using a checkerboard calibration board to acquire calibration images at multiple locations in the camera position area, and calculating the camera's intrinsic parameter matrix, distortion coefficients, and extrinsic parameter matrix. After calibration, for the target pixel coordinates detected in the video image, their spatial coordinates in the global coordinate system are calculated using the calibration parameters. For lidar point cloud data, each lidar is calibrated during installation, and the rotation matrix and translation vector of the lidar coordinate system relative to the global coordinate system are measured. For each data point in the point cloud, its polar coordinates in the lidar coordinate system are converted to rectangular coordinates, and then transformed to the global coordinate system through a coordinate transformation matrix. For UWB positioning data, the positioning calculation server directly outputs the three-dimensional coordinates in the global coordinate system; For millimeter-wave radar data, the transformation relationship from the radar coordinate system to the global coordinate system is established by calibrating the installation position, and the target distance and azimuth measured by the radar are converted into global coordinates; The fourth step is multi-source data association. After spatial coordinate unification and time alignment, the multi-source data are associated by timestamp to form a time-space synchronized multi-source dataset. Specifically, with a time window of 0.1 seconds, data from all data sources whose timestamps fall within this window are grouped together to form a multi-source observation data set for that moment. The fusion positioning module uses a weighted fusion method to calculate the fused position coordinates of the target. For an aircraft target, at the same time t, it may simultaneously obtain a position coordinate from the UWB positioning system, one or more position coordinates from the lidar detection system, a position coordinate from the video recognition system, and a position coordinate from the millimeter-wave radar. The fusion positioning module first assesses the confidence of each data source. The confidence assessment is based on the following factors: the statistical value of the positioning accuracy of the data source in the historical period, i.e., the root mean square of the deviation between the historical position output of the data source and the fused position; current environmental conditions, such as video confidence automatically decreasing at night and in rainy or foggy weather, lidar confidence decreasing in heavy rain, and UWB positioning confidence potentially decreasing due to multipath effects when the aircraft is close to metal obstacles; the confidence weights of each data source are dynamically adjusted through a pre-trained weight allocation model, with the total weights being 1. For a given time t, assuming there are n data sources detecting the same target, and the target location coordinates output by each data source are P1, P2, ..., Pn, and the confidence weights of each data source are w1, w2, ..., wn, then the fused location coordinate Pf is calculated by weighted summation: Pf = (w1 × P1 + w2 × P2 + ... + wn * Pn) / (w1 + w2 + ... + wn); when a data source does not detect the target at that time, its weight is reset to zero, and the fusion result is calculated by the other data sources. Experimental data show that after adopting the above weighted fusion method, the average positioning error of the fusion positioning is 0.12 meters and the maximum positioning error is 0.28 meters, which is much lower than the average positioning error of 0.21 meters for single UWB positioning and the average positioning error of 0.35 meters for single lidar positioning. The stability index of fusion positioning (standard deviation of positioning error) is also significantly better than any single data source.

[0036] Example 3: This example details the specific implementation of the target recognition and tracking module; The target recognition and tracking module employs different recognition and tracking strategies for different types of targets.

[0037] For aircraft targets, the first step is to use UWB positioning data to provide the position coordinates of the UWB tags installed on the aircraft. These coordinates serve as a rough prior to the aircraft's position. Since UWB tags are usually installed near the aircraft's nose landing gear or on the fuselage belly, these coordinates are not the geometric center of the aircraft. The system database pre-stores the offset of the UWB tag installation position relative to the fuselage geometric center for each aircraft type. This offset is obtained through actual measurement. After obtaining a rough prior location, the lidar point cloud data is searched within a preset range around the location (e.g., a spherical area with a radius of 15 meters centered on the UWB positioning point). The lidar data processing unit performs cluster analysis on the point cloud data falling within the search area, and uses the density-based DBSCAN clustering algorithm to identify point cloud clusters belonging to the aircraft fuselage. Since the aircraft fuselage is large and has a continuous surface, the number of points in its point cloud clusters is much larger than that of other targets such as ground vehicles and personnel. The size and shape characteristics of the point cloud clusters can distinguish it from surrounding targets. After identifying the point cloud clusters of the aircraft fuselage, the geometric parameters of the aircraft are further calculated: the length, width, and height of the fuselage are calculated using the extreme values ​​of the point cloud cluster coordinates; the centroid coordinates of the point cloud cluster are used as the position coordinates of the geometric center of the fuselage; and the principal direction of the point cloud clusters is calculated using the principal component analysis algorithm, which is then used as the aircraft's nose heading angle. When the aircraft is stationary, the point cloud data from the lidar can still stably identify the fuselage outline and heading, effectively solving the problem of difficulty in determining the heading under low speed and stationary conditions using traditional methods. Meanwhile, millimeter-wave radar data provides the aircraft's velocity vector and heading angle parameters; millimeter-wave radar measures the target's radial velocity through the Doppler effect, calculates the target's velocity vector through a multi-target tracking algorithm, and correlates and matches the measurement results with the aircraft identified by lidar; By fusing the coarse position from UWB positioning, the precise geometric center and orientation identified by lidar, and the velocity vector measured by millimeter-wave radar, a complete motion state information of the aircraft target is finally formed, including: three-dimensional coordinates, fuselage length, fuselage width, fuselage height, nose heading angle, velocity vector, and acceleration. For ground vehicle targets, UWB positioning data is prioritized as the primary positioning source. Vehicles equipped with UWB tags can be identified by their unique identifiers, providing their type and identity information. The UWB positioning data directly provides the vehicle's three-dimensional coordinates. Simultaneously, a video recognition system performs target detection and license plate recognition, correlating the video recognition results with the UWB tag information to confirm the vehicle type and correct for potential UWB positioning errors. For vehicles without UWB tags, target detection and continuous tracking are performed using video recognition, with millimeter-wave radar data used for auxiliary positioning.

[0038] Example 4: This example details the specific implementation of the dynamic electronic fence management module, such as... Figure 3 As shown; The dynamic electronic fence management module dynamically generates aircraft safety protection zones based on the aircraft type and current operational phase. These safety protection zones are presented as semi-transparent boundaries in the 3D digital twin scene. Aircraft types are categorized into four classes: C, D, E, and F, based on fuselage length and wingspan. Class C aircraft, such as the Boeing 737 and Airbus A320 series, have a fuselage length of approximately 30-40 meters and a wingspan of approximately 30-36 meters; Class D aircraft, such as the Boeing 767 and Airbus A330 series, have a fuselage length of approximately 50-60 meters and a wingspan of approximately 45-50 meters; Class E aircraft, such as the Boeing 777 and Airbus A350 series, have a fuselage length of approximately 60-70 meters and a wingspan of approximately 60-65 meters; and Class F aircraft, such as the Airbus A380, have a fuselage length of approximately 73 meters and a wingspan of approximately 80 meters. Different types correspond to different basic parameters for the size of the safety protection zone: the basic outward extension distance for Class C safety protection zones is 5 meters, for Class D it is 6 meters, for Class E it is 7 meters, and for Class F it is 8 meters. Aircraft operation is divided into three phases: entry, parking support, and departure / pushback. The shape and range parameters of the safety protection zone differ in each phase.

[0039] The safety protection zone during the approach phase is a conical area extending forward along the taxiway. When the aircraft turns from the taxiway into the parking position taxiway, the approach phase begins, and the system tracks the aircraft's position and speed in real time. The safety protection zone extends forward along the taxiway with the front of the aircraft's nose as the apex, forming a fan-shaped or conical area. The angle of the fan is adjusted according to the width parameter of the parking position, usually set to 60 to 90 degrees. The length of the conical area is dynamically calculated based on the aircraft's speed and braking distance; the faster the speed, the longer the conical area. The safety protection zone during the parking support phase is a closed area extending outward from the outer perimeter of the aircraft fuselage. The projection of this closed area onto a plane is a rounded rectangle or ellipse containing the aircraft fuselage projection. The minimum distance from the zone boundary to the fuselage is the base outward extension distance. Further subdivisions are established: a cabin door protection zone, located outside each aircraft door, a circular area with a radius of 3 meters centered on the door; an engine hazard zone, located in front of the aircraft engine intake and behind the exhaust nozzle, distributed in a fan shape; a service passageway protection zone, extending along both sides of the aircraft fuselage, with a width 1.5 times the base outward extension distance; and a wingtip protection zone, located on the outer sides of the aircraft wingtips, extending outward by a distance twice the base outward extension distance. Different intrusion thresholds and warning rules are set for different sub-zones. The safety protection zone during the takeoff phase is the warning zone on both sides of the takeoff ramp behind the aircraft; the safety protection zone is a long strip-shaped warning zone extending a certain distance on both sides of the centerline of the takeoff ramp, extending in front of the takeoff ramp to the taxiway interface. The boundary parameters of the safety protection zone are dynamically updated based on the real-time monitored actual position of the aircraft, moving with the aircraft's movement. The dynamic configuration information of the safety protection zone is pushed to the vehicle-mounted terminal of the ground vehicle through the 5G-AeroMACS network. The vehicle-mounted terminal receives the electronic fence information in real time and displays the boundary of the safety protection zone near the current vehicle location with a highlighted border on the vehicle navigation interface. When the vehicle approaches the boundary of the safety protection zone, the vehicle-mounted terminal emits a beeping sound and flashing lights as a warning. When the vehicle intrudes into the safety protection zone, the vehicle-mounted terminal issues a continuous alarm and automatically records the intrusion event.

[0040] Example 5: This example details the specific implementation of the collision risk analysis module; The collision risk analysis module performs the following risk calculation steps.

[0041] The first step is to acquire the fused positioning information and motion parameters of all aircraft and ground vehicle targets. Each target includes the following information: target type identifier, unique identification number, current 3D coordinates, current velocity vector, current acceleration, and target size parameters; The second step is to calculate the collision risk index for each pair of targets that may collide. The system uses a preset time window (e.g., the next 5 seconds) as the prediction range, assumes that all targets maintain uniform or uniformly accelerated motion under the current velocity vector and acceleration, and calculates the predicted trajectory points of each target in the future time series. For any pair of targets i and j, the predicted trajectory points at time t are denoted as follows: and The Euclidean distance between the two is The safe distance threshold S is dynamically set based on the target type and size; the calculation starts from the current moment. The moment when the value is first less than S is denoted as the estimated collision time. ;like If it is already less than S at the current moment, then The value is 0; simultaneously, the minimum predicted distance is calculated. That is, within the prediction time window Minimum value; Collision risk index Taking all factors into consideration and Two factors are used in the calculation; collision risk index Taking all factors into consideration and Two factors are used in the calculation; when When the distance is greater than the safe distance threshold S, =0; when When less than or equal to S, Calculation and They are inversely proportional, taking into account the relative speeds of the two targets; specifically, Where α is the risk adjustment coefficient and β is the target type weighting coefficient. It is the relative velocity; The third step is to issue graded warnings based on the comparison between the collision risk index Rij and the preset thresholds. Three warning thresholds are set: the first threshold is 0.3, the second threshold is 0.6, and the third threshold is 0.9. When Rij exceeds the first threshold, a level one warning is generated; when it exceeds the second threshold, a level two warning is generated; and when it exceeds the third threshold, a level three warning is generated. In addition to the collision risk between moving targets, the collision risk analysis module also calculates the collision risk between the aircraft and the fixed facilities at the parking position. The position coordinates of the fixed facilities are pre-entered into the system database, including the position of the boarding bridge, the position of the stop line, and the position of the parking sign. When the aircraft is in the parking or parking phase, the system calculates the minimum distance between various parts of the aircraft fuselage and the fixed facilities in real time. When the distance is less than the safety margin, an early warning is triggered.

[0042] Example 6: This example details the specific implementation of the monitoring status management module.

[0043] The monitoring status management module establishes an independent monitoring file for each parked aircraft. The monitoring file is managed by flight number and is created from the moment the aircraft enters the parking area until the aircraft has completely taxied away from the parking position. The file is archived and saved for 90 days. The guardianship file includes the following information.

[0044] The aircraft basic information section records the flight number, scheduled departure time, scheduled arrival time, aircraft type, registration number, gate number, estimated arrival time, and estimated departure time. This information is automatically imported from the airport operations database. The operational status recording section records the start and end times of various support processes, including refueling, water replenishment, sewage disposal, catering, baggage loading and unloading, passenger boarding bridge docking, and jet bridge docking. Operational status is obtained through several methods: automatic synchronization via data interface with the airport ground service system; detection of vehicle arrivals and departures via video recognition system; and manual confirmation by on-site personnel via handheld terminals. The system sets a standard operating time for each support process; when the operating time of a particular process exceeds 1.5 times the standard time, the system automatically generates a notification. The personnel and vehicle entry / exit record section records the time and path of all ground support personnel and vehicles approaching the aircraft entering and exiting the security protection area. The system identifies the vehicle and personnel entering the security protection area through UWB tags and records the entry time, entry location, stay time, and departure time for each entry event. For personnel and vehicles without UWB tags, the system uses video recognition and facial recognition technology for detection and recording. All entry / exit records are compared with the shift work plan, and unauthorized entry behavior is automatically marked as a violation event.

[0045] The risk event log section records all early warning events generated by the system, including the early warning type, warning level, time of occurrence, target identifier, location coordinates, risk index value, and handling result. Each risk event record is associated with the corresponding time period's video surveillance recording index, supporting one-click playback. For Level 3 emergency early warning events, the system automatically extracts 30-second video clips before and after the early warning and saves them separately. The aircraft entry and exit time node recording section records the time when the aircraft enters the parking space, the time when the aircraft completes its entry, the time when the aircraft pushes out of the parking space, and the time when the aircraft leaves the parking space. All records in the monitoring file are associated with the corresponding video surveillance recording index for the time period, allowing for review of on-site videos by time. Example 7: This example details the specific implementation of the early warning and handling module and the visualization display module; The early warning and response module generates early warning information including early warning level, early warning type, early warning target identifier, early warning location coordinates, and recommended response measures; the early warning information is in the format of structured data, which facilitates unified parsing and display across different terminals; The warning levels are divided into three levels: Level 1, Level 2, and Level 3, based on the collision risk index. The warning types include aircraft-to-aircraft collision warning, aircraft-to-vehicle collision warning, vehicle-to-vehicle collision warning, aircraft-to-fixed-facilities collision warning, electronic fence intrusion warning, support operation overtime warning, drone intrusion warning, bird strike risk warning, and non-cooperative object intrusion warning. Recommended response measures vary depending on the type and level of the warning. For example, for a Level 1 inter-aircraft collision warning, the recommended response is to monitor the relative positions of the two aircraft and notify the relevant crews to take precautions. For a Level 2 inter-aircraft vehicle collision warning, the recommended response is to notify the relevant vehicles to immediately slow down or stop to avoid the collision. For a Level 3 drone intrusion warning, the recommended response is to immediately activate the drone countermeasures procedure and notify the control tower to suspend takeoffs and landings. The early warning information is pushed to the following terminals via the 5G-AeroMACS network: the large screen display terminal of the apron operation monitoring center, the mobile terminals of relevant aircraft crews, the vehicle-mounted terminals of relevant ground vehicles, and the handheld mobile terminals of relevant monitoring personnel. The push strategy is set according to the early warning level: Level 1 early warnings are only pushed to the large screen of the monitoring center and the handheld terminals of relevant monitoring personnel; Level 2 early warnings are pushed to the vehicle-mounted terminals of relevant ground vehicles in addition to the push range of Level 1 early warnings; Level 3 early warnings are pushed to the mobile terminals of aircraft crews in addition to the push range of Level 2 early warnings, and are also pushed to the apron duty manager and the operation command center. After the warning information is pushed out, the system continuously tracks the status changes of the warning event; when the position and motion state of the target involved in the warning event changes, causing the collision risk index to drop below the warning threshold and remain stable for more than 2 seconds, the system automatically lifts the warning status; when the warning event continues to worsen, and the collision risk index continues to rise or exceeds the upper limit of the current warning level, the system automatically upgrades the warning level and expands the push range. The visualization layer constructs a 3D virtual scene of apron operation through a 3D digital twin engine; the 3D digital twin engine operates based on a high-precision GIS map of the apron and a library of 3D digital models of aircraft and vehicles; the high-precision GIS map of the apron has a scale accuracy of 1:200 and includes the 3D coordinates and geometric information of all ground and fixed facilities such as gate markings, taxiways, boarding bridges, gate signs, stop lines, signs, and streetlights; The 3D digital model library pre-stores 3D models of all mainstream aircraft types currently in civil aviation operations, as well as 3D models of various ground support vehicles; each 3D model is labeled with the coordinates of key parts; The target fusion position coordinates and nose-facing angle output by the fusion positioning module are matched and bound to the corresponding 3D model. Specifically, for each target, the system retrieves its corresponding 3D model from the 3D digital model library, aligns the origin of the model's local coordinate system with the target's fusion position coordinates, aligns the model's orientation with the target's nose-facing angle, and matches the model's scaling ratio with the target's actual size parameters. After alignment, the 3D model is presented in the virtual scene with the correct posture and position. The system drives the position and posture updates of the 3D model at a refresh rate of more than 30 frames per second. The visualization layer also provides the following interactive functions: scene perspective switching, target information query, historical trajectory playback, and video linkage.

[0046] Example 8: Low-Speed ​​Small Target Recognition and Differentiation Experiment This embodiment verifies the invention's ability to identify and distinguish between drones and birds through testing in a real airport environment; Experimental scenario setup: The experiment was conducted at the edge of the airspace of a large airport, covering a circular area with a radius of 5 kilometers centered on the airport runway center. Four sets of passive radar detection devices were deployed within the experimental area (located on the top of the airport terminal building and the control tower), and 12 sets of acoustic array detection devices were deployed (located in the apron areas and at the edge of the airspace). During the experiment, under the premise of ensuring flight safety, a total of 80 sorties of various types of drones (DJI Mavic 3, DJI Inspire 2, DJI Matrice 300) were launched. At the same time, the natural bird flocks (mainly pigeons and sparrows) at the airport were used as bird samples, and a total of 156 bird activities were recorded.

[0047] Experimental Method: Each time a drone was launched or bird activity was observed, the experimenter recorded the actual target type, target model, flight trajectory, and flight time. The system automatically recorded the detection data from the passive radar and acoustic array and output the identification results. The system output was compared with the actual records to calculate the identification accuracy, false alarm rate, and false negative rate.

[0048] Experimental results are shown in the table below.

[0049]

[0050] The recognition accuracy reflects the correctness of the system's judgment on low, slow, and small target types; the overall recognition accuracy of 91.9% is much higher than that of existing methods based on radar Doppler features (usually 70%-80%); the recognition accuracy of UAVs (95.0%) is higher than that of birds (90.4%), mainly because the acoustic signatures of some small birds are weaker; the false alarm rate of 4.2% and the false alarm rate of 3.8% indicate that the system has high practical value.

[0051] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-source fusion-based apron safety monitoring and aircraft surveillance system, characterized in that, include: The multi-source sensing layer includes UWB positioning base stations, lidar, video cameras, millimeter-wave radar, passive radar detection devices, acoustic array detection devices, thermal imaging cameras, and distributed fiber optic sensing devices. The data transmission layer is used to transmit data from the perception layer to the central processing platform in real time. The data fusion processing layer includes a spatiotemporal alignment module, a target recognition and tracking module, a fusion localization module, a low-speed, small target recognition and tracking module, and a non-cooperative object detection and tracking module; The intelligent analysis and decision-making layer includes a dynamic electronic fence management module, a collision risk analysis module, a monitoring status management module, an early warning and handling module, and an automatic monitoring module for post-flight periods or nighttime shutdown periods; The visualization layer includes a 3D digital twin engine and a user terminal; The low-speed small target identification and tracking module distinguishes between drones and birds by fusing passive radar detection data with acoustic array acoustic signature features. The non-cooperative object detection and tracking module detects and tracks personnel who are not carrying electronic tags by fusing thermal imaging data, millimeter-wave radar data, and fiber optic sensing data. The automatic monitoring module is automatically activated after flights are grounded or during nighttime shutdown periods. It increases the frequency and sensitivity of sensor sampling and provides fully automatic unattended monitoring of parked aircraft.

2. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The low-slow-speed small target recognition and tracking module compares the collected voiceprint features with the pre-trained classification model. The classification model outputs the probability value of the target belonging to a drone or a bird. When the probability of a drone exceeds a preset threshold, it is determined to be a drone threat and triggers a drone intrusion warning. When the probability of a bird exceeds a preset threshold, it is determined to be a bird risk and a bird heat map is recorded.

3. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The non-cooperative object detection and tracking module performs detection and positioning by fusing human attribute recognition data sources and motion positioning data sources. The human attribute recognition data source includes at least one of a thermal imaging camera and a video camera. The thermal imaging camera is used to detect human thermal radiation characteristics, and the video camera is used to perform human detection. The motion positioning data source includes at least one of a millimeter-wave radar and a distributed fiber optic sensing device. The millimeter-wave radar is used to detect moving target parameters, and the distributed fiber optic sensing device is used to detect ground vibration signals. The non-cooperative object detection and tracking module spatiotemporally correlates the human attribute recognition results with the motion positioning results. After successful fusion, it assigns a temporary tracking ID to the non-cooperative object and continuously records its motion trajectory. When the non-cooperative object enters the aircraft security protection area or stays in the prohibited area for more than a preset threshold, it generates a non-cooperative object intrusion warning.

4. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The automatic monitoring module for the post-flight period or nighttime shutdown period is automatically activated after the daily flight operation ends, enters the silent period baseline learning mode to record background environmental data of the apron area, and establishes an enhanced safety protection zone for parked aircraft. The enhanced safety protection zone is generated based on the safety protection zone during the parking support phase. The safety protection zone during the parking support phase is a closed area that extends outward from the outer perimeter of the aircraft fuselage by a preset distance. The linear distance between the outer boundary of the enhanced safety protection zone and the outer perimeter of the aircraft fuselage is 1.5 times the preset distance. The entry of any personnel or vehicle into the enhanced security protection area will immediately trigger a Level 3 warning and automatically generate a monitoring report for the post-flight period or the nighttime shutdown period.

5. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The dynamic electronic fence management module dynamically generates aircraft safety protection zones based on the aircraft type and the current operational phase. The operational phases include the entry phase, the parking support phase, and the departure and pushback phase. The safety protection zone for the entry phase is a conical area extending forward along the taxiway. The safety protection zone for the parking support phase is a closed area extending outward from the outer perimeter of the aircraft fuselage at a preset distance and is divided into zones including a cabin door protection zone and an engine danger zone. The safety protection zone for the departure and pushback phase is the warning zone on both sides of the pushback passage behind the aircraft.

6. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 5, characterized in that, The dynamic configuration information of the security protection area is pushed to the vehicle terminal of the ground vehicle through the 5G-AeroMACS network. When the vehicle approaches the boundary of the security protection area, the vehicle terminal issues an audible and visual alert. When the vehicle intrudes into the security protection area, the vehicle terminal issues a continuous alarm and automatically records the intrusion event.

7. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The collision risk analysis module acquires the fused positioning information and motion parameters of all aircraft targets and ground vehicle targets. For each pair of targets that may collide, it calculates the minimum expected distance and the expected collision time, and generates a collision risk index through comprehensive evaluation. When the collision risk index exceeds the first threshold, a level 1 warning is generated; when it exceeds the second threshold, a level 2 warning is generated; and when it exceeds the third threshold, a level 3 warning is generated.

8. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The monitoring status management module establishes an independent monitoring file for each parked aircraft. The monitoring file includes basic aircraft information, support operation status records, personnel and vehicle entry and exit records, risk event records, and time node records of aircraft entry and exit. All records are associated with the corresponding video surveillance recording index for the time period.

9. The system according to claim 1, characterized in that, The early warning and response module generates early warning information including early warning level, early warning type, early warning target identifier, early warning location coordinates, and recommended response measures. This information is pushed to the display terminal of the apron operation monitoring center, the aircraft crew mobile terminal, the ground vehicle vehicle terminal, and the handheld mobile terminal of the monitoring personnel via the 5G-AeroMACS network. After the early warning information is pushed, the module continuously tracks the changes in the status of the early warning event and automatically cancels or upgrades the early warning.

10. The apron safety monitoring and aircraft surveillance system based on multi-source fusion according to claim 1, characterized in that, The 3D digital twin engine constructs a 3D virtual scene of apron operation based on a high-precision GIS map of the apron and a library of 3D digital models of aircraft and vehicles. The target fusion position coordinates and orientation angles output by the fusion positioning module are matched and bound with the corresponding 3D models, driving the 3D models to be presented in the virtual scene with the correct posture and position. The visualization display layer also provides scene perspective switching, target information query, historical trajectory playback and video linkage calling functions.

Images