Aircraft visual surveillance method, system and electronic equipment

Through the combination of monocular camera and ADS-B information, the virtual scene and interval contraction method are used to realize accurate positioning of the three-dimensional attitude and position of the aircraft in complex airport scenes, and solve the problem that the three-dimensional contour information cannot be provided in the prior art. The problem provides a lower cost and efficient monitoring solution.

CN118379351BActive Publication Date: 2024-11-22THE SECOND RES INST OF CIVIL AVIATION ADMINISTRATION OF CHINA
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202410565043.8
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2024-05-09
Publication Date
2024-11-22
Estimated Expiration
2044-05-09

AI Technical Summary

Technical Problem

The existing civil aircraft surveillance technology cannot accurately provide three-dimensional contour information, and is costly. The monocular vision method has limited scope in three-dimensional surveillance, making it difficult to meet the surveillance needs of complex airport scene activity areas.

Method used

Two-dimensional images are obtained through a monocular camera, combined with ADS-B information and a prior three-dimensional model, and iteratively adjusts the model pose using virtual scenes and interval shrinkage methods to achieve matching the three-dimensional model and two-dimensional image, and obtain the three-dimensional pose and position of the aircraft.

Benefits of technology

It provides a lower cost and better effect aircraft visual surveillance solution, which can quickly and accurately locate aircraft in complex airport scenes, solve the problem that a single image cannot accurately locate posture, reduce the number of boundary similarity comparisons, and improve surveillance accuracy.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118379351B_ABST
    Figure CN118379351B_ABST
Patent Text Reader

Abstract

The present invention provides an aircraft visual monitoring method, system and electronic equipment. In the airport scene of fixed video monitoring, the monocular video image is combined with ADS-B and prior three-dimensional model information. First, the ADS-B system is connected to search for the prior three-dimensional model, the model is mapped into two-dimensional information and the posture is solved in an iterative manner with the monitoring data. Then, the three-dimensional posture and position of the real aircraft are obtained by matching the position with the two-dimensional information, so as to solve the problem of aircraft attitude positioning under visual monitoring and realize a low-cost and high-precision aircraft visual monitoring system. In addition, in the process of attitude estimation, according to the finiteness and continuity of rotation, an interval contraction method is proposed to solve complex multi-level nonlinear problems. By continuously iteratively reducing the search interval, reducing the number of boundary similarity comparisons, and quickly converging to obtain rotation variables with controllable accuracy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of computer technology applications, and in particular to an aircraft visual monitoring method, system and electronic equipment. Background Art

[0002] Modern civil aviation aircraft surveillance is a complex and evolving field involving a variety of technologies and methods. The main purpose of these technologies is to ensure flight safety, improve flight efficiency, and support air traffic management. The current mainstream civil aviation aircraft surveillance technologies include surface detection radar surveillance systems, automatic dependent surveillance-broadcast (ADS-B) systems, and multi-lateration positioning systems (MLAT).

[0003] Surface detection radar is a traditional surveillance method that can work in all weather conditions and provide air traffic controllers with real-time aircraft position information, but the radar has limited signal range at the airport surface and is less effective; ADS-B provides more accurate aircraft position information and updates quickly, thereby improving pilots' situational awareness, but it relies on aircraft transmitters. If the aircraft is not equipped or the equipment fails, the system will not be able to track it; MLAT receives signals through multiple receiving stations and can provide very accurate aircraft position information within a certain area, but it is expensive and depends on the status of the receiving station. In addition, emerging technologies include remote tower technology, which uses video sources as a means of surveillance, but is currently used only for target detection rather than real coordinate monitoring. LiDAR surveillance technology has high accuracy but limited surveillance range and is still in the research stage. The above technologies are expensive and only detect aircraft position information without accurate three-dimensional contour information, that is, one-dimensional point traces.

[0004] With the development of the civil aviation economy, there are more and more large and busy civil aviation airports and multi-runway airports, the scale and traffic volume of airports are increasing, and the layout of runway and apron areas is becoming more and more complex, resulting in increasing operational conflicts in airport surface activity areas. There is an urgent need for the joint application of multiple airport surface surveillance methods to enhance the surveillance capabilities of the airport surface, further optimize ground taxiing, and improve operational efficiency.

[0005] Considering the cost and regional adaptability issues, video surveillance is widely used in the transportation field and is easy to promote and use. However, in terms of three-dimensional monitoring technology, the binocular vision method has a limited monitoring range, and the monocular vision method cannot obtain 3D target information. In civil aviation traffic scenarios, the aircraft models are basically fixed and known, and the existing monitoring system can provide certain data support, which can provide sufficient information constraints for monocular 3D target monitoring. Research on 3D target monitoring systems with monocular vision as the main means can better meet the monitoring needs of the civil aviation industry in terms of cost, installation, performance, etc. Summary of the invention

[0006] In view of the above technical problems, the technical solution adopted by the present invention is:

[0007] According to a first aspect of the present invention, a method for visually monitoring an aircraft is provided, wherein the application scenario is an airport scene monitored by a monocular camera, and the method comprises the following steps:

[0008] S100, obtaining a boundary of an aircraft in a real two-dimensional image captured by a current monocular camera in a set real scene as the real two-dimensional boundary of the aircraft.

[0009] S200, using the currently received ADS-B information to search the prior model library, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene, and the virtual camera is used to capture a virtual two-dimensional image of the 3D model, and is set at the origin of a virtual coordinate system corresponding to the virtual scene.

[0010] S300, adjusting the posture rotation amount of the 3D model according to the set method, and obtaining the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the optimal matching degree as the target posture amount of the 3D model; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image.

[0011] S400 , based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera, obtaining a corresponding virtual ground boundary in the virtual scene as a target virtual ground boundary.

[0012] S500: Match the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image.

[0013] S600 , based on the local coordinate system of the reference point of the 3D model and the correspondence between the target virtual ground boundary point and the reference point, obtain the coordinates of the feature point of the 3D model.

[0014] According to a second aspect of the present invention, there is provided an aircraft visual monitoring system, the application scenario of which is an airport scene monitored by a monocular camera, the system comprising:

[0015] The first acquisition module is used to acquire the boundary of the aircraft in the real two-dimensional image captured by the current monocular camera in the set real scene as the real two-dimensional boundary of the aircraft.

[0016] The second acquisition module is used to search the prior model library using the currently received ADS-B information, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene. The virtual camera is used to capture a virtual two-dimensional image of the 3D model and is set at the origin of a virtual coordinate system corresponding to the virtual scene.

[0017] The third acquisition module is used to adjust the posture rotation amount of the 3D model according to a set method, and obtain the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the optimal matching degree as the target posture amount of the 3D model; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image.

[0018] The fourth acquisition module is used to acquire a corresponding virtual ground boundary in the virtual scene based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera as a target virtual ground boundary.

[0019] The matching module is used to match the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image.

[0020] The fifth acquisition module is used to acquire the coordinates of the feature points of the 3D model based on the local coordinate system of the reference point of the 3D model and the correspondence between the target virtual ground boundary point and the reference point.

[0021] According to a third aspect of the present invention, there is provided an electronic device, comprising a processor and a memory;

[0022] The processor is used to execute the steps of the aforementioned method by calling the program or instruction stored in the memory.

[0023] The present invention has at least the following beneficial effects:

[0024] The aircraft visual monitoring method and system provided by the embodiment of the present invention is aimed at the airport scene monitored by a monocular camera, combines the ADS-B system with the prior three-dimensional model, and uses the iterative method of mapping the three-dimensional model and the two-dimensional image to solve the problem that a single image cannot accurately determine the attitude and positioning, and provide a lower cost and better solution for more general scenarios. In addition, in the attitude estimation process, according to the finiteness and continuity of the rotation, the interval contraction method is used to solve complex multi-level nonlinear problems, and the search interval is narrowed by continuous iteration, the number of boundary similarity comparisons is reduced, and rapid convergence is achieved to obtain rotation variables with controllable accuracy.

[0025] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0027] Figure 1 A flowchart of a method for visually monitoring an aircraft provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0029] The embodiment of the present invention provides an aircraft visual monitoring method, the application scenario of which is an airport scene monitored by a fixed video such as a monocular camera. The aircraft may be an airplane such as a civil aircraft.

[0030] The method provided by the present invention combines ADS-B with prior three-dimensional model information. The calculation process mainly includes two steps. The first step is to calculate the attitude of the aircraft. The attitude parameters of the three-dimensional model in the virtual scene are used as variables. The boundary contour is directly extracted through the perspective imaging model of the camera and the boundary perception algorithm. The virtual two-dimensional boundary operator is generated by using the normalized central moment. After that, the boundary operator is extracted from the independent target in the two-dimensional image collected in real time. The extracted target is matched with the virtual model one by one. The similarity between the real boundary operator and the virtual boundary operator is evaluated. The interval contraction method is used to continuously iterate the attitude parameters of the model in the virtual scene to obtain the virtual model with the smallest boundary operator evaluation error. The attitude variables of the virtual model at this time are output, and the three-dimensional attitude result of the target can be obtained through a single two-dimensional image; the second step is to calculate the position of the aircraft. The real coordinates of the bottom boundary of the target are calculated by the collinear equation through the boundary pixels of the target, and then the local coordinates of the bottom boundary relative to the imaging center are obtained by using the real coordinates of the imaging center. The attitude quantity and local coordinates calculated above can be restored to the target position in the virtual scene. At this point, the coordinates and orientation of the aircraft contour can be obtained by extracting the aircraft model or geometric features.

[0031] Furthermore, in the embodiment of the present invention, the method includes the following conditions and reasonable assumptions:

[0032] (1) The airport surface surveillance area is a local area. Within the required error range, it can be assumed to be an ideal plane, that is, the aircraft only moves in the local plane.

[0033] (2) The camera imaging is a central perspective projection, and the calibration results of the internal and external parameters are within the controllable error.

[0034] (3) In the collinearity equation in the extrinsic calibration and mapping stage, the Z value is the ideal plane elevation to satisfy the solution of the underdetermined equation with a single image.

[0035] (4) The three-dimensional standard model is a priori processing data and can be directly applied to the algorithm.

[0036] (5) In the virtual scene and the real scene, elements such as the camera position, camera plane, and the object being photographed all correspond one to one.

[0037] Furthermore, if Figure 1 As shown, the aircraft visual monitoring method provided by the embodiment of the present invention may include the following steps:

[0038] S100, obtaining a boundary of an aircraft in a real two-dimensional image captured by a current monocular camera in a set real scene as the real two-dimensional boundary of the aircraft.

[0039] In an embodiment of the present invention, in a real scene, a monocular camera will collect a two-dimensional image of the monitoring area in real time. The boundary of the aircraft in the real two-dimensional image can be obtained by using an existing target detection method, for example, using the Yolo-v8 method to detect the aircraft target in the current scene and extract its outline, i.e., the boundary.

[0040] Furthermore, in the embodiment of the present invention, the coordinates of any bottom point of the boundary of the aircraft can be extracted through the camera collinearity equation to link the ADS-B information.

[0041] In the embodiment of the present invention, if the extracted boundary effect is not good, a matting method may be used to correct it to obtain a boundary with better effect.

[0042] It is known to those skilled in the art that a two-dimensional image may include one or more aircraft.

[0043] In the embodiment of the present invention, the real scene is an airport scene.

[0044] Furthermore, in the embodiment of the present invention, the internal and external parameters of the monocular camera are calibrated internal and external parameters. The calibration method can be an existing method, for example, the internal and external parameters of the camera can be calibrated by high-precision tools such as calibration targets and total stations. Specifically, the internal parameter calibration matrix K to be solved is:

[0045]

[0046] where f x =αf,f y =βf, α, β is the scaling ratio of the camera coordinate system and the pixel coordinate system on the X-axis and Y-axis, c x ,c y is the origin offset, and f is the focal length of the camera.

[0047] The solved external parameter calibration matrix R is:

[0048]

[0049] Where R 3×3 is a 3×3 rotation matrix, and t is a 3×1 translation matrix.

[0050] S200, using the currently received ADS-B information to search the prior model library, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene, and the virtual camera is used to capture a virtual two-dimensional image of the 3D model, and is set at the origin of a virtual coordinate system corresponding to the virtual scene.

[0051] In an embodiment of the present invention, the 3D model is pre-stored in a prior model library. Each aircraft corresponds to a 3D model. Specifically, the 3D model corresponding to the aircraft can be retrieved from the prior model library based on the location and flight number provided by the ADS-B information.

[0052] In the embodiment of the present invention, the coordinate axes of the virtual coordinate system are parallel to the world coordinate system. The elevation of the ideal plane in the virtual scene, that is, the plane on which the 3D model moves, that is, the distance from the virtual camera is Z0, to simulate the movement of the aircraft target in the local plane.

[0053] In the embodiment of the present invention, the set initial posture may be the true north direction of the world coordinate system; and the set position may be the center of the monitoring area.

[0054] S300, adjusting the posture rotation amount of the 3D model according to the set method, and obtaining the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the optimal matching degree as the target posture amount of the 3D model, that is, the posture of the current 3D model is the posture corresponding to the target posture amount; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image.

[0055] Furthermore, S300 may specifically include:

[0056] S301, set the interval iteration number n = 1; S302, obtain the posture rotation amount α of the current 3D model n =θ min +λ·(n-1), and rotate the current 3D model by α n Then perform boundary extraction to obtain the nth virtual two-dimensional boundary A n v;θ min is the minimum angle value, the initial value is 0; λ is the angle step size.

[0057] In the embodiment of the present invention, the virtual two-dimensional boundary can be obtained by using the same method as the method for extracting the two-dimensional boundary of a real image.

[0058] In the embodiment of the present invention, λ may be obtained based on calculation accuracy. In an exemplary embodiment, λ may be set to 5 to 10°, preferably 10°.

[0059] S303, obtain A n v and A R The similarity between n v , A R ), set C = C + 1; if D (A n v , A R )≤△, or C≥C0, execute S306, otherwise, execute S604; C is the iteration counter, and the initial value of C is 0; C0 is the set number threshold; △ is the set similarity threshold; A R It is a real two-dimensional boundary.

[0060] In the embodiment of the present invention, △ can be set based on actual needs and can be a positive number close to 0. C0 can be an empirical value.

[0061] In the embodiment of the present invention, D(A n v , A R ) meets the following conditions:

[0062] D(A n v , A R )=|MR -M n V ∣,M R Based on A R The obtained real two-dimensional boundary operator, M n V Based on A n v is the virtual two-dimensional boundary operator obtained, where || means taking the absolute value. The two-dimensional boundary operator has nothing to do with scaling and translation but is related to rotation.

[0063] In the embodiment of the present invention, the two-dimensional boundary operator is calculated by normalized central moment of the third order or less. The formula principle of the normalized central distance is:

[0064]

[0065] Among them, x0, y0 are the centers of the boundary point set, satisfying the following conditions:

[0066]

[0067]

[0068]

[0069]

[0070] p,q represent the order of two-dimensional coordinates x,y respectively; μ pq is the mixing central moment of p+q; μ′ pq is the normalized mixing central moment; P(x,y) is a two-dimensional grayscale density function, representing the grayscale feature; X represents the x-coordinate set of the boundary, and Y represents the y-coordinate set of the boundary; A1 is the mean of X, and A2 is the mean of Y.

[0071] According to the above content, it can be seen that the two-dimensional boundary operator M satisfies the following conditions:

[0072]

[0073] S304, if Set n=n+1, execute S602; if Setting θ min =(n i -1)×λ,θ max =(n i +1)×λ, execute S305; θ max is the maximum angle value, the initial value is 360°; n i is min(D(A 1 v , A R ), D(A 2v , A R ), ..., D(A n v , A R )) corresponds to the number of interval iterations; min() is the minimum value.

[0074] S305, Settings Execute S301.

[0075] S306: Taking the current posture rotation amount as the target posture amount.

[0076] The technical effect of S300 is that it can solve complex multi-level nonlinear problems by continuously iterating to narrow the search interval, reducing the number of boundary similarity comparisons, and quickly converge to obtain rotation variables with controllable accuracy.

[0077] S400 , based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera, obtaining a corresponding virtual ground boundary in the virtual scene as a target virtual ground boundary.

[0078] In the embodiment of the present invention, for the real two-dimensional image collected in reality, the internal and external parameters of the monocular camera are calculated to establish the collinearity equation, and the A R The local coordinates of the bottom ground boundary of the target virtual scene are calculated to obtain the target virtual bottom ground boundary. Specifically, the target virtual bottom ground boundary satisfies the following conditions:

[0079] Among them, X S ,Y S ,Z S They are the coordinates of the monocular camera on the X-axis, Y-axis, and Z-axis in the corresponding real coordinate system; u and v are the coordinates of a single pixel point in the real bottom boundary, V and Y V are respectively the coordinates of the pixel point corresponding to a single pixel point (u, v) in the virtual scene on the X-axis and Y-axis in the virtual coordinate system; Z0 is the distance between the virtual camera and the plane in which the 3D model moves.

[0080] S500: Match the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image.

[0081] In an embodiment of the present invention, the current ground-attached boundary of the 3D model is matched with the target virtual ground-attached boundary by translating the current ground-attached boundary of the 3D model to a position corresponding to the target virtual ground-attached boundary so as to overlap with the target virtual ground-attached boundary. The specific translation method may be the prior art and will not be described in detail in the present invention.

[0082] S600 , based on the local coordinate system of the reference point of the 3D model and the correspondence between the target virtual ground boundary point and the reference point, obtain the coordinates of the feature point of the 3D model.

[0083] In the embodiment of the present invention, the reference point of the 3D model may be any feature point on the corresponding aircraft, for example, the nose point. Since the 3D model is rigid, the local coordinate relationship of each position point relative to a certain reference point is known. Therefore, based on the coordinates of each point on the virtual ground boundary of the 3D model target, the coordinates of any feature point on the 3D model can be obtained. The feature points can be set based on actual needs, for example, they can include wings, tail fins, engines, etc.

[0084] At this point, the position and attitude of the aircraft corresponding to the current real two-dimensional image can be obtained.

[0085] Based on the same technical concept, the present invention provides an aircraft visual monitoring system, the application scenario of which is an airport scene monitored by a monocular camera, and the system comprises:

[0086] The first acquisition module is used to acquire the boundary of the aircraft in the real two-dimensional image captured by the current monocular camera in the set real scene as the real two-dimensional boundary of the aircraft.

[0087] The second acquisition module is used to search the prior model library using the currently received ADS-B information, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene. The virtual camera is used to capture a virtual two-dimensional image of the 3D model and is set at the origin of a virtual coordinate system corresponding to the virtual scene.

[0088] The third acquisition module is used to adjust the posture rotation amount of the 3D model according to a set method, and obtain the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the optimal matching degree as the target posture amount of the 3D model; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image.

[0089] The fourth acquisition module is used to acquire a corresponding virtual ground boundary in the virtual scene based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera as a target virtual ground boundary.

[0090] The matching module is used to match the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image.

[0091] The fifth acquisition module is used to acquire the coordinates of the feature points of the 3D model based on the local coordinate system of the reference point of the 3D model and the correspondence between the target virtual ground boundary point and the reference point.

[0092] The system can be used to perform Figure 1 Therefore, for the functions that can be realized by each functional module of the system, please refer to Figure 1 The description of the illustrated embodiment is not repeated in detail.

[0093] An embodiment of the present invention further provides an electronic device, comprising a processor and a memory; the processor is used to execute the steps of the aforementioned method by calling a program or instruction stored in the memory.

[0094] In summary, the aircraft visual monitoring solution provided by the embodiment of the present invention, for the airport scene monitored by a monocular camera, combines ADS-B information with a priori model library and uses the method of solving the attitude first and then matching the position, which solves the problem that a single image cannot accurately determine the attitude and position of the aircraft, and provides a lower cost and better solution for more general scenarios. In addition, in the attitude estimation process, according to the finiteness and continuity of the rotation, the interval contraction method is used to solve complex multi-level nonlinear problems, and the search interval is narrowed by continuous iteration, the number of boundary similarity comparisons is reduced, and rapid convergence is achieved to obtain rotation variables with controllable accuracy.

[0095] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for visually monitoring an aircraft, characterized in that: The application scenario is an airport scene monitored by a monocular camera. The method comprises the following steps: S100, obtaining a boundary of an aircraft in a real two-dimensional image captured by a current monocular camera in a set real scene as a real two-dimensional boundary of the aircraft; S200, using the currently received ADS-B information to search the prior model library, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene, and the virtual camera is used to capture a virtual two-dimensional image of the 3D model, and is set at the origin of a virtual coordinate system corresponding to the virtual scene; S300, adjusting the posture rotation amount of the 3D model according to a setting method, and obtaining the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the best matching degree as the target posture amount of the 3D model; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image; S400, based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera, obtaining a corresponding virtual ground boundary in the virtual scene as a target virtual ground boundary; S500, matching the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image; S600, acquiring coordinates of feature points of the 3D model based on a local coordinate system of the reference point of the 3D model and a correspondence between the target virtual ground boundary point and the reference point; S300 specifically includes: S301, set the interval iteration number n=1; S302, obtaining the posture rotation amount α of the current 3D model n =θ min +λ×(n-1), and rotate the current 3D model by α n Then perform boundary extraction to obtain the nth virtual two-dimensional boundary A n v;θ min is the minimum angle value, the initial value is 0; λ is the angle step; S303, obtain A n v and A R The similarity between n v , A R ), set C=C+1; if D(A n v , A R )≤△, or C≥C0, execute S306, otherwise, execute S304; C is the iteration counter, and the initial value of C is 0; C0 is the set number threshold; △ is the set similarity threshold; A R It is the real two-dimensional boundary; S304, if 0≤n≤(θ max -θ min ) / λ, set n=n+1, and execute S302; if n>(θ max -θ min ) / λ, set θ min =(n i -1)×λ,θ max =(n i +1)×λ, execute S305; θ max is the maximum angle value, the initial value is 360°; n i is min(D(A 1 v , A R ), D (A 2 v , A R ), ..., D (A n v , A R )) corresponds to the number of interval iterations; min() is the minimum value; S305, set λ=λ×△d1 / △d2, where △d1 is the difference between the maximum angle value and the minimum angle value before updating, and △d2 is the difference between the maximum angle value and the minimum angle value after updating, and execute S301; S306: Taking the current posture rotation amount as the target posture amount.

2. The method according to claim 1, characterized in that D(A n v , A R ) meets the following conditions: D(A n v , A R )=|M R -M n V ∣,M R Based on A R The obtained real two-dimensional boundary operator, M n V Based on A n v is the virtual two-dimensional boundary operator obtained, and || indicates taking the absolute value.

3. The method according to claim 2, characterized in that The two-dimensional boundary operator is calculated by normalized central moment below the third order.

4. The method according to claim 1, characterized in that The target virtual bottom ground boundary meets the following conditions: , where X S , Y S , Z S They are the coordinates of the monocular camera on the X-axis, Y-axis, and Z-axis in the corresponding real coordinate system; u and v are the coordinates of a single pixel point in the real bottom boundary, V and Y V are the coordinates of the pixel point corresponding to a single pixel point (u, v) in the virtual scene on the X-axis and Y-axis in the virtual coordinate system; Z0 is the distance between the virtual camera and the plane in which the 3D model moves, K is the solved intrinsic calibration matrix, and R is the solved extrinsic calibration matrix.

5. The method according to claim 1, characterized in that The initial posture is set to the true north direction of the world coordinate system.

6. The method according to claim 1, characterized in that The set position is the center of the monitoring area.

7. The method according to claim 1, characterized in that The actual two-dimensional boundary is obtained by using the Yolo-v8 method.

8. An aircraft visual surveillance system, characterized in that: The application scenario is an airport scene monitored by a monocular camera. The system includes: A first acquisition module is used to acquire the boundary of the aircraft in the real two-dimensional image captured by the current monocular camera in the set real scene as the real two-dimensional boundary of the aircraft; A second acquisition module is used to search the prior model library using the currently received ADS-B information, obtain the 3D model corresponding to the aircraft, and place the obtained 3D model at a set position in the constructed virtual scene with a set initial posture; the virtual scene is constructed based on the real scene, and a virtual camera corresponding to the monocular camera is provided in the virtual scene, and the virtual camera is used to capture a virtual two-dimensional image of the 3D model, and is set at the origin of a virtual coordinate system corresponding to the virtual scene; A third acquisition module is used to adjust the posture rotation amount of the 3D model according to a set method, and obtain the posture rotation amount that makes the matching degree between the virtual two-dimensional boundary corresponding to the 3D model and the real two-dimensional boundary the best matching degree as the target posture amount of the 3D model; the virtual two-dimensional boundary is obtained based on the virtual two-dimensional image; a fourth acquisition module, configured to acquire a virtual ground boundary corresponding to the virtual scene based on the real ground boundary in the real two-dimensional boundary and the coordinates of the monocular camera, as a target virtual ground boundary; A matching module, used for matching the current ground-attached boundary of the 3D model with the target virtual ground-attached boundary to obtain a 3D model that matches the real two-dimensional image; A fifth acquisition module, configured to acquire coordinates of feature points of the 3D model based on a local coordinate system of the reference point of the 3D model and a correspondence between the target virtual ground boundary point and the reference point; The third acquisition module is specifically used to perform the following operations: S301, set the interval iteration number n=1; S302, obtaining the posture rotation amount α of the current 3D model n =θ min +λ×(n-1), and rotate the current 3D model by α n Then perform boundary extraction to obtain the nth virtual two-dimensional boundary A n v;θ min is the minimum angle value, the initial value is 0; λ is the angle step; S303, obtain A n v and A R The similarity between n v , A R ), set C=C+1; if D(A n v , A R )≤△, or C≥C0, execute S306, otherwise, execute S304; C is the iteration counter, and the initial value of C is 0; C0 is the set number threshold; △ is the set similarity threshold; A R It is the real two-dimensional boundary; S304, if 0≤n≤(θ max -θ min ) / λ, set n=n+1, and execute S302; if n>(θ max -θ min ) / λ, set θ min =(n i -1)×λ,θ max =(n i +1)×λ, execute S305; θ max is the maximum angle value, the initial value is 360°; n i is min(D(A 1 v , A R ), D (A 2 v , A R ), ..., D (A n v , A R )) corresponds to the number of interval iterations; min() is the minimum value; S305, set λ=λ×△d1 / △d2, where △d1 is the difference between the maximum angle value and the minimum angle value before updating, and △d2 is the difference between the maximum angle value and the minimum angle value after updating, and execute S301; S306: Taking the current posture rotation amount as the target posture amount.

9. An electronic device, characterized in that: including a processor and a memory; The processor is used to execute the steps of the method according to any one of claims 1 to 7 by calling the program or instruction stored in the memory.