Airborne environment monitoring and obstacle early warning system and implementation method
By employing a multi-sensor collaborative layout and data synchronization, the system addresses the issues of insufficient coverage and modal fusion in airborne environmental monitoring systems, achieving high-precision obstacle identification and early warning, and supporting pilots' safe flight in complex environments.
Patent Information
- Application Number
- CN202511407099.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-29
- Publication Date
- 2025-12-23
AI Technical Summary
Existing airborne environmental monitoring and obstacle warning systems in the civil aviation field suffer from problems such as insufficient coverage of single sensors, poor robustness, lack of modal fusion and multi-dimensional risk assessment capabilities, and inability to adapt to complex weather and aircraft attitude changes, leading to perception failure and lack of deep integration with avionics systems.
It adopts a multi-sensor collaborative layout, including a forward-looking sensing module, a side-view camera module, a millimeter-wave radar system, and an airborne computing platform. Combined with a data synchronization and coordinate transformation module, it achieves unified alignment and fusion of sensor data. By reconstructing features through the BEV coordinate system, it dynamically adjusts the perception strategy, supports obstacle detection and early warning during the flight phase, and performs downgraded fusion when modes are missing.
In complex weather and highly dynamic environments, it achieves high-precision obstacle recognition and early warning, maintains a sensing frequency above 5Hz, reduces false alarm rate, improves system availability, and supports coordinated control of avionics systems.
Smart Images

Figure CN121191367A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of civil aviation auxiliary flight systems, specifically to an airborne environmental monitoring and obstacle warning system and its implementation method. Background Technology
[0002] Currently, autonomous driving technology has achieved mature applications of multi-sensor fusion in the automotive field, using BEV perception frameworks for real-time modeling and obstacle recognition in front, to the sides, and behind. Typical systems combine visible light cameras, millimeter-wave radar, lidar, infrared sensors, and rely on high-performance neural networks to complete dynamic obstacle detection and early warning. However, in the civil aviation field, most airborne perception systems are still at the stage of radar imaging or infrared-assisted observation, and have not yet formed structured 3D environment modeling and intelligent decision support. Especially in scenarios such as low altitude, taxiing, and complex airports, pilots still mainly rely on visual observation and traditional navigation systems for judgment, lacking high-precision obstacle recognition assistance tools.
[0003] Existing airborne environmental monitoring and obstacle warning systems typically employ a single sensor, resulting in insufficient coverage and robustness. They are ill-suited for complex weather and highly dynamic environments, lack modal fusion and multi-dimensional risk assessment capabilities, and most systems fail to consider the degradation of perception capabilities under aircraft attitude changes and modal loss. Consequently, differentiated perception strategy configurations are not established for taxiing and flight phases. Furthermore, they lack the ability to deeply integrate with existing avionics systems (such as FMS), and the image processing module and point cloud processing link have not established a unified coordinate and fusion system in the aviation field. Summary of the Invention
[0004] To address the aforementioned technical problems, an airborne environmental monitoring and obstacle warning system and its implementation method are provided. This technical solution solves the problems of single-sensor solutions proposed in the background technology failing to detect under complex weather and flight attitude changes, and lacking modal fusion and flight phase adaptation capabilities.
[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: An airborne environmental monitoring and obstacle warning system includes: The hardware components include a forward-looking sensing module, a side-facing camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module. The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Image data is projected from pixel space to spatial coordinates through intrinsic and extrinsic parameter matrices. Point cloud data is aligned through filtering, coordinate transformation, and interpolation. Time synchronization is centered on the airborne master clock and achieves full-modal data alignment through timestamps. The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to BEV coordinate system, feature fusion, target detection, attitude correction, early warning logic and configuration parameter module, wherein the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
[0006] Preferably, the hardware component module specifically includes: The forward-looking sensing module is installed in front of the cockpit and includes a visible light camera, an infrared camera, a low-light camera, and a lidar. The side camera module is installed on both sides of the wing and tail section, providing a frontal and rearward field of view; The millimeter-wave radar system is deployed in multiple directions on the aircraft body, providing environmental blind spot coverage and structural redundancy; The airborne computing platform is an automotive-grade SoC system packaged to aerospace standards, integrating a neural network inference module; The cockpit display system is used to display alarm information, including a graphical interface and a voice broadcast device; The positioning and attitude perception module integrates IMU, GPS and attitude solver for flight attitude correction and coordinate projection.
[0007] Preferably, the feature extraction in the perception algorithm process specifically includes: Feature extraction is performed on image data using convolutional neural networks; Feature extraction is performed on point cloud data using a point cloud feature extraction network. During the preprocessing stage, the health status of the sensor determines whether it will participate in subsequent calculations; if the sensor fails or malfunctions, it is excluded.
[0008] Preferably, the feature fusion in the perception algorithm process involves converting image features and point cloud features to the BEV coordinate system, and then fusing feature vectors within the same grid. The fusion method includes accumulation, averaging, or finding the maximum value.
[0009] Preferably, the target detection in the perception algorithm process specifically includes: During the ground taxiing phase, detect people, vehicles, aircraft, and other obstacles; During the flight phase, detect aircraft, birds, and buildings; An object detection model is used for obstacle recognition, and post-processing is performed, including duplicate box suppression.
[0010] Preferably, the early warning logic in the perception algorithm process specifically includes: The obstacle hazard level is calculated using a multidimensional weighted fusion model, with the following formula: Where M is the obstacle hazard level, D is the distance, V is the speed, S is the volume, T is the category, and a1 to a4 are adjustable parameters; Alarm levels are categorized based on the degree of danger, and the alerts are output to the cockpit display system for red-yellow-green color prompts and voice announcements. At the same time, the alerts are output to the flight management system to support route replanning.
[0011] Preferably, the state adaptation module receives aircraft state information output by the air-to-ground judgment module from the configuration parameter module and dynamically switches the perception strategy. During the taxiing phase, the target detection range is X meters in all directions (front, back, left, and right), and during the flight phase, the target detection range is Y meters in the forward direction, where X and Y are arbitrary positive numbers.
[0012] Preferably, the sensing algorithm module supports a modality loss sensing and degradation fusion mechanism. When a single sensor fails, the remaining sensor data is used to complete the sensing task, achieving degraded operation.
[0013] A method for airborne environmental monitoring and obstacle warning includes: the hardware components include a forward-looking sensing module, a side-view camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module; The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Image data is projected from pixel space to spatial coordinates through intrinsic and extrinsic parameter matrices. Point cloud data is aligned through filtering, coordinate transformation, and interpolation. Time synchronization is centered on the airborne master clock and achieves full-modal data alignment through timestamps. The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to BEV coordinate system, feature fusion, target detection, attitude correction, early warning logic, and configuration parameter module, wherein the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
[0014] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention proposes an airborne environmental monitoring and obstacle warning system and its implementation method. By aligning the BEV coordinate system with the body coordinate system in real time, the system achieves dynamic coordinate binding through IMU / GPS fusion calculation (200Hz). Then, based on the ARINC629 bus, the system receives landing gear compression signals and throttle position, triggering millisecond-level switching between the detection range and the target database (ground mode focuses on vehicles / pedestrians, flight mode focuses on birds / buildings), completing seamless strategy switching. Finally, image features are mapped to the BEV through perspective inverse transformation, and point cloud features are aligned through voxel interpolation. This optimization of heterogeneous features improves the detection rate of small targets, reduces the false alarm rate due to dynamic range constraints, and achieves a fusion latency of <120ms, meeting the 10Hz aviation safety standard. This invention proposes an airborne environmental monitoring and obstacle warning system and its implementation method. By reconstructing feature weights using a multi-head self-attention mechanism, the system automatically enhances the confidence of the remaining modes when a single sensor fails, performs a downgrade fusion algorithm, and calculates the danger level in real time using a multi-dimensional weight model for dynamic risk assessment. Finally, when a red zone alarm (M>0.8) is triggered, a voice alarm and graphic flashing are activated, and the obstacle coordinates are pushed to the FMS to generate an avoidance path, forming avionics coordinated control. This allows the system to maintain a perception frequency of >5Hz (traditional systems ≤2Hz) even in scenarios where a single sensor fails, such as dense fog or strong light. It also improves availability when both sensors fail and achieves end-to-end protection from perception (BEV fusion) to decision-making (risk classification) and execution (FMS obstacle avoidance). Attached Figure Description
[0015] Figure 1 This is a schematic diagram of the process of the present invention; Figure 2 This is a schematic diagram of the perception fusion process in this invention; Figure 3 This is a flowchart illustrating the hardware components of the present invention. Figure 4 This is a schematic diagram of the feature extraction process in the perception algorithm of this invention; Figure 5 This is a schematic diagram of the target detection process in the perception algorithm of this invention. Detailed Implementation
[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.
[0017] Reference Figure 1 As shown, an airborne environmental monitoring and obstacle warning system and its implementation method include: The hardware components include a forward-looking sensing module, a side-facing camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module. The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Image data is projected from pixel space to spatial coordinates through intrinsic and extrinsic parameter matrices. Point cloud data is aligned through filtering, coordinate transformation, and interpolation. Time synchronization is centered on the airborne master clock and achieves full-modal data alignment through timestamps. The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to BEV coordinate system, feature fusion, target detection, attitude correction, early warning logic and configuration parameter module, wherein the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
[0018] The hardware modules adopt a multi-sensor collaborative layout: the forward-looking module (visible light / infrared / low light camera + lidar deployed in front of the cockpit) solves the problem of perception in complex lighting environments; the lateral module (mounted on both sides of the wing and tail section) achieves 270° field of view coverage; the millimeter-wave radar (deployed in multiple directions) provides all-weather detection redundancy; the onboard computing platform uses an aerospace-grade automotive-grade SoC with integrated dedicated neural network accelerator to process high-concurrency data; and the positioning module integrates IMU / GPS to achieve centimeter-level attitude correction.
[0019] The data synchronization and coordinate transformation module establishes a unified spatiotemporal reference: the pixel coordinates of multi-view images are projected to the body coordinate system through the extrinsic parameter matrix; Kalman filtering is applied to the lidar point cloud to eliminate noise, and then the coordinate system is aligned through spatial interpolation; nanosecond-level timestamps are generated using the airborne master clock, and the synchronization error is controlled within ±3ms, solving the data misalignment problem caused by high-speed movement in aviation scenarios.
[0020] The perception algorithm module adopts a BEV (Bird's Eye View) fusion architecture: in the feature extraction stage, CNN processes images / PointNet processes point clouds; heterogeneous features are transformed into a BEV grid that coincides with the body coordinate system; weighted fusion (average / maximum value strategy) is implemented on features within the same grid; and differentiated detection models are called based on the flight stage (vehicles / pedestrians are detected during taxiing, and birds / buildings are identified during flight).
[0021] The state adaptation module enables dynamic strategy switching: based on air-to-ground signal analysis (from landing gear sensors or FMS), the detection range is constrained to a spherical area with a radius of X meters during taxiing, and expanded to a forward conical airspace of Y meters during flight. Compared to traditional fixed threshold systems, the false collision rate is reduced.
[0022] The hardware components specifically include: The forward-looking sensing module is installed in front of the cockpit and includes a visible light camera, an infrared camera, a low-light camera, and a lidar. The side camera module is installed on both sides of the wing and tail section, providing a frontal and rearward field of view; The millimeter-wave radar system is deployed in multiple directions on the aircraft body, providing environmental blind spot coverage and structural redundancy; The airborne computing platform is an automotive-grade SoC system packaged to aerospace standards, integrating a neural network inference module; The cockpit display system is used to display alarm information, including a graphical interface and a voice broadcast device; The positioning and attitude perception module integrates IMU, GPS and attitude solver for flight attitude correction and coordinate projection.
[0023] The forward-looking module uses a wide-angle (120° FOV) + telephoto (30° FOV) camera group, and the vibration damping design of the bracket suppresses the impact of engine vibration. The side module is installed on the wingtip and uses a fisheye lens (190° FOV) to cover the blind spot of ground taxiing. At the same time, the millimeter-wave radar is deployed at the four corners to form a detection overlap area. Even when a single radar fails, it still maintains 80% coverage. The airborne computing platform is connected to the avionics system through the ARINC 429 bus and supports the DO-178C aviation standard. The positioning module integrates RTK-GPS (accuracy 2cm) and fiber optic IMU, and the attitude calculation frequency reaches 200Hz.
[0024] The feature extraction process in the perception algorithm specifically includes: Feature extraction is performed on image data using convolutional neural networks; Feature extraction is performed on point cloud data using a point cloud feature extraction network. During the preprocessing stage, the health status of the sensor determines whether it will participate in subsequent calculations; if the sensor fails or malfunctions, it is excluded.
[0025] The image CNN adopts the EfficientNet-B4 architecture, with an input resolution of 1280×720 and an output of 512-dimensional feature vectors. The point cloud network uses the PointPillars structure to convert the point cloud into a pseudo-image format to accelerate inference. The sensor health monitoring obtains the BIT (Built-in Test) status code through the CAN bus. When a fault occurs, the sensor data stream is automatically blocked, and the system switches to degraded mode to perform feature extraction.
[0026] The feature fusion process in the perception algorithm involves converting image features and point cloud features to the BEV coordinate system and then fusing feature vectors within the same grid. The fusion methods include accumulation, averaging, or finding the maximum value.
[0027] The body coordinate system is divided into 0.2m×0.2m grid units. Image features are projected onto the grid through inverse perspective transformation, and point cloud features are mapped through voxelization. Then, channel-level fusion is performed on features within the same grid: when the confidence difference is >0.3, the maximum value is taken; otherwise, a weighted average is used. Compared with the traditional pre-fusion scheme, BEV fusion improves the detection rate of small targets and achieves the effect of BEV feature fusion.
[0028] The target detection in the perception algorithm process specifically includes: During the ground taxiing phase, detect people, vehicles, aircraft, and other obstacles; During the flight phase, detect aircraft, birds, and buildings; An object detection model is used for obstacle recognition, and post-processing is performed, including duplicate box suppression.
[0029] During the taxiing phase, the model is optimized to detect low-lying targets (height < 1.5m) and outputs bounding boxes and velocity vectors. During the flight phase, the model focuses on high-speed targets (relative speed > 50km / h) and adopts an improved YOLOv4 architecture. In post-processing, soft-NMS (non-maximum suppression) is implemented, and the overlap threshold is dynamically adjusted (0.7 for taxiing phase / 0.5 for flight phase).
[0030] The warning logic in the perception algorithm process specifically includes: The obstacle hazard level is calculated using a multidimensional weighted fusion model, with the following formula: Where M is the obstacle hazard level, D is the distance, V is the speed, S is the volume, T is the category, and a1 to a4 are adjustable parameters; Alarm levels are categorized based on the degree of danger, and the alerts are output to the cockpit display system for red-yellow-green color prompts and voice announcements. At the same time, the alerts are output to the flight management system to support route replanning.
[0031] In the danger level formula M = 0.6D + 0.2V + 0.1S + 0.1T of the core algorithm of the warning logic: D is the normalized distance parameter (0 - 1.0), V takes the absolute value of the relative speed (km / h), S is classified into four levels according to the obstacle object integration (1 - 4), and T is assigned a weight according to the category (bird = 1.0, cloud = 0.3). Warning classification: Red (M > 0.8) triggers voice warning + graphic flashing, yellow (0.5 < M ≤ 0.8) graphic highlighting, green (M ≤ 0.5) only displays the icon.
[0032] The state adaptation module receives the aircraft state information output by the air-ground judgment module through the configuration parameter module and dynamically switches the perception strategy. The target detection range is X meters in all directions during the taxiing phase, and the target detection range is Y meters forward during the flight phase. Both X and Y are arbitrary positive numbers.
[0033] Receive air-ground signals (landing gear compression state / throttle position) through the ARINC 629 bus. The detection range X = 200 meters (configurable) during the taxiing phase, and Y = 5000 meters (configurable) during the flight phase. A 1-second delay is set for the strategy switch to prevent false triggering. The range parameters are stored in the non-volatile memory to form a dynamic state adaptation mechanism.
[0034] The perception algorithm module supports the modality missing perception and degradation fusion mechanism. When a single sensor fails, the remaining sensor data is used to complete the perception task, enabling degraded operation.
[0035] When the lidar fails, monocular depth estimation + millimeter-wave data compensation is adopted. When the visible light fails, infrared / low-light fusion imaging is started. It can still maintain a detection frequency of more than 5Hz in the degraded mode, improving the availability compared with traditional systems.
[0036] A method for realizing airborne environmental monitoring and obstacle warning, including: The hardware composition module includes a front-view sensing module, a side camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module; The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Among them, the image data completes the projection from the pixel space to the space coordinate through the internal parameter matrix and the external parameter matrix. The point cloud data is aligned through filtering, coordinate transformation, and interpolation. The time synchronization is centered on the airborne main clock, and all-modal data alignment is achieved through timestamps; The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to the BEV coordinate system, feature fusion, target detection, attitude correction, warning logic, and configuration parameter module, where the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
[0037] Time synchronization adopts a PTP (Precise Time Protocol) master-slave clock architecture, BEV coordinate transformation uses a quaternion rotation matrix to avoid gimbal lock problems, and a self-attention mechanism is added in the feature fusion stage to enhance the feature weights of key areas. Compared with existing aviation perception systems, this solution reduces computational latency.
[0038] In summary, the advantages of this invention are: improving the aircraft's ability to perceive the surrounding environment during taxiing and flight, assisting pilots in identifying obstacles, and preventing collision accidents.
[0039] 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 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 claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.
Claims
1. An airborne environmental monitoring and obstacle early warning system, characterized in that, include: The hardware components include a forward-looking sensing module, a side-facing camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module. The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Image data is projected from pixel space to spatial coordinates through intrinsic and extrinsic parameter matrices. Point cloud data is aligned through filtering, coordinate transformation, and interpolation. Time synchronization is centered on the airborne master clock and achieves full-modal data alignment through timestamps. The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to BEV coordinate system, feature fusion, target detection, attitude correction, early warning logic and configuration parameter module, wherein the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
2. The airborne environmental monitoring and obstacle early warning system according to claim 1, characterized in that, The hardware components specifically include: The forward-looking sensing module is installed in front of the cockpit and includes a visible light camera, an infrared camera, a low-light camera, and a lidar. The side camera module is installed on both sides of the wing and tail section, providing a frontal and rearward field of view; The millimeter-wave radar system is deployed in multiple directions on the aircraft body, providing environmental blind spot coverage and structural redundancy; The airborne computing platform is an automotive-grade SoC system packaged to aerospace standards, integrating a neural network inference module; The cockpit display system is used to display alarm information, including a graphical interface and a voice broadcast device; The positioning and attitude perception module integrates IMU, GPS and attitude solver for flight attitude correction and coordinate projection.
3. The airborne environmental monitoring and obstacle early warning system according to claim 2, characterized in that, The feature extraction process in the perception algorithm specifically includes: Feature extraction is performed on image data using convolutional neural networks; Feature extraction is performed on point cloud data using a point cloud feature extraction network. During the preprocessing stage, the health status of the sensor determines whether it will participate in subsequent calculations; if the sensor fails or malfunctions, it is excluded.
4. The airborne environmental monitoring and obstacle early warning system according to claim 3, characterized in that, The feature fusion process in the perception algorithm involves converting image features and point cloud features to the BEV coordinate system and then fusing feature vectors within the same grid. The fusion methods include accumulation, averaging, or finding the maximum value.
5. The airborne environmental monitoring and obstacle early warning system according to claim 4, characterized in that, The target detection in the perception algorithm process specifically includes: During the ground taxiing phase, detect people, vehicles, aircraft, and other obstacles; During the flight phase, detect aircraft, birds, and buildings; An object detection model is used for obstacle recognition, and post-processing is performed, including duplicate box suppression.
6. The airborne environmental monitoring and obstacle early warning system according to claim 5, characterized in that, The warning logic in the perception algorithm process specifically includes: The obstacle hazard level is calculated using a multidimensional weighted fusion model, with the following formula: Where M is the obstacle hazard level, D is the distance, V is the speed, S is the volume, T is the category, and a1 to a4 are adjustable parameters; Alarm levels are categorized based on the degree of danger, and the alerts are output to the cockpit display system for red-yellow-green color prompts and voice announcements. At the same time, the alerts are output to the flight management system to support route replanning.
7. The airborne environmental monitoring and obstacle early warning system according to claim 6, characterized in that, The state adaptation module receives aircraft state information output by the air-to-ground judgment module from the configuration parameter module and dynamically switches the perception strategy. During the taxiing phase, the target detection range is X meters in all directions (front, back, left, and right), and during the flight phase, the target detection range is Y meters in the forward direction, where X and Y are arbitrary positive numbers.
8. The airborne environmental monitoring and obstacle warning system according to claim 7, characterized in that, The perception algorithm module supports modal missing perception and degradation fusion mechanism. When a single sensor fails, the remaining sensor data is used to complete the perception task, realizing degraded operation.
9. A method for airborne environmental monitoring and obstacle early warning, characterized in that, Specifically, it includes: The hardware components include a forward-looking sensing module, a side-view camera module, a millimeter-wave radar system, an airborne computing platform, a cockpit display system, and a positioning and attitude perception module. The data synchronization and coordinate transformation module aligns all sensor data to the aircraft body coordinate system. Image data is projected from pixel space to spatial coordinates through intrinsic and extrinsic parameter matrices. Point cloud data is aligned through filtering, coordinate transformation, and interpolation. Time synchronization is centered on the airborne master clock and achieves full-modal data alignment through timestamps. The perception algorithm module includes sensor data acquisition and preprocessing, feature extraction, coordinate transformation to BEV coordinate system, feature fusion, target detection, attitude correction, early warning logic, and configuration parameter module, wherein the BEV coordinate system coincides with the body coordinate system; The state adaptation module dynamically adjusts the perception strategy based on the flight phase. When the aircraft is in the ground taxiing phase, it prioritizes ground obstacles and sets the detection range. When it is in the flight phase, it prioritizes aerial targets and sets the detection range.
Citation Information
Cited By
Airport runway multi-sensor cooperative target intrusion identification method and device
CN122200373A
