Four-way vision cooperation-based airport two-way taxiway hierarchical early warning method and system

Abstract

The application provides a kind of based on four-way vision coordination's airport two-way taxiway grading early warning method, system, the grading early warning method includes: in the two-way taxiway of airport two sides each disposes two road side perception equipment, respectively undertake far end, near end distance detection;Far end equipment detects that aircraft enters far end early warning threshold, issues prohibition advance signal and links same side near end equipment and starts monitoring;When near end equipment detects that aircraft reaches near end early warning threshold, output keeps waiting signal;Near end equipment sends release waiting signal after aircraft disappears in field of vision, restores vehicle traffic, relies on far and near end equipment and completes grading early warning control in coordination.The application can realize two-way taxiing full coverage and grading early warning control, adopts pure vision cooperative perception scheme, compared with radar-vision fusion scheme, hardware cost is reduced by more than 60%, and target recognition accuracy can reach 95% under complex weather such as rain and fog, with the obvious advantages of low cost, high reliability and easy deployment.

Description

Technical Field

[0001] This invention relates to the field of airport traffic control and aviation safety technology, and more specifically, to a graded early warning method and system for airport two-way taxiways based on four-way visual collaboration. Background Technology

[0002] Within the airport's flight area, before landing or preparing for takeoff, aircraft must taxi between the runway and their parking positions. Simultaneously, dozens of ground support vehicles, including passenger boarding stairs, shuttle buses, refueling trucks, towing vehicles, and de-icing vehicles, frequently move between the taxiways and the apron, creating a high-density, dynamic, and interconnected operational scenario involving aircraft and ground vehicles within a limited space. Particularly at the intersections of taxiways and vehicle lanes, where aircraft taxiing paths and vehicle routes physically intersect, these areas represent one of the most concentrated safety risks in airport ground operations. According to airport operational safety management regulations, ground vehicles must give way to aircraft at least 50 meters to one side when encountering a taxiing or towing aircraft, and are prohibited from passing within 200 meters in front of or 50 meters behind a taxiing aircraft. However, in actual operation, the implementation of the above safety requirements mainly relies on human observation and driver visual judgment. Affected by factors such as obstructed vision, weather conditions and differences in personnel experience, the risk of misjudgment and omission is high. At best, it will cause vehicles to avoid obstacles in time and delay the operation. At worst, it will cause collisions between aircraft and vehicles at intersections, posing serious safety hazards.

[0003] To address the aforementioned issues, some existing technologies have proposed sensor-based automated alarm solutions. Patent application number 202311266858.8 discloses an aircraft taxiing alarm method based on vehicle-road cooperation, employing a heterogeneous sensor fusion scheme of millimeter-wave radar and a high-definition gimbal camera. However, radar point clouds and image pixels are heterogeneous data, inherently prone to errors in timestamp alignment and spatial coordinate mapping, easily leading to target location drift. Simultaneously, millimeter-wave radar is susceptible to multipath interference in the complex electromagnetic environment of airports, affecting detection accuracy and stability. Patent application number 202111011103.4 discloses a real-time monitoring and early warning method for airport taxiways, using a deep learning model to detect left- and right-hand aircraft in video footage. However, this type of solution uses a single camera setup, resulting in a limited field of view. It only uses pixel area thresholds for region judgment, and the alarm logic is single-level, failing to cover the entire bidirectional taxiing process or achieve accurate physical ranging and hierarchical control.

[0004] Therefore, there is an urgent need for an aircraft taxiing safety alarm solution that can balance low-cost deployment with high-reliability control, providing a complete safety loop for ground vehicles from remote prohibition of entry to forced stop at the near end. Summary of the Invention

[0005] In view of this, the present invention aims to propose a graded early warning method and system for airport two-way taxiways based on four-way visual collaboration, so as to solve the problems of high risk of misjudgment and missed judgment by manual observation in the prior art, high hardware cost of radar-visual heterogeneous fusion scheme and data synchronization error.

[0006] To achieve the above objectives, the technical solution of the present invention is implemented as follows:

[0007] This invention provides a graded early warning method for airport bidirectional taxiways based on four-way visual collaboration, comprising:

[0008] Two roadside sensing devices are deployed on each side of the airport's two-way taxiway, with two devices on each side corresponding to the detection of the far-end warning distance and the near-end warning distance, respectively.

[0009] The roadside sensing device that detects the distance to the far-end warning will issue a prohibition signal when the aircraft enters the far-end warning threshold, and at the same time trigger the nearby sensing device on the same side to enter the linkage monitoring mode.

[0010] The roadside sensing device that detects the near-end warning distance issues a hold signal when it continuously detects an aircraft entering the near-end warning threshold.

[0011] When there are no aircraft targets within the monitoring field of view, the roadside sensing device that detects the near-end warning distance issues a wait-out signal, and the vehicle resumes normal traffic.

[0012] Furthermore, the specific steps include:

[0013] Step S1: Deploy two roadside sensing devices on each side of the airport taxiway. The two devices on each side are used to detect the far-end warning distance and the near-end warning distance, respectively.

[0014] Step S2: When the intelligent transfer platform enters the designated position in the intersection area of ​​the taxiway, the system automatically triggers the platform parking inspection process;

[0015] Step S3: If none of the four roadside sensing devices detect the aircraft entering the warning distance, the intelligent transfer platform resumes normal operation after receiving the permission signal; if any roadside sensing device detects the aircraft entering the corresponding warning distance, proceed to step S4.

[0016] Step S4: When the roadside sensing device detecting the far-end warning distance detects an aircraft entering the far-end warning distance, it sends a prohibition signal to the intelligent transmission platform and triggers the roadside sensing device on the same side detecting the near-end warning distance to enter the linkage monitoring mode; when the roadside sensing device detecting the near-end warning distance detects an aircraft entering the near-end warning distance, it sends a hold signal to the intelligent transmission platform; when the roadside sensing device detecting the near-end warning distance confirms during continuous monitoring that there is no longer an aircraft target in its field of view, it sends a release signal to the intelligent transmission platform.

[0017] This invention achieves full coverage of the bidirectional taxiing trajectory by deploying two roadside sensing devices on each side of the taxiway, eliminating blind spots in the warning system. At the same time, it constructs a complete safety control closed loop of "remote entry prohibition - near-end mandatory waiting - release after departure" through a dual-level detection mechanism of remote and near ends, ensuring that the warning signal is continuous and uninterrupted.

[0018] It should be noted that in step S4, the four roadside sensing devices continuously monitor the taxiway intersection area. Once any device detects an aircraft entering its corresponding warning distance, it triggers the tiered warning and coordinated control process. Specifically, for any taxiing direction, the roadside sensing device located on the side of the aircraft's approach direction is used to detect the far-end warning distance, while the roadside sensing device located on the side of the aircraft's departure direction is used to detect the near-end warning distance. This includes the following sub-steps:

[0019] Step S41: Remote early warning triggered

[0020] When the roadside sensing device that detects the far-end warning distance identifies that the aircraft has entered the far-end warning distance, it sends a prohibition signal to the intelligent transmission platform. At the same time, it synchronizes the aircraft's position, speed, and heading detection data to the roadside sensing device on the same side that detects the near-end warning distance, triggering the same-side linkage monitoring mode.

[0021] Step S42: Near-end early warning triggered

[0022] When the roadside sensing device that detects the near-end warning distance identifies that the aircraft has entered the near-end warning distance, it sends a hold signal to the intelligent transmission platform.

[0023] Step S43: Warning lifted

[0024] The roadside sensing device, which detects the near-end warning distance, continuously monitors the aircraft taxiing. When it confirms that the aircraft has completely passed the intersection and there are no more aircraft targets in its field of vision, it sends a wait-out signal to the intelligent transfer platform.

[0025] It should be noted that once an aircraft enters the detection field of any roadside sensing device, it is continuously and uninterruptedly tracked and monitored. From the moment the aircraft enters the far-end warning distance until it has completely taxied past the intersection, it remains within the detection field of the corresponding roadside sensing device, with continuous and uninterrupted monitoring signals. Once the aircraft has completely passed the intersection and continues taxiing forward, and the aircraft target is no longer visible in the detection field of the roadside sensing device detecting the near-end warning distance, the system immediately sends a release signal to the intelligent transfer platform. Upon receiving this signal, the platform resumes normal operation, thus completing a full alarm control process. Throughout this process, the triggering and cancellation of all warning signals are based on the continuous monitoring results of the roadside sensing devices on the aircraft target, rather than relying on intermittent sampling or single-frame judgments. This effectively ensures the continuity and reliability of the warning response and avoids misjudgments or missed judgments due to monitoring gaps.

[0026] In this invention, four roadside sensing devices are installed near the taxiway intersection area. The first and second roadside sensing devices are deployed on the right side, while the third and fourth roadside sensing devices are deployed on the left side. The devices on both sides are symmetrically installed, with devices on the same side arranged forward and backward along the aircraft's taxiing direction, and the detection fields of adjacent devices are seamlessly connected. By adjusting the orientation angle, field of view, and camera calibration parameters of the devices, the detection fields of the first and third roadside sensing devices cover the location of the far-end warning distance, used to detect whether the aircraft has entered the far-end warning distance; the detection fields of the second and fourth roadside sensing devices cover the location of the near-end warning distance, used to detect whether the aircraft has entered the near-end warning distance.

[0027] Taking an aircraft taxiing forward (from right to left) as an example, the two devices on the same side are installed on opposite sides of the intersection:

[0028] The first roadside sensing device (right-side far end) is installed at the front end of the intersection (i.e. the side facing the direction the aircraft is approaching). Its detection field of view faces the direction the aircraft is approaching and covers the location of the far warning line. It is used to detect whether the aircraft has entered the far warning distance.

[0029] The second roadside sensing device (right-side near end) is installed at the rear of the intersection (i.e., the side in which the aircraft is leaving). Its detection field of view faces the center area of ​​the intersection, covering the location of the near-end warning line, and is used to detect whether the aircraft has entered the near-end warning distance.

[0030] In this invention, the response time from the moment the aircraft enters the far-end warning distance or the near-end warning distance until the intelligent transmission platform receives the corresponding signal is no more than 2 seconds.

[0031] Furthermore, the roadside sensing device uses a purely visual perception method to collect aircraft taxiing image data, identifies aircraft targets and extracts their contours through multi-scale edge detection and a lightweight deep learning model, and calculates the instantaneous speed and orientation of the aircraft by combining a physical mapping algorithm of pixel displacement and camera calibration parameters.

[0032] This invention employs a pure visual perception approach, reducing hardware costs by more than 60% compared to radar solutions. It also avoids timestamp alignment errors and spatial coordinate mapping deviations caused by the fusion of heterogeneous radar and visual data. By combining multi-scale edge detection with a lightweight deep learning model, it can still achieve robust recognition and contour extraction of aircraft targets in complex lighting environments such as rain, fog, insect interference, and low light at night, demonstrating better environmental adaptability than traditional pure visual solutions. Furthermore, the physical mapping algorithm between pixel displacement and camera calibration parameters further enables accurate physical distance calculation without the need for radar ranging.

[0033] Furthermore, the far-end warning distance is the sum of the far-end fixed reference value and the safety distance, and the near-end warning distance is the near-end fixed reference value.

[0034] In this invention, the distal fixed reference value is 200m, and the proximal fixed reference value is 50m.

[0035] Furthermore, the safety distance is calculated according to the following formula 1:

[0036]

[0037] Where S is the safety distance in meters (m); L is the distance from the platform parking checkpoint to the edge of the taxiway in meters (m); W is the width of the taxiway in the direction of aircraft taxiing in meters (m); V0 is the operating speed of the intelligent transfer platform in meters (m / s); and V1 is the aircraft taxiing speed in meters (m / s).

[0038] This invention multiplies the time required for the intelligent transfer platform to start from the parking checkpoint and completely pass through the taxiway by the aircraft's taxiing speed. This ensures that the aircraft will not enter the restricted area during the entire process of the platform passing through the intersection. It enables the remote warning distance to be adaptively adjusted according to the actual airport operating parameters (taxiway width, vehicle speed, and aircraft speed), avoiding the problem of insufficient safety margin or overly conservative approach that may exist with fixed distance thresholds under different airports and operating conditions.

[0039] In this invention, the platform parking checkpoint is a preset location where the intelligent transfer platform triggers the parking check process, and the distance L from the edge of the taxiway is generally set to 50m.

[0040] Taking Beijing Capital International Airport as an example, the taxiway width W is 44m, the aircraft taxiing speed V1 is 50km / h (approximately 13.89m / s), and the intelligent transfer platform operating speed V0 is 25km / h (approximately 6.94m / s). Substituting these values ​​into Equation 1, the calculated safe distance S is approximately 188m, and the remote warning distance is 388m.

[0041] Furthermore, the four roadside sensing devices have the same hardware structure. Through software configuration, one of the roadside sensing devices is set as the master and the other three are set as slaves. The slaves report the sensing data to the master, and the master sends the warning signal to the intelligent transmission platform.

[0042] This invention adopts a master-slave architecture, where the slave devices are only responsible for sensing and data reporting, while the master device performs unified data fusion, decision-making, and signal distribution. This avoids channel contention and data conflicts caused by multiple devices simultaneously sending wireless signals to the vehicle, ensuring the reliability of communication and the stability of early warning response time.

[0043] Furthermore, the host and the intelligent transmission platform exchange early warning signals via LoRa low-power wide-area communication.

[0044] LoRa communication exhibits excellent penetration and diffraction capabilities in the large, open environments of airports, offering wide coverage and low power consumption, making it suitable for the collaborative networking needs of multiple devices across a large area in airport flight zones. This invention replaces traditional DSRC or C-V2X vehicle-to-everything (V2X) communication solutions with LoRa low-power wide-area communication. Vehicles do not require dedicated OBU equipment, and existing airport vehicles (especially older vehicles) can be connected to the system without modification, significantly reducing the system deployment threshold and the cost of promotion and application.

[0045] Furthermore, the intelligent transfer platform is an unmanned vehicle, and the warning signals include a permission signal, a prohibition signal, a hold signal, and a release signal.

[0046] Furthermore, the four roadside sensing devices are symmetrically deployed on both sides of the taxiway, with the detection ranges of adjacent roadside sensing devices on the same side seamlessly connected, covering the two-way taxiing trajectory of the aircraft and eliminating blind spots in early warning.

[0047] Furthermore, the roadside sensing device is equipped with an industrial-grade high-definition camera.

[0048] This invention is equipped with an industrial-grade high-definition camera, which features high resolution, low-light imaging, and wide dynamic range. It can acquire clear images under complex lighting conditions such as strong light, backlight, and nighttime at airports, providing a high-quality image data source for subsequent aircraft target identification and distance calculation, and ensuring reliable operation of the system around the clock.

[0049] The present invention also provides a graded early warning detection system that executes the graded early warning method described in the above technical solution, including four roadside sensing devices and a vehicle-mounted receiving terminal;

[0050] The roadside sensing devices are symmetrically arranged on the left and right sides of the airport taxiway. The right side has the first and second roadside sensing devices, and the left side has the third and fourth roadside sensing devices. The detection fields of adjacent devices are seamlessly connected, covering the two-way taxiing trajectory of the aircraft.

[0051] A single roadside sensing device integrates the following functional modules:

[0052] The image acquisition module is used to acquire real-time image data streams of the entire taxiway using a high-definition vision camera;

[0053] The aircraft target recognition module is used to identify aircraft targets and extract fuselage contours using a multi-scale edge detection algorithm and a lightweight deep learning model.

[0054] The motion parameter calculation module is used to calculate the instantaneous taxiing speed and heading of the aircraft using a pixel displacement physical mapping algorithm and camera calibration parameters.

[0055] A dual-channel communication module is used to enable low-power collaborative networking between roadside sensing devices, and to interact with the vehicle-mounted receiving terminal via LoRa communication to exchange warning signals and traffic control signals.

[0056] Of the four roadside sensing devices, one is configured as the master unit via software, and the other three are configured as slave units. The master unit sends a prohibition signal, a hold signal, a release signal, and a passage permission signal to the vehicle-mounted receiving terminal based on the sensing data reported by each slave unit.

[0057] Compared with existing technologies, the hierarchical early warning method and system for airport bidirectional taxiways based on four-way visual collaboration described in this invention have the following advantages:

[0058] This invention employs four roadside sensing devices symmetrically deployed left and right. This symmetrical layout eliminates blind spots in bidirectional taxiing warnings, and the seamless connection of the detection ranges of adjacent devices ensures uninterrupted alarm coverage, unlike traditional single-direction or limited-device systems that cannot cover bidirectional scenarios. Simultaneously, it utilizes a pure vision-based solution, integrating multi-scale edge detection and a lightweight deep learning model. Even in complex lighting conditions such as rain, fog, and insect interference, it achieves robust aircraft target recognition and contour extraction. A physical mapping algorithm combining pixel displacement and camera calibration parameters accurately calculates the aircraft's instantaneous speed and orientation. A dual-channel mechanism enables low-power collaborative networking between roadside devices and low-latency warning interaction with the vehicle terminal, ensuring a warning response time of no more than 2 seconds. Furthermore, this invention constructs a hierarchical control strategy that triggers a prohibition signal from the far end and a hold signal from the near end, achieving dual safety guarantees for human-machine interaction and vehicle control, covering bidirectional taxiing scenarios, and ensuring vehicles receive advance information on intersection risks and adjust their travel paths in a timely manner. Attached Figure Description

[0059] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0060] Figure 1 This diagram illustrates the safe taxiing distance for aircraft.

[0061] Figure 2 This is a schematic diagram of the deployment of four roadside sensing devices of the present invention in the intersection area of ​​airport taxiways, wherein Figure (a) is a schematic diagram of an aircraft taxiing along the first direction, and Figure (b) is a schematic diagram of an aircraft taxiing along the second direction;

[0062] Figure 3 This is a diagram showing the detection range of the roadside sensing device in the first direction according to the present invention. Detailed Implementation

[0063] The present invention will be further described below with reference to specific embodiments. First, it should be noted that the data in the following experimental examples were obtained by the inventors through numerous experiments. Due to space limitations, only a portion of these data is shown in the specification, and those skilled in the art can understand and implement the present invention based on this data. These embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. Furthermore, it should be understood that after reading the contents of this invention, those skilled in the art can make various modifications or alterations to the invention, and these modifications or alterations also fall within the scope of protection of this application.

[0064] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0065] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0066] like Figure 1 As shown, the aircraft taxis from left to right along the taxiway. The safe avoidance zone is within 200 meters in front of it, 50 meters behind it, and 50 meters to both sides. Ground vehicles are not allowed to pass within 200 meters in front of the aircraft, not to follow within 50 meters behind it, and must give way 50 meters away from the side of the aircraft. However, the implementation of the above safety requirements faces severe challenges: existing airports mainly rely on manual observation and driver visual judgment. Due to factors such as obstructed vision, weather conditions, and differences in personnel experience, the risk of misjudgment and omission is high. Specifically, the existing technology still has the following shortcomings: (1) The detection method is inefficient. Traditional control relies on manual observation or fixed monitoring, which cannot capture the aircraft's taxiing position and speed in real time, resulting in delayed avoidance instructions; (2) The performance of roadside equipment is insufficient. Existing roadside sensing equipment is mostly single-function design, with high power consumption, unstable communication, and low detection accuracy, which cannot meet the needs of close-range avoidance; (3) The alarm logic is simple. It can only realize a simple alarm of whether there is an aircraft or not, and lacks a mechanism to dynamically adjust the warning level according to the aircraft's taxiing distance.

[0067] It should be noted that the intelligent transfer platform described in this embodiment is an unmanned ground support vehicle; the vehicle positioning module or roadside sensing equipment is used to identify whether the intelligent transfer platform has reached the preset parking checkpoint.

[0068] This embodiment provides a graded early warning method for airport bidirectional taxiways based on four-way visual collaboration, including:

[0069] Two roadside sensing devices are deployed on each side of the airport's two-way taxiway, with two devices on each side corresponding to the detection of the far-end warning distance and the near-end warning distance, respectively.

[0070] The roadside sensing device that detects the distance to the far-end warning will issue a prohibition signal when the aircraft enters the far-end warning threshold, and at the same time trigger the nearby sensing device on the same side to enter the linkage monitoring mode.

[0071] The roadside sensing device that detects the near-end warning distance issues a hold signal when it continuously detects an aircraft entering the near-end warning threshold.

[0072] When there are no aircraft targets within the monitoring field of view, the roadside sensing device that detects the near-end warning distance issues a wait-out signal, and the vehicle resumes normal traffic.

[0073] It should be noted that this embodiment does not limit the specific installation location of the roadside sensing device, as long as its detection field of view can cover the location of the corresponding far-end warning distance / near-end warning distance.

[0074] In this embodiment, the roadside sensing device acquires aircraft taxiing image data using a purely visual sensing method, identifies the aircraft target and extracts its contour through multi-scale edge detection and a lightweight deep learning model, and calculates the instantaneous speed and orientation of the aircraft by combining a physical mapping algorithm of pixel displacement and camera calibration parameters.

[0075] It should be noted that the roadside sensing device uses multi-scale edge detection combined with a lightweight deep learning model to complete aircraft target detection and fuselage contour extraction, achieving the following technical effects: It stably outputs the coordinates of the aircraft detection box and contour feature points based on real-world taxiway images, providing reliable input for subsequent pixel displacement physical ranging; the algorithm maintains stable detection accuracy under complex airport conditions such as rain, fog, insect occlusion, low light at night, and backlight; and the model inference overhead is adapted to embedded computing platforms, meeting the needs of synchronous real-time processing by multiple devices. This invention does not impose a unique limitation on the basic algorithms for image detection and edge extraction; any existing image algorithms that can achieve the above effects are applicable.

[0076] In a preferred embodiment, a lightweight improved YOLOv5s network is used as the target detection infrastructure: the backbone network is replaced with a lightweight MobileNetV3 or ShuffleNetV2 structure, and depthwise separable convolutions are used to replace standard convolutions, reducing the number of model parameters and computational load. Preferably, the number of model parameters is controlled within 5MB to adapt to real-time inference on roadside embedded platforms.

[0077] The backbone network incorporates a multi-scale feature fusion module (Feature Pyramid FPN) to extract edge features in layers: shallow high-resolution feature maps preserve spatial details for fine contour extraction of aircraft at close range; deep low-resolution feature maps carry global semantic information to identify small-sized aircraft targets at a distance; and deep and shallow features are fused through a top-down feature fusion path to output aircraft detection boxes and fuselage contour key points that fuse multi-scale edge features.

[0078] The supporting lightweight optimization strategies include: depthwise separable convolution: splitting standard convolution into depthwise convolution and pointwise convolution, which can reduce the number of model parameters by about 80%; structured model pruning: removing low-contribution convolution channels and connections to further compress the model size; and lightweight replacement of activation functions: replacing high-overhead activation functions with low-computation ReLU activation functions to reduce embedded computing power consumption.

[0079] Offline training process: Collect no less than 5,000 real-world images of airport taxiways, covering typical operating conditions of aircraft activity areas such as clear weather, rain and fog, low light at night, backlight, and partial occlusion of jet bridges. Label the rectangular detection boxes of aircraft and key points of the fuselage outline in the images. After iterative training until the loss function converges, convert the model into a dedicated format for embedded inference and deploy it to the embedded computing unit of a four-way roadside perception device. The optimal inference frame rate is no less than 15 frames / second to achieve real-time image frame parsing.

[0080] The four symmetrically deployed roadside sensing devices independently operate the aforementioned lightweight recognition model, collecting features of taxiing aircraft in both directions to ensure no blind spots in bidirectional taxiing scenarios.

[0081] It should be noted that by constructing a physical mapping algorithm using pixel displacement changes combined with camera calibration parameters, the instantaneous taxiing speed and heading of the aircraft are calculated. The core technical advantage is that it completes the transformation from pixel coordinates to real-world coordinates on the taxiway plane solely based on visual images, calculating the longitudinal physical distance between the aircraft and the intersection with vehicles, as well as the actual displacement between frames. No additional ranging sensors such as radar or lidar are required, achieving high-precision ranging and speed measurement based solely on visual perception. This invention does not impose a single limitation on the basic algorithm for pixel-physical space coordinate mapping and motion parameter calculation.

[0082] In a preferred embodiment, the following steps are included:

[0083] After the roadside sensing equipment was installed, the camera was calibrated using Zhang Zhengyou's calibration method. A checkerboard calibration board was placed at multiple different positions and orientations within the taxiway detection area, and at least 15 calibration images were captured. The pixel coordinates of the checkerboard corners were extracted using a corner detection algorithm, and combined with the known physical dimensions of the checkerboard, the camera's intrinsic parameter matrix (including the focal length f) was calculated. x f y Principal point coordinates c x c y The system measures the camera's external parameters and distortion coefficients (radial distortion k1, k2, k3, tangential distortion p1, p2); it also measures the camera's mounting height H and pitch angle θ (the angle between the camera's optical axis and the horizontal plane) relative to the taxiway plane, storing these values ​​as fixed external parameters of the camera in the device.

[0084] Based on the pinhole camera model, a mapping relationship from the pixel coordinate system to the world coordinate system is established. For any pixel (u, v) in the image, combined with camera intrinsic parameters, installation height H, and pitch angle θ, it is mapped to world coordinates (X, Y, Z=0) on the taxiway plane through inverse perspective projection transformation, where X is the longitudinal distance along the taxiway direction (i.e., the distance between the aircraft and the intersection), and Y is the lateral distance perpendicular to the taxiway direction. Based on this, the system can obtain the physical distance X between the aircraft and the intersection in real time.

[0085] For the same aircraft target detected in two consecutive frames, its pixel coordinates (u1, v1) and (u2, v2) in the two frames are obtained respectively. Using the above mapping relationship, the aircraft's position in the world coordinate system (X1, Y1) and (X2, Y2) in the two frames are calculated respectively. Then, the actual displacement of the aircraft between the two frames is ΔD = |X2 - X1|. Combining the inter-frame time interval Δt (determined by the video frame rate), the instantaneous taxiing speed of the aircraft V = ΔD / Δt can be calculated. The aircraft's heading is determined by the displacement direction—if X2 > X1, the aircraft taxis in the forward direction of the taxiway; otherwise, it taxis in the reverse direction.

[0086] Through the above mapping, the system obtains the physical distance X between the aircraft and the intersection in real time. When X ≤ the far-end warning distance, a far-end warning is triggered, and a prohibition signal is sent to the intelligent transfer platform; when X ≤ the near-end warning distance, a near-end warning is triggered, and a hold-and-wait signal is sent to the intelligent transfer platform.

[0087] Example 1

[0088] A hierarchical early warning method for airport bidirectional taxiways based on four-way visual collaboration includes the following steps:

[0089] Step S1: Equipment Installation

[0090] like Figures 2-3 As shown, a field survey was conducted on the airport's aircraft taxiway and the area accessible by the intelligent transfer platform. Based on the taxiway width and the intelligent transfer platform's route, the specific deployment locations of four roadside sensing devices were determined: the first roadside sensing device A and the second roadside sensing device B were deployed on the right side of the taxiway, and the third roadside sensing device D and the fourth roadside sensing device C were deployed on the left side, symmetrically installed on both sides. During deployment, the detection fields of adjacent devices on the same side (A and B, D and C) were seamlessly connected according to the device's detection angle. Specifically, the detection fields of the first roadside sensing device A and the third roadside sensing device D cover the far-end warning distance (the distance from their installation location to the edge of the intersection is the far-end warning distance), while the detection fields of the second roadside sensing device B and the fourth roadside sensing device C cover the near-end warning distance (the distance from their installation location to the edge of the intersection is the near-end warning distance), accurately covering the intersection of the taxiway and the intelligent transfer platform's access area.

[0091] Step S2: Parking Inspection

[0092] When a vehicle enters the designated location in the intersection area of ​​the skid lanes, the system automatically triggers the platform's parking inspection process, suspending the vehicle's operation and waiting for feedback from the roadside sensing equipment.

[0093] Step S3: Determine passage conditions

[0094] If none of the four roadside sensing devices detect the aircraft entering the warning distance, the platform will resume normal operation after receiving the permission signal; if any roadside sensing device detects the aircraft entering the corresponding warning distance, the platform will remain in a stopped and waiting state.

[0095] Step S4: Implementation of Tiered Early Warning and Control Measures

[0096] Step S41: Remote warning trigger. When the aircraft taxis along the first direction to the remote warning distance, the first roadside sensing device A on the right accurately identifies the aircraft target through multi-scale edge detection and a lightweight deep learning model. It also calculates the instantaneous speed and bearing of the aircraft using a physical mapping algorithm of pixel displacement and camera calibration parameters. After confirming that the aircraft has entered the remote warning distance, it sends a prohibition signal to the intelligent transmission platform. At the same time, it synchronizes the aircraft's position, speed, and heading detection data to the second roadside sensing device B on the same side that detects the near-end warning distance, triggering the linkage monitoring mode to ensure continuous and uninterrupted alarms in subsequent stages.

[0097] Step S42: Near-end warning triggered. When the aircraft continues to taxi to within 50m of the near-end warning distance, the second roadside sensing device B continuously monitors the aircraft and simultaneously sends a hold signal to the intelligent transfer platform to further strengthen safety control and prevent the platform from being accidentally activated when the aircraft is taxiing close to it.

[0098] Step S43: Warning Cancellation. The second roadside sensing device B continuously monitors the taxiway intersection area, with its detection field covering the entire intersection area. The aircraft remains within the device's monitoring field of view throughout its taxiing through the intersection. After the aircraft continues taxiing and passes through the intersection, the second roadside sensing device B uses speed and orientation detection technology to confirm that there is no longer an aircraft target within its detection field of view, indicating that the aircraft has completely passed the critical intersection area. It then sends a cancellation signal to the intelligent transfer platform.

[0099] When the aircraft taxis along the second direction, the third roadside sensing device D and the fourth roadside sensing device C on the left side perform the same functions as the first roadside sensing device A and the second roadside sensing device B on the right side: triggering remote warnings, triggering near-end warnings, and canceling warnings.

[0100] During implementation, the timeliness and accuracy of alarm signals at each stage must be tested by simulating a two-way aircraft taxiing scenario. This ensures that roadside sensing devices detecting far-end warning distances trigger the "No Proceed" signal 100% of the time when an aircraft enters the far-end warning distance, and that roadside sensing devices detecting near-end warning distances trigger the "Hold" signal 100% of the time when an aircraft enters within 50 meters. Furthermore, the "Release" signal should only be triggered after the aircraft has completely left the detection field of view, with no false alarms or missed alarms. Simultaneously, the solution must be verified to comply with the safety requirements for intelligent traffic control at airports as stipulated in the "Regulations on the Management of Special Equipment for Civil Airports," ensuring a high degree of consistency between the alarm logic and airport operational safety management regulations.

[0101] While the present invention has been disclosed above, it is not limited thereto. Any person skilled in the art can make various alterations and modifications without departing from the spirit and scope of the invention; therefore, the scope of protection of the present invention should be determined by the scope defined in the claims.

Claims

1. A hierarchical early warning method for airport bidirectional taxiways based on four-way visual collaboration, characterized in that, include: Two roadside sensing devices are deployed on each side of the airport's two-way taxiway, with two devices on each side corresponding to the detection of the far-end warning distance and the near-end warning distance, respectively. The roadside sensing device that detects the distance to the far-end warning will issue a prohibition signal when the aircraft enters the far-end warning threshold, and at the same time trigger the nearby sensing device on the same side to enter the linkage monitoring mode. The roadside sensing device that detects the near-end warning distance issues a hold signal when it continuously detects an aircraft entering the near-end warning threshold. When there are no aircraft targets within the monitoring field of view, the roadside sensing device that detects the near-end warning distance issues a wait-out signal, and the vehicle resumes normal traffic.

2. The graded early warning method according to claim 1, characterized in that, Specifically, the following steps are included: Step S1: Deploy two roadside sensing devices on each side of the airport taxiway. The two devices on each side are used to detect the far-end warning distance and the near-end warning distance, respectively. Step S2: When the intelligent transfer platform enters the designated position in the intersection area of ​​the taxiway, the system automatically triggers the platform parking inspection process; Step S3: If none of the four roadside sensing devices detect the aircraft entering the warning distance, the intelligent transfer platform will resume normal operation after receiving the permission signal. If any roadside sensing device detects that an aircraft has entered the corresponding warning distance, proceed to step S4; Step S4: When the roadside sensing device detecting the far-end warning distance detects an aircraft entering the far-end warning distance, it sends a prohibition signal to the intelligent transmission platform and triggers the roadside sensing device on the same side detecting the near-end warning distance to enter the linkage monitoring mode; when the roadside sensing device detecting the near-end warning distance continuously monitors an aircraft entering the near-end warning distance, it sends a hold signal to the intelligent transmission platform; when the roadside sensing device detecting the near-end warning distance confirms during continuous monitoring that there is no longer an aircraft target in its field of view, it sends a release signal to the intelligent transmission platform.

3. The graded early warning method according to claim 1, characterized in that, The roadside sensing device uses a purely visual perception method to collect aircraft taxiing image data. It identifies aircraft targets and extracts their contours through multi-scale edge detection and a lightweight deep learning model. It then uses a physical mapping algorithm combining pixel displacement and camera calibration parameters to calculate the aircraft's instantaneous speed and orientation.

4. The graded early warning method according to claim 1, characterized in that, The far-end warning distance is the sum of the far-end fixed reference value and the safety distance, and the near-end warning distance is the near-end fixed reference value.

5. The graded early warning method according to claim 4, characterized in that, The safety distance is calculated according to the following formula: ； Where S is the safety distance in meters (m); L is the distance from the platform parking checkpoint to the edge of the taxiway in meters (m); W is the width of the taxiway in the direction of aircraft taxiing in meters (m); V0 is the operating speed of the intelligent transfer platform in meters (m / s); and V1 is the aircraft taxiing speed in meters (m / s).

6. The graded early warning method according to claim 1, characterized in that, The four roadside sensing devices have the same hardware structure. Through software configuration, one of the roadside sensing devices is set as the master and the other three are set as slaves. The slaves report the sensing data to the master, and the master sends the warning signal to the intelligent transmission platform.

7. The graded early warning method according to claim 6, characterized in that, The host and the intelligent transmission platform exchange early warning signals via LoRa low-power wide-area communication.

8. The graded early warning method according to claim 1, characterized in that, The four roadside sensing devices are symmetrically deployed on both sides of the taxiway. The detection ranges of adjacent roadside sensing devices on the same side are seamlessly connected, covering the two-way taxiing trajectory of the aircraft and eliminating blind spots in the early warning system.

9. The graded early warning method according to claim 2, characterized in that, The roadside sensing device is equipped with an industrial-grade high-definition camera.

10. A graded early warning detection system, characterized in that, The graded early warning method according to any one of claims 1 to 9 includes four roadside sensing devices and a vehicle-mounted receiving terminal; The roadside sensing devices are symmetrically arranged on the left and right sides of the airport taxiway. The right side has the first and second roadside sensing devices, and the left side has the third and fourth roadside sensing devices. The detection fields of adjacent devices are seamlessly connected, covering the two-way taxiing trajectory of the aircraft. A single roadside sensing device integrates the following functional modules: The image acquisition module is used to acquire real-time image data streams of the entire taxiway using a high-definition vision camera; The aircraft target recognition module is used to identify aircraft targets and extract fuselage contours using a multi-scale edge detection algorithm and a lightweight deep learning model. The motion parameter calculation module is used to calculate the instantaneous taxiing speed and heading of the aircraft using a pixel displacement physical mapping algorithm and camera calibration parameters. A dual-channel communication module is used to enable low-power collaborative networking between roadside sensing devices, and to interact with the vehicle-mounted receiving terminal via LoRa communication to exchange warning signals and traffic control signals. Of the four roadside sensing devices, one is configured as the master unit via software, and the other three are configured as slave units. The master unit sends a prohibition signal, a hold signal, a release signal, and a passage permission signal to the vehicle-mounted receiving terminal based on the sensing data reported by each slave unit.

Images