Airport FOD inspection and operation safety assurance system based on unmanned system

Abstract

This invention discloses an unmanned airport surface FOD inspection and operational safety assurance system. It combines drone technology with a deep learning target detection algorithm to detect foreign objects in airport surfaces. The drone can be equipped with a combination of high-definition cameras, lidar, and infrared sensors, enabling real-time inspection of the airport surface along a specific flight trajectory within the airport airspace. This system can detect FOD over large areas and long distances on airport runways or taxiways.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an airport field FOD inspection and operation safety assurance system based on an unmanned system. Background Art

[0002] The airport surface is a critical area for aircraft takeoff, landing, and taxiing. Any foreign object on the surface can seriously impact the safe operation of aircraft and lead to serious consequences. Detection is typically performed using fixed detection base stations on both sides of the runway, primarily using millimeter-wave radar detection, supplemented by video image recognition. Multi-point video image recognition technology can also be used. While these systems can actively detect FOD, their accuracy and range are affected by the system's installation method and location. Achieving high-precision FOD detection over a wide area requires the simultaneous deployment of a large number of detection points, significantly increasing system costs. Furthermore, the complex electromagnetic environment and surface layout of airports impose significant limitations on the installation and use of detection points, hindering the application and widespread adoption of such systems. Currently, most airports still rely on manual inspections by runway inspectors, requiring runway closures for inspections. This is not only inefficient and unreliable, but also consumes valuable runway time. Furthermore, the current lack of widespread use of airport runway foreign object detection systems is due to the high cost of existing systems and the lack of mature, readily available technologies or products.

[0003] Therefore, it is necessary to research and develop a low-cost, large-detection-range, convenient and practical airport runway FOD inspection and operation safety assurance system, so that airport security personnel can achieve large-scale remote monitoring of the foreign object safety situation in the runway area, mark the detected foreign objects at fixed points, and send the corresponding information to the airport operation monitoring system for recording; according to the marked information, the best path for fixed-point removal of foreign objects can be planned, so as to minimize the workload of airport security personnel, reduce the time and manpower costs spent by the airport on FOD detection and cleaning, and improve the efficiency of airport operation; at the same time, the accuracy of FOD detection can be improved, and the safety risks of airport operation can be reduced.

[0004] At the same time, from a management perspective, managers can analyze the safety situation of flights and airport operations based on FOD detection data, and provide reference for optimizing corresponding operating rules and regulations, which is of great significance to the high-quality development of civil aviation. Summary of the Invention

[0005] The purpose of the present invention is to provide an airport FOD inspection and operation safety assurance system based on an unmanned system to solve the above-mentioned problems existing in the prior art.

[0006] The unmanned system-based airport FOD inspection and operation safety assurance system includes detection equipment, FOD detection module, task scheduling module and wireless data transmission module:

[0007] The detection equipment includes a high-definition camera, a laser radar, and an infrared sensor; the detection equipment is fixed on the drone;

[0008] The task scheduling module is used to adjust the shooting position of the detection device to obtain a second detection path; the second detection path has multiple shooting positions as nodes of the path; the second detection path is the shortest path and the images captured by all nodes can capture the complete surface of the detection object; the detection object represents the object that needs to be inspected;

[0009] The detection device is used to obtain a three-mode image set corresponding to the node on the second detection path; the three-mode image set includes a laser depth image, an infrared image, and a camera image;

[0010] The FOD detection module is used to obtain multiple foreign body locations based on the three-mode image set through an image recognition network;

[0011] The task scheduling module is used to perform path planning based on multiple foreign object positions to obtain a foreign object detection path; the foreign object detection path represents a path for clearing foreign objects.

[0012] Optionally, the task scheduling module includes a constraint module, an algorithm module and a deep learning module;

[0013] The constraint condition module is used to obtain multiple constraint data;

[0014] The algorithm module has multiple path detection algorithms built in;

[0015] The deep learning module has a built-in path detection network;

[0016] Multiple path detection algorithms are used to train the path detection network.

[0017] Optionally, adjusting the shooting position of the detection device to obtain the second detection path includes:

[0018] Marking locations where the drone is restricted from flying based on the plurality of constraint data to obtain a three-dimensional area image; locations where the grayscale value is 1 in the three-dimensional area image represent locations where the drone can fly, and locations where the grayscale value is 0 represent locations where the drone is restricted from flying;

[0019] Based on the three-dimensional area image, a three-dimensional detection object position image and an initial detection path are obtained; the three-dimensional detection object position image represents an image that marks the position of the detection object and the flight restriction position; the initial detection path represents an initially set path for obtaining the complete surface of the detection object;

[0020] Using the trained path detection network, multiple path detection algorithms are identified based on the three-dimensional detection object position image to obtain a second detection path;

[0021] The second detection path is used to replace the initial detection path.

[0022] Optionally, the adopting of multiple path detection algorithms to train the path detection network includes:

[0023] Acquire a training three-dimensional detection object position image; the training three-dimensional detection object position image represents a three-dimensional detection object position image used for training at a historical time point;

[0024] Based on the training three-dimensional detection object position image, a first detection path is obtained through multiple path detection algorithms; the first detection path represents the shortest path among the multiple path detection algorithms that can obtain the complete surface of the detection object;

[0025] using the path detection algorithm corresponding to the first detection path as a determination algorithm;

[0026] Based on the training three-dimensional detection object position image, a first detection algorithm is obtained through a path detection network; the first detection algorithm represents an algorithm predicted to be suitable for the three-dimensional detection object position image;

[0027] The determination algorithm and the first detection algorithm are used to calculate the loss and train the path detection network.

[0028] Optionally, obtaining the first detection path by using multiple path detection algorithms based on the training three-dimensional detection object position image includes:

[0029] The input of the path detection algorithm is the position of the detection object in the training three-dimensional detection object position image and multiple constraint data in the corresponding constraint condition module;

[0030] Multiple path detection algorithms correspondingly obtain multiple training detection paths; the training detection paths represent paths corresponding to the path detection algorithms that can obtain the complete surface of the detection object;

[0031] Among the multiple training detection paths, the training detection path with the least number of nodes is used as the first detection path.

[0032] Optionally, obtaining a first detection algorithm based on the training three-dimensional detection object position image through a path detection network includes:

[0033] The input of the path detection network is a training three-dimensional detection object position image;

[0034] The output of the path detection network is a first detection algorithm.

[0035] Optionally, the step of determining a plurality of path detection algorithms based on the three-dimensional detection object position image using a trained path detection network to obtain the second detection path includes:

[0036] Inputting the three-dimensional detection object position image into the trained path detection network to obtain a second detection algorithm;

[0037] The plurality of constraint data in the constraint condition module and the position of the detection object in the three-dimensional detection object position image are input into the second detection algorithm to obtain a second detection path.

[0038] Optionally, the number of output neurons of the path detection network is equal to the number of path detection algorithms built into the algorithm module.

[0039] Optionally, obtaining multiple foreign body positions based on the three-mode image set through an image recognition network includes:

[0040] The image recognition network includes a first foreign body detection network, a second foreign body detection network and a third foreign body detection network;

[0041] Inputting the laser depth image, the infrared image and the camera image into a first foreign body detection network, a second foreign body detection network and a third foreign body detection network respectively to obtain a first foreign body feature, a second foreign body feature and a third foreign body feature;

[0042] Superimposing the first foreign body feature, the second foreign body feature, and the third foreign body feature to obtain an overlapping feature;

[0043] The overlapping features are input into a fully connected neural network to obtain multiple foreign body locations.

[0044] Optionally, obtaining a three-dimensional detection object position image and an initial detection path based on the three-dimensional area image includes:

[0045] Acquire multiple laser view images; the laser view images represent images of the detected object acquired by the UAV's laser radar at multiple view angles;

[0046] Arrange multiple laser view images corresponding to the drone's location in order from early to late time points to obtain the initial detection path;

[0047] Reconstructing the multiple laser view angle images in three dimensions to obtain a detection object position image; the detection object position image represents the position of the detection object in three dimensions;

[0048] The position corresponding to the detection object in the detection object position image is marked on the three-dimensional region image to obtain a three-dimensional detection object position image.

[0049] Compared with the prior art, the embodiments of the present invention achieve the following beneficial effects:

[0050] Through the training of multiple dynamic programming path detection algorithms, a path detection network suitable for the current constraint data and detection objects can be identified, making it possible to find the optimal path more accurately and quickly.

[0051] The combination of drone technology and deep learning target detection algorithms can realize the detection of foreign objects in airport scenes. Drones can be equipped with detection devices composed of multiple sensors such as high-definition cameras, lidar and infrared sensors to conduct real-time detection of airport scenes in the airport airspace according to a certain flight trajectory. It can realize FOD detection in airport runways or taxiways over large areas and long distances. At the same time, drones can be maneuvered and flexibly moved to facilitate rapid entry and exit. This avoids the need to close airport runways for manual inspections under traditional methods, which takes up a lot of time and the resulting reduction in airport operating efficiency. It also avoids the impact of interrupted flight operations and the resulting economic losses to airlines.

[0052] This system enables intelligent runway FOD inspections using drones and multiple sensors, significantly improving FOD detection efficiency and accuracy. Optimized drone inspection route planning reduces energy consumption and improves operational efficiency. Furthermore, the system's drone inspection flight paths not only enhance the automation of mission execution but also ensure safe separation between drones and aircraft during mixed operations, preventing potential conflicts on the same flights. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 This is a flowchart of a method implemented in an airport FOD inspection and operation safety assurance system based on an unmanned system provided by an embodiment of the present invention.

[0054] Figure 2 This is a structural diagram of the algorithm module and deep learning module in an airport FOD inspection and operation safety assurance system based on an unmanned system provided by an embodiment of the present invention.

[0055] Figure 3 The present invention provides an unmanned system-based airport FOD inspection and operation safety assurance system with a workflow diagram.

[0056] Figure 4 This is a principle block diagram of an airport FOD inspection and operation safety assurance system based on an unmanned system provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0057] The present invention will be described in detail below with reference to the accompanying drawings.

[0058] Example

[0059] The embodiment of the present invention provides an airport surface FOD inspection and operation safety assurance system based on an unmanned system, including detection equipment, a FOD detection module, a task scheduling module, and a wireless data transmission module:

[0060] Among them, the wireless data transmission module is used to transmit the drone detection information to the FOD detection module.

[0061] The detection equipment includes a high-definition camera, a laser radar, and an infrared sensor; the detection equipment is fixed on the drone.

[0062] The task scheduling module is used to:

[0063] S101: Adjust the shooting position of the detection device to obtain a second detection path; the second detection path consists of multiple shooting positions as nodes of the path; the second detection path is the shortest path and the images shot by all nodes can obtain the complete surface of the detection object; the detection object represents an object that needs to be inspected.

[0064] The shooting position refers to the position where the detection equipment shoots the laser depth image, infrared image and camera image.

[0065] In this embodiment, the number of nodes in the path is set to n, the length of the side connecting the nodes is set to m, and the value of 0.7n+0.5m is used as the shortest path. n and m are positive integers.

[0066] The reason why the images taken by all nodes can capture the complete surface of the inspection object is that each node can capture a part of the inspection object when shooting, and the images taken in all directions by all nodes can capture the entire area of the inspection object's surface.

[0067] The unmanned airport surface FOD inspection and operational safety assurance system also includes a positioning module. This module is used on the drone platform to control detection equipment to a shooting location to perform airport surface scanning and data collection. In this embodiment, the positioning module is a Beidou-Inertial Navigation (BIGN) positioning module.

[0068] On the ground control console, the FOD detection module and task scheduling module perform image processing and path planning. The data visualization platform displays the work dynamics in real time, including images and paths.

[0069] The ground control console and the UAV platform transmit data via a wireless data transmission module. In this embodiment, the wireless data transmission module is a 5G / wireless data transmission module.

[0070] The task scheduling module includes a cleaning path planning component, an airport operation electronic status component, and a flight path planning component. The cleaning path planning component is used to determine the path for workers to clean foreign objects. The flight path planning component is used to determine the flight path of the drone. The airport operation electronic status component is used to determine whether the detected object in the airport is stationary or in motion.

[0071] The principle block diagram of the airport FOD inspection and operation safety assurance system based on the unmanned system is as follows: Figure 4 shown.

[0072] The second detection path is obtained by a flight path planning component.

[0073] Wherein, the second detection path is based on Figure 3 The strategy of maximum coverage and fastest speed optimization obtains the path. In the strategy optimization, maximum coverage is represented by obtaining the complete surface of the detection object. Path generation is performed using path mapping and instruction generation, wherein the path mapping is used to map to a second detection path, and the instruction generation is used to generate instructions for obtaining the second detection path.

[0074] The detection device is used to:

[0075] S102: Acquire a three-mode image set corresponding to a node on a second detection path; the three-mode image set includes a laser depth image, an infrared image, and a camera image.

[0076] Among them, during the execution of FOD inspection flight missions, the detection equipment carried by the drone platform records the current scene in real time, and transmits the generated laser depth images, infrared images and camera images back to the ground control console in real time through the 5G wireless data transmission module.

[0077] Among them, such as Figure 3 As shown, flight control instructions are used to capture video images, infrared images, and laser radar depth images. The flight control instructions are instructions for controlling the drone to fly to a shooting position for shooting.

[0078] The FOD detection module is used to:

[0079] S103: Obtain multiple foreign body positions based on the three-mode image set through an image recognition network.

[0080] Among them, the location of foreign objects is detected through the image recognition network and displayed on the data visualization platform.

[0081] Among them, such as Figure 3 As shown, the three-mode image set is used as image data for target detection based on deep learning. In this embodiment, an image recognition network is used for target detection.

[0082] The task scheduling module is used to:

[0083] S104: performing path planning based on multiple foreign object positions to obtain a foreign object detection path; the foreign object detection path represents a path for removing foreign objects.

[0084] Among them, through Figure 3 As shown, the exact algorithm, heuristic algorithm, multi-stage algorithm, continuous approximation algorithm, and meta-heuristic algorithm in the algorithm selection are used to find the path with the shortest distance between the positions of foreign objects and perform path mapping.

[0085] Among them, such as Figure 3 As shown, the foreign object detection path of the path mapping is output as navigation information.

[0086] The flowchart of the method implemented in the airport surface FOD inspection and operation safety assurance system based on the unmanned system is as follows: Figure 1 shown.

[0087] Optionally, the task scheduling module includes a constraint module, an algorithm module and a deep learning module.

[0088] The structural diagram of the algorithm module and deep learning module is as follows: Figure 2 shown. Figure 2 The position of the object to be detected in the three-dimensional object position image and the corresponding constraint data in the constraint condition module represent the data input during detection. Therefore, the corresponding first training detection path, second training detection path, and first detection path represent the paths obtained during detection.

[0089] The constraint condition module is used to obtain a plurality of constraint data.

[0090] The constraint data is obtained by the airport operation electronic situation component.

[0091] In this embodiment, the constraint data includes safety restrictions, task constraints, and special instructions.

[0092] The operational safety restriction information is stored in a database on the ground control console. The constraint data includes electronic maps, no-fly zones, and flight range as safety constraints. Mission constraints include mission time, fleet size, geo-fences, and flight dynamics. Flight dynamics indicate whether an airport is closed. Geo-fences indicate areas drones cannot pass through. Special commands indicate commands to control drones due to mission interruption or emergency evacuation. Fleet size indicates the number of fleets at the airport. Special commands include mission interruption and emergency evacuation commands. Upon receiving a special command, the drone is controlled to cease operations.

[0093] The algorithm module has multiple path detection algorithms built in.

[0094] Path detection algorithms include exact algorithms, heuristic algorithms, multi-stage algorithms, continuous approximation algorithms, and meta-heuristic algorithms. In this embodiment, branch-cutting algorithms, mileage-saving algorithms, artificial bee colony algorithms, continued fraction approximation algorithms, and cluster-path algorithms are used.

[0095] The deep learning module has a built-in path detection network.

[0096] Multiple path detection algorithms are used to train the path detection network.

[0097] Optionally, adjusting the shooting position of the detection device to obtain the second detection path includes:

[0098] According to multiple constraint data, the positions where the drone is restricted from flying are marked to obtain a three-dimensional area image; the positions where the grayscale value is 1 in the three-dimensional area image represent positions where the drone can fly, and the positions where the grayscale value is 0 represent positions where flight is restricted.

[0099] Wherein, the three-dimensional region image is a grayscale image.

[0100] Based on the three-dimensional area image, a three-dimensional detection object position image and an initial detection path are obtained; the three-dimensional detection object position image represents an image marking the position of the detection object and the restricted flight position; the initial detection path represents the path that is initially set and obtains the complete surface of the detection object.

[0101] Through the trained path detection network, based on the three-dimensional detection object position image, multiple path detection algorithms are judged to obtain the second detection path.

[0102] The second detection path is used to replace the initial detection path.

[0103] Optionally, the adopting of multiple path detection algorithms to train the path detection network includes:

[0104] A training three-dimensional detection object position image is obtained; the training three-dimensional detection object position image represents a three-dimensional detection object position image used for training at a historical time point.

[0105] Based on the training three-dimensional detection object position image, a first detection path is obtained through multiple path detection algorithms. The first detection path represents the shortest path among the multiple path detection algorithms that can obtain the complete surface of the detection object.

[0106] The first detection path is a path obtained during training.

[0107] The path detection algorithm corresponding to the first detection path is used as the determination algorithm.

[0108] The determination algorithm is represented by the label of the corresponding path detection network.

[0109] Based on the training three-dimensional detection object position image, a first detection algorithm is obtained through a path detection network; the first detection algorithm represents a predicted algorithm suitable for the three-dimensional detection object position image.

[0110] The first detection algorithm is represented by the label of the corresponding path detection network.

[0111] The determination algorithm and the first detection algorithm are used to calculate the loss and train the path detection network.

[0112] In this embodiment, the cross entropy loss function is used to calculate the loss value.

[0113] Optionally, obtaining the first detection path by using multiple path detection algorithms based on the training three-dimensional detection object position image includes:

[0114] The input of the path detection algorithm is the position of the detection object in the training three-dimensional detection object position image and multiple constraint data in the corresponding constraint condition module.

[0115] The constraint data in the constraint condition module is also the constraint data used for training at a historical time point.

[0116] Multiple path detection algorithms correspond to obtaining multiple training detection paths; the training detection paths represent paths corresponding to the path detection algorithms that can obtain the complete surface of the detection object.

[0117] Among the multiple training detection paths, the shortest training detection path is used as the first detection path.

[0118] In this embodiment, the branch cutting algorithm is used as the first path detection algorithm, and the optimal path output is used as the first training detection path. The mileage saving algorithm is used as the second path detection algorithm, and the optimal path output is used as the second training detection path. The artificial bee colony algorithm is used as the third path detection algorithm, and the optimal path output is used as the third training detection path. The continued fraction approximation algorithm is used as the fourth path detection algorithm, and the optimal path output is used as the fourth training detection path. The clustering-path algorithm is used as the fifth path detection algorithm, and the optimal path output is used as the fifth training detection path. The shortest path among the first training detection path, the second training detection path, the third training detection path, the fourth training detection path, and the fifth training detection path is used as the first detection path.

[0119] Optionally, obtaining a first detection algorithm based on the training three-dimensional detection object position image through a path detection network includes:

[0120] The input of the path detection network is a three-dimensional image of the position of the detected object;

[0121] The output of the path detection network is a first detection algorithm.

[0122] In this embodiment, the path detection network is a three-dimensional convolutional neural network (3D Convolutional Neural Networks, 3D CNN).

[0123] Optionally, the step of determining a plurality of path detection algorithms based on the three-dimensional detection object position image using a trained path detection network to obtain the second detection path includes:

[0124] Inputting the three-dimensional detection object position image into the trained path detection network to obtain a second detection algorithm;

[0125] The plurality of constraint data in the constraint condition module and the position of the detection object in the three-dimensional detection object position image are input into the second detection algorithm to obtain a second detection path.

[0126] Optionally, the number of output neurons of the path detection network is equal to the number of path detection algorithms built into the algorithm module.

[0127] Optionally, obtaining multiple foreign body positions based on the three-mode image set through an image recognition network includes:

[0128] The image recognition network includes a first foreign object detection network, a second foreign object detection network and a third foreign object detection network.

[0129] The laser depth image, infrared image and camera image are respectively input into a first foreign body detection network, a second foreign body detection network and a third foreign body detection network to obtain a first foreign body feature, a second foreign body feature and a third foreign body feature.

[0130] In this embodiment, the first foreign object detection network, the second foreign object detection network, and the third foreign object detection network are Yolov8 networks. Because the input data is different, the parameters of the networks trained by the first foreign object detection network, the second foreign object detection network, and the third foreign object detection network are different.

[0131] In this embodiment, Figure 4As shown, the FOD detection module integrates video image recognition, LiDAR depth image recognition, and infrared pattern recognition components, enabling intelligent analysis of different types of image data transmitted back. This data is used to determine the presence of foreign object debris (FOD) on the airport floor. If FOD is present, the captured location of the transmitted image data is extracted. If FOD is not present, a signal indicating its absence is transmitted. The video image recognition component in this system integrates the YOLOv8 object detection algorithm.

[0132] Through the above method, because the YOLOv8 target detection algorithm has an efficient detection effect and a wide range of applications, it can accurately identify and detect foreign objects on the airport scene.

[0133] The first foreign body feature, the second foreign body feature and the third foreign body feature are superimposed to obtain an overlapping feature.

[0134] The overlapping features are input into a fully connected neural network to obtain multiple foreign body locations.

[0135] Among them, in this embodiment, the fully connected neural network (FCN).

[0136] Optionally, obtaining a three-dimensional detection object position image and an initial detection path based on the three-dimensional area image includes:

[0137] Acquire multiple laser perspective images; the laser perspective images represent images containing the detection object acquired by the UAV's laser radar at multiple perspectives.

[0138] Among them, multiple laser view angle images can obtain the complete surface of the inspection object;

[0139] Arrange multiple laser view images corresponding to the location of the drone in the order of time points from early to late to obtain the initial detection path.

[0140] The multiple laser view angle images are reconstructed in three dimensions to obtain a detection object position image; the detection object position image represents the position of the detection object in three dimensions.

[0141] In this embodiment, multiple laser view images are input into Neural Radiance Fields (NeRF) for three-dimensional reconstruction.

[0142] The position corresponding to the detection object in the detection object position image is marked on the three-dimensional region image to obtain a three-dimensional detection object position image.

[0143] The three-dimensional area image and the detection object position image have the same size.

[0144] The value 2 is used for marking, and the position with a gray value of 2 in the three-dimensional detection object position image represents the position of the detection object.

[0145] The three-dimensional detected object position image is a grayscale image.

[0146] The algorithm and display provided herein are not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing this type of system. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages can be utilized to realize the content of the present invention described herein, and the above description of specific languages is for the purpose of disclosing the best mode of the present invention.

[0147] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

Claims

1. An airport FOD inspection and operation safety assurance system based on an unmanned system, characterized by: Including detection equipment, FOD detection module, task scheduling module and wireless data transmission module: The detection equipment includes a high-definition camera, a laser radar, and an infrared sensor; the detection equipment is fixed on the drone; The task scheduling module is used to adjust the shooting position of the detection device to obtain a second detection path; the second detection path has multiple shooting positions as nodes of the path; the second detection path is the shortest path and the images captured by all nodes can capture the complete surface of the detection object; The detection object represents an object that needs to be inspected; The detection device is used to obtain a three-mode image set corresponding to the node on the second detection path; the three-mode image set includes a laser depth image, an infrared image, and a camera image; The FOD detection module is used to obtain multiple foreign body locations based on the three-mode image set through an image recognition network; The task scheduling module is used to perform path planning based on multiple foreign body positions to obtain a foreign body detection path; the foreign body detection path represents a path for clearing foreign bodies; The task scheduling module includes a constraint module, an algorithm module and a deep learning module; The constraint condition module is used to obtain multiple constraint data; The algorithm module has multiple path detection algorithms built in; The deep learning module has a built-in path detection network; The path detection network is trained using multiple path detection algorithms, including: Acquire a training three-dimensional detection object position image; the training three-dimensional detection object position image represents a three-dimensional detection object position image used for training at a historical time point; Based on the training three-dimensional detection object position image, a first detection path is obtained through multiple path detection algorithms; the first detection path represents the shortest path among the multiple path detection algorithms that can obtain the complete surface of the detection object; using the path detection algorithm corresponding to the first detection path as a determination algorithm; Based on the training three-dimensional detection object position image, a first detection algorithm is obtained through a path detection network; the first detection algorithm represents an algorithm predicted to be suitable for the three-dimensional detection object position image; The determination algorithm and the first detection algorithm are used to calculate the loss and train the path detection network.

2. The airport FOD inspection and operation safety assurance system based on an unmanned system according to claim 1 is characterized in that: The adjusting the shooting position of the detection device to obtain the second detection path includes: Marking locations where the drone is restricted from flying based on the plurality of constraint data to obtain a three-dimensional area image; locations where the grayscale value is 1 in the three-dimensional area image represent locations where the drone can fly, and locations where the grayscale value is 0 represent locations where the drone is restricted from flying; Based on the three-dimensional area image, a three-dimensional detection object position image and an initial detection path are obtained; the three-dimensional detection object position image represents an image that marks the position of the detection object and the flight restriction position; the initial detection path represents an initially set path for obtaining the complete surface of the detection object; Using the trained path detection network, multiple path detection algorithms are identified based on the three-dimensional detection object position image to obtain a second detection path; The second detection path is used to replace the initial detection path.

3. The airport FOD inspection and operation safety assurance system based on an unmanned system according to claim 1 is characterized in that: The method of obtaining a first detection path based on the training three-dimensional detection object position image by using multiple path detection algorithms includes: The input of the path detection algorithm is the position of the detection object in the training three-dimensional detection object position image and multiple constraint data in the corresponding constraint condition module; Multiple path detection algorithms correspondingly obtain multiple training detection paths; the training detection paths represent paths corresponding to the path detection algorithms that can obtain the complete surface of the detection object; Among the multiple training detection paths, the training detection path with the least number of nodes is used as the first detection path.

4. The airport FOD inspection and operation safety assurance system based on an unmanned system according to claim 1 is characterized in that: The first detection algorithm is obtained by using a path detection network based on the training three-dimensional detection object position image, including: The input of the path detection network is a training three-dimensional detection object position image; The output of the path detection network is a first detection algorithm.

5. The unmanned system-based airport FOD inspection and operation safety assurance system according to claim 2 is characterized in that: The method of obtaining a second detection path by using a trained path detection network and discriminating multiple path detection algorithms based on a three-dimensional detection object position image includes: Inputting the three-dimensional detection object position image into the trained path detection network to obtain a second detection algorithm; The plurality of constraint data in the constraint condition module and the position of the detection object in the three-dimensional detection object position image are input into the second detection algorithm to obtain a second detection path.

6. The airport FOD inspection and operation safety assurance system based on an unmanned system according to claim 1 is characterized in that: The number of output neurons of the path detection network is equal to the number of path detection algorithms built into the algorithm module.

7. The unmanned system-based airport FOD inspection and operation safety assurance system according to claim 1 is characterized in that: The method of obtaining multiple foreign body positions based on the three-mode image set through the image recognition network includes: The image recognition network includes a first foreign body detection network, a second foreign body detection network and a third foreign body detection network; Inputting the laser depth image, the infrared image and the camera image into a first foreign body detection network, a second foreign body detection network and a third foreign body detection network respectively to obtain a first foreign body feature, a second foreign body feature and a third foreign body feature; Superimposing the first foreign body feature, the second foreign body feature, and the third foreign body feature to obtain an overlapping feature; The overlapping features are input into a fully connected neural network to obtain multiple foreign body locations.

8. The unmanned system-based airport FOD inspection and operation safety assurance system according to claim 2 is characterized in that: The obtaining of a three-dimensional detection object position image and an initial detection path based on the three-dimensional area image includes: Acquire multiple laser view images; the laser view images represent images of the detected object acquired by the UAV's laser radar at multiple view angles; Arrange multiple laser view images corresponding to the drone's location in order from early to late time points to obtain the initial detection path; Reconstructing the multiple laser view angle images in three dimensions to obtain a detection object position image; the detection object position image represents the position of the detection object in three dimensions; The position corresponding to the detection object in the detection object position image is marked on the three-dimensional region image to obtain a three-dimensional detection object position image.

Images