An aircraft identification system, method, and storage medium

Abstract

The application provides an aircraft identification system, method and storage medium, belonging to the field of aircraft identification, comprising a plurality of satellites, a plurality of ground gateway stations and a cloud server, receiving satellite broadcast signals reflected by an aircraft as original satellite signals, and preprocessing the original satellite signals to obtain preprocessed satellite signals; the ground gateway station is used for receiving the preprocessed satellite signals sent by the plurality of satellites; the static hardware fingerprint feature is extracted from all the preprocessed satellite signals. The application solves the pain points of insufficient ground mobile network coverage, lack of broadcast capability, easy congestion and easy interference, improves the accuracy of the aircraft identification result, solves the problem of mutual amplification of positioning error and feature misalignment, avoids the unexplainable deviation of the model, and also realizes high-precision tracking and identity recognition of the aircraft.

Description

Technical Field

[0001] This invention relates to the field of aircraft identification technology, specifically to an aircraft identification system, method, and storage medium. Background Technology

[0002] The most common method for aircraft operation identification is manual identification. Manual identification (whether the aircraft is identified by humans or the identified aircraft is manned) requires a large number of human resources and is not suitable for scenarios with a large number of aircraft, such as the low-altitude economy. It is mainly used for manned aircraft with extremely high safety requirements. Unmanned / automatic identification can be widely used in unmanned / manned aircraft and is an important safety foundation in the low-altitude economy.

[0003] As shown in Table 1, Table 1 lists the key characteristics and advantages and disadvantages of mainstream UAV operation recognition technologies.

[0004] Table 1 Currently, drones widely use the first and second methods in Table 1 for identification. Wide-area low-altitude scenarios mainly rely on the second method, while the other methods also heavily rely on ground mobile network signal transmission. In addition to being unusable in many areas such as remote areas, sea, and medium and high altitudes, there is also a significant safety hazard of low-altitude network paralysis when the ground network fails or is congested. Summary of the Invention

[0005] The technical problem to be solved by the present invention is to address the shortcomings of the prior art by providing an aircraft identification system, method and storage medium.

[0006] The technical solution of this invention to solve the above-mentioned technical problems is as follows: an aircraft identification system, comprising multiple satellites, multiple ground gateway stations, and a cloud server. The satellite is used to receive satellite broadcast signals reflected by aircraft as raw satellite signals, and to preprocess the raw satellite signals to obtain preprocessed satellite signals; The ground gateway station is used to receive preprocessed satellite signals transmitted by multiple satellites; Static hardware fingerprint features are extracted from all the preprocessed satellite signals; The static hardware fingerprint features and the preprocessed satellite signals corresponding to the ground gateway stations are calculated and analyzed to obtain the propagation delay of the signals corresponding to each ground gateway station, the delay change rate corresponding to each ground gateway station, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. The cloud server is used to receive satellite parameters sent from all of the satellites; The satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays are analyzed for identification to obtain the identification data of the aircraft to be identified. The identification data of the aircraft to be identified is analyzed, and the analysis results are used as the aircraft identification results.

[0007] Another technical solution of the present invention to solve the above-mentioned technical problems is as follows: An aircraft identification method, comprising the following steps: The satellite broadcast signal reflected by the aircraft is received as the raw satellite signal, and the raw satellite signal is preprocessed to obtain the preprocessed satellite signal; Receive preprocessed satellite signals from multiple satellites and extract static hardware fingerprint features from all of the preprocessed satellite signals; The static hardware fingerprint features and the preprocessed satellite signals corresponding to the ground gateway stations are calculated and analyzed to obtain the propagation delay of the signals corresponding to each ground gateway station, the delay change rate corresponding to each ground gateway station, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. Receive satellite parameters transmitted from all the satellites, and perform identity recognition analysis on the satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays to obtain the identity data of the aircraft to be identified; The identification data of the aircraft to be identified is analyzed, and the analysis results are used as the aircraft identification results.

[0008] The beneficial effects of this invention are as follows: Preprocessing the original satellite signals yields preprocessed satellite signals; static hardware fingerprint features are extracted from all preprocessed satellite signals; the static hardware fingerprint features and the preprocessed satellite signals are analyzed to obtain signal propagation delay, delay rate of change, aircraft three-dimensional position, and aircraft three-dimensional velocity; identity analysis of satellite parameters, aircraft three-dimensional position, aircraft three-dimensional velocity, static hardware fingerprint features, signal propagation delay, and delay rate of change yields the aircraft identification data to be identified; the aircraft identification data is then analyzed, and the analysis results are used as the aircraft identification result. This solves the pain points of insufficient terrestrial mobile network coverage, lack of broadcast capabilities, easy congestion, and susceptibility to interference, improves the accuracy of aircraft identification results, solves the problem of mutual amplification between positioning errors and feature inaccuracies, avoids unexplainable biases in the model, and achieves high-precision tracking and identification of aircraft. Attached Figure Description

[0009] Figure 1This is a schematic diagram of a module of an aircraft identification system provided in an embodiment of the present invention; Figure 2 This is a diagram of a three-level collaborative architecture of an aircraft identification system provided in an embodiment of the present invention, consisting of a satellite, a base station, and a cloud. Figure 3 This is a flowchart illustrating an aircraft identification method provided in an embodiment of the present invention. Detailed Implementation

[0010] The principles and features of the present invention are described below with reference to the accompanying drawings. The examples given are only for explaining the present invention and are not intended to limit the scope of the present invention.

[0011] Figure 1 This is a block diagram of an aircraft identification system provided in an embodiment of the present invention.

[0012] like Figure 1 As shown, an aircraft identification system includes multiple satellites, multiple ground gateway stations, and a cloud server. The satellite is used to receive satellite broadcast signals reflected by aircraft as raw satellite signals, and to preprocess the raw satellite signals to obtain preprocessed satellite signals; The ground gateway station is used to receive preprocessed satellite signals transmitted by multiple satellites; Static hardware fingerprint features are extracted from all the preprocessed satellite signals; The static hardware fingerprint features and the preprocessed satellite signals corresponding to the ground gateway stations are calculated and analyzed to obtain the propagation delay of the signals corresponding to each ground gateway station, the delay change rate corresponding to each ground gateway station, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. The cloud server is used to receive satellite parameters sent from all of the satellites; The satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays are analyzed for identification to obtain the identification data of the aircraft to be identified. The identification data of the aircraft to be identified is analyzed, and the analysis results are used as the aircraft identification results.

[0013] It should be understood that the satellites may be low-Earth orbit communication constellations, high-Earth orbit broadcast satellite signals, and the BeiDou satellite navigation system.

[0014] Specifically, the invention prioritizes the use of downlink broadcast signals (such as pilot signals and synchronization signals) (i.e., original satellite signals) from autonomous low-Earth orbit communication constellations (i.e., satellites) as the illuminator of opportunity. This invention is completely autonomous and controllable, does not rely on GPS or other sensitive signals from other countries, and has high security. It can also integrate high-Earth orbit broadcast satellite signals and signals such as B2b and B3I from the BeiDou Navigation Satellite System (BDS) as supplementary illuminators to form multi-band, multi-standard fusion sensing, thereby improving the system's anti-interference capability and positioning accuracy in complex electromagnetic environments.

[0015] In the above embodiments, preprocessed satellite signals are obtained by preprocessing the original satellite signals. Static hardware fingerprint features are extracted from all preprocessed satellite signals. The signal propagation delay, delay change rate, aircraft three-dimensional position, and aircraft three-dimensional velocity are obtained by solving and analyzing the static hardware fingerprint features and preprocessed satellite signals. The identity data of the aircraft to be identified is obtained by identity recognition analysis of the satellite parameters, aircraft three-dimensional position, aircraft three-dimensional velocity, static hardware fingerprint features, signal propagation delay, and delay change rate. The identity data of the aircraft to be identified is then used for identification analysis, and the analysis results are used as the aircraft identification results. This solves the pain points of insufficient coverage of terrestrial mobile networks, lack of broadcast capabilities, easy congestion, and susceptibility to interference. It improves the accuracy of aircraft identification results, solves the problem of mutual amplification of positioning errors and feature inaccuracies, avoids unexplainable biases in the model, and achieves high-precision tracking and identification of aircraft.

[0016] Optionally, as an embodiment of the present invention, the process of preprocessing the original satellite signal to obtain the preprocessed satellite signal in the satellite includes: The original satellite signal is filtered to obtain the filtered satellite signal; The filtered satellite signal is compressed to obtain a preprocessed satellite signal.

[0017] It should be understood that by carrying a lightweight signal preprocessing unit on a low-Earth orbit satellite, the received reflected signal (i.e., the original satellite signal) is initially filtered and compressed, and only the effective signal feature fragments are transmitted back, which greatly reduces the load on the satellite-to-ground transmission link and ground gateway station.

[0018] In the above embodiments, the original satellite signal is preprocessed to obtain the preprocessed satellite signal, which greatly reduces the load on the satellite-to-ground transmission link and ground gateway station.

[0019] Optionally, as an embodiment of the present invention, the process of extracting static hardware fingerprint features from all the preprocessed satellite signals in the ground gateway station includes: Multiple constellation diagram distortions, multiple phase noise spectra, and multiple power amplifier nonlinear characteristics are extracted from each of the preprocessed satellite signals. A static hardware fingerprint is generated based on all the constellation diagram distortions, all the phase noise spectra, and all the power amplifier nonlinear characteristics.

[0020] In the above embodiments, static hardware fingerprint features are extracted from all preprocessed satellite signals, which solves the problems of insufficient coverage of terrestrial mobile networks, lack of broadcast capabilities, easy congestion and easy interference, and improves the accuracy of aircraft identification results.

[0021] Optionally, as an embodiment of the present invention, in the ground gateway station, the process of calculating and analyzing the static hardware fingerprint features and the multiple preprocessed satellite signals corresponding to the multiple ground gateway stations to obtain the multiple signal propagation delays corresponding to each of the ground gateway stations, the multiple delay change rates corresponding to each of the ground gateway stations, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft includes: The time difference is calculated for each of the preprocessed satellite signals corresponding to each of the ground gateway stations to obtain the propagation delay of each of the signals corresponding to each of the ground gateway stations. Frequency difference calculations are performed on multiple preprocessed satellite signals corresponding to each of the ground gateway stations to obtain multiple time delay change rates corresponding to each of the ground gateway stations. A multi-source joint observation model is constructed, wherein the multi-source joint observation model is as follows: , in, For the first Time of the first The first ground signal station The observation vectors corresponding to each satellite For the first Time of the first The first ground signal station The signal propagation delay corresponding to each satellite For the first The three-dimensional position of the aircraft at any given moment. For the first Time of the first The first ground signal station The rate of change of time delay corresponding to each satellite For the first The three-dimensional velocity of the aircraft at any given moment. For the first Time of the first The first ground signal station The signal feature vectors corresponding to each satellite This is a static hardware fingerprint feature. For the first The initial dynamic security fingerprint characteristics at any given moment. For the first The observation noise vector at time step; The multi-source joint observation model is solved using the static hardware fingerprint features, the propagation delays of all the signals, and the rates of change of all the delays to obtain the three-dimensional position and three-dimensional velocity of the aircraft.

[0022] It should be understood that high-performance domestically produced computing clusters are deployed at ground gateway stations to run core passive positioning and fingerprint extraction algorithms. By utilizing data received from multiple gateway stations (i.e., pre-processed satellite signals), time difference of address (TDOA) / frequency difference of address (FDOA) is jointly calculated to achieve high-precision positioning.

[0023] Specifically, when the signal from the i-th satellite (the illumination source) is reflected by the target and received by the j-th ground station, its observation equation not only includes geometric relationships but also embeds identity information, as shown in the following formula: , The parameter logic is explained as follows: These are the signal propagation delay (TDOA basics) and the rate of change of delay (FDOA basics) (i.e., the delay of signal propagation), which are the target position. With speed The function, which is the location observation. A feature vector (i.e., a signal feature vector) is extracted from the physical layer features of the received signal (such as the signal constellation diagram distortion matrix and the phase noise spectrum of a specific frequency band). Crucially, the mathematical expression of this feature is modeled as a static hardware fingerprint. With dynamic security fingerprint Composite functions. For example: Matrix A and B describe the modulation effect of the channel on the fingerprint, which is the identity observation. : Observation noise (i.e., observation noise vector).

[0024] In the above embodiments, static hardware fingerprint features and preprocessed satellite signals are analyzed to obtain signal propagation delay, delay change rate, aircraft three-dimensional position and aircraft three-dimensional velocity. This solves the pain points of insufficient ground mobile network coverage, lack of broadcasting capability, easy congestion and easy interference, improves the accuracy of aircraft identification results, and also solves the problem of mutual amplification of positioning error and feature inaccuracy.

[0025] Optionally, as an embodiment of the present invention, the satellite parameters include a message sequence number and a timestamp; In the cloud server, the process of performing identity recognition analysis on the satellite parameters, the aircraft's three-dimensional position, the aircraft's three-dimensional velocity, the static hardware fingerprint features, all the signal propagation delays, and all the delay change rates to obtain the aircraft's identity data includes: The static hardware fingerprint feature is compared with the preset fingerprint feature in the preset fingerprint feature database. If the comparison is successful, the corresponding device key is extracted from the preset fingerprint feature database. The target dynamic security fingerprint feature is obtained by calculating the device key, the static hardware fingerprint feature, the message sequence number, and the timestamp using the first formula. The first formula is: , in, For the first The target's dynamic security fingerprint characteristics at any given time. It is a Chinese national standard SM4 encryption function. This is a static hardware fingerprint feature. XOR, For the first The timestamp of the moment For splicing, For the first The message sequence number at that moment. For device key; The original joint state vector is obtained by calculating the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, and the target dynamic security fingerprint features using the second equation. The second equation is: , in, For the first The original joint state vector at time t, For transpose, For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first Dynamic security fingerprint characteristics of the target at any given time; The set of all said signal propagation delays is obtained by combining them, and the set of all said delay change rates is obtained by combining them. The updated joint state vector is obtained by calculating the three-dimensional position of the aircraft, the original joint state vector, the three-dimensional velocity of the aircraft, the static hardware fingerprint feature, and the target dynamic security fingerprint feature using the third equation. The third equation is: , in, , in, For the first The joint state vector updated at time step [time]. Let be the motion state evolution function. For the first The original joint state vector at time t, This is the process noise vector. For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first Dynamic security fingerprint characteristics of the target at any given time; The updated multi-source joint observation vector is obtained by calculating the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the signal propagation delay set, the delay change rate set, the static hardware fingerprint features, and the target dynamic security fingerprint features using the fourth formula. The updated joint state vector and the updated multi-source joint observation vector are then used as the identification data of the aircraft to be identified. The fourth formula is: , in, , in, , in, For the first The updated multi-source joint observation vector at each time step. For joint observation function, For the first The original joint state vector at time t, For the first The observation noise vector at time step, For the first The set of signal propagation delays at each moment. For the first The set of rates of change of time delay at any given moment. For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first The target's dynamic security fingerprint characteristics at any given time. For the first Identity observation characteristics at any given moment and All of these are channel modulation matrices.

[0026] It should be understood that in aerospace surveillance scenarios, for non-cooperative targets, under low signal-to-noise ratio and dynamic channel conditions, the traditional serial approach of "first locating, then extracting fingerprints" has a fatal flaw: the positioning error and feature inaccuracy amplify each other, leading to overall performance failure. This invention can solve the problem of synchronously and collaboratively optimizing the accuracy of position estimation and identity authentication at the signal processing level. This invention proposes a joint state estimation framework that constructs the target's motion state, hardware fingerprint, and national cryptographic security features into a unified state vector, and performs Bayesian optimal estimation in a unified observation model.

[0027] Specifically, the joint state vector is defined as follows: , Explanation of parameter logic: : Used to describe the motion state, representing the target's position and velocity vectors in three-dimensional space at time k. This is the target for localization. This describes a static hardware fingerprint, a static feature vector determined by inherent defects in the transmitter hardware (such as power amplifier nonlinearity, local oscillator phase noise, etc.). It is relatively stable throughout the device's lifespan and is the source of traditional RF fingerprints. : Used to describe dynamic security fingerprints, which are dynamic security feature vectors generated in real time by lightweight national cryptographic algorithms (such as SM4). Their generation model is as follows: = , in, This is the device's unique key. Indicates XOR, This indicates concatenation; the vector is bound to the hardware and changes dynamically with time and serial number, thus addressing the vulnerability of static hardware fingerprints being reverse-engineered.

[0028] It should be understood that the following state-space model can be solved using sequence estimation algorithms such as extended Kalman filtering or particle filtering: (State evolution equations, including motion patterns). (Joint observation equation, as shown above). In each iterative estimation, the algorithm utilizes all observation information. (Including latency, frequency shift, and physical layer features), while updating the location ,speed and fingerprints The optimal estimate .

[0029] Specifically, the technical effects of this invention are as follows: Synergistic gain: This model is achieved through This approach strongly couples identity and location information at the observation level. Under low signal-to-noise ratio conditions, it... Preliminary estimates help to more accurately separate and estimate h with Conversely, accurate fingerprint estimation provides additional observation constraints, which can effectively correct for... and The measurement ambiguity in the signal processing improves positioning accuracy. This is an inherent synergy between "positioning" and "authentication" at the signal processing level, which is incomparable to a simple series of modules.

[0030] Enhanced security: Dynamic security fingerprint The introduction of this technology upgrades radio frequency fingerprints from "observable and imitable" hardware features to "verifiable and counterfeit-proof" cryptographic-physical layer hybrid features. Estimating these features as state variables is a concrete manifestation of the deep integration of security mechanisms into the perception signal processing flow, which is not a conventional method in this field.

[0031] In the above embodiments, satellite parameters, aircraft three-dimensional position, aircraft three-dimensional velocity, static hardware fingerprint features, signal propagation delay, and delay change rate are analyzed for identification to obtain the identification data of the aircraft to be identified, thereby improving positioning accuracy and enhancing security.

[0032] Optionally, as an embodiment of the present invention, the process of identifying and analyzing the aircraft identity data to be identified in the cloud server, and using the analysis result as the aircraft identification result, includes: Import multiple aircraft identity training data and the real labels corresponding to each aircraft identity training data, and construct a training model; The training model is analyzed based on all the aircraft identity training data and all the real labels to obtain a compliance identification model; The compliance identification model is used to identify the aircraft identity data to be identified in accordance with regulations, thereby obtaining regulatory instructions. The aircraft identity data to be identified and the regulatory instructions are then used as the aircraft identification result.

[0033] In the above embodiments, the identification and analysis of the aircraft identity data to be identified are performed, and the analysis results are used as the aircraft identification results. This ensures that the model's decisions under any data are in line with the top-level design, realizes precise and explicit intervention in the model optimization path, and solves the risk of federated learning getting out of control in safety-critical areas.

[0034] Optionally, as an embodiment of the present invention, the training model includes a rule-based teacher model and a master federation model, and the aircraft identity training data includes multi-source joint observation training vectors; In the cloud server, the process of performing model analysis on the trained model based on all the aircraft identity training data and all the real labels to obtain a compliance identification model includes: The multi-source joint observation training vector set is obtained by combining all the aforementioned multi-source joint observation training vectors. The rule-based teacher model is used to evaluate each of the multi-source joint observation training vectors and each of the real labels to obtain multiple violation evaluation values ​​corresponding to each of the multi-source joint observation training vectors. The main federation model is used to predict each of the multi-source joint observation training vectors to obtain the three-dimensional position vector of the aircraft corresponding to each of the multi-source joint observation training vectors. The rule-based loss function is obtained by calculating the multi-source joint observation training vector set, all violation evaluation values, all aircraft 3D position vectors, and all real labels using the fifth equation. The fifth equation is: , in, , in, For the rule loss function, For the first A multi-source joint observation training vector, For the first The true labels corresponding to the multi-source joint observation training vectors. For multi-source joint observation training vector set, For the first The weight coefficients corresponding to each spatial rule. For the first The multi-source joint observation training vector is the first The violation assessment value corresponding to each airspace rule. This is the scaling factor. For the first The multi-source joint observation training vector is the first The rule violation degree corresponding to each spatial rule Let be the expected function. The total number of moments. It is a rectified linear unit function. To set the radius of the control zone, For the first Time of the first The three-dimensional position vector of the aircraft corresponding to the multi-source joint observation training vector. The coordinates of the center of the pre-defined control area, For the first The observation noise vector at time step, The distance is Euclidean. Obtain the global model parameters from the previous iteration and the original model parameters of multiple nodes in the current iteration. Then, calculate the global model parameters for the current iteration using the sixth equation, which is: = + , in, , in, For the first Global model parameters for each iteration. For the first Global model parameters for each iteration. For federal learning rate, For multi-source joint observation training vector set, For the first A multi-source joint observation training vector, For the first The node of the first A multi-source joint observation training vector, For the first The iteration number is... The amount of parameter updates for each node. The constraint strength coefficient is the rule constraint coefficient. For gradient operators, For parameters of the federal model, For the rule loss function, For the first The iteration number is... The original model parameters of each node; The training model is updated with parameters based on the global model parameters at the current iteration number to obtain a compliance identification model.

[0035] It should be understood that the core technical problem solved by the present invention is as follows: In a distributed, data-driven federated learning framework, the global model may learn patterns that violate airspace rules (such as intruding into no-fly zones) from local data, resulting in uncontrollable and unreliable model behavior. This invention aims to solve how to transform deterministic and discontinuous regulatory rules into differentiable constraints that can drive the optimization direction of federated learning.

[0036] Specifically, each regulatory rule (For example, "Aircraft are prohibited from entering a circular airspace centered at c with radius r") can be transformed into a differentiable violation function. (i.e., the degree of rule violation), as follows: , Explanation of parameter logic: The model predicts the trajectory of the target at K future time points. The rectified linear unit function outputs a positive value (violation depth) when the predicted location is within a no-fly zone (distance less than r), and 0 otherwise. This function transforms the discrete "violation" judgment into a continuously differentiable numerical value. Time weighting can be used to emphasize compliance at the current moment.

[0037] Specifically, a lightweight rule-based teacher model is trained in the cloud. Its input is the scene context, and its output is a rule violation assessment. Then, through knowledge distillation technology, it... The "understanding" of complex rules is condensed into a rule-aware loss function module. in The parameters of the federated model to be trained are as follows: , Explanation of parameter logic: Training data synthesized to cover various violation scenarios. The evaluation output of the rule-based teacher model for the m-th rule. The predicted output of the main federated model (student), the loss function forces the main model to... The rule violation degree generated by the prediction results Authoritative evaluation of teacher models Maintaining consistency allows rule knowledge to be "distilled" into the design of the loss function.

[0038] It should be understood that during the t-th round of federated learning aggregation, the key steps performed by the cloud server are: 1. Receive local parameter updates from each node: 2. Calculate the global update after injecting the regular gradient: = + . , Explanation of parameter logic: Standard federated aggregation is the first item: weighted average local update; innovative injection is the second item. This is a mandatory, rule-based global gradient correction. Regardless of the local data distribution, this correction term always applies to the global model parameters. To move towards more compliant rules.

[0039] It should be understood that the technical effects of the present invention are as follows: Paradigm shift: This mechanism transforms the application of knowledge distillation from "model compression" to "behavioral norms." It creates a new paradigm where prior rules drive the evolution of AI models, ensuring that the model's decisions, regardless of the data, align with the top-level design.

[0040] Precise and controllable: through differentiable functions and forced gradient injection term This enables precise and explicit intervention in the optimization path of AI models, addressing the risk of federated learning "getting out of control" in safety-critical areas.

[0041] In the above embodiments, the training model is analyzed based on all aircraft identity training data and all real labels to obtain a compliance identification model. This ensures that the model's decisions under any data are in line with the top-level design, achieving precise and explicit intervention in the model optimization path and solving the risk of federated learning getting out of control in safety-critical areas.

[0042] Alternatively, as another embodiment of the present invention, such as Figure 2 As shown, this invention no longer treats satellites as independent communication or navigation nodes, but rather deeply integrates them into the larger framework of the PNT (Positioning, Navigation, and Timing) system. By utilizing autonomous and controllable high and low Earth orbit communication constellations and aerospace resources such as the BeiDou system, an integrated aerospace sensing network with inherent security capabilities is constructed to achieve high-precision tracking and identification of aircraft without cooperative targets.

[0043] Technical solution refinement: (1) Autonomous and diversified sensing signal sources: Primary signal source: The downlink broadcast signals (such as pilot signals and synchronization signals) of the autonomous low-Earth orbit communication constellation are used as the illuminator of opportunity. This approach is completely autonomous and controllable, without relying on GPS or other sensitive signals from other countries, ensuring high security.

[0044] Backup / enhanced signal sources: By integrating high-orbit broadcast satellite signals and BeiDou Navigation Satellite System (BDS) B2b, B3I and other signals as supplementary illumination sources, multi-band and multi-standard fusion sensing is formed, which improves the system's anti-interference capability and positioning accuracy in complex electromagnetic environments.

[0045] (2) "Star-Station-Cloud" collaborative processing architecture: 1. Lightweight onboard processing: A lightweight signal preprocessing unit is carried on the low-orbit satellite to perform preliminary filtering and compression on the received reflected signals, and only the effective signal feature segments are transmitted back, which greatly reduces the load on the satellite-to-ground transmission link and ground gateway station.

[0046] 2. Centralized Processing at Ground Gateway Stations: High-performance domestically produced computing clusters are deployed at ground gateway stations to run core passive positioning and fingerprint extraction algorithms. Data received from multiple gateway stations is used to jointly calculate Time Difference of Occurrence (TDOA) and Frequency Difference of Occurrence (FDOA) to achieve high-precision positioning.

[0047] 3. Cloud-based fingerprint database and authentication: An "Aircraft Radio Frequency Fingerprint Feature Database" is established in the cloud. This database is created when regulatory authorities collect and store initial fingerprints during device network access licensing. The system uses AI to compare the extracted fingerprint features with the data in the database in real time to complete identity authentication.

[0048] Secure RF fingerprint extraction based on national cryptographic algorithms: This is a core innovation in engineering and security that distinguishes it from academic research. We do not simply analyze hardware flaws, but rather introduce lightweight national cryptographic algorithms to perform controllable and verifiable obfuscation and enhancement of certain inherent physical layer characteristics of the terminal's transmitted signals.

[0049] Security Mechanism: During terminal manufacturing, a unique digital certificate is bound to the RF hardware characteristics, generating a "physical layer signature" using a specific national cryptographic algorithm. This signature contains both hardware uniqueness and cryptographic strength, making the fingerprint both identifiable and resistant to forgery. Even if the signal is intercepted, an adversary cannot reverse engineer or imitate the fingerprint.

[0050] Optionally, as another embodiment of the present invention, the cloud platform of the present invention can use knowledge distillation technology to generate a "rule model" and a "security vector" from prior knowledge such as national laws and regulations, airspace control rules, and typical violation patterns, and forcibly and encryptedly inject them into the global federated model. This ensures that the evolved AI model not only learns data features but also deeply embeds regulatory rules, and its decision-making logic is always aligned with national security policies, avoiding the model from "learning incorrectly" or producing unexplainable deviations.

[0051] Optionally, as another embodiment of the present invention, in addition to flight data, the system encourages access to compliant third-party data sources (such as airport radar, ground camera monitoring systems, etc.) to conduct "cross-modal learning" within a federated framework. For example, if a visual model identifies a new drone model, this "knowledge" can be shared with the radio frequency sensing model through federation, and the latter can quickly learn to identify the model at the signal level. This cross-modal intelligent resonance effect enables the sensing capability of the entire national low-altitude monitoring system to grow exponentially across the network.

[0052] Alternatively, as another embodiment of the present invention, the key points of the present invention are as follows: (I) This invention completely breaks away from the traditional framework of "target-cooperative" supervision, creatively upgrading my country's high and low orbit communication constellations and BeiDou and other aerospace systems from "communication and navigation infrastructure" to the core pillars of the "national aerospace security awareness network." Through the "integrated communication, sensing, and computing" architecture and "national cryptographic enhanced radio frequency fingerprinting" technology, we have achieved high-precision and high-confidence tracking and identification of frequency-using equipment even when it is "silent" (does not actively transmit) or even "flies illegally," providing a brand-new, autonomous, and controllable solution to address the major national need for low-altitude security supervision. This technology has irreplaceable strategic value for maintaining national air defense security, combating illegal flights, and ensuring the success of major events.

[0053] (II) The "Star-Station-Cloud" collaborative knowledge federation framework constructed in this invention successfully resolves the inherent contradiction between "data privacy protection" and "national regulatory effectiveness," achieving national-level collective intelligence aggregation and model evolution while absolutely guaranteeing enterprise data sovereignty and citizen privacy. This framework is not only a technological platform but also a core infrastructure for national low-altitude digital governance, providing a crucial "AI brain" for ensuring the rapid development of my country's low-altitude economy under safe and controllable conditions, possessing significant strategic importance and industry-leading value. (III) This invention achieves the fusion of perception and security mechanisms at the signal layer through a location-security fingerprint joint estimation model; and achieves the optimization and unification of data-driven and rule-driven approaches at the algorithm layer through a federated distillation injection mechanism of rule loss. The parameters of the two sets of mathematical models each have their own responsibilities and work together to solve systemic problems that cannot be addressed by a single technology, resulting in a technical effect of "1+1>2", which has outstanding substantive features and significant progress.

[0054] (iv) It pioneered a satellite broadcast identification mechanism, breaking through the coverage limitations of traditional wireless networks and achieving wide-area coverage at the national level and even across continents.

[0055] Optionally, as another embodiment of the present invention, the satellite broadcasting technology of the present invention provides an innovative solution for spacecraft global operational identification that overcomes the limitations of terrestrial networks, especially in addressing pain points such as insufficient terrestrial mobile network coverage, lack of broadcasting capabilities, susceptibility to congestion, and vulnerability to interference. The following systematically analyzes the purpose of satellite broadcasting in the field of UAV identification, starting from the core issues: I. Low-cost, wide-area coverage: Overcoming geographical limitations, high-orbit satellites facilitate wide-area coverage at the national or even intercontinental levels. - Case Application: When cross-border logistics drones (such as SF Express drones) fly over mountainous or ocean areas without 4G / 5G signals, they send their location, identification code and monitoring data to the regulatory platform via satellite link.

[0056] - Marine scientific research drones transmit their flight trajectories back in real time via satellite links, preventing them from escaping regulatory oversight.

[0057] II. High-efficiency broadcast synchronization: One-to-many efficient distribution, low-cost realization of synchronous data reception by massive aircraft. - Dynamic airspace management: The control center broadcasts electronic fence updates (such as the coordinates of temporary no-fly zones) to all drones in a specific airspace via satellite, without the need to establish connections one by one.

[0058] Effect: Command issuance latency is reduced from minutes (cellular network) to seconds, avoiding the risk of mass intrusion and significantly saving network costs.

[0059] - Collision Avoidance: The UAV receives the positions of surrounding aircraft from satellite broadcasts (ADS-B Lite data) to achieve direct collision avoidance between aircraft without the involvement of ground stations, ensuring flight safety.

[0060] III. Excellent congestion resistance: High-capacity broadcast channel As shown in Table 2, Table 2 is a comparison of channel models.

[0061] Table 2 IV. High Security and Emergency Backup - Disaster recovery value: When earthquakes or floods paralyze ground communications, drones can broadcast their location and distress codes via satellite systems.

[0062] Figure 3 This is a flowchart illustrating an aircraft identification method provided in an embodiment of the present invention.

[0063] Alternatively, as another embodiment of the present invention, such as Figure 3 As shown, an aircraft identification method includes the following steps: S1: Receive the satellite broadcast signal reflected by the aircraft as the original satellite signal, and preprocess the original satellite signal to obtain the preprocessed satellite signal; S2: Receive preprocessed satellite signals transmitted by multiple satellites and extract static hardware fingerprint features from all the preprocessed satellite signals; S3: Perform calculation and analysis on the static hardware fingerprint features and the multiple preprocessed satellite signals corresponding to the multiple ground gateway stations to obtain the multiple signal propagation delays corresponding to each of the ground gateway stations, the multiple delay change rates corresponding to each of the ground gateway stations, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. S4: Receive satellite parameters sent from all the satellites, and perform identity recognition analysis on the satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays to obtain the identity data of the aircraft to be identified; S5: Perform identification and analysis on the aircraft identity data to be identified, and use the analysis results as the aircraft identification results.

[0064] Optionally, another embodiment of the present invention provides an aircraft identification system, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the aircraft identification method as described above. This system can be a computer or similar system.

[0065] Optionally, another embodiment of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the aircraft identification method as described above.

[0066] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0067] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the above-described apparatus and unit can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0068] In the several embodiments provided in this application, it should be understood that the disclosed apparatus and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative. For instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed.

[0069] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of the embodiments of the present invention, depending on actual needs.

[0070] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0071] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. This is understood to mean that the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0072] The above are merely specific embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An aircraft identification system, characterized in that, include: Multiple satellites, multiple ground gateway stations, and cloud servers, The satellite is used to receive satellite broadcast signals reflected by aircraft as raw satellite signals, and to preprocess the raw satellite signals to obtain preprocessed satellite signals; The ground gateway station is used to receive preprocessed satellite signals transmitted by multiple satellites; Static hardware fingerprint features are extracted from all the preprocessed satellite signals; The static hardware fingerprint features and the preprocessed satellite signals corresponding to the ground gateway stations are calculated and analyzed to obtain the propagation delay of the signals corresponding to each ground gateway station, the delay change rate corresponding to each ground gateway station, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. The cloud server is used to receive satellite parameters sent from all of the satellites; The satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays are analyzed for identification to obtain the identification data of the aircraft to be identified. The identification data of the aircraft to be identified is analyzed, and the analysis results are used as the aircraft identification results.

2. The aircraft identification system according to claim 1, characterized in that, In the aforementioned satellite, the process of preprocessing the original satellite signal to obtain the preprocessed satellite signal includes: The original satellite signal is filtered to obtain the filtered satellite signal; The filtered satellite signal is compressed to obtain a preprocessed satellite signal.

3. The aircraft identification system according to claim 1, characterized in that, In the ground gateway station, the process of extracting static hardware fingerprint features from all the preprocessed satellite signals includes: Multiple constellation diagram distortions, multiple phase noise spectra, and multiple power amplifier nonlinear characteristics are extracted from each of the preprocessed satellite signals. A static hardware fingerprint is generated based on all the constellation diagram distortions, all the phase noise spectra, and all the power amplifier nonlinear characteristics.

4. The aircraft identification system according to claim 1, characterized in that, In the aforementioned ground gateway station, the process of calculating and analyzing the static hardware fingerprint features and multiple preprocessed satellite signals corresponding to multiple ground gateway stations to obtain multiple signal propagation delays, multiple delay change rates, aircraft three-dimensional position, and aircraft three-dimensional velocity includes: The time difference is calculated for each of the preprocessed satellite signals corresponding to each of the ground gateway stations to obtain the propagation delay of each of the signals corresponding to each of the ground gateway stations. Frequency difference calculations are performed on multiple preprocessed satellite signals corresponding to each of the ground gateway stations to obtain multiple time delay change rates corresponding to each of the ground gateway stations. A multi-source joint observation model is constructed, wherein the multi-source joint observation model is as follows: ， in, For the first Time of the first The first ground signal station The observation vectors corresponding to each satellite For the first Time of the first The first ground signal station The signal propagation delay corresponding to each satellite For the first The three-dimensional position of the aircraft at any given moment. For the first Time of the first The first ground signal station The rate of change of time delay corresponding to each satellite For the first The three-dimensional velocity of the aircraft at any given moment. For the first Time of the first The first ground signal station The signal feature vectors corresponding to each satellite This is a static hardware fingerprint feature. For the first The initial dynamic security fingerprint characteristics at any given moment. For the first The observation noise vector at time step; The multi-source joint observation model is solved using the static hardware fingerprint features, the propagation delays of all the signals, and the rates of change of all the delays to obtain the three-dimensional position and three-dimensional velocity of the aircraft.

5. The aircraft identification system according to claim 4, characterized in that, The satellite parameters include the message sequence number and timestamp; In the cloud server, the process of performing identity recognition analysis on the satellite parameters, the aircraft's three-dimensional position, the aircraft's three-dimensional velocity, the static hardware fingerprint features, all the signal propagation delays, and all the delay change rates to obtain the aircraft's identity data includes: The static hardware fingerprint feature is compared with the preset fingerprint feature in the preset fingerprint feature database. If the comparison is successful, the corresponding device key is extracted from the preset fingerprint feature database. The target dynamic security fingerprint feature is obtained by calculating the device key, the static hardware fingerprint feature, the message sequence number, and the timestamp using the first formula. The first formula is: ， in, For the first The target's dynamic security fingerprint characteristics at any given time. It is a Chinese national standard SM4 encryption function. This is a static hardware fingerprint feature. XOR, For the first The timestamp of the moment For splicing, For the first The message sequence number at that moment. For device key; The original joint state vector is obtained by calculating the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, and the target dynamic security fingerprint features using the second equation. The second equation is: ， in, For the first The original joint state vector at time t, For transpose, For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first Dynamic security fingerprint characteristics of the target at any given time; The set of all said signal propagation delays is obtained by combining them, and the set of all said delay change rates is obtained by combining them. The updated joint state vector is obtained by calculating the three-dimensional position of the aircraft, the original joint state vector, the three-dimensional velocity of the aircraft, the static hardware fingerprint feature, and the target dynamic security fingerprint feature using the third equation. The third equation is: ， in, , in, For the first The joint state vector updated at time step [time]. Let be the motion state evolution function. For the first The original joint state vector at time t, This is the process noise vector. For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first Dynamic security fingerprint characteristics of the target at any given time; The updated multi-source joint observation vector is obtained by calculating the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the signal propagation delay set, the delay change rate set, the static hardware fingerprint features, and the target dynamic security fingerprint features using the fourth formula. The updated joint state vector and the updated multi-source joint observation vector are then used as the identification data of the aircraft to be identified. The fourth formula is: ， in, , in, , in, For the first The updated multi-source joint observation vector at each time step. For joint observation function, For the first The original joint state vector at time t, For the first The observation noise vector at time step, For the first The set of signal propagation delays at each moment. For the first The set of rates of change of time delay at any given moment. For the first The three-dimensional position of the aircraft at any given moment. For the first The three-dimensional velocity of the aircraft at any given moment. This is a static hardware fingerprint feature. For the first The target's dynamic security fingerprint characteristics at any given time. For the first Identity observation characteristics at any given moment and All of these are channel modulation matrices.

6. The aircraft identification system according to claim 1, characterized in that, In the cloud server, the process of identifying and analyzing the aircraft identity data to be identified, and using the analysis results as the aircraft identification result, includes: Import multiple aircraft identity training data and the real labels corresponding to each aircraft identity training data, and construct a training model; The training model is analyzed based on all the aircraft identity training data and all the real labels to obtain a compliance identification model; The compliance identification model is used to identify the aircraft identity data to be identified in accordance with regulations, thereby obtaining regulatory instructions. The aircraft identity data to be identified and the regulatory instructions are then used as the aircraft identification result.

7. The aircraft identification system according to claim 6, characterized in that, The training model includes a rule-based teacher model and a main federation model, and the aircraft identity training data includes multi-source joint observation training vectors. In the cloud server, the process of performing model analysis on the trained model based on all the aircraft identity training data and all the real labels to obtain a compliance identification model includes: The multi-source joint observation training vector set is obtained by combining all the aforementioned multi-source joint observation training vectors. The rule-based teacher model is used to evaluate each of the multi-source joint observation training vectors and each of the real labels to obtain multiple violation evaluation values ​​corresponding to each of the multi-source joint observation training vectors. The main federation model is used to predict each of the multi-source joint observation training vectors to obtain the three-dimensional position vector of the aircraft corresponding to each of the multi-source joint observation training vectors. The rule-based loss function is obtained by calculating the multi-source joint observation training vector set, all violation evaluation values, all aircraft 3D position vectors, and all real labels using the fifth equation. The fifth equation is: ， in, , in, For the rule loss function, For the first A multi-source joint observation training vector, For the first The true labels corresponding to the multi-source joint observation training vectors. For multi-source joint observation training vector set, For the first The weight coefficients corresponding to each spatial rule. For the first The multi-source joint observation training vector is the first The violation assessment value corresponding to each airspace rule. This is the scaling factor. For the first The multi-source joint observation training vector is the first The rule violation degree corresponding to each spatial rule Let be the expected function. The total number of moments. For rectified linear unit functions, To set the radius of the control zone, For the first Time of the first The three-dimensional position vector of the aircraft corresponding to the multi-source joint observation training vector. The coordinates of the center of the pre-defined control area, For the first The observation noise vector at time step, The distance is Euclidean. Obtain the global model parameters from the previous iteration and the original model parameters of multiple nodes in the current iteration. Then, calculate the global model parameters for the current iteration using the sixth equation, which combines the global model parameters from the previous iteration, the rule loss function, and the original model parameters of multiple nodes in the current iteration. The sixth equation is: = + ， in, , in, For the first Global model parameters for each iteration. For the first Global model parameters for each iteration. For federal learning rate, For multi-source joint observation training vector set, For the first A multi-source joint observation training vector, For the first The node of the first A multi-source joint observation training vector, For the first The iteration number is... The amount of parameter updates for each node. The constraint strength coefficient is the rule constraint coefficient. For gradient operators, For parameters of the federal model, For the rule loss function, For the first The iteration number is... The original model parameters of each node; The training model is updated with parameters based on the global model parameters at the current iteration number to obtain a compliance identification model.

8. An aircraft identification method, characterized in that, Includes the following steps: The satellite broadcast signal reflected by the aircraft is received as the raw satellite signal, and the raw satellite signal is preprocessed to obtain the preprocessed satellite signal; Receive preprocessed satellite signals from multiple satellites and extract static hardware fingerprint features from all of the preprocessed satellite signals; The static hardware fingerprint features and the preprocessed satellite signals corresponding to the ground gateway stations are calculated and analyzed to obtain the propagation delay of the signals corresponding to each ground gateway station, the delay change rate corresponding to each ground gateway station, the three-dimensional position of the aircraft, and the three-dimensional velocity of the aircraft. Receive satellite parameters transmitted from all the satellites, and perform identity recognition analysis on the satellite parameters, the three-dimensional position of the aircraft, the three-dimensional velocity of the aircraft, the static hardware fingerprint features, the propagation delay of all the signals, and the rate of change of all the delays to obtain the identity data of the aircraft to be identified; The identification data of the aircraft to be identified is analyzed, and the analysis results are used as the aircraft identification results.

9. An aircraft identification device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the aircraft identification method as described in claim 8.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the aircraft identification method as described in claim 8.

Images