A Cross-Satellite Airport Aircraft Target Tracking Method Incorporating Road Constraints

Abstract

The present invention discloses a method for tracking aircraft targets across satellite airports in combination with road constraints. The method includes the following steps: Step S101: Use a target detector to detect aircraft targets in the airport images collected by the satellite constellation, obtain the detection box data of the aircraft targets in the scene, and calculate the time difference between the data collected by adjacent satellites in the constellation; Step S102: Establish a trajectory prediction model; use the detection box data obtained in Step S101, and respectively adopt a re-identification network and a multi-modal trajectory prediction network to extract appearance and motion information of the aircraft targets in the airport area; Step S103: Use an optimized matching algorithm to achieve cross-frame target matching, and combine airport road constraints to correct the rationality of the matched trajectories and draw the final motion trajectory of the aircraft targets. This method can reduce tracking errors and improve the accuracy and stability of tracking aircraft targets across satellites.

Description

Technical Field

[0001] The present invention relates to the field of satellite surveillance and target tracking, and particularly to a method for implementing aircraft target tracking on a dataset of cross-satellite imaging and introducing airport road constraints to optimize the tracking path, which is used to improve the accuracy and stability of cross-satellite aircraft target tracking. Background Art

[0002] With the large-scale deployment of low-earth-orbit commercial Internet satellites and the increase in the number of remote sensing satellites, multiple satellites working together can achieve high-frequency revisit of ground areas. In the field of satellite surveillance and target tracking, the existing technologies mainly rely on the data processing of a single imaging platform and perform target tracking based on traditional Kalman filtering. However, they face problems such as insufficient accuracy of the tracking path, great difficulty in integrating cross-satellite data, limited data acquisition duration, and inability to provide long-term continuous airport surveillance capabilities. Especially under the satellite imaging conditions at different times and different perspectives, the robustness of target tracking is poor. The existing methods do not fully utilize the road information in the airport scene for prediction, resulting in limited accuracy of cross-satellite target matching and difficulty in realizing cross-satellite target tracking. Due to the technical problems of large tracking errors and insufficient accuracy and stability in the cross-satellite target tracking method of the existing technology, the present invention provides a solution to the above technical problems. Summary of the Invention

[0003] Aiming at the problems existing in the existing technology, the purpose of the present invention is to provide a cross-satellite airport aircraft target tracking method combined with road constraints, which is used to reduce tracking errors and improve the accuracy and stability of cross-satellite aircraft target tracking.

[0004] To achieve the above purpose, the present invention provides a cross-satellite airport aircraft target tracking method combined with road constraints, and the method includes the following steps:

[0005] Step S101: Use a target detector to detect aircraft targets in the airport images collected by the satellite constellation, obtain the detection box data of the aircraft targets in the scene, and calculate the time difference between the data collected by adjacent satellites in the constellation at the same time;

[0006] Step S102: Establish a trajectory prediction model; use the detection box data obtained in Step S101, and respectively adopt a re-identification network and a multi-modal trajectory prediction network to extract appearance and motion information of the aircraft targets in the airport area;

[0007] Step S103: Use an optimized matching algorithm to achieve cross-frame target matching, and combine airport road constraints to reasonably correct the matched trajectory, and draw the final motion trajectory of the aircraft target.

[0008] Further, in Step S101, it specifically includes:

[0009] Step S201: Use a target detector to identify airplanes within the field of view, generate rotated bounding boxes and crop the image data, and combine channel and spatial attention modules to enhance the features extracted by the backbone network;

[0010] Step S202: Calculate and give the imaging time difference between adjacent satellites in the satellite constellation based on the imaging time information contained in the image.

[0011] Furthermore, the target detector is the YOLOv7 target detection network.

[0012] Furthermore, in step S102, it further includes collecting ADS-B data of airport airplanes, cleaning the data, and ignoring the trajectories of stationary airplanes.

[0013] Furthermore, step S102 further includes:

[0014] Step S301: Collect historical motion trajectory data of airplanes through ADS-B and preprocess the data, including extracting the airplane target trajectories from satellite images, and performing data cleaning, outlier removal, and normalization processing;

[0015] Step S302: Use a multi-modal trajectory prediction network based on LSTM to predict future trajectories.

[0016] Furthermore, in step S302, a multi-modal prediction method is used to generate multiple possible trajectories, and Gaussian noise is added during the training process to simulate possible motion changes of the airplanes, improving the adaptability of the model to uncertain motion patterns.

[0017] Furthermore, step S102 further includes:

[0018] Step S303: Determine the prediction steps of the trajectory prediction model according to the calculated time difference, combine the trajectory prediction end point, give a candidate range, perform cropping, and obtain the cropped target image;

[0019] Step S304: Use a lightweight re-identification network to extract the appearance features of the target.

[0020] Furthermore, the lightweight re-identification network consists of four convolutional blocks and a downsampling module, takes the cropped target image as the input, and outputs a 128×1 feature vector.

[0021] Furthermore, use the airport road node map to correct the trajectory, match the starting point and the ending point with the nearest road nodes, and then form a tracking trajectory that conforms to the roads within the airport.

[0022] The beneficial effects of the present invention are as follows:

[0023] The present invention proposes a method for tracking aircraft targets across satellites in combination with road constraints. A multi-modal trajectory prediction model is trained through a multi-layer stacked Long Short-Term Memory (LSTM) network, and an optimized cross-frame matching algorithm is combined. The trajectory is corrected using airport road constraints, effectively improving the tracking accuracy of airport aircraft targets on cross-satellite datasets. Road constraints not only include traditional airport road node information but also aircraft movement trajectory constraints. Since the movement trajectory of an aircraft in an airport is naturally restricted by roads, the trajectory itself contains road information. This method can not only effectively reduce tracking errors but also provide new technical means and application prospects for the field of satellite remote sensing data processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flowchart of the method for tracking aircraft targets across satellites in combination with road constraints according to the present invention;

[0025] Figure 2 is a structural diagram of a multi-modal trajectory network;

[0026] Figure 3 is a schematic diagram of trajectory prediction screening;

[0027] Figure 4 is an effect diagram of target tracking. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0028] The technical solutions of the present invention will be described clearly and completely below in conjunction with the accompanying drawings. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present invention belong to the scope of protection of the present invention.

[0029] In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "left", "right", "vertical", "horizontal", "inner", "outer", etc. is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, and does not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be construed as a limitation of the present invention. In addition, the terms "first", "second", "third" are only used for descriptive purposes and cannot be construed as indicating or implying relative importance.

[0030] In the description of the present invention, it should be noted that unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific situations.

[0031] The following will be combined with Figures 1-4 to detail the specific implementation manners of the present invention. It should be understood that the specific implementation manners described herein are only used to illustrate and explain the present invention and are not used to limit the present invention.

[0032] The present invention proposes a new aircraft target tracking method, which comprehensively utilizes trajectory prediction, time difference calculation, and an optimized matching strategy to improve tracking accuracy and robustness, and can achieve precise and stable tracking of aircraft targets on a cross-satellite imaging dataset.

[0033] The core of the tracking method of the present invention is to use historical aircraft motion data containing trajectory information for deep learning modeling, construct a trajectory prediction model through a multi-layer stacked LSTM network to predict the possible positions of the aircraft at future moments, and perform cross-frame target tracking in combination with an optimized matching algorithm; finally, correct the tracking trajectory through airport road constraints to ensure the rationality of the trajectory and improve the accuracy of tracking. As Figure 1 shown, this method includes the following steps:

[0034] Step S101: Use a YOLOv7 object detector to detect aircraft targets in the airport images collected by the satellite constellation, obtain the detection box information of the aircraft targets in the scene, and at the same time calculate the time difference between the data collected by adjacent satellites in the constellation.

[0035] Step S102: Collect ADS-B (Automatic Dependent Surveillance-Broadcast) data of airport aircraft, clean the data, and ignore the trajectories of stationary aircraft. Train a multi-modal trajectory prediction network based on historical aircraft motion trajectory data, learn the motion patterns of aircraft targets in the airport, and establish a trajectory prediction model. Use the detection box data obtained in Step S101, and respectively use a re-identification network and a multi-modal trajectory prediction network to extract appearance and motion information of aircraft targets in the airport area.

[0036] Step S103: Use an optimized matching algorithm to achieve cross-frame target matching, and combine airport road constraints to correct the rationality of the matched trajectory, draw the final motion trajectory of the aircraft target, and ensure that the trajectory accurately reflects the dynamic motion characteristics of the aircraft target and conforms to airport road constraints.

[0037] The detailed steps of the relevant modules are introduced below:

[0038] Regarding step S101, the detailed steps of the method provided in this solution are as follows (S201 - S203):

[0039] Step S201: Use a low - earth orbit satellite constellation to collect satellite remote - sensing image data of the airport area and perform pre - processing (including orthorectification, registration, etc.). (This part corresponds to the satellite acquisition image frame in Figure 1 , and a single satellite provides a single image data.) Use the YOLOv7 object detector to identify the airplanes within the field of view, generate rotated bounding boxes, and crop the image data. Combine the channel and spatial attention modules to enhance the features extracted by the backbone network, enabling the detection network to focus on the key parts of the feature map that contain the targets and improving the recognition ability for small and weak targets.

[0040] Step S202: Calculate and give the imaging time difference between adjacent satellites in the satellite constellation according to the imaging time information contained in the image, and introduce a time variable for the trajectory prediction network to improve the accuracy of predicting the moving trajectory of the target during the target tracking process.

[0041] Regarding step S102, the detailed steps of the method provided in this solution are as follows (S301 - S304):

[0042] Step S301: Collect the historical motion trajectory data of airplanes through ADS - B (Automatic Dependent Surveillance - Broadcast) and pre - process the data, including extracting the airplane target trajectories from satellite images and performing data cleaning, outlier removal, and normalization to ensure data consistency. The road constraints include trajectory constraints and road network constraints. The historical motion trajectory data of airplanes contains the static information of the airport runway and the dynamic information of the airplane movement, which can provide more effective information for target tracking. The dynamic information includes motion speed, direction, etc.

[0043] Step S302: Adopt a multi - modal trajectory prediction network based on LSTM, such as Figure 2As shown, to optimize the network structure, the present invention adopts an encoder-decoder structure. The encoder maps historical observation trajectories to trajectory features (saved as LSTM hidden states), and the decoder takes the existing historical trajectory features (the output of the encoder) and random variables as inputs to achieve multi-modal trajectory prediction. Multiple layers of LSTM are used in the encoder to fit the complex movement trajectories of aircraft in airport scenarios. At the end of the decoder, multi-round linear mapping is adopted to obtain multiple possible movement trajectories and extrapolate the positions at subsequent multiple moments to adapt to time-series trajectory data and improve the long-term prediction ability. To improve the prediction accuracy, interpolation technology is used to generate finer-grained intermediate positions to make the trajectory data smoother and more continuous. Recurrent units are used to capture historical movement patterns to predict future trajectories. A multi-modal prediction method is adopted to generate multiple possible trajectories, and Gaussian noise is added during the training process to simulate possible movement changes of the aircraft and improve the adaptability of the model to uncertain movement patterns. During the training process, supervision is carried out by combining the loss of the predicted end point and the entire trajectory to improve the accuracy of the predicted trajectory end point.

[0044] The composite loss function supervises the trajectory prediction network by combining the absolute distance of the final predicted point and the matching distance of the entire trajectory. Since the time interval of satellite constellation imaging data is relatively long, multiple rounds of position prediction are required to estimate the target position. Therefore, more emphasis is placed on the accuracy of the final predicted point. Endpoint loss is defined as:

[0045] ;

[0046] where and are the ground truth and the predicted final position of the sample. is the number of predicted trajectories, that is, the number of endpoint points of all trajectories. The mean absolute distance loss of the endpoints of multiple predicted trajectories is obtained by calculating the absolute distance between the predicted position and the true position.

[0047] Overall trajectory loss combines L1 (mean absolute error) and L2 (mean square error) losses to explain the error at each time step during the trajectory prediction process:

[0048] ;

[0049] In the formula is the number of time steps of the input trajectory, is the number of time steps of the predicted trajectory, and are the weights of L1 and L2 losses.

[0050] The composite loss function adds a weight to the endpoint loss to emphasize its importance:

[0051] ;

[0052] This loss is used to supervise the trajectory prediction network as a measurement metric. When the value is large, the network prediction result is inaccurate. During the training process, the parameters are gradually adjusted to reduce the loss, and finally the network can make the prediction result closer to the true value.

[0053] Step S303: According to the calculated time difference, determine the prediction steps of the trajectory prediction model to make the prediction step size approximate the time span of adjacent frames, while avoiding errors caused by excessive extrapolation. Input the position information of the aircraft target in the previous few frames of images into the trajectory prediction model to generate a set of candidate positions for multiple possible future aircraft targets. Combine the trajectory prediction end point to give a candidate range, which is determined by the time difference of adjacent data. As Figure 3 shown, obtain the extrapolated trajectory of the aircraft target at the previous moment, and obtain the candidate range at the trajectory end point, and exclude the detection targets outside the range to reduce the computational consumption in the subsequent matching process.

[0054] Step S304: Adopt a lightweight re-identification network to extract the appearance features of the target. This network is mainly used for pedestrian re-identification and is used to calculate the appearance similarity of targets in different images. In this invention, the network is retrained with aircraft data to obtain a re-identification network suitable for aircraft targets. This network consists of four convolutional blocks and a downsampling module, takes the cropped target image as input, and outputs a 128×1 feature vector. This feature vector contains the specific appearance features of the aircraft target and is robust to illumination changes, perspective changes, and scale changes.

[0055] Regarding step S103, the detailed steps of the method provided by this solution are as follows (S401~S403):

[0056] In cross-frame matching, considering the characteristic of low time resolution of satellite constellation data acquisition, moving targets will have a large range of motion, resulting in a large difference between moving targets and stationary targets. Therefore, different matching schemes need to be adopted for different targets.

[0057] Step S401: The displacement of static targets between adjacent frames is small, and their position changes are mainly due to angle or georegistration errors. Calculate the overlap degree of detection frames in adjacent frames using the Intersection over Union (IoU), and screen out the set of candidate static targets. Perform preliminary matching on targets with IoU higher than the set threshold, and use the Hungarian algorithm to optimize the combination of all candidate matching pairs. For the airport queuing phenomenon, IoU matching may misidentify different targets as the same target. At this time, use the appearance features of the target extracted by the re-identification network to calculate the cosine similarity of targets in adjacent frames:

[0058] ;

[0059] Among them, and respectively represent the appearance feature vectors of the target. If the cosine similarity is lower than the set threshold, it is considered that the matching pair is inconsistent in appearance, and this matching pair will be excluded, and the unmatched target will be moved to the moving target matching stage.

[0060] Step S402: Based on the LSTM trajectory prediction model, generate multiple possible motion trajectories. Calculate the Euclidean distance between the predicted trajectory and the detected target in the current frame:

[0061] ;

[0062] Among them, is the detection position of the target in the current frame, is the predicted position of the target generated by the trajectory prediction network. If the distance exceeds the set threshold, it is considered that the target is far from the predicted trajectory, and this matching pair is excluded. For the target pairs that pass the position screening, further calculate the cosine distance of the appearance features:

[0063] ;

[0064] For multiple possible matching results, adopt the Hungarian algorithm to generate the optimal matching combination according to the cosine distance to complete the matching of moving targets.

[0065] Step S403: After the matching is completed, combine the matching results to update the tracking trajectory of each target. Since the confidence of the detection box in the satellite remote sensing video is usually higher than that of the prediction box, a weighted strategy is introduced when updating the target position:

[0066] ;

[0067] Among them, and respectively represent the coordinate positions of the detection box and the prediction box, and respectively represent the position weights of the detection box and the prediction box, and the weight of the detection box is increased to improve the accuracy of position update. Then, use the airport road node map to correct the trajectory, match the starting point and the ending point with the nearest road nodes, and then use the Dijkstra algorithm to calculate the shortest path between two points in the airport road node map to form a tracking trajectory that conforms to the roads in the airport and avoid the target trajectory from appearing in abnormal areas (such as non - taxiways). The road node map contains the static information of the roads, and the obtained tracking trajectory is corrected by means of hard constraints to make it conform to the actual movement of the aircraft.

[0068] Through the above steps, the present invention achieves high-precision aircraft target tracking on the satellite constellation imaging dataset, and optimizes the tracking path by introducing airport road constraints, effectively improving the tracking accuracy and the accuracy of trajectory drawing in the case of a large time difference between adjacent data.

[0069] As Figure 4 shown in the target tracking effect diagram after satellite image processing, the present invention can achieve multi-target tracking of airport scenes from the satellite perspective. The figure shows the tracking frame of the target's position in the current frame, as well as the position punctuation marks and trajectories of the moving target in multiple past frames. The trajectory line of the moving target and the color of the position dots in the past frames are the same as its tracking frame. It can be seen that the target in the figure undergoes long-distance non-linear motion in adjacent frames. The method of the present invention can correctly match it and correct the trajectory in combination with the road network to obtain a historical motion trajectory that conforms to the road.

[0070] The technical advantages of the present invention are as follows:

[0071] 1. Combine trajectory prediction to implement a two-stage tracking network (the two-stage tracking network refers to the two stages of detection and then tracking, which is a tracking network based on detection (Tracking based on detection, TBD)), improve the performance of continuous target tracking, and reduce the tracking error caused by the time interval.

[0072] 2. Propose an improved multi-level matching algorithm (S401 and S402 correspond to the specific work of multi-level matching), which can effectively reduce the influence of target displacement caused by a large time difference between adjacent image frames and optimize the multi-target tracking effect.

[0073] 3. By introducing a road node map, avoid the situation where the tracking trajectory is linear and does not conform to the airport runway in long-time-difference data, and effectively improve the accuracy of airport dynamic activity monitoring.

[0074] The present invention introduces airport road constraints in the cross-satellite imaging dataset and uses the airport road information in different satellite imaging data to optimize the aircraft target tracking path. Compared with traditional methods, the present invention can more accurately limit the trajectory range of the aircraft target, make the tracking path of the target more reasonable, and effectively improve the target matching accuracy and trajectory continuity in complex environments.

[0075] The airport road information not only provides clear spatial constraints but also naturally contains trajectory information, enabling the present method to still achieve high-precision target tracking in satellite image data with low time resolution.

[0076] The present invention constructs an LSTM trajectory prediction model through deep learning, and combines airport road constraint information to construct a multi-modal trajectory prediction method. Compared with the traditional Kalman filter, this method has higher stability and adaptability in long-term span prediction.

[0077] Trajectory prediction combined with road constraints can effectively reduce prediction errors and improve the rationality of the tracking trajectory when the aircraft target enters a complex airport environment (such as taxiway intersections or takeoff and landing path changes).

[0078] In the target matching process, the present invention adopts a two-layer optimization strategy for stationary and moving targets. For stationary targets, it combines IoU matching and appearance feature similarity for matching to reduce mis-matching caused by aircraft staying or moving briefly.

[0079] For moving targets, it adopts a matching mechanism based on LSTM trajectory prediction and combines airport road constraints for correction to improve the robustness of matching and avoid target loss due to a long time interval.

[0080] Relying on the collaborative imaging ability of the low-earth orbit satellite constellation, this method can perform long-term (the imaging surveillance ability of the low-earth orbit satellite constellation cannot reach the high frequency of video data. The present invention is more for realizing long-term regional surveillance to compensate for the short duration of satellite staring video data) and multi-angle imaging surveillance of the airport area. By using trajectory prediction and road constraint optimization matching between multiple imaging data, it can effectively reduce the loss of targets caused by too long inter-frame time spans.

[0081] This method is applicable to large-scale airport target monitoring and can provide accurate dynamic tracking data support for airport activity analysis, flight trajectory monitoring, and airspace safety management.

[0082] Any process or method description in the flowchart of the present invention or described in other ways herein can be understood as representing a module, segment, or part of code including one or more executable instructions for implementing a specific logical function or process, which can be implemented in any computer-readable medium for an instruction execution system, apparatus, or device. The computer-readable medium can be any medium including storage, communication, propagation, or transmission of a program for use by an instruction execution system, apparatus, or device, including read-only memory, magnetic disks, or optical discs, etc.

[0083] In the description of this specification, the descriptions referring to terms such as "embodiment", "example", etc. mean that the specific features, structures, materials or characteristics described in connection with that embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. In addition, those skilled in the art can combine or combine different embodiments or examples described in this specification and the features therein without contradiction.

[0084] Although the above has shown and described embodiments of the present invention, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can perform update operations such as changes, modifications, substitutions, and variations on the above embodiments within the scope of the present invention.

Claims

1. A cross-satellite airport aircraft target tracking method combined with road constraints, characterized in that: The method comprises the following steps: Step S101: Using a target detector to detect aircraft targets in the airport image collected by the satellite constellation, obtaining detection frame data of the aircraft targets in the scene, and calculating the time difference between the data collected by adjacent satellites in the constellation; Step S102: Establishing a trajectory prediction model; using the detection box data obtained in step S101, a re-identification network and a multimodal trajectory prediction network are used to extract the appearance and motion information of the aircraft targets in the airport area; Step S103: Using an optimized matching algorithm to achieve cross-frame target matching, and combining airport road constraints to make rational corrections to the matched trajectory, and draw the final aircraft target motion trajectory; In step S102, a composite loss function is used to supervise the trajectory prediction network by combining the absolute distance of the final prediction point and the matching distance of the entire trajectory; End point loss Defined as: in, and are the ground truth value of the sample and the predicted final position, respectively, is the number of predicted trajectories. The average absolute distance loss of the end points of multiple predicted trajectories is obtained by calculating the absolute distance between the predicted position and the true position. Overall trajectory loss The mean absolute error L1 and mean squared error L2 losses are combined to explain the error at each time step in the trajectory prediction process: in, is the number of time steps of the input trajectory, is the number of time steps for predicting the trajectory, and are the weights of L1 and L2 losses respectively; represents the end point of the aircraft trajectory predicted by LSTM at time t for the i-th trajectory, Indicates the end point of the real aircraft's movement at that moment; Adding weights to the comprehensive loss function To the end loss, to emphasize its importance: This loss is used to supervise the trajectory prediction network. As a measurement indicator, when the value is large, the network prediction results are inaccurate. During the training process, the parameters are gradually adjusted to reduce the loss. Ultimately, the trajectory prediction network can make the prediction results closer to the true value.

2. The cross-satellite airport aircraft target tracking method combined with road constraints according to claim 1 is characterized in that: Step S101 specifically includes: Step S201: Use the object detector to identify the aircraft in the field of view, generate a rotated bounding box and crop the image data, and combine the channel and spatial attention modules to enhance the features extracted by the backbone network.

3. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 1, characterized in that: Step S101 further includes: Step S202: calculating and providing the imaging time difference of adjacent satellites in the satellite constellation according to the imaging time information contained in the image.

4. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 2, characterized in that: The target detector is the YOLOv7 target detection network.

5. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 1, characterized in that: Step S102 further includes collecting ADS-B data of aircraft at the airport and cleaning the data, ignoring stationary aircraft trajectories.

6. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 1, characterized in that: Step S102 further includes: Step S301: collecting historical aircraft trajectory data through ADS-B and preprocessing the data, including extracting aircraft target trajectories from satellite images, and performing data cleaning, outlier removal, and normalization. Step S302: Use the LSTM-based multimodal trajectory prediction network to predict future trajectories.

7. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 5, characterized in that: In step S302, a multimodal prediction method is used to generate multiple trajectories, and Gaussian noise is added during the training process to simulate the movement changes of the aircraft and improve the adaptability of the model to uncertain movement patterns.

8. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 5, characterized in that: Step S102 further includes: Step S303: Determine the number of prediction steps of the trajectory prediction model based on the calculated time difference, give a candidate range based on the trajectory prediction endpoint, perform cropping, and obtain the cropped target image; Step S304: A lightweight re-identification network is used to extract the appearance features of the target.

9. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 7, characterized in that: The lightweight re-ID network consists of four convolutional blocks and a downsampling module, takes the cropped target image as input, and outputs a 128×1 feature vector.

10. The method for tracking aircraft targets across satellite airports in combination with road constraints according to claim 9, characterized in that: The trajectory is corrected using the airport road node graph, the start and end points are matched with the nearest neighbor road nodes, and then a tracking trajectory that conforms to the roads within the airport is formed.

Images